M9AXF0CZ

1Download document: http://vializer.com/8tSg
2
3commit 890da9cf098364b11a7f7f5c22fa652531624d03
4Author: Linus Torvalds <torvalds@linux-foundation.org>
5Date:   Thu Nov 2 14:06:32 2017 -0700
6
7    Revert "x86: do not use cpufreq_quick_get() for /proc/cpuinfo "cpu MHz""
8
9    This reverts commit 51204e0639c49ada02fd823782ad673b6326d748.
10
11    There wasn't really any good reason for it, and people are complaining
12    (rightly) that it broke existing practice.
13
14    Cc: Len Brown <len.brown@intel.com>
15    Cc: Thomas Gleixner <tglx@linutronix.de>
16    Cc: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
17    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
18commit 80c094a47dd4ea63375e3f60b5e076064f16e857
19Author: Linus Torvalds <torvalds@linux-foundation.org>
20Date:   Thu Oct 26 19:35:35 2017 +0200
21
22    Revert "apparmor: add base infastructure for socket mediation"
23
24    This reverts commit 651e28c5537abb39076d3949fb7618536f1d242e.
25
26    This caused a regression:
27     "The specific problem is that dnsmasq refuses to start on openSUSE Leap
28      42.2.  The specific cause is that and attempt to open a PF_LOCAL socket
29      gets EACCES.  This means that networking doesn't function on a system
30      with a 4.14-rc2 system."
31
32    Sadly, the developers involved seemed to be in denial for several weeks
33    about this, delaying the revert.  This has not been a good release for
34    the security subsystem, and this area needs to change development
35    practices.
36
37    Reported-and-bisected-by: James Bottomley <James.Bottomley@hansenpartnership.com>
38    Tracked-by: Thorsten Leemhuis <regressions@leemhuis.info>
39    Cc: John Johansen <john.johansen@canonical.com>
40    Cc: Vlastimil Babka <vbabka@suse.cz>
41    Cc: Seth Arnold <seth.arnold@canonical.com>
42    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
43commit e49aa15ef6c179f69e5578a271801f31a09e9a3f
44Author: Linus Torvalds <torvalds@linux-foundation.org>
45Date:   Thu Sep 28 13:20:32 2017 -0700
46
47    Revert "Bluetooth: Add option for disabling legacy ioctl interfaces"
48
49    This reverts commit dbbccdc4ced015cdd4051299bd87fbe0254ad351.
50
51    It turns out that the "legacy" users aren't so legacy at all, and that
52    turning off the legacy ioctl will break the current Qt bluetooth stack
53    for bluetooth LE devices that were released just a couple of months ago.
54
55    So it's simply not true that this was a legacy interface that hasn't
56    been needed and is only limited to old legacy BT devices.  Because I
57    actually read Kconfig help messages, and actively try to turn off
58    features that I don't need, I turned the option off.
59
60    Then I spent _way_ too much time debugging BLE issues until I realized
61    that it wasn't the Qt and subsurface development that had broken one of
62    my dive computer BLE downloads, but simply my broken kernel config.
63
64    Maybe in a decade it will be true that this is a legacy interface.  And
65    maybe with a better help-text and correct dependencies, this kind of
66    legacy removal might be acceptable.  But as things are right now both
67    the commit message and the Kconfig help text were misleading, and the
68    Kconfig option had the wrong dependenencies.
69
70    There's no reason to keep that broken Kconfig option in the tree.
71
72    Cc: Marcel Holtmann <marcel@holtmann.org>
73    Cc: Johan Hedberg <johan.hedberg@intel.com>
74    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
75commit f007cad159e99fa2acd3b2e9364fbb32ad28b971
76Author: Linus Torvalds <torvalds@linux-foundation.org>
77Date:   Sun Sep 10 21:19:06 2017 -0700
78
79    Revert "firmware: add sanity check on shutdown/suspend"
80
81    This reverts commit 81f95076281fdd3bc382e004ba1bce8e82fccbce.
82
83    It causes random failures of firmware loading at resume time (well,
84    random for me, it seems to be more reliable for others) because the
85    firmware disabling is not actually synchronous with any particular
86    resume event, and at least the btusb driver that uses a workqueue to
87    load the firmware at resume seems to occasionally hit the "firmware
88    loading is disabled" logic because the firmware loader hasn't gotten the
89    resume event yet.
90
91    Some kind of sanity check for not trying to load firmware when it's not
92    possible might be a good thing, but this commit was not it.
93
94    Greg seems to have silently suffered the same issue, and pointed to the
95    likely culprit, and Gabriel C verified the revert fixed it for him too.
96
97    Reported-by: Linus Torvalds <torvalds@linux-foundation.org>
98    Pointed-at-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
99    Tested-by: Gabriel C <nix.or.die@gmail.com>
100    Cc: Luis R. Rodriguez <mcgrof@kernel.org>
101    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
102commit 785373b4c38719f4af6775845df6be1dfaea120f
103Author: Linus Torvalds <torvalds@linux-foundation.org>
104Date:   Tue Aug 29 09:11:06 2017 -0700
105
106    Revert "rmap: do not call mmu_notifier_invalidate_page() under ptl"
107
108    This reverts commit aac2fea94f7a3df8ad1eeb477eb2643f81fd5393.
109
110    It turns out that that patch was complete and utter garbage, and broke
111    KVM, resulting in odd oopses.
112
113    Quoting Andrea Arcangeli:
114     "The aforementioned commit has 3 bugs.
115
116      1) mmu_notifier_invalidate_range cannot be used in replacement of
117         mmu_notifier_invalidate_range_start/end.
118
119         For KVM mmu_notifier_invalidate_range is a noop and rightfully so.
120
121         A MMU notifier implementation has to implement either
122         ->invalidate_range method or the invalidate_range_start/end
123         methods, not both. And if you implement invalidate_range_start/end
124         like KVM is forced to do, calling mmu_notifier_invalidate_range in
125         common code is a noop for KVM.
126
127         For those MMU notifiers that can get away only implementing
128         ->invalidate_range, the ->invalidate_range is implicitly called by
129         mmu_notifier_invalidate_range_end(). And only those secondary MMUs
130         that share the same pagetable with the primary MMU (like AMD
131         iommuv2) can get away only implementing ->invalidate_range.
132
133         So all cases (THP on/off) are broken right now.
134
135         To fix this is enough to replace mmu_notifier_invalidate_range with
136         mmu_notifier_invalidate_range_start;mmu_notifier_invalidate_range_end.
137         Either that or call multiple mmu_notifier_invalidate_page like
138         before.
139
140      2) address + (1UL << compound_order(page) is buggy, it should be
141         PAGE_SIZE << compound_order(page), it's bytes not pages, 2M not
142         512.
143
144      3) The whole invalidate_range thing was an attempt to call a single
145         invalidate while walking multiple 4k ptes that maps the same THP
146         (after a pmd virtual split without physical compound page THP
147         split).
148
149         It's unclear if the rmap_walk will always provide an address that
150         is 2M aligned as parameter to try_to_unmap_one, in presence of THP.
151         I think it needs also an address &= (PAGE_SIZE <<
152         compound_order(page)) - 1 to be safe"
153
154    In general, we should stop making excuses for horrible MMU notifier
155    users.  It's much more important that the core VM is sane and safe, than
156    letting MMU notifiers sleep.
157
158    So if some MMU notifier is sleeping under a spinlock, we need to fix the
159    notifier, not try to make excuses for that garbage in the core VM.
160
161    Reported-and-tested-by: Bernhard Held <berny156@gmx.de>
162    Reported-and-tested-by: Adam Borowski <kilobyte@angband.pl>
163    Cc: Andrea Arcangeli <aarcange@redhat.com>
164    Cc: Radim Krčmář <rkrcmar@redhat.com>
165    Cc: Wanpeng Li <kernellwp@gmail.com>
166    Cc: Paolo Bonzini <pbonzini@redhat.com>
167    Cc: Takashi Iwai <tiwai@suse.de>
168    Cc: Nadav Amit <nadav.amit@gmail.com>
169    Cc: Mike Galbraith <efault@gmx.de>
170    Cc: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
171    Cc: Jérôme Glisse <jglisse@redhat.com>
172    Cc: axie <axie@amd.com>
173    Cc: Andrew Morton <akpm@linux-foundation.org>
174    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
175commit 143c97cc652949893c8056c679012f0aeccb80e5
176Author: Linus Torvalds <torvalds@linux-foundation.org>
177Date:   Wed Aug 23 18:16:11 2017 -0700
178
179    Revert "pty: fix the cached path of the pty slave file descriptor in the master"
180
181    This reverts commit c8c03f1858331e85d397bacccd34ef409aae993c.
182
183    It turns out that while fixing the ptmx file descriptor to have the
184    correct 'struct path' to the associated slave pty is a really good
185    thing, it breaks some user space tools for a very annoying reason.
186
187    The problem is that /dev/ptmx and its associated slave pty (/dev/pts/X)
188    are on different mounts.  That was what caused us to have the wrong path
189    in the first place (we would mix up the vfsmount of the 'ptmx' node,
190    with the dentry of the pty slave node), but it also means that now while
191    we use the right vfsmount, having the pty master open also keeps the pts
192    mount busy.
193
194    And it turn sout that that makes 'pbuilder' very unhappy, as noted by
195    Stefan Lippers-Hollmann:
196
197     "This patch introduces a regression for me when using pbuilder
198      0.228.7[2] (a helper to build Debian packages in a chroot and to
199      create and update its chroots) when trying to umount /dev/ptmx (inside
200      the chroot) on Debian/ unstable (full log and pbuilder configuration
201      file[3] attached).
202
203      [...]
204      Setting up build-essential (12.3) ...
205      Processing triggers for libc-bin (2.24-15) ...
206      I: unmounting dev/ptmx filesystem
207      W: Could not unmount dev/ptmx: umount: /var/cache/pbuilder/build/1340/dev/ptmx: target is busy
208              (In some cases useful info about processes that
209               use the device is found by lsof(8) or fuser(1).)"
210
211    apparently pbuilder tries to unmount the /dev/pts filesystem while still
212    holding at least one master node open, which is arguably not very nice,
213    but we don't break user space even when fixing other bugs.
214
215    So this commit has to be reverted.
216
217    I'll try to figure out a way to avoid caching the path to the slave pty
218    in the master pty.  The only thing that actually wants that slave pty
219    path is the "TIOCGPTPEER" ioctl, and I think we could just recreate the
220    path at that time.
221
222    Reported-by: Stefan Lippers-Hollmann <s.l-h@gmx.de>
223    Cc: Eric W Biederman <ebiederm@xmission.com>
224    Cc: Christian Brauner <christian.brauner@canonical.com>
225    Cc: Al Viro <viro@zeniv.linux.org.uk>
226    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
227commit 558e8e27e73f53f8a512485be538b07115fe5f3c
228Author: Linus Torvalds <torvalds@linux-foundation.org>
229Date:   Thu Feb 16 12:19:18 2017 -0800
230
231    Revert "nohz: Fix collision between tick and other hrtimers"
232
233    This reverts commit 24b91e360ef521a2808771633d76ebc68bd5604b and commit
234    7bdb59f1ad47 ("tick/nohz: Fix possible missing clock reprog after tick
235    soft restart") that depends on it,
236
237    Pavel reports that it causes occasional boot hangs for him that seem to
238    depend on just how the machine was booted.  In particular, his machine
239    hangs at around the PCI fixups of the EHCI USB host controller, but only
240    hangs from cold boot, not from a warm boot.
241
242    Thomas Gleixner suspecs it's a CPU hotplug interaction, particularly
243    since Pavel also saw suspend/resume issues that seem to be related.
244    We're reverting for now while trying to figure out the root cause.
245
246    Reported-bisected-and-tested-by: Pavel Machek <pavel@ucw.cz>
247    Acked-by: Frederic Weisbecker <fweisbec@gmail.com>
248    Cc: Wanpeng Li <wanpeng.li@hotmail.com>
249    Cc: Peter Zijlstra <peterz@infradead.org>
250    Cc: Rik van Riel <riel@redhat.com>
251    Cc: Thomas Gleixner <tglx@linutronix.de>
252    Cc: stable@kernel.org  # reverted commits were marked for stable
253    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
254commit d966564fcdc19e13eb6ba1fbe6b8101070339c3d
255Author: Linus Torvalds <torvalds@linux-foundation.org>
256Date:   Wed Feb 8 18:08:29 2017 -0800
257
258    Revert "x86/ioapic: Restore IO-APIC irq_chip retrigger callback"
259
260    This reverts commit 020eb3daaba2857b32c4cf4c82f503d6a00a67de.
261
262    Gabriel C reports that it causes his machine to not boot, and we haven't
263    tracked down the reason for it yet.  Since the bug it fixes has been
264    around for a longish time, we're better off reverting the fix for now.
265
266    Gabriel says:
267     "It hangs early and freezes with a lot RCU warnings.
268
269      I bisected it down to :
270
271      > Ruslan Ruslichenko (1):
272      >       x86/ioapic: Restore IO-APIC irq_chip retrigger callback
273
274      Reverting this one fixes the problem for me..
275
276      The box is a PRIMERGY TX200 S5 , 2 socket , 2 x E5520 CPU(s) installed"
277
278    and Ruslan and Thomas are currently stumped.
279
280    Reported-and-bisected-by: Gabriel C <nix.or.die@gmail.com>
281    Cc: Ruslan Ruslichenko <rruslich@cisco.com>
282    Cc: Thomas Gleixner <tglx@linutronix.de>
283    Cc: stable@kernel.org   # for the backport of the original commit
284    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
285commit cdb98c2698b4af287925abcba4d77d92af82a0c3
286Author: Linus Torvalds <torvalds@linux-foundation.org>
287Date:   Tue Dec 13 19:53:37 2016 -0800
288
289    Revert "nvme: add support for the Write Zeroes command"
290
291    This reverts commit 6d31e3ba232ea22458b2f36b6d3f2f9f11bf3fa4.
292
293    This causes bootup problems for me both on my laptop and my desktop.
294    What they have in common is that they have NVMe disks with dm-crypt, but
295    it's not the same controller, so it's not controller-specific.
296
297    Jens does not see it on his machine (also NVMe), so it's presumably
298    something that triggers just on bootup.  Possibly related to dm-crypt
299    and the fact that I mark my luks volume with "allow-discards" in
300    /etc/crypttab.
301
302    It's 100% repeatable for me, which made it fairly straightforward to
303    bisect the problem to this commit. Small mercies.
304
305    So we don't know what the reason is yet, but the revert is needed to get
306    things going again.
307
308    Acked-by: Jens Axboe <axboe@fb.com>
309    Cc: Chaitanya Kulkarni <chaitanya.kulkarni@hgst.com>
310    Cc: Christoph Hellwig <hch@lst.de>
311    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
312commit 2b41226b39b654a5e20bce5a7332f307fdb9156b
313Author: Linus Torvalds <torvalds@linux-foundation.org>
314Date:   Fri Dec 9 10:41:42 2016 -0800
315
316    Revert "radix tree test suite: fix compilation"
317
318    This reverts commit 53855d10f4567a0577360b6448d52a863929775b.
319
320    It shouldn't have come in yet - it depends on the changes in linux-next
321    that will come in during the next merge window.  As Matthew Wilcox says,
322    the test suite is broken with the current state without the revert.
323
324    Requested-by: Matthew Wilcox <mawilcox@microsoft.com>
325    Cc: Andrew Morton <akpm@linux-foundation.org>
326    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
327commit f27c2f69cc8edc03ea8086f974811b9b45b2f3a5
328Author: Linus Torvalds <torvalds@linux-foundation.org>
329Date:   Wed Dec 7 08:39:00 2016 -0800
330
331    Revert "default exported asm symbols to zero"
332
333    This reverts commit 8ab2ae655bfe384335c5b6b0d6041e0ddce26b00.
334
335    I loved that commit because of how it explained what the problem with
336    newer versions of binutils were, but the actual patch itself turns out
337    to not work very well.
338
339    It has two problems:
340
341     - a zero CRC value isn't actually right.  It happens to work for the
342       case where both sides of the equation fail at giving the symbol a
343       crc, but there are cases where the users of the exported symbol get
344       the right crc (due to seeing the C declarations), but the actual
345       exporting itself does not (due to the whole weak asm symbol issue).
346
347       So then the module load fails after all - we did have a crc for the
348       symbol, but we couldn't match it with the loaded module.
349
350     - it seems that the alpha assembler has special semantics for the
351       '.set' directive, and on alpha it doesn't actually set the value of
352       the specified symbol at all, it is instead used to set various
353       assembly modes (eg ".set noat" and ".set noreorder").
354
355       So using ".set" to set the symbol value would just cause build
356       failures on alpha.
357
358    I'm sure we'll find some other workaround for these issues (hopefully
359    that involves getting rid of modversions entirely some day, but people
360    are also talking about just using smarter tools).  But for now we'll
361    just fall back on commit faaae2a58143 ("Re-enable CONFIG_MODVERSIONS in
362    a slightly weaker form") that just let's a missing crc through.
363
364    Reported-by: Jan Stancek <jstancek@redhat.com>
365    Reported-by: Philip Müller <philm@manjaro.org>
366    Reported-by: Guenter Roeck <linux@roeck-us.net>
367    Cc: Arnd Bergmann <arnd@arndb.de>
368    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
369commit f5c9f9c72395c3291c2e35c905dedae2b98475a4
370Author: Linus Torvalds <torvalds@linux-foundation.org>
371Date:   Mon Nov 14 09:31:52 2016 -0800
372
373    Revert "printk: make reading the kernel log flush pending lines"
374
375    This reverts commit bfd8d3f23b51018388be0411ccbc2d56277fe294.
376
377    It turns out that this flushes things much too aggressiverly, and causes
378    lines to break up when the system logger races with new continuation
379    lines being printed.
380
381    There's a pending patch to make printk() flushing much more
382    straightforward, but it's too invasive for 4.9, so in the meantime let's
383    just not make the system message logging flush continuation lines.
384    They'll be flushed by the final newline anyway.
385
386    Suggested-by: Petr Mladek <pmladek@suse.com>
387    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
388commit 38f7bd94a97b542de86a2be9229289717e33a7a4
389Author: Linus Torvalds <torvalds@linux-foundation.org>
390Date:   Thu Sep 1 14:56:49 2016 -0700
391
392    Revert "af_unix: Fix splice-bind deadlock"
393
394    This reverts commit c845acb324aa85a39650a14e7696982ceea75dc1.
395
396    It turns out that it just replaces one deadlock with another one: we can
397    still get the wrong lock ordering with the readlock due to overlayfs
398    calling back into the filesystem layer and still taking the vfs locks
399    after the readlock.
400
401    The proper solution ends up being to just split the readlock into two
402    pieces: the bind lock (taken *outside* the vfs locks) and the IO lock
403    (taken *inside* the filesystem locks).  The two locks are independent
404    anyway.
405
406    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
407    Reviewed-by: Shmulik Ladkani <shmulik.ladkani@gmail.com>
408    Signed-off-by: David S. Miller <davem@davemloft.net>
409commit a0cba2179ea4c1820fce2ee046b6ed90ecc56196
410Author: Linus Torvalds <torvalds@linux-foundation.org>
411Date:   Tue Aug 9 10:48:18 2016 -0700
412
413    Revert "printk: create pr_<level> functions"
414
415    This reverts commit 874f9c7da9a4acbc1b9e12ca722579fb50e4d142.
416
417    Geert Uytterhoeven reports:
418     "This change seems to have an (unintendent?) side-effect.
419
420      Before, pr_*() calls without a trailing newline characters would be
421      printed with a newline character appended, both on the console and in
422      the output of the dmesg command.
423
424      After this commit, no new line character is appended, and the output
425      of the next pr_*() call of the same type may be appended, like in:
426
427        - Truncating RAM at 0x0000000040000000-0x00000000c0000000 to -0x0000000070000000
428        - Ignoring RAM at 0x0000000200000000-0x0000000240000000 (!CONFIG_HIGHMEM)
429        + Truncating RAM at 0x0000000040000000-0x00000000c0000000 to -0x0000000070000000Ignoring RAM at 0x0000000200000000-0x0000000240000000 (!CONFIG_HIGHMEM)"
430
431    Joe Perches says:
432     "No, that is not intentional.
433
434      The newline handling code inside vprintk_emit is a bit involved and
435      for now I suggest a revert until this has all the same behavior as
436      earlier"
437
438    Reported-by: Geert Uytterhoeven <geert@linux-m68k.org>
439    Requested-by: Joe Perches <joe@perches.com>
440    Cc: Andrew Morton <akpm@linux-foundation.org>
441    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
442commit 96b585267f552d4b6a28ea8bd75e5ed03deb6e71
443Author: Linus Torvalds <torvalds@linux-foundation.org>
444Date:   Wed Aug 3 22:20:22 2016 -0400
445
446    Revert "ACPI / hotplug / PCI: Runtime resume bridge before rescan"
447
448    This reverts commit 16468c783cb4cf72475dcda23fabecb4a4bb0e17.
449
450    Bisection showed that it was the root cause for a resume hang on a
451    bog-standard all-Intel laptop (Sony Vaio Pro 11), and reverting fixes
452    the hang.
453
454    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
455commit 20d00ee829428ea6aab77fa3acca048a6f57d3bc
456Author: Linus Torvalds <torvalds@linux-foundation.org>
457Date:   Fri Jul 29 12:17:52 2016 -0700
458
459    Revert "vfs: add lookup_hash() helper"
460
461    This reverts commit 3c9fe8cdff1b889a059a30d22f130372f2b3885f.
462
463    As Miklos points out in commit c1b2cc1a765a, the "lookup_hash()" helper
464    is now unused, and in fact, with the hash salting changes, since the
465    hash of a dentry name now depends on the directory dentry it is in, the
466    helper function isn't even really likely to be useful.
467
468    So rather than keep it around in case somebody else might end up finding
469    a use for it, let's just remove the helper and not trick people into
470    thinking it might be a useful thing.
471
472    For example, I had obviously completely missed how the helper didn't
473    follow the normal dentry hashing patterns, and how the hash salting
474    patch broke overlayfs.  Things would quietly build and look sane, but
475    not work.
476
477    Suggested-by: Miklos Szeredi <mszeredi@redhat.com>
478    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
479commit 4046d6e81f33b7ef50d6668b78076d54c5e066b6
480Author: Linus Torvalds <torvalds@linux-foundation.org>
481Date:   Thu Apr 14 11:18:57 2016 -0700
482
483    Revert "x86: remove the kernel code/data/bss resources from /proc/iomem"
484
485    This reverts commit c4004b02f8e5b9ce357a0bb1641756cc86962664.
486
487    Sadly, my hope that nobody would actually use the special kernel entries
488    in /proc/iomem were dashed by kexec.  Which reads /proc/iomem explicitly
489    to find the kernel base address.  Nasty.
490
491    Anyway, that means we can't do the sane and simple thing and just remove
492    the entries, and we'll instead have to mask them out based on permissions.
493
494    Reported-by: Zhengyu Zhang <zhezhang@redhat.com>
495    Reported-by: Dave Young <dyoung@redhat.com>
496    Reported-by: Freeman Zhang <freeman.zhang1992@gmail.com>
497    Reported-by: Emrah Demir <ed@abdsec.com>
498    Reported-by: Baoquan He <bhe@redhat.com>
499    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
500commit 9f2394c9be47a754bae9e4b6d382bdd4d77d0a11
501Author: Linus Torvalds <torvalds@linux-foundation.org>
502Date:   Sun Apr 10 16:52:24 2016 -0700
503
504    Revert "ext4: allow readdir()'s of large empty directories to be interrupted"
505
506    This reverts commit 1028b55bafb7611dda1d8fed2aeca16a436b7dff.
507
508    It's broken: it makes ext4 return an error at an invalid point, causing
509    the readdir wrappers to write the the position of the last successful
510    directory entry into the position field, which means that the next
511    readdir will now return that last successful entry _again_.
512
513    You can only return fatal errors (that terminate the readdir directory
514    walk) from within the filesystem readdir functions, the "normal" errors
515    (that happen when the readdir buffer fills up, for example) happen in
516    the iterorator where we know the position of the actual failing entry.
517
518    I do have a very different patch that does the "signal_pending()"
519    handling inside the iterator function where it is allowable, but while
520    that one passes all the sanity checks, I screwed up something like four
521    times while emailing it out, so I'm not going to commit it today.
522
523    So my track record is not good enough, and the stars will have to align
524    better before that one gets committed.  And it would be good to get some
525    review too, of course, since celestial alignments are always an iffy
526    debugging model.
527
528    IOW, let's just revert the commit that caused the problem for now.
529
530    Reported-by: Greg Thelen <gthelen@google.com>
531    Cc: Theodore Ts'o <tytso@mit.edu>
532    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
533commit 1701f680407c6cc05121566cef5293d33881fe71
534Author: Linus Torvalds <torvalds@linux-foundation.org>
535Date:   Fri Mar 25 09:02:13 2016 -0700
536
537    Revert "ppdev: use new parport device model"
538
539    This reverts commit e7223f18603374d235d8bb0398532323e5f318b9.
540
541    It causes problems when a ppdev tries to register before the parport
542    driver has been registered with the device model. That will trigger the
543
544            BUG_ON(!drv->bus->p);
545
546    at drivers/base/driver.c:153. The call chain is
547
548      kernel_init ->
549        kernel_init_freeable ->
550          do_one_initcall ->
551            ppdev_init ->
552              __parport_register_driver ->
553                driver_register *BOOM*
554
555    Reported-by: kernel test robot <fengguang.wu@intel.com>
556    Reported-by: Ross Zwisler <zwisler@gmail.com>
557    Reported-by: Petr Mladek <pmladek@suse.com>
558    Cc: Sudip Mukherjee <sudipm.mukherjee@gmail.com>
559    Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
560    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
561commit 256faedcfd646161477d47a1a78c32a562d2e845
562Author: Linus Torvalds <torvalds@linux-foundation.org>
563Date:   Mon Mar 7 13:15:09 2016 -0800
564
565    Revert "drm/radeon: call hpd_irq_event on resume"
566
567    This reverts commit dbb17a21c131eca94eb31136eee9a7fe5aff00d9.
568
569    It turns out that commit can cause problems for systems with multiple
570    GPUs, and causes X to hang on at least a HP Pavilion dv7 with hybrid
571    graphics.
572
573    This got noticed originally in 4.4.4, where this patch had already
574    gotten back-ported, but 4.5-rc7 was verified to have the same problem.
575
576    Alexander Deucher says:
577     "It looks like you have a muxed system so I suspect what's happening is
578      that one of the display is being reported as connected for both the
579      IGP and the dGPU and then the desktop environment gets confused or
580      there some sort problem in the detect functions since the mux is not
581      switched to the dGPU.  I don't see an easy fix unless Dave has any
582      ideas.  I'd say just revert for now"
583
584    Reported-by: Jörg-Volker Peetz <jvpeetz@web.de>
585    Acked-by: Alexander Deucher <Alexander.Deucher@amd.com>
586    Cc: Dave Airlie <airlied@gmail.com>
587    Cc: stable@kernel.org  # wherever dbb17a21c131 got back-ported
588    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
589commit 00a3d660cbac05af34cca149cb80fb611e916935
590Author: Linus Torvalds <torvalds@linux-foundation.org>
591Date:   Wed Oct 7 08:32:38 2015 +0100
592
593    Revert "fs: do not prefault sys_write() user buffer pages"
594
595    This reverts commit 998ef75ddb5709bbea0bf1506cd2717348a3c647.
596
597    The commit itself does not appear to be buggy per se, but it is exposing
598    a bug in ext4 (and Ted thinks ext3 too, but we solved that by getting
599    rid of it).  It's too late in the release cycle to really worry about
600    this, even if Dave Hansen has a patch that may actually fix the
601    underlying ext4 problem.  We can (and should) revisit this for the next
602    release.
603
604    The problem is that moving the prefaulting later now exposes a special
605    case with partially successful writes that isn't handled correctly.  And
606    the prefaulting likely isn't normally even that much of a performance
607    issue - it looks like at least one reason Dave saw this in his
608    performance tests is that he also ran them on Skylake that now supports
609    the new SMAP code, which makes the normally very cheap user space
610    prefaulting noticeably more expensive.
611
612    Bisected-and-acked-by: Ted Ts'o <tytso@mit.edu>
613    Analyzed-and-acked-by: Dave Hansen <dave.hansen@linux.intel.com>
614    Cc: Andrew Morton <akpm@linux-foundation.org>
615    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
616commit 0ba13fd19d39b7cb672bcec052bc813389c079a4
617Author: Linus Torvalds <torvalds@linux-foundation.org>
618Date:   Fri Sep 11 13:26:39 2015 -0700
619
620    Revert "writeback: plug writeback at a high level"
621
622    This reverts commit d353d7587d02116b9732d5c06615aed75a4d3a47.
623
624    Doing the block layer plug/unplug inside writeback_sb_inodes() is
625    broken, because that function is actually called with a spinlock held:
626    wb->list_lock, as pointed out by Chris Mason.
627
628    Chris suggested just dropping and re-taking the spinlock around the
629    blk_finish_plug() call (the plgging itself can happen under the
630    spinlock), and that would technically work, but is just disgusting.
631
632    We do something fairly similar - but not quite as disgusting because we
633    at least have a better reason for it - in writeback_single_inode(), so
634    it's not like the caller can depend on the lock being held over the
635    call, but in this case there just isn't any good reason for that
636    "release and re-take the lock" pattern.
637
638    [ In general, we should really strive to avoid the "release and retake"
639      pattern for locks, because in the general case it can easily cause
640      subtle bugs when the caller caches any state around the call that
641      might be invalidated by dropping the lock even just temporarily. ]
642
643    But in this case, the plugging should be easy to just move up to the
644    callers before the spinlock is taken, which should even improve the
645    effectiveness of the plug.  So there is really no good reason to play
646    games with locking here.
647
648    I'll send off a test-patch so that Dave Chinner can verify that that
649    plug movement works.  In the meantime this just reverts the problematic
650    commit and adds a comment to the function so that we hopefully don't
651    make this mistake again.
652
653    Reported-by: Chris Mason <clm@fb.com>
654    Cc: Josef Bacik <jbacik@fb.com>
655    Cc: Dave Chinner <david@fromorbit.com>
656    Cc: Neil Brown <neilb@suse.de>
657    Cc: Jan Kara <jack@suse.cz>
658    Cc: Christoph Hellwig <hch@lst.de>
659    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
660commit ed596cde9425509ec6ce88e19f03e9b13b6f518b
661Author: Linus Torvalds <torvalds@linux-foundation.org>
662Date:   Thu Aug 13 08:25:20 2015 -0700
663
664    Revert x86 sigcontext cleanups
665
666    This reverts commits 9a036b93a344 ("x86/signal/64: Remove 'fs' and 'gs'
667    from sigcontext") and c6f2062935c8 ("x86/signal/64: Fix SS handling for
668    signals delivered to 64-bit programs").
669
670    They were cleanups, but they break dosemu by changing the signal return
671    behavior (and removing 'fs' and 'gs' from the sigcontext struct - while
672    not actually changing any behavior - causes build problems).
673
674    Reported-and-tested-by: Stas Sergeev <stsp@list.ru>
675    Acked-by: Andy Lutomirski <luto@amacapital.net>
676    Cc: Ingo Molnar <mingo@kernel.org>
677    Cc: stable@vger.kernel.org
678    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
679commit d725e66c06ab440032f49ef17e960896d0ec6d49
680Author: Linus Torvalds <torvalds@linux-foundation.org>
681Date:   Tue Jul 21 16:06:53 2015 -0700
682
683    Revert "fsnotify: fix oops in fsnotify_clear_marks_by_group_flags()"
684
685    This reverts commit a2673b6e040663bf16a552f8619e6bde9f4b9acf.
686
687    Kinglong Mee reports a memory leak with that patch, and Jan Kara confirms:
688
689     "Thanks for report! You are right that my patch introduces a race
690      between fsnotify kthread and fsnotify_destroy_group() which can result
691      in leaking inotify event on group destruction.
692
693      I haven't yet decided whether the right fix is not to queue events for
694      dying notification group (as that is pointless anyway) or whether we
695      should just fix the original problem differently...  Whenever I look
696      at fsnotify code mark handling I get lost in the maze of locks, lists,
697      and subtle differences between how different notification systems
698      handle notification marks :( I'll think about it over night"
699
700    and after thinking about it, Jan says:
701
702     "OK, I have looked into the code some more and I found another
703      relatively simple way of fixing the original oops.  It will be IMHO
704      better than trying to fixup this issue which has more potential for
705      breakage.  I'll ask Linus to revert the fsnotify fix he already merged
706      and send a new fix"
707
708    Reported-by: Kinglong Mee <kinglongmee@gmail.com>
709    Requested-by: Jan Kara <jack@suse.cz>
710    Cc: Andrew Morton <akpm@linux-foundation.org>
711    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
712commit 01e2d0627a9a6edb24c37db45db5ecb31e9de808
713Author: Linus Torvalds <torvalds@linux-foundation.org>
714Date:   Sun Jul 12 15:00:20 2015 -0700
715
716    Revert "drm/i915: Use crtc_state->active in primary check_plane func"
717
718    This reverts commit dec4f799d0a4c9edae20512fa60b0a36f3299ca2.
719
720    Jörg Otte reports a NULL pointder dereference due to this commit, as
721    'crtc_state' very much can be NULL:
722
723            crtc_state = state->base.state ?
724                    intel_atomic_get_crtc_state(state->base.state, intel_crtc) : NULL;
725
726    So the change to test 'crtc_state->base.active' cannot possibly be
727    correct as-is.
728
729    There may be some other minimal fix (like just checking crtc_state for
730    NULL), but I'm just reverting it now for the rc2 release, and people
731    like Daniel Vetter who actually know this code will figure out what the
732    right solution is in the longer term.
733
734    Reported-and-bisected-by: Jörg Otte <jrg.otte@gmail.com>
735    Cc: Ander Conselvan de Oliveira <ander.conselvan.de.oliveira@intel.com>
736    Cc: Jani Nikula <jani.nikula@linux.intel.com>
737    Cc: Daniel Vetter <daniel.vetter@intel.com>
738    CC: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
739    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
740commit ebeaa8ddb3663b5c6cfc205605c35116381550c5
741Author: Linus Torvalds <torvalds@linux-foundation.org>
742Date:   Sat Jun 27 09:41:50 2015 -0700
743
744    Revert "jbd2: speedup jbd2_journal_dirty_metadata()"
745
746    This reverts commit 2143c1965a761332ae417b22fd477b636e4f54ec.
747
748    This commit seems to be the cause of the following jbd2 assertion
749    failure:
750
751       ------------[ cut here ]------------
752       kernel BUG at fs/jbd2/transaction.c:1325!
753       invalid opcode: 0000 [#1] SMP
754       Modules linked in: bnep bluetooth fuse ip6t_rpfilter ip6t_REJECT nf_reject_ipv6 nf_conntrack_ipv6 ...
755       CPU: 7 PID: 5509 Comm: gcc Not tainted 4.1.0-10944-g2a298679b411 #1
756       Hardware name:                  /DH87RL, BIOS RLH8710H.86A.0327.2014.0924.1645 09/24/2014
757       task: ffff8803bf866040 ti: ffff880308528000 task.ti: ffff880308528000
758       RIP: jbd2_journal_dirty_metadata+0x237/0x290
759       Call Trace:
760         __ext4_handle_dirty_metadata+0x43/0x1f0
761         ext4_handle_dirty_dirent_node+0xde/0x160
762         ? jbd2_journal_get_write_access+0x36/0x50
763         ext4_delete_entry+0x112/0x160
764         ? __ext4_journal_start_sb+0x52/0xb0
765         ext4_unlink+0xfa/0x260
766         vfs_unlink+0xec/0x190
767         do_unlinkat+0x24a/0x270
768         SyS_unlink+0x11/0x20
769         entry_SYSCALL_64_fastpath+0x12/0x6a
770       ---[ end trace ae033ebde8d080b4 ]---
771
772    which is not easily reproducible (I've seen it just once, and then Ted
773    was able to reproduce it once).  Revert it while Ted and Jan try to
774    figure out what is wrong.
775
776    Cc: Jan Kara <jack@suse.cz>
777    Acked-by: Theodore Ts'o <tytso@mit.edu>
778    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
779commit f3ca10dde49043fbbb055854278b26954b549b36
780Author: Linus Torvalds <torvalds@linux-foundation.org>
781Date:   Wed Apr 22 09:44:36 2015 -0700
782
783    Revert "mm: avoid tail page refcounting on non-THP compound pages"
784
785    This reverts commit 8d63d99a5dfbdb997d12dd3c07b2070ca723db3b.
786
787    It causes in VM mapping refcount errors:
788
789      page:ffffea0010a15040 count:0 mapcount:1 mapping:          (null) index:0x0
790      flags: 0x8000000000008014(referenced|dirty|tail)
791      page dumped because: VM_BUG_ON_PAGE(page_mapcount(page) != 0)
792      ------------[ cut here ]------------
793      kernel BUG at mm/swap.c:134!
794
795    as reported by Borislav Petkov
796
797    Reported-and-tested-by: Borislav Petkov <bp@alien8.de>
798    Cc: Kirill A. Shutemov <kirill@shutemov.name>
799    Cc: Hugh Dickins <hughd@google.com>
800    Cc: Andrea Arcangeli <aarcange@redhat.com>
801    Cc: Andrew Morton <akpm@linux-foundation.org>
802    Cc: linux-mm@kvack.org
803    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
804commit 8f443e2372ba23d51ee365974f54507acd6f69d1
805Author: Linus Torvalds <torvalds@linux-foundation.org>
806Date:   Tue Apr 21 09:17:28 2015 -0700
807
808    Revert "ocfs2: incorrect check for debugfs returns"
809
810    This reverts commit e2ac55b6a8e337fac7cc59c6f452caac92ab5ee6.
811
812    Huang Ying reports that this causes a hang at boot with debugfs disabled.
813
814    It is true that the debugfs error checks are kind of confusing, and this
815    code certainly merits more cleanup and thinking about it, but there's
816    something wrong with the trivial "check not just for NULL, but for error
817    pointers too" patch.
818
819    Yes, with debugfs disabled, we will end up setting the o2hb_debug_dir
820    pointer variable to an error pointer (-ENODEV), and then continue as if
821    everything was fine.  But since debugfs is disabled, all the _users_ of
822    that pointer end up being compiled away, so even though the pointer can
823    not be dereferenced, that's still fine.
824
825    So it's confusing and somewhat questionable, but the "more correct"
826    error checks end up causing more trouble than they fix.
827
828    Reported-by: Huang Ying <ying.huang@intel.com>
829    Acked-by: Andrew Morton <akpm@linux-foundation.org>
830    Acked-by: Chengyu Song <csong84@gatech.edu>
831    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
832commit 12522eeac88165d0db56a6a19ab607addf728995
833Author: Linus Torvalds <torvalds@linux-foundation.org>
834Date:   Sat Apr 11 13:46:07 2015 -0700
835
836    Revert "dmaengine: Add a warning for drivers not using the generic slave caps retrieval"
837
838    This reverts commit ecc19d17868be9c9f8f00ed928791533c420f3e0.
839
840    It added a new warning to try to encourage driver writers to set the
841    device capabities properly, but drivers haven't been updated and in the
842    meantime it just generaters a scary message that users cannot actually
843    do anything about.
844
845    Warnings like these are appropriate if you actually expect to fix the
846    code that causes them.  They are not appropriate for releases.
847
848    Requested-by: Peter Hurley <peter@hurleysoftware.com>
849    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
850commit 8329aa9fff3fca84009e6a444d8d160193643bac
851Author: Linus Torvalds <torvalds@linux-foundation.org>
852Date:   Fri Feb 13 10:26:18 2015 -0800
853
854    Revert "x86/apic: Only disable CPU x2apic mode when necessary"
855
856    This reverts commit 5fcee53ce705d49c766f8a302c7e93bdfc33c124.
857
858    It causes the suspend to fail on at least the Chromebook Pixel, possibly
859    other platforms too.
860
861    Joerg Roedel points out that the logic should probably have been
862
863                    if (max_physical_apicid > 255 ||
864                        !(IS_ENABLED(CONFIG_HYPERVISOR_GUEST) &&
865                          hypervisor_x2apic_available())) {
866
867    instead, but since the code is not in any fast-path, so we can just live
868    without that optimization and just revert to the original code.
869
870    Acked-by: Joerg Roedel <joro@8bytes.org>
871    Acked-by: Jiang Liu <jiang.liu@linux.intel.com>
872    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
873commit 7f4054836d811c650c51f9c93088f8ebd61b0020
874Author: Linus Torvalds <torvalds@linux-foundation.org>
875Date:   Wed Dec 31 12:59:34 2014 -0800
876
877    Revert "Input: atmel_mxt_ts - use deep sleep mode when stopped"
878
879    This reverts commit 9d469d033d135d80742a4e39e6bbb4519dd5eee1.
880
881    It breaks the Chromebook Pixel touchpad (and touchscreen).
882
883    Reported-by: Dirk Hohndel <dirk@hohndel.org>
884    Bisected-by: Linus Torvalds <torvalds@linux-foundation.org>
885    Cc: Nick Dyer <nick.dyer@itdev.co.uk>
886    Cc: Benson Leung <bleung@chromium.org>
887    Cc: Yufeng Shen <miletus@chromium.org>
888    Cc: Dmitry Torokhov <dmitry.torokhov@gmail.com>
889    Cc: stable@vger.kernel.org  # v3.16+
890    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
891commit f3670394c29ff3730638762c1760fd2f624e6d7b
892Author: Linus Torvalds <torvalds@linux-foundation.org>
893Date:   Mon Sep 22 23:05:49 2014 -0700
894
895    Revert "x86/efi: Fixup GOT in all boot code paths"
896
897    This reverts commit 9cb0e394234d244fe5a97e743ec9dd7ddff7e64b.
898
899    It causes my Sony Vaio Pro 11 to immediately reboot at startup.
900
901    Acked-by: Ingo Molnar <mingo@kernel.org>
902    Cc: Peter Anvin <hpa@zytor.com>
903    Cc: Maarten Lankhorst <maarten.lankhorst@canonical.com>
904    Cc: Ard Biesheuvel <ard.biesheuvel@linaro.org>
905    Cc: Matt Fleming <matt.fleming@intel.com>
906    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
907commit 155134fef2b6c7426c3f25ffe84fb3043167c860
908Author: Linus Torvalds <torvalds@linux-foundation.org>
909Date:   Sun Aug 10 21:24:59 2014 -0700
910
911    Revert "proc: Point /proc/{mounts,net} at /proc/thread-self/{mounts,net} instead of /proc/self/{mounts,net}"
912
913    This reverts commits 344470cac42e and e81324407269.
914
915    It turns out that the exact path in the symlink matters, if for somewhat
916    unfortunate reasons: some apparmor configurations don't allow dhclient
917    access to the per-thread /proc files.  As reported by Jörg Otte:
918
919      audit: type=1400 audit(1407684227.003:28): apparmor="DENIED"
920        operation="open" profile="/sbin/dhclient"
921        name="/proc/1540/task/1540/net/dev" pid=1540 comm="dhclient"
922        requested_mask="r" denied_mask="r" fsuid=0 ouid=0
923
924    so we had better revert this for now.  We might be able to work around
925    this in practice by only using the per-thread symlinks if the thread
926    isn't the thread group leader, and if the namespaces differ between
927    threads (which basically never happens).
928
929    We'll see. In the meantime, the revert was made to be intentionally easy.
930
931    Reported-by: Jörg Otte <jrg.otte@gmail.com>
932    Acked-by: Eric W. Biederman <ebiederm@xmission.com>
933    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
934commit 27d438c56009e5ae632de36fe70985d1aab5e344
935Author: Linus Torvalds <torvalds@linux-foundation.org>
936Date:   Fri Aug 8 11:48:26 2014 -1000
937
938    Revert "drm/i915: Enable PSR by default."
939
940    This reverts commit b6d547791fd3ef4ccc89ad2556ab01045640aef7.
941
942    The panel self refresh clearly isn't stable yet, and causes my laptop
943    (Haswell ULT in a Sony Vaio Pro) to have the screen lock up.  Maybe it
944    doesn't ever get out of self-refresh, or maybe there are gremlins in the
945    machine that get unhappy.  Regardless, it's broken, and it gets
946    reverted.
947
948    Cc: Rodrigo Vivi <rodrigo.vivi@intel.com>
949    Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
950    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
951commit 1d8fcba1de632d7a43349788ad534c5a32c5a44c
952Author: Linus Torvalds <torvalds@linux-foundation.org>
953Date:   Wed Jul 30 08:56:23 2014 -0700
954
955    Revert "cdc_subset: deal with a device that needs reset for timeout"
956
957    This reverts commit 20fbe3ae990fd54fc7d1f889d61958bc8b38f254.
958
959    As reported by Stephen Rothwell, it causes compile failures in certain
960    configurations:
961
962      drivers/net/usb/cdc_subset.c:360:15: error: 'dummy_prereset' undeclared here (not in a function)
963        .pre_reset = dummy_prereset,
964                     ^
965      drivers/net/usb/cdc_subset.c:361:16: error: 'dummy_postreset' undeclared here (not in a function)
966        .post_reset = dummy_postreset,
967                      ^
968
969    Reported-by: Stephen Rothwell <sfr@canb.auug.org.au>
970    Acked-by: David Miller <davem@davemloft.net>
971    Cc: Oliver Neukum <oneukum@suse.de>
972    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
973commit bb077d600689dbf9305758efed1e16775db1c84c
974Author: Linus Torvalds <torvalds@linux-foundation.org>
975Date:   Sun Jun 8 10:09:49 2014 -0700
976
977    Revert "x86/smpboot: Initialize secondary CPU only if master CPU will wait for it"
978
979    This reverts commit 3e1a878b7ccdb31da6d9d2b855c72ad87afeba3f.
980
981    It came in very late, and already has one reported failure: Sitsofe
982    reports that the current tree fails to boot on his EeePC, and bisected
983    it down to this.  Rather than waste time trying to figure out what's
984    wrong, just revert it.
985
986    Reported-by: Sitsofe Wheeler <sitsofe@gmail.com>
987    Cc: Igor Mammedov <imammedo@redhat.com>
988    Cc: Toshi Kani <toshi.kani@hp.com>
989    Cc: Thomas Gleixner <tglx@linutronix.de>
990    Acked-by: Ingo Molnar <mingo@kernel.org>
991    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
992commit 29b1deb2a48a9dd02b93597aa4c055a24c0e989f
993Author: Linus Torvalds <torvalds@linux-foundation.org>
994Date:   Sun Dec 15 11:17:45 2013 -0800
995
996    Revert "selinux: consider filesystem subtype in policies"
997
998    This reverts commit 102aefdda4d8275ce7d7100bc16c88c74272b260.
999
1000    Tom London reports that it causes sync() to hang on Fedora rawhide:
1001
1002      https://bugzilla.redhat.com/show_bug.cgi?id=1033965
1003
1004    and Josh Boyer bisected it down to this commit.  Reverting the commit in
1005    the rawhide kernel fixes the problem.
1006
1007    Eric Paris root-caused it to incorrect subtype matching in that commit
1008    breaking fuse, and has a tentative patch, but by now we're better off
1009    retrying this in 3.14 rather than playing with it any more.
1010
1011    Reported-by: Tom London <selinux@gmail.com>
1012    Bisected-by: Josh Boyer <jwboyer@fedoraproject.org>
1013    Acked-by: Eric Paris <eparis@redhat.com>
1014    Cc: James Morris <jmorris@namei.org>
1015    Cc: Anand Avati <avati@redhat.com>
1016    Cc: Paul Moore <paul@paul-moore.com>
1017    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
1018commit 4c1cc40a2d49500d84038ff751bc6cd183e729b5
1019Author: Linus Torvalds <torvalds@linux-foundation.org>
1020Date:   Sat Nov 23 16:38:17 2013 -0800
1021
1022    Revert "KEYS: verify a certificate is signed by a 'trusted' key"
1023
1024    This reverts commit 09fbc47373826d67531380662b516de2da120545, which
1025    caused the following build errors:
1026
1027      crypto/asymmetric_keys/x509_public_key.c: In function ‘x509_key_preparse’:
1028      crypto/asymmetric_keys/x509_public_key.c:237:35: error: ‘system_trusted_keyring’ undeclared (first use in this function)
1029       ret = x509_validate_trust(cert, system_trusted_keyring);
1030                                       ^
1031      crypto/asymmetric_keys/x509_public_key.c:237:35: note: each undeclared identifier is reported only once for each function it appears in
1032
1033    reported by Jim Davis. Mimi says:
1034
1035     "I made the classic mistake of requesting this patch to be upstreamed
1036      at the last second, rather than waiting until the next open window.
1037
1038      At this point, the best course would probably be to revert the two
1039      commits and fix them for the next open window"
1040
1041    Reported-by: Jim Davis <jim.epost@gmail.com>
1042    Acked-by: Mimi Zohar <zohar@linux.vnet.ibm.com>
1043    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
1044commit 34ef7bd3823bf4401bf8f1f855e1bc77b82b1a43
1045Author: Linus Torvalds <torvalds@linux-foundation.org>
1046Date:   Sat Nov 23 16:36:35 2013 -0800
1047
1048    Revert "ima: define '_ima' as a builtin 'trusted' keyring"
1049
1050    This reverts commit 217091dd7a7a1bdac027ddb7c5a25f6ac0b8e241, which
1051    caused the following build error:
1052
1053      security/integrity/digsig.c:70:5: error: redefinition of ‘integrity_init_keyring’
1054      security/integrity/integrity.h:149:12: note: previous definition of ‘integrity_init_keyring’ w
1055      security/integrity/integrity.h:149:12: warning: ‘integrity_init_keyring’ defined but not used
1056
1057    reported by Krzysztof Kolasa. Mimi says:
1058
1059     "I made the classic mistake of requesting this patch to be upstreamed
1060      at the last second, rather than waiting until the next open window.
1061
1062      At this point, the best course would probably be to revert the two
1063      commits and fix them for the next open window"
1064
1065    Reported-by: Krzysztof Kolasa <kkolasa@winsoft.pl>
1066    Acked-by: Mimi Zohar <zohar@linux.vnet.ibm.com>
1067    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
1068commit 8b2e9b712f6139df9c754af0d67fecc4bbc88545
1069Author: Linus Torvalds <torvalds@linux-foundation.org>
1070Date:   Wed Nov 20 14:41:47 2013 -0800
1071
1072    Revert "mm: create a separate slab for page->ptl allocation"
1073
1074    This reverts commit ea1e7ed33708c7a760419ff9ded0a6cb90586a50.
1075
1076    Al points out that while the commit *does* actually create a separate
1077    slab for the page->ptl allocation, that slab is never actually used, and
1078    the code continues to use kmalloc/kfree.
1079
1080    Damien Wyart points out that the original patch did have the conversion
1081    to use kmem_cache_alloc/free, so it got lost somewhere on its way to me.
1082
1083    Revert the half-arsed attempt that didn't do anything.  If we really do
1084    want the special slab (remember: this is all relevant just for debug
1085    builds, so it's not necessarily all that critical) we might as well redo
1086    the patch fully.
1087
1088    Reported-by: Al Viro <viro@zeniv.linux.org.uk>
1089    Acked-by: Andrew Morton <akpm@linux-foundation.org>
1090    Cc: Kirill A Shutemov <kirill.shutemov@linux.intel.com>
1091    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
1092commit a1212d278c05ca0a38f5cbd7ae90ac2e367228a8
1093Author: Linus Torvalds <torvalds@linux-foundation.org>
1094Date:   Thu Nov 7 20:47:28 2013 +0900
1095
1096    Revert "sysfs: drop kobj_ns_type handling"
1097
1098    This reverts commit cb26a311578e67769e92a39a0a63476533cb7e12.
1099
1100    It mysteriously causes NetworkManager to not find the wireless device
1101    for me.  As far as I can tell, Tejun *meant* for this commit to not make
1102    any semantic changes, but there clearly are some.  So revert it, taking
1103    into account some of the calling convention changes that happened in
1104    this area in subsequent commits.
1105
1106    Cc: Tejun Heo <tj@kernel.org>
1107    Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
1108    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
1109commit b04c99e3b845892d754ee8052d6324c39c4040de
1110Author: Linus Torvalds <torvalds@linux-foundation.org>
1111Date:   Sat Sep 7 09:48:41 2013 -0700
1112
1113    Revert "Input: introduce BTN/ABS bits for drums and guitars"
1114
1115    This reverts commits 61e00655e9cb, 73f8645db191 and 8e22ecb603c8:
1116      "Input: introduce BTN/ABS bits for drums and guitars"
1117      "HID: wiimote: add support for Guitar-Hero drums"
1118      "HID: wiimote: add support for Guitar-Hero guitars"
1119
1120    The extra new ABS_xx values resulted in ABS_MAX no longer being a
1121    power-of-two, which broke the comparison logic.  It also caused the
1122    ioctl numbers to overflow into the next byte, causing problems for that.
1123
1124    We'll try again for 3.13.
1125
1126    Reported-by: Markus Trippelsdorf <markus@trippelsdorf.de>
1127    Reported-by: Linus Torvalds <torvalds@linux-foundation.org>
1128    Acked-by: David Herrmann <dh.herrmann@gmail.com>
1129    Acked-by: Dmitry Torokhov <dmitry.torokhov@gmail.com>
1130    Cc: Benjamin Tissoires <benjamin.tissoires@gmail.com>
1131    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
1132commit f0cc6ffb8ce8961db587e5072168cac0cbc25f05
1133Author: Linus Torvalds <torvalds@linux-foundation.org>
1134Date:   Wed Aug 28 09:18:05 2013 -0700
1135
1136    Revert "fs: Allow unprivileged linkat(..., AT_EMPTY_PATH) aka flink"
1137
1138    This reverts commit bb2314b47996491bbc5add73633905c3120b6268.
1139
1140    It wasn't necessarily wrong per se, but we're still busily discussing
1141    the exact details of this all, so I'm going to revert it for now.
1142
1143    It's true that you can already do flink() through /proc and that flink()
1144    isn't new.  But as Brad Spengler points out, some secure environments do
1145    not mount proc, and flink adds a new interface that can avoid path
1146    lookup of the source for those kinds of environments.
1147
1148    We may re-do this (and even mark it for stable backporting back in 3.11
1149    and possibly earlier) once the whole discussion about the interface is done.
1150
1151    Cc: Andy Lutomirski <luto@amacapital.net>
1152    Cc: Al Viro <viro@zeniv.linux.org.uk>
1153    Cc: Oleg Nesterov <oleg@redhat.com>
1154    Cc: Brad Spengler <spender@grsecurity.net>
1155    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
1156commit 5ea80f76a56605a190a7ea16846c82aa63dbd0aa
1157Author: Linus Torvalds <torvalds@linux-foundation.org>
1158Date:   Thu Aug 22 09:13:06 2013 -0700
1159
1160    Revert "x86 get_unmapped_area(): use proper mmap base for bottom-up direction"
1161
1162    This reverts commit df54d6fa54275ce59660453e29d1228c2b45a826.
1163
1164    The commit isn't necessarily wrong, but because it recalculates the
1165    random mmap_base every time, it seems to confuse user memory allocators
1166    that expect contiguous mmap allocations even when the mmap address isn't
1167    specified.
1168
1169    In particular, the MATLAB Java runtime seems to be unhappy. See
1170
1171      https://bugzilla.kernel.org/show_bug.cgi?id=60774
1172
1173    So we'll want to apply the random offset only once, and Radu has a patch
1174    for that.  Revert this older commit in order to apply the other one.
1175
1176    Reported-by: Jeff Shorey <shoreyjeff@gmail.com>
1177    Cc: Radu Caragea <sinaelgl@gmail.com>
1178    Cc: Andrew Morton <akpm@linux-foundation.org>
1179    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
1180commit 370905069ce6515f38d5de0a5b1c899cbe58fe22
1181Author: Linus Torvalds <torvalds@linux-foundation.org>
1182Date:   Thu Aug 8 09:06:37 2013 -0700
1183
1184    Revert "slub: do not put a slab to cpu partial list when cpu_partial is 0"
1185
1186    This reverts commit 318df36e57c0ca9f2146660d41ff28e8650af423.
1187
1188    This commit caused Steven Rostedt's hackbench runs to run out of memory
1189    due to a leak.  As noted by Joonsoo Kim, it is buggy in the following
1190    scenario:
1191
1192     "I guess, you may set 0 to all kmem caches's cpu_partial via sysfs,
1193      doesn't it?
1194
1195      In this case, memory leak is possible in following case.  Code flow of
1196      possible leak is follwing case.
1197
1198       * in __slab_free()
1199       1. (!new.inuse || !prior) && !was_frozen
1200       2. !kmem_cache_debug && !prior
1201       3. new.frozen = 1
1202       4. after cmpxchg_double_slab, run the (!n) case with new.frozen=1
1203       5. with this patch, put_cpu_partial() doesn't do anything,
1204          because this cache's cpu_partial is 0
1205       6. return
1206
1207      In step 5, leak occur"
1208
1209    And Steven does indeed have cpu_partial set to 0 due to RT testing.
1210
1211    Joonsoo is cooking up a patch, but everybody agrees that reverting this
1212    for now is the right thing to do.
1213
1214    Reported-and-bisected-by: Steven Rostedt <rostedt@goodmis.org>
1215    Acked-by: Joonsoo Kim <iamjoonsoo.kim@lge.com>
1216    Acked-by: Pekka Enberg <penberg@kernel.org>
1217    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
1218commit 6d128e1e72bf082542e85f72e6b7ddd704193588
1219Author: Linus Torvalds <torvalds@linux-foundation.org>
1220Date:   Wed Jul 10 19:02:51 2013 -0700
1221
1222    Revert "Makefile: Fix install error with make -j option"
1223
1224    This reverts commit d2aae8477cd00325bb7c7c7e95be488088900c48.
1225
1226    It is completely and utterly broken.  Module install should not build
1227    any files, and adding broken dependencies to "help" it build files is
1228    complete and utter sh*t.
1229
1230    The kernel should not be built by root, and "make install" and "make
1231    module_install" (that for obvious reasons need to be run as root)
1232    absolutely must not build any files.  They should only ever copy the
1233    already-built files over.
1234
1235    So having dependencies for the install targets is wrong, wrong, wrong.
1236
1237    If you try to install a kernel without building it first, you *should*
1238    get errors. The build system shouldn't try to help root build the files.
1239
1240    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
1241commit 0a82a8d132b26d438eb90b3ab35a7016e7227a1d
1242Author: Linus Torvalds <torvalds@linux-foundation.org>
1243Date:   Thu Apr 18 09:00:26 2013 -0700
1244
1245    Revert "block: add missing block_bio_complete() tracepoint"
1246
1247    This reverts commit 3a366e614d0837d9fc23f78cdb1a1186ebc3387f.
1248
1249    Wanlong Gao reports that it causes a kernel panic on his machine several
1250    minutes after boot. Reverting it removes the panic.
1251
1252    Jens says:
1253     "It's not quite clear why that is yet, so I think we should just revert
1254      the commit for 3.9 final (which I'm assuming is pretty close).
1255
1256      The wifi is crap at the LSF hotel, so sending this email instead of
1257      queueing up a revert and pull request."
1258
1259    Reported-by: Wanlong Gao <gaowanlong@cn.fujitsu.com>
1260    Requested-by: Jens Axboe <axboe@kernel.dk>
1261    Cc: Tejun Heo <tj@kernel.org>
1262    Cc: Steven Rostedt <rostedt@goodmis.org>
1263    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
1264commit 9360b53661a2c7754517b2925580055bacc8ec38
1265Author: Linus Torvalds <torvalds@linux-foundation.org>
1266Date:   Mon Dec 17 11:29:09 2012 -0800
1267
1268    Revert "bdi: add a user-tunable cpu_list for the bdi flusher threads"
1269
1270    This reverts commit 8fa72d234da9b6b473bbb1f74d533663e4996e6b.
1271
1272    People disagree about how this should be done, so let's revert this for
1273    now so that nobody starts using the new tuning interface.  Tejun is
1274    thinking about a more generic interface for thread pool affinity.
1275
1276    Requested-by: Tejun Heo <tj@kernel.org>
1277    Acked-by: Jeff Moyer <jmoyer@redhat.com>
1278    Acked-by: Jens Axboe <axboe@kernel.dk>
1279    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
1280commit 11520e5e7c1855fc3bf202bb3be35a39d9efa034
1281Author: Linus Torvalds <torvalds@linux-foundation.org>
1282Date:   Sat Dec 15 15:15:24 2012 -0800
1283
1284    Revert "x86-64/efi: Use EFI to deal with platform wall clock (again)"
1285
1286    This reverts commit bd52276fa1d4 ("x86-64/efi: Use EFI to deal with
1287    platform wall clock (again)"), and the two supporting commits:
1288
1289      da5a108d05b4: "x86/kernel: remove tboot 1:1 page table creation code"
1290
1291      185034e72d59: "x86, efi: 1:1 pagetable mapping for virtual EFI calls")
1292
1293    as they all depend semantically on commit 53b87cf088e2 ("x86, mm:
1294    Include the entire kernel memory map in trampoline_pgd") that got
1295    reverted earlier due to the problems it caused.
1296
1297    This was pointed out by Yinghai Lu, and verified by me on my Macbook Air
1298    that uses EFI.
1299
1300    Pointed-out-by: Yinghai Lu <yinghai@kernel.org>
1301    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
1302commit be354f40812314dee2b1e3aa272528c056bb827d
1303Author: Linus Torvalds <torvalds@linux-foundation.org>
1304Date:   Sat Dec 15 12:29:54 2012 -0800
1305
1306    Revert "x86, mm: Include the entire kernel memory map in trampoline_pgd"
1307
1308    This reverts commit 53b87cf088e2ea68d7c59619d0214cc15bb76133.
1309
1310    It causes odd bootup problems on x86-64.  Markus Trippelsdorf gets a
1311    repeatable oops, and I see a non-repeatable oops (or constant stream of
1312    messages that scroll off too quickly to read) that seems to go away with
1313    this commit reverted.
1314
1315    So we don't know exactly what is wrong with the commit, but it's
1316    definitely problematic, and worth reverting sooner rather than later.
1317
1318    Bisected-by: Markus Trippelsdorf <markus@trippelsdorf.de>
1319    Cc: H Peter Anvin <hpa@zytor.com>
1320    Cc: Jan Beulich <jbeulich@suse.com>
1321    Cc: Matt Fleming <matt.fleming@intel.com>
1322    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
1323commit 17bc14b767cf0692420c43dbe5310ae98a5a7836
1324Author: Linus Torvalds <torvalds@linux-foundation.org>
1325Date:   Fri Dec 14 07:20:43 2012 -0800
1326
1327    Revert "sched: Update_cfs_shares at period edge"
1328
1329    This reverts commit f269ae0469fc882332bdfb5db15d3c1315fe2a10.
1330
1331    It turns out it causes a very noticeable interactivity regression with
1332    CONFIG_SCHED_AUTOGROUP (test-case: "make -j32" of the kernel in a
1333    terminal window, while scrolling in a browser - the autogrouping means
1334    that the two end up in separate cgroups, and the browser should be
1335    smooth as silk despite the high load).
1336
1337    Says Paul Turner:
1338     "It seems that the update-throttling on the wake-side is reducing the
1339      interactive tasks' ability to preempt.  While I suspect the right
1340      longer term answer here is force these updates only in the
1341      cross-cgroup case; this is less trivial.  For this release I believe
1342      the right answer is either going to be a revert or restore the updates
1343      on the enqueue-side."
1344
1345    Reported-by: Linus Torvalds <torvalds@linux-foundation.org>
1346    Bisected-by: Mike Galbraith <efault@gmx.de>
1347    Acked-by: Paul Turner <pjt@google.com>
1348    Acked-by: Ingo Molnar <mingo@kernel.org>
1349    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
1350commit caf491916b1c1e939a2c7575efb7a77f11fc9bdf
1351Author: Linus Torvalds <torvalds@linux-foundation.org>
1352Date:   Mon Dec 10 10:51:16 2012 -0800
1353
1354    Revert "revert "Revert "mm: remove __GFP_NO_KSWAPD""" and associated damage
1355
1356    This reverts commits a50915394f1fc02c2861d3b7ce7014788aa5066e and
1357    d7c3b937bdf45f0b844400b7bf6fd3ed50bac604.
1358
1359    This is a revert of a revert of a revert.  In addition, it reverts the
1360    even older i915 change to stop using the __GFP_NO_KSWAPD flag due to the
1361    original commits in linux-next.
1362
1363    It turns out that the original patch really was bogus, and that the
1364    original revert was the correct thing to do after all.  We thought we
1365    had fixed the problem, and then reverted the revert, but the problem
1366    really is fundamental: waking up kswapd simply isn't the right thing to
1367    do, and direct reclaim sometimes simply _is_ the right thing to do.
1368
1369    When certain allocations fail, we simply should try some direct reclaim,
1370    and if that fails, fail the allocation.  That's the right thing to do
1371    for THP allocations, which can easily fail, and the GPU allocations want
1372    to do that too.
1373
1374    So starting kswapd is sometimes simply wrong, and removing the flag that
1375    said "don't start kswapd" was a mistake.  Let's hope we never revisit
1376    this mistake again - and certainly not this many times ;)
1377
1378    Acked-by: Mel Gorman <mgorman@suse.de>
1379    Acked-by: Johannes Weiner <hannes@cmpxchg.org>
1380    Cc: Rik van Riel <riel@redhat.com>
1381    Cc: Andrew Morton <akpm@linux-foundation.org>
1382    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
1383commit 31f8d42d44b48ba72b586ca03e810cbbd21ea16b
1384Author: Linus Torvalds <torvalds@linux-foundation.org>
1385Date:   Mon Dec 10 10:47:45 2012 -0800
1386
1387    Revert "mm: avoid waking kswapd for THP allocations when compaction is deferred or contended"
1388
1389    This reverts commit 782fd30406ecb9d9b082816abe0c6008fc72a7b0.
1390
1391    We are going to reinstate the __GFP_NO_KSWAPD flag that has been
1392    removed, the removal reverted, and then removed again.  Making this
1393    commit a pointless fixup for a problem that was caused by the removal of
1394    __GFP_NO_KSWAPD flag.
1395
1396    The thing is, we really don't want to wake up kswapd for THP allocations
1397    (because they fail quite commonly under any kind of memory pressure,
1398    including when there is tons of memory free), and these patches were
1399    just trying to fix up the underlying bug: the original removal of
1400    __GFP_NO_KSWAPD in commit c654345924f7 ("mm: remove __GFP_NO_KSWAPD")
1401    was simply bogus.
1402
1403    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
1404commit 37407ea7f93864c2cfc03edf8f37872ec539ea2b
1405Author: Linus Torvalds <torvalds@linux-foundation.org>
1406Date:   Sun Sep 16 12:29:43 2012 -0700
1407
1408    Revert "sched: Improve scalability via 'CPU buddies', which withstand random perturbations"
1409
1410    This reverts commit 970e178985cadbca660feb02f4d2ee3a09f7fdda.
1411
1412    Nikolay Ulyanitsky reported thatthe 3.6-rc5 kernel has a 15-20%
1413    performance drop on PostgreSQL 9.2 on his machine (running "pgbench").
1414
1415    Borislav Petkov was able to reproduce this, and bisected it to this
1416    commit 970e178985ca ("sched: Improve scalability via 'CPU buddies' ...")
1417    apparently because the new single-idle-buddy model simply doesn't find
1418    idle CPU's to reschedule on aggressively enough.
1419
1420    Mike Galbraith suspects that it is likely due to the user-mode spinlocks
1421    in PostgreSQL not reacting well to preemption, but we don't really know
1422    the details - I'll just revert the commit for now.
1423
1424    There are hopefully other approaches to improve scheduler scalability
1425    without it causing these kinds of downsides.
1426
1427    Reported-by: Nikolay Ulyanitsky <lystor@gmail.com>
1428    Bisected-by: Borislav Petkov <bp@alien8.de>
1429    Acked-by: Mike Galbraith <efault@gmx.de>
1430    Cc: Andrew Morton <akpm@linux-foundation.org>
1431    Cc: Thomas Gleixner <tglx@linutronix.de>
1432    Cc: Ingo Molnar <mingo@kernel.org>
1433    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
1434commit bc51b0c22cebf5c311a6f1895fcca9f78efd0478
1435Author: Linus Torvalds <torvalds@linux-foundation.org>
1436Date:   Tue Jul 10 12:49:32 2012 -0700
1437
1438    Revert "of: match by compatible property first"
1439
1440    This reverts commit 107a84e61cdd3406c842a0e4be7efffd3a05dba6.
1441
1442    Meelis Roos reports a regression since 3.5-rc5 that stops Sun Fire V100
1443    and Sun Netra X1 sparc64 machines from booting, hanging after enabling
1444    serial console.  He bisected it to commit 107a84e61cdd.
1445
1446    Rob Herring explains:
1447     "The problem is match combinations of compatible plus name and/or type
1448      fail to match correctly.  I have a fix for this, but given how late it
1449      is for 3.5 I think it is best to revert this for now.  There could be
1450      other cases that rely on the current although wrong behavior.  I will
1451      post an updated version for 3.6."
1452
1453    Bisected-and-reported-by: Meelis Roos <mroos@linux.ee>
1454    Requested-by: Rob Herring <rob.herring@calxeda.com>
1455    Cc: Thierry Reding <thierry.reding@avionic-design.de>
1456    Cc: Grant Likely <grant.likely@secretlab.ca>
1457    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
1458commit 8f53369b753f5f4c7684c2eb0b592152abb1dd00
1459Author: Linus Torvalds <torvalds@linux-foundation.org>
1460Date:   Fri Jun 8 14:53:06 2012 -0700
1461
1462    Revert "drm/i915/crt: Do not rely upon the HPD presence pin"
1463
1464    This reverts commit 9e612a008fa7fe493a473454def56aa321479495.
1465
1466    It incorrectly finds VGA connectors where none are attached, apparently
1467    not noticing that nothing replied to the EDID queries, and happily using
1468    the default EDID modes that have nothing to do with actual hardware.
1469
1470    That in turn then causes X to fall down to the lowest common
1471    denominator, which is usually the default 1024x768 mode that is in the
1472    default EDID and pretty much anything supports).
1473
1474    I'd suggest that if not relying on the HDP pin, the code should at least
1475    check whether it gets valid EDID data back, rather than just assume
1476    there's something on the VGA connector.
1477
1478    Cc: Dave Airlie <airlied@linux.ie>
1479    Cc: Chris Wilson <chris@chris-wilson.co.uk>
1480    Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
1481    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
1482commit 32ba9c3fcab960f0b0d332c86ebcd2c4870d9bb8
1483Author: Linus Torvalds <torvalds@linux-foundation.org>
1484Date:   Fri Jun 8 10:34:03 2012 -0700
1485
1486    Revert "vfs: stop d_splice_alias creating directory aliases"
1487
1488    This reverts commit 7732a557b1342c6e6966efb5f07effcf99f56167 (and commit
1489    3f50fff4dace23d3cfeb195d5cd4ee813cee68b7, which was a follow-up
1490    cleanup).
1491
1492    We're chasing an elusive bug that Dave Jones can apparently reproduce
1493    using his system call fuzzer tool, and that looks like some kind of
1494    locking ordering problem on the directory i_mutex chain.  Our i_mutex
1495    locking is rather complex, and depends on the topological ordering of
1496    the directories, which is why we have been very wary of splicing
1497    directory entries around.
1498
1499    Of course, we really don't want to ever see aliased unconnected
1500    directories anyway, so none of this should ever happen, but this revert
1501    aims to basically get us back to a known older state.
1502
1503    Bruce points to some of the previous discussion at
1504
1505           http://marc.info/?i=<20110310105821.GE22723@ZenIV.linux.org.uk>
1506
1507    and in particular a long post from Neil:
1508
1509           http://marc.info/?i=<20110311150749.2fa2be66@notabene.brown>
1510
1511    It should be noted that it's possible that Dave's problems come from
1512    other changes altohgether, including possibly just the fact that Dave
1513    constantly is teachning his fuzzer new tricks.  So what appears to be a
1514    new bug could in fact be an old one that just gets newly triggered, but
1515    reverting these patches as "still under heavy discussion" is the right
1516    thing regardless.
1517
1518    Requested-by: Al Viro <viro@zeniv.linux.org.uk>
1519    Acked-by: J. Bruce Fields <bfields@fieldses.org>
1520    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
1521commit 48d212a2eecaca2e1875925837ad27b2f43f48a3
1522Author: Linus Torvalds <torvalds@linux-foundation.org>
1523Date:   Thu Jun 7 17:54:07 2012 -0700
1524
1525    Revert "mm: correctly synchronize rss-counters at exit/exec"
1526
1527    This reverts commit 40af1bbdca47e5c8a2044039bb78ca8fd8b20f94.
1528
1529    It's horribly and utterly broken for at least the following reasons:
1530
1531     - calling sync_mm_rss() from mmput() is fundamentally wrong, because
1532       there's absolutely no reason to believe that the task that does the
1533       mmput() always does it on its own VM.  Example: fork, ptrace, /proc -
1534       you name it.
1535
1536     - calling it *after* having done mmdrop() on it is doubly insane, since
1537       the mm struct may well be gone now.
1538
1539     - testing mm against NULL before you call it is insane too, since a
1540    NULL mm there would have caused oopses long before.
1541
1542    .. and those are just the three bugs I found before I decided to give up
1543    looking for me and revert it asap.  I should have caught it before I
1544    even took it, but I trusted Andrew too much.
1545
1546    Cc: Konstantin Khlebnikov <khlebnikov@openvz.org>
1547    Cc: Markus Trippelsdorf <markus@trippelsdorf.de>
1548    Cc: Hugh Dickins <hughd@google.com>
1549    Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
1550    Cc: Oleg Nesterov <oleg@redhat.com>
1551    Cc: Andrew Morton <akpm@linux-foundation.org>
1552    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
1553commit 68e3e92620c323703bc7db75c2ba15239ee85c39
1554Author: Linus Torvalds <torvalds@linux-foundation.org>
1555Date:   Sun Jun 3 20:05:57 2012 -0700
1556
1557    Revert "mm: compaction: handle incorrect MIGRATE_UNMOVABLE type pageblocks"
1558
1559    This reverts commit 5ceb9ce6fe9462a298bb2cd5c9f1ca6cb80a0199.
1560
1561    That commit seems to be the cause of the mm compation list corruption
1562    issues that Dave Jones reported.  The locking (or rather, absense
1563    there-of) is dubious, as is the use of the 'page' variable once it has
1564    been found to be outside the pageblock range.
1565
1566    So revert it for now, we can re-visit this for 3.6.  If we even need to:
1567    as Minchan Kim says, "The patch wasn't a bug fix and even test workload
1568    was very theoretical".
1569
1570    Reported-and-tested-by: Dave Jones <davej@redhat.com>
1571    Acked-by: Hugh Dickins <hughd@google.com>
1572    Acked-by: KOSAKI Motohiro <kosaki.motohiro@gmail.com>
1573    Acked-by: Minchan Kim <minchan@kernel.org>
1574    Cc: Bartlomiej Zolnierkiewicz <b.zolnierkie@samsung.com>
1575    Cc: Kyungmin Park <kyungmin.park@samsung.com>
1576    Cc: Andrew Morton <akpm@linux-foundation.org>
1577    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
1578commit 2e321806b681b1920b6dfa7d81bbe3d312fe1d19
1579Author: Linus Torvalds <torvalds@linux-foundation.org>
1580Date:   Mon May 21 18:48:10 2012 -0700
1581
1582    Revert "vfs: remove unnecessary d_unhashed() check from __d_lookup_rcu"
1583
1584    This reverts commit 8c01a529b861ba97c7d78368e6a5d4d42e946f75.
1585
1586    It turns out the d_unhashed() check isn't unnecessary after all: while
1587    it's true that unhashing will increment the sequence numbers, that does
1588    not necessarily invalidate the RCU lookup, because it might have seen
1589    the dentry pointer (before it got unhashed), but by the time it loaded
1590    the sequence number, it could have seen the *new* sequence number (after
1591    it got unhashed).
1592
1593    End result: we might look up an unhashed dentry that is about to be
1594    freed, with the sequence number never indicating anything bad about it.
1595    So checking that the dentry is still hashed (*after* reading the sequence
1596    number) is indeed the proper fix, and was never unnecessary.
1597
1598    Reported-by: Dave Jones <davej@redhat.com>
1599    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
1600commit fcbf94b9dedd2ce08e798a99aafc94fec8668161
1601Author: Linus Torvalds <torvalds@linux-foundation.org>
1602Date:   Sat Apr 28 08:29:56 2012 -0700
1603
1604    Revert "autofs: work around unhappy compat problem on x86-64"
1605
1606    This reverts commit a32744d4abae24572eff7269bc17895c41bd0085.
1607
1608    While that commit was technically the right thing to do, and made the
1609    x86-64 compat mode work identically to native 32-bit mode (and thus
1610    fixing the problem with a 32-bit systemd install on a 64-bit kernel), it
1611    turns out that the automount binaries had workarounds for this compat
1612    problem.
1613
1614    Now, the workarounds are disgusting: doing an "uname()" to find out the
1615    architecture of the kernel, and then comparing it for the 64-bit cases
1616    and fixing up the size of the read() in automount for those.  And they
1617    were confused: it's not actually a generic 64-bit issue at all, it's
1618    very much tied to just x86-64, which has different alignment for an
1619    'u64' in 64-bit mode than in 32-bit mode.
1620
1621    But the end result is that fixing the compat layer actually breaks the
1622    case of a 32-bit automount on a x86-64 kernel.
1623
1624    There are various approaches to fix this (including just doing a
1625    "strcmp()" on current->comm and comparing it to "automount"), but I
1626    think that I will do the one that teaches pipes about a special "packet
1627    mode", which will allow user space to not have to care too deeply about
1628    the padding at the end of the autofs packet.
1629
1630    That change will make the compat workaround unnecessary, so let's revert
1631    it first, and get automount working again in compat mode.  The
1632    packetized pipes will then fix autofs for systemd.
1633
1634    Reported-and-requested-by: Michael Tokarev <mjt@tls.msk.ru>
1635    Cc: Ian Kent <raven@themaw.net>
1636    Cc: stable@kernel.org # for 3.3
1637    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
1638commit 19244ad06b70ed84931df868583547ce1cd3a186
1639Author: Linus Torvalds <torvalds@linux-foundation.org>
1640Date:   Fri Apr 20 11:19:35 2012 -0700
1641
1642    Revert "ACPI: ignore FADT reset-reg-sup flag"
1643
1644    This reverts commit cf450136bfde77c7f95065c91bffded4aa7fa731.
1645
1646    It breaks reboot on at least one Thinkpad T43, as reported by Jörg Otte:
1647     "On reboot it shuts down as normal.
1648      The last lines displayed are:
1649
1650      >Unmounting temporary filesystems.. [OK]
1651      >Deactivating swap...               [OK]
1652      >Unmounting local filesystems...    [OK]
1653      >Will now restart
1654      >    Restarting system
1655
1656      Then I hear it accessing the cd-drive, but then it's being stuck."
1657
1658    Jörg bisected the regression to this commit.
1659
1660    That commit fixes another machine (see
1661
1662      https://bugzilla.kernel.org/show_bug.cgi?id=11533
1663
1664    for details) that has a BIOS bug and doesn't support ACPI reset.
1665    However, at least one of those other reporters no longer even has the
1666    machine in question, and had a different workaround to begin with.
1667    Besides, it clearly was a buggy BIOS.  Let's not break the correct case
1668    to fix that case.
1669
1670    Reported-and-bisected-by: Jörg Otte <jrg.otte@googlemail.com>
1671    Cc: linux-acpi@vger.kernel.org
1672    Cc: Len Brown <lenb@kernel.org>
1673    Cc: Peter Anvin <hpa@zytor.com>
1674    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
1675commit f3718a818f7fb3636130d5f34bde8df34f45c5e5
1676Author: Linus Torvalds <torvalds@linux-foundation.org>
1677Date:   Wed Apr 4 08:16:25 2012 -0700
1678
1679    Revert "nouveau/bios: Fix tracking of BIOS image data"
1680
1681    This reverts commit d06221c0617ab6d0bc41c4980cefdd9c8cc9a1c1.
1682
1683    It turns out to trigger the "BUG_ON(!PageCompound(page))" in kfree(),
1684    apparently because the code ends up trying to free somethng that was
1685    never kmalloced in the first place.
1686
1687    BenH points out that the patch was untested and wasn't meant to go into
1688    the upstream kernel that quickly in the first place.
1689
1690    Backtrace:
1691      bios_shadow
1692      bios_shadow_prom
1693      nv_mask
1694      init_io
1695      bios_shadow
1696      nouveau_bios_init
1697      NVReadVgaCrtc
1698      NVSetOwner
1699      nouveau_card_init
1700      nouveau_load
1701
1702    Reported-by: Meelis Roos <mroos@linux.ee>
1703    Requested-by: Dave Airlie <airlied@gmail.com>
1704    Acked-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
1705    Cc: Ben Skeggs <bskeggs@redhat.com>
1706    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
1707commit 6268b325c3066234e7bddb99d2b98bcedb0c0033
1708Author: Linus Torvalds <torvalds@linux-foundation.org>
1709Date:   Thu Mar 29 17:00:56 2012 -0700
1710
1711    Revert "ext4: don't release page refs in ext4_end_bio()"
1712
1713    This reverts commit b43d17f319f2c502b17139d1cf70731b2b62c644.
1714
1715    Dave Jones reports that it causes lockups on his laptop, and his debug
1716    output showed a lot of processes hung waiting for page_writeback (or
1717    more commonly - processes hung waiting for a lock that was held during
1718    that writeback wait).
1719
1720    The page_writeback hint made Ted suggest that Dave look at this commit,
1721    and Dave verified that reverting it makes his problems go away.
1722
1723    Ted says:
1724     "That commit fixes a race which is seen when you write into fallocated
1725      (and hence uninitialized) disk blocks under *very* heavy memory
1726      pressure.  Furthermore, although theoretically it could trigger under
1727      normal direct I/O writes, it only seems to trigger if you are issuing
1728      a huge number of AIO writes, such that a just-written page can get
1729      evicted from memory, and then read back into memory, before the
1730      workqueue has a chance to update the extent tree.
1731
1732      This race has been around for a little over a year, and no one noticed
1733      until two months ago; it only happens under fairly exotic conditions,
1734      and in fact even after trying very hard to create a simple repro under
1735      lab conditions, we could only reproduce the problem and confirm the
1736      fix on production servers running MySQL on very fast PCIe-attached
1737      flash devices.
1738
1739      Given that Dave was able to hit this problem pretty quickly, if we
1740      confirm that this commit is at fault, the only reasonable thing to do
1741      is to revert it IMO."
1742
1743    Reported-and-tested-by: Dave Jones <davej@redhat.com>
1744    Acked-by: Theodore Ts'o <tytso@mit.edu>
1745    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
1746commit 4293f20c19f44ca66e5ac836b411d25e14b9f185
1747Author: Linus Torvalds <torvalds@linux-foundation.org>
1748Date:   Wed Mar 7 08:21:19 2012 -0800
1749
1750    Revert "CPU hotplug, cpusets, suspend: Don't touch cpusets during suspend/resume"
1751
1752    This reverts commit 8f2f748b0656257153bcf0941df8d6060acc5ca6.
1753
1754    It causes some odd regression that we have not figured out, and it's too
1755    late in the -rc series to try to figure it out now.
1756
1757    As reported by Konstantin Khlebnikov, it causes consistent hangs on his
1758    laptop (Thinkpad x220: 2x cores + HT).  They can be avoided by adding
1759    calls to "rebuild_sched_domains();" in cpuset_cpu_[in]active() for the
1760    CPU_{ONLINE/DOWN_FAILED/DOWN_PREPARE}_FROZEN cases, but it's not at all
1761    clear why, and it makes no sense.
1762
1763    Konstantin's config doesn't even have CONFIG_CPUSETS enabled, just to
1764    make things even more interesting.  So it's not the cpusets, it's just
1765    the scheduling domains.
1766
1767    So until this is understood, revert.
1768
1769    Bisected-reported-and-tested-by: Konstantin Khlebnikov <khlebnikov@openvz.org>
1770    Acked-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
1771    Acked-by: Ingo Molnar <mingo@elte.hu>
1772    Acked-by: Srivatsa S. Bhat <srivatsa.bhat@linux.vnet.ibm.com>
1773    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
1774commit 951880e634a79884236a575b896abf55c39ae0bf
1775Author: Linus Torvalds <torvalds@linux-foundation.org>
1776Date:   Tue Jan 17 10:19:41 2012 -0800
1777
1778    Revert "capabitlies: ns_capable can use the cap helpers rather than lsm call"
1779
1780    This reverts commit d2a7009f0bb03fa22ad08dd25472efa0568126b9.
1781
1782    J. R. Okajima explains:
1783
1784     "After this commit, I am afraid access(2) on NFS may not work
1785      correctly.  The scenario based upon my guess.
1786       - access(2) overrides the credentials.
1787       - calls inode_permission() -- ... -- generic_permission() --
1788          ns_capable().
1789       - while the old ns_capable() calls security_capable(current_cred()),
1790         the new ns_capable() calls has_ns_capability(current) --
1791         security_capable(__task_cred(t)).
1792
1793      current_cred() returns current->cred which is effective (overridden)
1794      credentials, but __task_cred(current) returns current->real_cred (the
1795      NFSD's credential).  And the overridden credentials by access(2) lost."
1796
1797    Requested-by: J. R. Okajima <hooanon05@yahoo.co.jp>
1798    Acked-by: Eric Paris <eparis@redhat.com>
1799    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
1800commit f423fc627b05f47bc9305f9661630fce30f208f9
1801Author: Linus Torvalds <torvalds@linux-foundation.org>
1802Date:   Wed Jan 4 07:57:22 2012 -0800
1803
1804    Revert "rtc: Expire alarms after the time is set."
1805
1806    This reverts commit 93b2ec0128c431148b216b8f7337c1a52131ef03.
1807
1808    The call to "schedule_work()" in rtc_initialize_alarm() happens too
1809    early, and can cause oopses at bootup
1810
1811    Neil Brown explains why we do it:
1812
1813      "If you set an alarm in the future, then shutdown and boot again after
1814       that time, then you will end up with a timer_queue node which is in
1815       the past.
1816
1817       When this happens the queue gets stuck.  That entry-in-the-past won't
1818       get removed until and interrupt happens and an interrupt won't happen
1819       because the RTC only triggers an interrupt when the alarm is "now".
1820
1821       So you'll find that e.g.  "hwclock" will always tell you that
1822       'select' timed out.
1823
1824       So we force the interrupt work to happen at the start just in case."
1825
1826    and has a patch that convert it to do things in-process rather than with
1827    the worker thread, but right now it's too late to play around with this,
1828    so we just revert the patch that caused problems for now.
1829
1830    Reported-by: Sander Eikelenboom <linux@eikelenboom.it>
1831    Requested-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
1832    Requested-by: John Stultz <john.stultz@linaro.org>
1833    Cc: Neil Brown <neilb@suse.de>
1834    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
1835commit 157e8bf8b4823bfcdefa6c1548002374b61f61df
1836Author: Linus Torvalds <torvalds@linux-foundation.org>
1837Date:   Tue Jan 3 17:32:13 2012 -0800
1838
1839    Revert "rtc: Disable the alarm in the hardware"
1840
1841    This reverts commit c0afabd3d553c521e003779c127143ffde55a16f.
1842
1843    It causes failures on Toshiba laptops - instead of disabling the alarm,
1844    it actually seems to enable it on the affected laptops, resulting in
1845    (for example) the laptop powering on automatically five minutes after
1846    shutdown.
1847
1848    There's a patch for it that appears to work for at least some people,
1849    but it's too late to play around with this, so revert for now and try
1850    again in the next merge window.
1851
1852    See for example
1853
1854            http://bugs.debian.org/652869
1855
1856    Reported-and-bisected-by: Andreas Friedrich <afrie@gmx.net> (Toshiba Tecra)
1857    Reported-by: Antonio-M. Corbi Bellot <antonio.corbi@ua.es> (Toshiba Portege R500)
1858    Reported-by: Marco Santos <marco.santos@waynext.com> (Toshiba Portege Z830)
1859    Reported-by: Christophe Vu-Brugier <cvubrugier@yahoo.fr>  (Toshiba Portege R830)
1860    Cc: Jonathan Nieder <jrnieder@gmail.com>
1861    Requested-by: John Stultz <john.stultz@linaro.org>
1862    Cc: stable@kernel.org  # for the versions that applied this
1863    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
1864commit 3b87487ac5008072f138953b07505a7e3493327f
1865Author: Linus Torvalds <torvalds@linux-foundation.org>
1866Date:   Fri Dec 30 13:24:40 2011 -0800
1867
1868    Revert "clockevents: Set noop handler in clockevents_exchange_device()"
1869
1870    This reverts commit de28f25e8244c7353abed8de0c7792f5f883588c.
1871
1872    It results in resume problems for various people. See for example
1873
1874      http://thread.gmane.org/gmane.linux.kernel/1233033
1875      http://thread.gmane.org/gmane.linux.kernel/1233389
1876      http://thread.gmane.org/gmane.linux.kernel/1233159
1877      http://thread.gmane.org/gmane.linux.kernel/1227868/focus=1230877
1878
1879    and the fedora and ubuntu bug reports
1880
1881      https://bugzilla.redhat.com/show_bug.cgi?id=767248
1882      https://bugs.launchpad.net/ubuntu/+source/linux/+bug/904569
1883
1884    which got bisected down to the stable version of this commit.
1885
1886    Reported-by: Jonathan Nieder <jrnieder@gmail.com>
1887    Reported-by: Phil Miller <mille121@illinois.edu>
1888    Reported-by: Philip Langdale <philipl@overt.org>
1889    Reported-by: Tim Gardner <tim.gardner@canonical.com>
1890    Cc: Thomas Gleixner <tglx@linutronix.de>
1891    Cc: Greg KH <gregkh@suse.de>
1892    Cc: stable@kernel.org    # for stable kernels that applied the original
1893    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
1894commit ed4a51842a9d9e618d4f4c31349b15b974dba5df
1895Author: Linus Torvalds <torvalds@linux-foundation.org>
1896Date:   Fri Dec 16 12:58:39 2011 -0800
1897
1898    Revert "drm/i915: fix infinite recursion on unbind due to ilk vt-d w/a"
1899
1900    This reverts commit eb1711bb94991e93669c5a1b5f84f11be2d51ea1.
1901
1902    It blows up the i915 seqno tracking, resulting in the
1903
1904            BUG_ON(seqno == 0);
1905
1906    in i915_wait_request() triggering, which will cause lock-ups.
1907
1908    See for example
1909      https://bugs.launchpad.net/ubuntu/+source/linux/+bug/903010
1910      https://lkml.org/lkml/2011/12/14/395
1911
1912    Reported-requested-and-tested-by: Dirk Hohndel <dirk@hohndel.org>
1913    Reported-by: Richard Eames <Richard.Eames@flinders.edu.au>
1914    Reported-by: Rocko Requin <rockorequin@hotmail.com>
1915    Acked-by: Daniel Vetter <daniel.vetter@ffwll.ch>
1916    Cc: Dave Airlie <airlied@redhat.com>
1917    Cc: Chris Wilson <chris@chris-wilson.co.uk>
1918    Cc: Keith Packard <keithp@keithp.com>
1919    Cc: Eric Anholt <eric@anholt.net>
1920    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
1921commit b4bbb02934e4511d9083f15c23e90703482e84ad
1922Author: Linus Torvalds <torvalds@linux-foundation.org>
1923Date:   Tue Nov 22 15:09:20 2011 -0800
1924
1925    Revert "of/irq: of_irq_find_parent: check for parent equal to child"
1926
1927    This reverts commit dc9372808412edbc653a675a526c2ee6c0c14a91.
1928
1929    As requested by Ben Herrenschmidt:
1930      "This breaks some powerpc platforms at least.  The practice of having
1931       a node provide an explicit "interrupt-parent" property pointing to
1932       itself is an old trick that we've used in the past to allow a
1933       device-node to have interrupts routed to different controllers.
1934
1935       In that case, the node also contains an interrupt-map, so the node is
1936       its own parent, the interrupt resolution hits the map, which then can
1937       route each individual interrupt to a different parent."
1938
1939    Grant says:
1940      "Ah, nuts, yes that is broken then.  Yes, please revert the commit and
1941       Rob & I will come up with a better solution.
1942
1943       Rob, I think it can be done by explicitly checking for np ==
1944       desc->interrupt_parent in of_irq_init() instead of relying on
1945       of_irq_find_parent() returning NULL."
1946
1947    Requested-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
1948    Acked-by: Grant Likely <grant.likely@secretlab.ca>
1949    Cc: Rob Herring <rob.herring@calxeda.com>
1950    Cc: devicetree-discuss@lists.ozlabs.org
1951    Cc: linuxppc-dev <linuxppc-dev@lists.ozlabs.org>
1952    Cc: Tanmay Inamdar <tinamdar@apm.com>
1953    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
1954commit 5e442a493fc59fa536c76db1fff5b49ca36a88c5
1955Author: Linus Torvalds <torvalds@linux-foundation.org>
1956Date:   Wed Nov 9 18:16:00 2011 -0500
1957
1958    Revert "proc: fix races against execve() of /proc/PID/fd**"
1959
1960    This reverts commit aa6afca5bcaba8101f3ea09d5c3e4100b2b9f0e5.
1961
1962    It escalates of some of the google-chrome SELinux problems with ptrace
1963    ("Check failed: pid_ > 0.  Did not find zygote process"), and Andrew
1964    says that it is also causing mystery lockdep reports.
1965
1966    Reported-by: Alex Villacís Lasso <a_villacis@palosanto.com>
1967    Requested-by: James Morris <jmorris@namei.org>
1968    Requested-by: Andrew Morton <akpm@linux-foundation.org>
1969    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
1970commit 31555213f03bca37d2c02e10946296052f4ecfcd
1971Author: Linus Torvalds <torvalds@linux-foundation.org>
1972Date:   Sun Nov 6 22:22:16 2011 -0800
1973
1974    Revert "hvc_console: display printk messages on console."
1975
1976    This reverts commit 361162459f62dc0826b82c9690a741a940f457f0.
1977
1978    It causes an infinite loop when booting Linux under Xen, as so:
1979
1980      [    2.382984] console [hvc0] enabled
1981      [    2.382984] console [hvc0] enabled
1982      [    2.382984] console [hvc0] enabled
1983      ...
1984
1985    as reported by Konrad Rzeszutek Wilk.  And Rusty reports the same for
1986    lguest.  He goes on to say:
1987
1988       "This is not a concurrency problem: the issue seems to be that
1989        calling register_console() twice on the same struct console is a bad
1990        idea."
1991
1992    and Greg says he'll fix it up properly at some point later. Revert for now.
1993
1994    Reported-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
1995    Reported-by: Rusty Russell <rusty@ozlabs.org>
1996    Requested-by: Stephen Rothwell <sfr@canb.auug.org.au>
1997    Cc: Miche Baker-Harvey <miche@google.com>
1998    Cc: Greg Kroah-Hartman <gregkh@suse.de>
1999    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2000commit 4536e4d1d21c8172402a2217b0fa1880665ace36
2001Author: Linus Torvalds <torvalds@linux-foundation.org>
2002Date:   Thu Nov 3 07:44:04 2011 -0700
2003
2004    Revert "perf: Add PM notifiers to fix CPU hotplug races"
2005
2006    This reverts commit 144060fee07e9c22e179d00819c83c86fbcbf82c.
2007
2008    It causes a resume regression for Andi on his Acer Aspire 1830T post
2009    3.1.  The screen just stays black after wakeup.
2010
2011    Also, it really looks like the wrong way to suspend and resume perf
2012    events: I think they should be done as part of the CPU suspend and
2013    resume, rather than as a notifier that does smp_call_function().
2014
2015    Reported-by: Andi Kleen <andi@firstfloor.org>
2016    Acked-by: Ingo Molnar <mingo@elte.hu>
2017    Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
2018    Cc: Rafael J. Wysocki <rjw@sisk.pl>
2019    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2020commit 69dd3d8e29e294caaf63eb5e8a72d250279f9e5f
2021Author: Linus Torvalds <torvalds@linux-foundation.org>
2022Date:   Tue Aug 23 10:36:51 2011 -0700
2023
2024    Revert "irq: Always set IRQF_ONESHOT if no primary handler is specified"
2025
2026    This reverts commit f3637a5f2e2eb391ff5757bc83fb5de8f9726464.
2027
2028    It turns out that this breaks several drivers, one example being OMAP
2029    boards which use the on-board OMAP UARTs and the omap-serial driver that
2030    will not boot to userspace after the commit.
2031
2032    Paul Walmsley reports that enabling CONFIG_DEBUG_SHIRQ reveals 'IRQ
2033    handler type mismatch' errors:
2034
2035      IRQ handler type mismatch for IRQ 74
2036      current handler: serial idle
2037      ...
2038
2039    and the reason is that setting IRQF_ONESHOT will now result in those
2040    interrupt handlers having different IRQF flags, and thus being
2041    unsharable.  So the commit log in the reverted commit:
2042
2043                                "Since it is required for those users and
2044        there is no difference for others it makes sense to add this flag
2045        unconditionally."
2046
2047    is simply not true: there may not be any difference from a "actions at
2048    irq time", but there is a *big* difference wrt this flag testing irq
2049    management (see __setup_irq() in kernel/irq/manage.c).
2050
2051    One solution may be to stop verifying IRQF_ONESHOT in __setup_irq(), but
2052    right now the safe course of action is to revert the change.  Let's
2053    revisit this in a later merge window.
2054
2055    Reported-by: Paul Walmsley <paul@pwsan.com>
2056    Cc: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
2057    Requested-by: Alan Cox <alan@lxorguk.ukuu.org.uk>
2058    Acked-by: Thomas Gleixner <tglx@linutronix.de>
2059    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2060commit a9f729f0e28bb4e4ab0d9e9e3c1675fe4b910f47
2061Author: Linus Torvalds <torvalds@linux-foundation.org>
2062Date:   Thu Aug 11 08:58:41 2011 -0700
2063
2064    Revert "EDAC: Correct Kconfig dependencies"
2065
2066    This reverts commit af9d220bac41dc3201893e1601cc7c44f7da4498.
2067
2068    It turns out that one was meant to be applied on top of the edac.git
2069    tree in -next that has more i7core_edac changes, but that wasn't clear
2070    in the original email.
2071
2072    Reported-by: Stephen Rothwell <sfr@canb.auug.org.au>
2073    Acked-by: Borislav Petkov <borislav.petkov@amd.com>
2074    Cc: Randy Dunlap <rdunlap@xenotime.net>
2075    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2076commit 13fca640bb8ab611a50e0ba120b186faa2994d6c
2077Author: Linus Torvalds <torvalds@linux-foundation.org>
2078Date:   Wed Jun 15 21:53:52 2011 -0700
2079
2080    Revert "fs/exec.c: use BUILD_BUG_ON for VM_STACK_FLAGS & VM_STACK_INCOMPLETE_SETUP"
2081
2082    This reverts commit 7f81c8890c15a10f5220bebae3b6dfae4961962a.
2083
2084    It turns out that it's not actually a build-time check on x86-64 UML,
2085    which does some seriously crazy stuff with VM_STACK_FLAGS.
2086
2087    The VM_STACK_FLAGS define depends on the arch-supplied
2088    VM_STACK_DEFAULT_FLAGS value, and on x86-64 UML we have
2089
2090      arch/um/sys-x86_64/shared/sysdep/vm-flags.h:
2091
2092            #define VM_STACK_DEFAULT_FLAGS \
2093                    (test_thread_flag(TIF_IA32) ? vm_stack_flags32 : vm_stack_flags)
2094
2095            #define VM_STACK_DEFAULT_FLAGS vm_stack_flags
2096
2097    (yes, seriously: two different #define's for that thing, with the first
2098    one being inside an "#ifdef TIF_IA32")
2099
2100    It's possible that it is UML that should just be fixed in this area, but
2101    for now let's just undo the (very small) optimization.
2102
2103    Reported-by: Randy Dunlap <randy.dunlap@oracle.com>
2104    Acked-by: Andrew Morton <akpm@linux-foundation.org>
2105    Cc: Michal Hocko <mhocko@suse.cz>
2106    Cc: Richard Weinberger <richard@nod.at>
2107    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2108commit bb3d6bf1919a20c56b3257b4ec09e67a9226cfb2
2109Author: Linus Torvalds <torvalds@linux-foundation.org>
2110Date:   Sat Jun 4 07:00:50 2011 +0900
2111
2112    Revert "ASoC: Update cx20442 for TTY API change"
2113
2114    This reverts commit ed0bd2333cffc3d856db9beb829543c1dfc00982.
2115
2116    Since we reverted the TTY API change, we should revert the ASoC update
2117    to it too.
2118
2119    Cc: Mark Brown <broonie@opensource.wolfsonmicro.com>
2120    Cc: Liam Girdwood <lrg@ti.com>
2121    Cc: Greg Kroah-Hartman <gregkh@suse.de>
2122    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2123commit 55db4c64eddf37e31279ec15fe90314713bc9cfa
2124Author: Linus Torvalds <torvalds@linux-foundation.org>
2125Date:   Sat Jun 4 06:33:24 2011 +0900
2126
2127    Revert "tty: make receive_buf() return the amout of bytes received"
2128
2129    This reverts commit b1c43f82c5aa265442f82dba31ce985ebb7aa71c.
2130
2131    It was broken in so many ways, and results in random odd pty issues.
2132
2133    It re-introduced the buggy schedule_work() in flush_to_ldisc() that can
2134    cause endless work-loops (see commit a5660b41af6a: "tty: fix endless
2135    work loop when the buffer fills up").
2136
2137    It also used an "unsigned int" return value fo the ->receive_buf()
2138    function, but then made multiple functions return a negative error code,
2139    and didn't actually check for the error in the caller.
2140
2141    And it didn't actually work at all.  BenH bisected down odd tty behavior
2142    to it:
2143      "It looks like the patch is causing some major malfunctions of the X
2144       server for me, possibly related to PTYs.  For example, cat'ing a
2145       large file in a gnome terminal hangs the kernel for -minutes- in a
2146       loop of what looks like flush_to_ldisc/workqueue code, (some ftrace
2147       data in the quoted bits further down).
2148
2149       ...
2150
2151       Some more data: It -looks- like what happens is that the
2152       flush_to_ldisc work queue entry constantly re-queues itself (because
2153       the PTY is full ?) and the workqueue thread will basically loop
2154       forver calling it without ever scheduling, thus starving the consumer
2155       process that could have emptied the PTY."
2156
2157    which is pretty much exactly the problem we fixed in a5660b41af6a.
2158
2159    Milton Miller pointed out the 'unsigned int' issue.
2160
2161    Reported-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2162    Reported-by: Milton Miller <miltonm@bga.com>
2163    Cc: Stefan Bigler <stefan.bigler@keymile.com>
2164    Cc: Toby Gray <toby.gray@realvnc.com>
2165    Cc: Felipe Balbi <balbi@ti.com>
2166    Cc: Greg Kroah-Hartman <gregkh@suse.de>
2167    Cc: Alan Cox <alan@lxorguk.ukuu.org.uk>
2168    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2169commit 1fa7b6a29c61358cc2ca6f64cef4aa0e1a7ca74c
2170Author: Linus Torvalds <torvalds@linux-foundation.org>
2171Date:   Thu Jun 2 06:11:24 2011 +0900
2172
2173    Revert "mm: fail GFP_DMA allocations when ZONE_DMA is not configured"
2174
2175    This reverts commit a197b59ae6e8bee56fcef37ea2482dc08414e2ac.
2176
2177    As rmk says:
2178     "Commit a197b59ae6e8 (mm: fail GFP_DMA allocations when ZONE_DMA is not
2179      configured) is causing regressions on ARM with various drivers which
2180      use GFP_DMA.
2181
2182      The behaviour up until now has been to silently ignore that flag when
2183      CONFIG_ZONE_DMA is not enabled, and to allocate from the normal zone.
2184      However, as a result of the above commit, such allocations now fail
2185      which causes drivers to fail.  These are regressions compared to the
2186      previous kernel version."
2187
2188    so just revert it.
2189
2190    Requested-by: Russell King <linux@arm.linux.org.uk>
2191    Acked-by: Andrew Morton <akpm@linux-foundation.org>
2192    Cc: David Rientjes <rientjes@google.com>
2193    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2194commit 982b2035d9d7033f63db187bac55e9d8998b0266
2195Author: Linus Torvalds <torvalds@linux-foundation.org>
2196Date:   Thu May 12 12:19:43 2011 -0700
2197
2198    Revert "drm/i915: Only enable the plane after setting the fb base (pre-ILK)"
2199
2200    This reverts commit 49183b2818de6899383bb82bc032f9344d6791ff.
2201
2202    Quoth Franz Melchior:
2203
2204      "This patch introduces a bug on my infamous "Acer Travelmate
2205       5735Z-452G32Mnss": when KMS takes over, the frame buffer contents get
2206       completely garbled up on screen, with colored stripes and unreadable
2207       text (photo on request).  Only when X11 is started, the screen gets
2208       restored again.  Closing and re-opening the lid partly cures the
2209       mess, too: it makes the font readable, though horizontally stretched."
2210
2211    Acked-by: Keith Packard <keithp@keithp.com>
2212    Cc: Chris Wilson <chris@chris-wilson.co.uk>
2213    Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
2214    Cc: Jesse Barnes <jbarnes@virtuousgeek.org>
2215    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2216commit ce8453776d68982cfe93bcb28191af8ccad01f45
2217Author: Linus Torvalds <torvalds@linux-foundation.org>
2218Date:   Wed May 11 18:58:16 2011 -0700
2219
2220    Revert "Bluetooth: fix shutdown on SCO sockets"
2221
2222    This reverts commit f21ca5fff6e548833fa5ee8867239a8378623150.
2223
2224    Quoth Gustavo F. Padovan:
2225      "Commit f21ca5fff6e548833fa5ee8867239a8378623150 can cause a NULL
2226       dereference if we call shutdown in a bluetooth SCO socket and doesn't
2227       wait the shutdown completion to call close().  Please revert it.  I
2228       may have a fix for it soon, but we don't have time anymore, so revert
2229       is the way to go.  ;)"
2230
2231    Requested-by: Gustavo F. Padovan <padovan@profusion.mobi>
2232    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2233commit 71460af58f8565110160283849db4d6bf7e1efa1
2234Author: Linus Torvalds <torvalds@linux-foundation.org>
2235Date:   Tue Apr 19 10:54:44 2011 -0700
2236
2237    Revert "[media] V4L: videobuf, don't use dma addr as physical"
2238
2239    This reverts commit 35d9f510b67b10338161aba6229d4f55b4000f5b.
2240
2241    Quoth Jiri Slaby:
2242     "It fixes mmap when IOMMU is used on x86 only, but breaks architectures
2243      like ARM or PPC where virt_to_phys(dma_alloc_coherent) doesn't work.
2244      We need there dma_mmap_coherent or similar (the trickery what
2245      snd_pcm_default_mmap does but in some saner way).  But this cannot be
2246      done at this phase."
2247
2248    Requested-by: Jiri Slaby <jslaby@suse.cz>
2249    Cc: Russell King - ARM Linux <linux@arm.linux.org.uk>
2250    Cc: Janusz Krzysztofik <jkrzyszt@tis.icnet.pl>
2251    Acked-by: Mauro Carvalho Chehab <mchehab@redhat.com>
2252    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2253commit be85bccaa5aa5a11dcaf85f9e945ffefd253f631
2254Author: Linus Torvalds <torvalds@linux-foundation.org>
2255Date:   Tue Apr 12 13:35:56 2011 -0700
2256
2257    Revert "vfs: Export file system uuid via /proc/<pid>/mountinfo"
2258
2259    This reverts commit 93f1c20bc8cdb757be50566eff88d65c3b26881f.
2260
2261    It turns out that libmount misparses it because it adds a '-' character
2262    in the uuid string, which libmount then incorrectly confuses with the
2263    separator string (" - ") at the end of all the optional arguments.
2264
2265    Upstream libmount (in the util-linux tree) has been fixed, but until
2266    that fix actually percolates up to users, we'd better not expose this
2267    change in the kernel.
2268
2269    Let's revisit this later (possibly by exposing the UUID without any '-'
2270    characters in it, avoiding the user-space bug).
2271
2272    Reported-by: Dave Jones <davej@redhat.com>
2273    Cc: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
2274    Cc: Al Viro <viro@zeniv.linux.org.uk>
2275    Cc: Karel Zak <kzak@redhat.com>
2276    Cc: Ram Pai <linuxram@us.ibm.com>
2277    Cc: Miklos Szeredi <mszeredi@suse.cz>
2278    Cc: Eric Sandeen <sandeen@redhat.com>
2279    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2280commit 52d3c03675fdbe1965b9b1909072b40ad2f80063
2281Author: Linus Torvalds <torvalds@linux-foundation.org>
2282Date:   Mon Mar 14 15:17:07 2011 -0700
2283
2284    Revert "oom: oom_kill_process: fix the child_points logic"
2285
2286    This reverts the parent commit.  I hate doing that, but it's generating
2287    some discussion ("half of it is right"), and since I am planning on
2288    doing the 2.6.38 release later today we can punt it to stable if
2289    required. Let's not rock the boat right now.
2290
2291    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2292commit 8d1dc20e8d689c7e6a0a4d2c94e36a99d5793ecb
2293Author: Linus Torvalds <torvalds@linux-foundation.org>
2294Date:   Tue Mar 1 13:23:27 2011 -0800
2295
2296    Revert "TPM: Long default timeout fix"
2297
2298    This reverts commit c4ff4b829ef9e6353c0b133b7adb564a68054979.
2299
2300    Ted Ts'o reports:
2301
2302     "TPM is working for me so I can log into employer's network in 2.6.37.
2303      It broke when I tried 2.6.38-rc6, with the following relevant lines
2304      from my dmesg:
2305
2306      [   11.081627] tpm_tis 00:0b: 1.2 TPM (device-id 0x0, rev-id 78)
2307      [   25.734114] tpm_tis 00:0b: Operation Timed out
2308      [   78.040949] tpm_tis 00:0b: Operation Timed out
2309
2310      This caused me to get suspicious, especially since the _other_ TPM
2311      commit in 2.6.38 had already been reverted, so I tried reverting
2312      commit c4ff4b829e: "TPM: Long default timeout fix".  With this commit
2313      reverted, my TPM on my Lenovo T410 is once again working."
2314
2315    Requested-and-tested-by: Theodore Ts'o <tytso@mit.edu>
2316    Acked-by: Rajiv Andrade <srajiv@linux.vnet.ibm.com>
2317    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2318commit 78794b2cdeac37ac1fd950fc9c4454b56d88ac03
2319Author: Linus Torvalds <torvalds@linux-foundation.org>
2320Date:   Wed Feb 23 19:42:03 2011 -0800
2321
2322    Revert "Bluetooth: Enable USB autosuspend by default on btusb"
2323
2324    This reverts commit 556ea928f78a390fe16ae584e6433dff304d3014.
2325
2326    Jeff Chua reports that it can cause some bluetooth devices (he mentions
2327    an Bluetooth Intermec scanner) to just stop responding after a while
2328    with messages like
2329
2330      [ 4533.361959] btusb 8-1:1.0: no reset_resume for driver btusb?
2331      [ 4533.361964] btusb 8-1:1.1: no reset_resume for driver btusb?
2332
2333    from the kernel. See also
2334
2335      https://bugzilla.kernel.org/show_bug.cgi?id=26182
2336
2337    for other reports.
2338
2339    Reported-by: Jeff Chua <jeff.chua.linux@gmail.com>
2340    Reported-by: Andrew Meakovski <meako@bigmir.net>
2341    Reported-by: Jim Faulkner <jfaulkne@ccs.neu.edu>
2342    Acked-by: Greg KH <gregkh@suse.de>
2343    Acked-by: Matthew Garrett <mjg@redhat.com>
2344    Acked-by: Gustavo F. Padovan <padovan@profusion.mobi>
2345    Cc: stable@kernel.org (for 2.6.37)
2346    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2347commit e58713724059da7d2982d6ad945192c8fca5b729
2348Author: Linus Torvalds <torvalds@linux-foundation.org>
2349Date:   Sun Feb 20 10:03:12 2011 -0800
2350
2351    Revert "tpm_tis: Use timeouts returned from TPM"
2352
2353    This reverts commit 9b29050f8f75916f974a2d231ae5d3cd59792296.
2354
2355    It has caused hibernate regressions, for example Juri Sladby's report:
2356
2357      "I'm unable to hibernate 2.6.37.1 unless I rmmod tpm_tis:
2358       [10974.074587] Suspending console(s) (use no_console_suspend to debug)
2359       [10974.103073] tpm_tis 00:0c: Operation Timed out
2360       [10974.103089] legacy_suspend(): pnp_bus_suspend+0x0/0xa0 returns -62
2361       [10974.103095] PM: Device 00:0c failed to freeze: error -62"
2362
2363    and Rafael points out that some of the new conditionals in that commit
2364    seem to make no sense.  This commit needs more work and testing, let's
2365    revert it for now.
2366
2367    Reported-by: Norbert Preining <preining@logic.at>
2368    Reported-and-requested-by: Jiri Slaby <jirislaby@gmail.com>
2369    Cc: Stefan Berger <stefanb@linux.vnet.ibm.com>
2370    Cc: Guillaume Chazarain <guichaz@gmail.com>
2371    Cc: Rajiv Andrade <srajiv@linux.vnet.ibm.com>
2372    Acked-by: Rafael J. Wysocki <rjw@sisk.pl>
2373    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2374commit f00eaeea7a42b5ea327e9ce8839cb0b53d3bdb4e
2375Author: Linus Torvalds <torvalds@linux-foundation.org>
2376Date:   Sun Feb 13 07:50:50 2011 -0800
2377
2378    Revert "pci: use security_capable() when checking capablities during config space read"
2379
2380    This reverts commit 47970b1b2aa64464bc0a9543e86361a622ae7c03.
2381
2382    It turns out it breaks several distributions.  Looks like the stricter
2383    selinux checks fail due to selinux policies not being set to allow the
2384    access - breaking X, but also lspci.
2385
2386    So while the change was clearly the RightThing(tm) to do in theory, in
2387    practice we have backwards compatibility issues making it not work.
2388
2389    Reported-by: Dave Young <hidave.darkstar@gmail.com>
2390    Acked-by: David Airlie <airlied@linux.ie>
2391    Acked-by: Alex Riesen <raa.lkml@gmail.com>
2392    Cc: Eric Paris <eparis@redhat.com>
2393    Cc: Chris Wright <chrisw@sous-sol.org>
2394    Cc: James Morris <jmorris@namei.org>
2395    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2396commit 83896fb5e51594281720d145164f866ba769abd5
2397Author: Linus Torvalds <torvalds@linux-foundation.org>
2398Date:   Mon Jan 17 14:42:34 2011 -0800
2399
2400    Revert "mm: simplify code of swap.c"
2401
2402    This reverts commit d8505dee1a87b8d41b9c4ee1325cd72258226fbc.
2403
2404    Chris Mason ended up chasing down some page allocation errors and pages
2405    stuck waiting on the IO scheduler, and was able to narrow it down to two
2406    commits: commit 744ed1442757 ("mm: batch activate_page() to reduce lock
2407    contention") and d8505dee1a87 ("mm: simplify code of swap.c").
2408
2409    This reverts the second one.
2410
2411    Reported-and-debugged-by: Chris Mason <chris.mason@oracle.com>
2412    Cc: Mel Gorman <mel@csn.ul.ie>
2413    Cc: Andrew Morton <akpm@linux-foundation.org>
2414    Cc: Jens Axboe <jaxboe@fusionio.com>
2415    Cc: linux-mm <linux-mm@kvack.org>
2416    Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
2417    Cc: Andrea Arcangeli <aarcange@redhat.com>
2418    Cc: Shaohua Li <shaohua.li@intel.com>
2419    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2420commit 7a608572a282a74978e10fd6cd63090aebe29f5c
2421Author: Linus Torvalds <torvalds@linux-foundation.org>
2422Date:   Mon Jan 17 14:42:19 2011 -0800
2423
2424    Revert "mm: batch activate_page() to reduce lock contention"
2425
2426    This reverts commit 744ed1442757767ffede5008bb13e0805085902e.
2427
2428    Chris Mason ended up chasing down some page allocation errors and pages
2429    stuck waiting on the IO scheduler, and was able to narrow it down to two
2430    commits: commit 744ed1442757 ("mm: batch activate_page() to reduce lock
2431    contention") and d8505dee1a87 ("mm: simplify code of swap.c").
2432
2433    This reverts the first of them.
2434
2435    Reported-and-debugged-by: Chris Mason <chris.mason@oracle.com>
2436    Cc: Mel Gorman <mel@csn.ul.ie>
2437    Cc: Andrew Morton <akpm@linux-foundation.org>
2438    Cc: Jens Axboe <jaxboe@fusionio.com>
2439    Cc: linux-mm <linux-mm@kvack.org>
2440    Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
2441    Cc: Andrea Arcangeli <aarcange@redhat.com>
2442    Cc: Shaohua Li <shaohua.li@intel.com>
2443    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2444commit d8a3515e2a9523f8ed56d1f4537d16338bda2bc2
2445Author: Linus Torvalds <torvalds@linux-foundation.org>
2446Date:   Thu Jan 13 17:26:46 2011 -0800
2447
2448    Revert "gpiolib: annotate gpio-intialization with __must_check"
2449
2450    This reverts commit 0fdae42d361bbb431ca0ab0efed5126a94821177, which
2451    wasn't really supposed to go in, and causes lots of annoying warnings.
2452
2453    Quoth Andrew:
2454      "Complete brainfart - I meant to drop that patch ages ago."
2455
2456    Quoth Greg:
2457      "Ick, yeah, that patch isn't ok to go in as-is, all of the callers
2458       need to be fixed up first, which is what I thought we had agreed on..."
2459
2460    Reported-by: Stephen Rothwell <sfr@canb.auug.org.au>
2461    Acked-by: Andrew Morton <akpm@linux-foundation.org>
2462    Acked-by: Greg KH <greg@kroah.com>
2463    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2464commit 33e0d57f5d2f079104611be9f3fccc27ef2c6b24
2465Author: Linus Torvalds <torvalds@linux-foundation.org>
2466Date:   Fri Nov 19 11:54:40 2010 -0800
2467
2468    Revert "kernel: make /proc/kallsyms mode 400 to reduce ease of attacking"
2469
2470    This reverts commit 59365d136d205cc20fe666ca7f89b1c5001b0d5a.
2471
2472    It turns out that this can break certain existing user land setups.
2473    Quoth Sarah Sharp:
2474
2475     "On Wednesday, I updated my branch to commit 460781b from linus' tree,
2476      and my box would not boot.  klogd segfaulted, which stalled the whole
2477      system.
2478
2479      At first I thought it actually hung the box, but it continued booting
2480      after 5 minutes, and I was able to log in.  It dropped back to the
2481      text console instead of the graphical bootup display for that period
2482      of time.  dmesg surprisingly still works.  I've bisected the problem
2483      down to this commit (commit 59365d136d205cc20fe666ca7f89b1c5001b0d5a)
2484
2485      The box is running klogd 1.5.5ubuntu3 (from Jaunty).  Yes, I know
2486      that's old.  I read the bit in the commit about changing the
2487      permissions of kallsyms after boot, but if I can't boot that doesn't
2488      help."
2489
2490    So let's just keep the old default, and encourage distributions to do
2491    the "chmod -r /proc/kallsyms" in their bootup scripts.  This is not
2492    worth a kernel option to change default behavior, since it's so easily
2493    done in user space.
2494
2495    Reported-and-bisected-by: Sarah Sharp <sarah.a.sharp@linux.intel.com>
2496    Cc: Marcus Meissner <meissner@suse.de>
2497    Cc: Tejun Heo <tj@kernel.org>
2498    Cc: Eugene Teo <eugeneteo@kernel.org>
2499    Cc: Jesper Juhl <jj@chaosbits.net>
2500    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2501commit 6c2754c28f2388a276fe21edde826f2113c8f60e
2502Author: Linus Torvalds <torvalds@linux-foundation.org>
2503Date:   Sat Oct 23 08:14:12 2010 -0700
2504
2505    Revert "tty: Add a new file /proc/tty/consoles"
2506
2507    This reverts commit f4a3e0bceb57466c31757f25e4e0ed108d1299ec.  Jiri
2508    Sladby points out that the tty structure we're using may already be
2509    gone, and Al Viro doesn't hold back in complaining about the random
2510    loading of 'filp->private_data' which doesn't have to be a pointer at
2511    all, nor does checking the magic field for TTY_MAGIC prove anything.
2512
2513    Belated review by Al:
2514
2515     "a) global variable depending on stdin of the last opener? Affecting
2516         output of read(2)? Really?
2517
2518      b) iterator is broken; list should be locked in ->start(), unlocked in
2519         ->stop() and *NOT* unlocked/relocked in ->next()
2520
2521      c) ->show() ought to do nothing in case of ->device == NULL, instead
2522         of skipping those in ->next()/->start()
2523
2524      d) regardless of the merits of the bright idea about asterisk at that
2525         line in output *and* regardless of (a), the implementation is not
2526         only atrociously ugly, it's actually very likely to be a roothole.
2527         Verifying that Cthulhu knows what number happens to be address of a
2528         tty_struct by blindly dereferencing memory at that address...
2529         Ouch.
2530
2531      Please revert that crap."
2532
2533    And Christoph pipes in and NAK's the approach of walking fd tables etc
2534    too.  So it's pretty unanimous.
2535
2536    Noticed-by: Jri Slaby <jslaby@suse.cz>
2537    Requested-by: Al Viro <viro@zeniv.linux.org.uk>
2538    Cc: Greg Kroah-Hartman <gregkh@suse.de>
2539    Cc: Werner Fink <werner@suse.de>
2540    Cc: Alan Cox <alan@lxorguk.ukuu.org.uk>
2541    Cc: Christoph Hellwig <hch@infradead.org>
2542    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2543commit 2069601b3f0ea38170d4b509b89f3ca0a373bdc1
2544Author: Linus Torvalds <torvalds@linux-foundation.org>
2545Date:   Thu Aug 12 14:23:04 2010 -0700
2546
2547    Revert "fsnotify: store struct file not struct path"
2548
2549    This reverts commit 3bcf3860a4ff9bbc522820b4b765e65e4deceb3e (and the
2550    accompanying commit c1e5c954020e "vfs/fsnotify: fsnotify_close can delay
2551    the final work in fput" that was a horribly ugly hack to make it work at
2552    all).
2553
2554    The 'struct file' approach not only causes that disgusting hack, it
2555    somehow breaks pulseaudio, probably due to some other subtlety with
2556    f_count handling.
2557
2558    Fix up various conflicts due to later fsnotify work.
2559
2560    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2561commit 8d86dc6a5bcd0f1d5b9364d43843f1bb4b15f57a
2562Author: Linus Torvalds <torvalds@linux-foundation.org>
2563Date:   Tue Jun 8 20:16:28 2010 -0700
2564
2565    Revert "drm/i915: Don't enable pipe/plane/VCO early (wait for DPMS on)."
2566
2567    This reverts commit cfecde435dda78248d6fcdc424bed68d5db6be0b, since it
2568    seems to cause some systems to not come up with any video output at all
2569    (or video that only comes on when X starts up).
2570
2571    Fixes bugzilla:
2572
2573        http://bugzilla.kernel.org/show_bug.cgi?id=16163
2574
2575    Reported-and-tested-by: David John <davidjon@xenontk.org>
2576    Tested-by: Nick Bowler <nbowler@elliptictech.com>
2577    Acked-by: Carl Worth <cworth@cworth.org>
2578    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2579commit 386f40c86d6c8d5b717ef20620af1a750d0dacb4
2580Author: Linus Torvalds <torvalds@linux-foundation.org>
2581Date:   Sun Jun 6 20:44:04 2010 -0700
2582
2583    Revert "tty: fix a little bug in scrup, vt.c"
2584
2585    This reverts commit 962400e8fd29981a7b166e463dd143b6ac6a3e76, which was
2586    entirely bogus.
2587
2588    The code used to multiply the character offset by "vc->vc_cols", and
2589    that's actually correct, because 'd' itself is an 'unsigned short'.  So
2590    the pointer arithmetic already takes the size of a VGA character into
2591    account.  Changing it to use vc_size_row (which is just "vc_cols"
2592    shifted up to take the size of the character into account) ends up
2593    multiplying with the VGA character size twice.
2594
2595    This got reported as bugs for various other subsystems, because what it
2596    actually results in is writing the 16-bit vc_video_erase_char pattern
2597    (usually 0x0720: 0x07 is the default attribute, 0x20 is ASCII space)
2598    into some random other allocation.
2599
2600    So Markus ended up reporting this as a ext4 bug, while to Torsten Kaiser
2601    it looked like a problem with KMS or libata.  Jeff Chua saw it in
2602    different places.
2603
2604    And finally - Justin Mattock had slab poisoning enabled, and saw it as a
2605    slab poison overwritten.  And bisected and reverted this to verify the
2606    buggy commit.
2607
2608    Reported-by: Markus Trippelsdorf <markus@trippelsdorf.de>
2609    Reported-by: Torsten Kaiser <just.for.lkml@googlemail.com>
2610    Reported-by: Jeff Chua <jeff.chua.linux@gmail.com>
2611    Reported-by: Justin P. Mattock <justinmattock@gmail.com>
2612    Reported-bisected-and-tested-by: Justin P. Mattock <justinmattock@gmail.com>
2613    Acked-by: Dave Airlie <airlied@redhat.com>
2614    Cc: Frank Pan <frankpzh@gmail.com>
2615    Cc: Greg Kroah-Hartman <gregkh@suse.de>
2616    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2617commit 35926ff5fba8245bd1c6ac04155048f6f89232b1
2618Author: Linus Torvalds <torvalds@linux-foundation.org>
2619Date:   Sun May 30 09:00:03 2010 -0700
2620
2621    Revert "cpusets: randomize node rotor used in cpuset_mem_spread_node()"
2622
2623    This reverts commit 0ac0c0d0f837c499afd02a802f9cf52d3027fa3b, which
2624    caused cross-architecture build problems for all the wrong reasons.
2625    IA64 already added its own version of __node_random(), but the fact is,
2626    there is nothing architectural about the function, and the original
2627    commit was just badly done. Revert it, since no fix is forthcoming.
2628
2629    Requested-by: Stephen Rothwell <sfr@canb.auug.org.au>
2630    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2631commit 13da9e200fe4740b02cd51e07ab454627e228920
2632Author: Linus Torvalds <torvalds@linux-foundation.org>
2633Date:   Wed May 26 08:30:15 2010 -0700
2634
2635    Revert "endian: #define __BYTE_ORDER"
2636
2637    This reverts commit b3b77c8caef1750ebeea1054e39e358550ea9f55, which was
2638    also totally broken (see commit 0d2daf5cc858 that reverted the crc32
2639    version of it).  As reported by Stephen Rothwell, it causes problems on
2640    big-endian machines:
2641
2642    > In file included from fs/jfs/jfs_types.h:33,
2643    >                  from fs/jfs/jfs_incore.h:26,
2644    >                  from fs/jfs/file.c:22:
2645    > fs/jfs/endian24.h:36:101: warning: "__LITTLE_ENDIAN" is not defined
2646
2647    The kernel has never had that crazy "__BYTE_ORDER == __LITTLE_ENDIAN"
2648    model.  It's not how we do things, and it isn't how we _should_ do
2649    things.  So don't go there.
2650
2651    Requested-by: Stephen Rothwell <sfr@canb.auug.org.au>
2652    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2653commit 218ce7351413b8287a80fab1d7b94906a5559f01
2654Author: Linus Torvalds <torvalds@linux-foundation.org>
2655Date:   Tue May 25 16:48:30 2010 -0700
2656
2657    Revert "module: drop the lock while waiting for module to complete initialization."
2658
2659    This reverts commit 480b02df3aa9f07d1c7df0cd8be7a5ca73893455, since
2660    Rafael reports that it causes occasional kernel paging request faults in
2661    load_module().
2662
2663    Dropping the module lock and re-taking it deep in the call-chain is
2664    definitely not the right thing to do.  That just turns the mutex from a
2665    lock into a "random non-locking data structure" that doesn't actually
2666    protect what it's supposed to protect.
2667
2668    Requested-and-tested-by: Rafael J. Wysocki <rjw@sisk.pl>
2669    Cc: Rusty Russell <rusty@rustcorp.com.au>
2670    Cc: Brandon Philips <brandon@ifup.org>
2671    Cc: Andrew Morton <akpm@linux-foundation.org>
2672    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2673commit a69eee4988752c7196677958b4ed8f4c2b28499a
2674Author: Linus Torvalds <torvalds@linux-foundation.org>
2675Date:   Mon May 24 07:45:43 2010 -0700
2676
2677    Revert "ath9k: Group Key fix for VAPs"
2678
2679    This reverts commit 03ceedea972a82d343fa5c2528b3952fa9e615d5, since it
2680    breaks resume from suspend-to-ram on Rafael's Acer Ferrari One.
2681    NetworkManager thinks everything is ok, but it can't connect to the AP
2682    to get an IP address after the resume.
2683
2684    In fact, it even breaks resume for non-ath9k chipsets: reverting it also
2685    fixes Rafael's Toshiba Protege R500 with the iwlagn driver.  As Johannes
2686    says:
2687
2688      "Indeed, this patch needs to be reverted. That mac80211 change is wrong
2689       and completely unnecessary."
2690
2691    Reported-and-requested-by: Rafael J. Wysocki <rjw@sisk.pl>
2692    Acked-by: Johannes Berg <johannes@sipsolutions.net>
2693    Cc: Daniel Yingqiang Ma <yma.cool@gmail.com>
2694    Cc: John W. Linville <linville@tuxdriver.com>
2695    Cc: David Miller <davem@davemloft.net>
2696    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2697commit 769d9968e42c995eaaf61ac5583d998f32e0769a
2698Author: Linus Torvalds <torvalds@linux-foundation.org>
2699Date:   Wed May 12 18:39:45 2010 -0700
2700
2701    Revert "PCI: update bridge resources to get more big ranges in PCI assign unssigned"
2702
2703    This reverts commit 977d17bb1749517b353874ccdc9b85abc7a58c2a, because it
2704    can cause problems with some devices not getting any resources at all
2705    when the resource tree is re-allocated.
2706
2707    For an example of this, see
2708
2709            https://bugzilla.kernel.org/show_bug.cgi?id=15960
2710            (originally https://bugtrack.alsa-project.org/alsa-bug/view.php?id=4982)
2711            (lkml thread: http://lkml.org/lkml/2010/4/19/20)
2712
2713    where Peter Henriksson reported his Xonar DX sound card gone, because
2714    the IO port region was no longer allocated.
2715
2716    Reported-bisected-and-tested-by: Peter Henriksson <peter.henriksson@gmail.com>
2717    Requested-by: Andrew Morton <akpm@linux-foundation.org>
2718    Requested-by: Clemens Ladisch <clemens@ladisch.de>
2719    Acked-by: Jesse Barnes <jbarnes@virtuousgeek.org>
2720    Cc: Yinghai Lu <yinghai@kernel.org>
2721    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2722commit 4dc86ae1f925b2121d4e75058675895f83e54c71
2723Author: Linus Torvalds <torvalds@linux-foundation.org>
2724Date:   Fri Apr 9 10:05:33 2010 -0700
2725
2726    Revert "memory-hotplug: add 0x prefix to HEX block_size_bytes"
2727
2728    This reverts commit ba168fc37dea145deeb8fa9e7e71c748d2e00d74.
2729
2730    It changes user-visible sysfs interfaces, and breaks some existing user
2731    space applications which apparently rely on the fact that the output
2732    does not contain the "0x" prefix.
2733
2734    Requested-by: Heiko Carstens <heiko.carstens@de.ibm.com>
2735    Acked-by: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
2736    Acked-by: Wu Fengguang <fengguang.wu@intel.com>
2737    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2738commit b8fa05719ba4349be80ce929237249b57886a203
2739Author: Linus Torvalds <torvalds@linux-foundation.org>
2740Date:   Sun Mar 7 09:54:44 2010 -0800
2741
2742    Revert "lib: build list_sort() only if needed"
2743
2744    This reverts commit a069c266ae5fdfbf5b4aecf2c672413aa33b2504.
2745
2746    It turns ou that not only was it missing a case (XFS) that needed it,
2747    but perhaps more importantly, people sometimes want to enable new
2748    modules that they hadn't had enabled before, and if such a module uses
2749    list_sort(), it can't easily be inserted any more.
2750
2751    So rather than add a "select LIST_SORT" to the XFS case, just leave it
2752    compiled in.  It's not all _that_ big, after all, and the inconvenience
2753    isn't worth it.
2754
2755    Requested-by: Alexey Dobriyan <adobriyan@gmail.com>
2756    Cc: Christoph Hellwig <hch@infradead.org>
2757    Cc: Don Mullis <don.mullis@gmail.com>
2758    Cc: Andrew Morton <akpm@linux-foundation.org>
2759    Cc: Dave Chinner <david@fromorbit.com>
2760    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2761commit 2f99f5c8f05e02f3df1bb4d93b6704e6f5972872
2762Author: Linus Torvalds <torvalds@linux-foundation.org>
2763Date:   Wed Dec 23 15:04:53 2009 -0800
2764
2765    Revert "x86, ucode-amd: Ensure ucode update on suspend/resume after CPU off/online cycle"
2766
2767    This reverts commit 9f15226e75583547aaf542c6be4bdac1060dd425.  It's just
2768    wrong, and broke resume for Rafael even on a non-AMD CPU.
2769
2770    As Rafael says:
2771     "... it causes microcode_init_cpu() to be called during resume even for
2772      CPUs for which there's no microcode to apply.  That, in turn, results
2773      in executing request_firmware() (on Intel CPUs at least) which doesn't
2774      work at this stage of resume (we have device interrupts disabled, I/O
2775      devices are still suspended and so on).
2776
2777      If I'm not mistaken, the "if (uci->valid)" logic means "if that CPU is
2778      known to us" , so before commit 9f15226e755 microcode_resume_cpu() was
2779      called for all CPUs already in the system during suspend, which was
2780      the right thing to do.  The commit changed it so that the CPUs without
2781      microcode to apply are now treated as "unknown", which is not quite
2782      right.
2783
2784      The problem this commit attempted to solve has to be handled
2785      differently."
2786
2787    Bisected-and -requested-by: Rafael J. Wysocki <rjw@sisk.pl>
2788    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2789commit 83f57a11d84460dfe2afdb5a8bc759953428e38b
2790Author: Linus Torvalds <torvalds@linux-foundation.org>
2791Date:   Tue Dec 22 14:10:37 2009 -0800
2792
2793    Revert "time: Remove xtime_cache"
2794
2795    This reverts commit 7bc7d637452383d56ba4368d4336b0dde1bb476d, as
2796    requested by John Stultz. Quoting John:
2797
2798     "Petr Titěra reported an issue where he saw odd atime regressions with
2799      2.6.33 where there were a full second worth of nanoseconds in the
2800      nanoseconds field.
2801
2802      He also reviewed the time code and narrowed down the problem: unhandled
2803      overflow of the nanosecond field caused by rounding up the
2804      sub-nanosecond accumulated time.
2805
2806      Details:
2807
2808       * At the end of update_wall_time(), we currently round up the
2809      sub-nanosecond portion of accumulated time when storing it into xtime.
2810      This was added to avoid time inconsistencies caused when the
2811      sub-nanosecond portion was truncated when storing into xtime.
2812      Unfortunately we don't handle the possible second overflow caused by
2813      that rounding.
2814
2815       * Previously the xtime_cache code hid this overflow by normalizing the
2816      xtime value when storing into the xtime_cache.
2817
2818       * We could try to handle the second overflow after the rounding up, but
2819      since this affects the timekeeping's internal state, this would further
2820      complicate the next accumulation cycle, causing small errors in ntp
2821      steering. As much as I'd like to get rid of it, the xtime_cache code is
2822      known to work.
2823
2824       * The correct fix is really to include the sub-nanosecond portion in the
2825      timekeeping accessor function, so we don't need to round up at during
2826      accumulation. This would greatly simplify the accumulation code.
2827      Unfortunately, we can't do this safely until the last three
2828      non-GENERIC_TIME arches (sparc32, arm, cris) are converted  (those
2829      patches are in -mm) and we kill off the spots where arches set xtime
2830      directly. This is all 2.6.34 material, so I think reverting the
2831      xtime_cache change is the best approach for now.
2832
2833      Many thanks to Petr for both reporting and finding the issue!"
2834
2835    Reported-by: Petr Titěra <P.Titera@century.cz>
2836    Requested-by: john stultz <johnstul@us.ibm.com>
2837    Cc: Ingo Molnar <mingo@elte.hu>
2838    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2839commit b6e3224fb20954f155e41ec5709b2ab70b50ae2d
2840Author: Linus Torvalds <torvalds@linux-foundation.org>
2841Date:   Thu Dec 17 13:23:24 2009 -0800
2842
2843    Revert "task_struct: make journal_info conditional"
2844
2845    This reverts commit e4c570c4cb7a95dbfafa3d016d2739bf3fdfe319, as
2846    requested by Alexey:
2847
2848     "I think I gave a good enough arguments to not merge it.
2849      To iterate:
2850       * patch makes impossible to start using ext3 on EXT3_FS=n kernels
2851         without reboot.
2852       * this is done only for one pointer on task_struct"
2853
2854      None of config options which define task_struct are tristate directly
2855      or effectively."
2856
2857    Requested-by: Alexey Dobriyan <adobriyan@gmail.com>
2858    Acked-by: Andrew Morton <akpm@linux-foundation.org>
2859    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2860commit a2770d86b33024f71df269fde2de096df89d6a48
2861Author: Linus Torvalds <torvalds@linux-foundation.org>
2862Date:   Thu Dec 17 12:51:05 2009 -0800
2863
2864    Revert "fix mismerge with Trond's stuff (create_mnt_ns() export is gone now)"
2865
2866    This reverts commit e9496ff46a20a8592fdc7bdaaf41b45eb808d310. Quoth Al:
2867
2868     "it's dependent on a lot of other stuff not currently in mainline
2869      and badly broken with current fs/namespace.c.  Sorry, badly
2870      out-of-order cherry-pick from old queue.
2871
2872      PS: there's a large pending series reworking the refcounting and
2873      lifetime rules for vfsmounts that will, among other things, allow to
2874      rip a subtree away _without_ dissolving connections in it, to be
2875      garbage-collected when all active references are gone.  It's
2876      considerably saner wrt "is the subtree busy" logics, but it's nowhere
2877      near being ready for merge at the moment; this changeset is one of the
2878      things becoming possible with that sucker, but it certainly shouldn't
2879      have been picked during this cycle.  My apologies..."
2880
2881    Noticed-by: Eric Paris <eparis@redhat.com>
2882    Requested-by: Al Viro <viro@ZenIV.linux.org.uk>
2883    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2884commit d4da6c9ccf648f3f1cb5bf9d981a62c253d30e28
2885Author: Linus Torvalds <torvalds@linux-foundation.org>
2886Date:   Mon Nov 2 10:15:27 2009 -0800
2887
2888    Revert "ext4: Remove journal_checksum mount option and enable it by default"
2889
2890    This reverts commit d0646f7b636d067d715fab52a2ba9c6f0f46b0d7, as
2891    requested by Eric Sandeen.
2892
2893    It can basically cause an ext4 filesystem to miss recovery (and thus get
2894    mounted with errors) if the journal checksum does not match.
2895
2896    Quoth Eric:
2897
2898       "My hand-wavy hunch about what is happening is that we're finding a
2899        bad checksum on the last partially-written transaction, which is
2900        not surprising, but if we have a wrapped log and we're doing the
2901        initial scan for head/tail, and we abort scanning on that bad
2902        checksum, then we are essentially running an unrecovered filesystem.
2903
2904        But that's hand-wavy and I need to go look at the code.
2905
2906        We lived without journal checksums on by default until now, and at
2907        this point they're doing more harm than good, so we should revert
2908        the default-changing commit until we can fix it and do some good
2909        power-fail testing with the fixes in place."
2910
2911    See
2912
2913            http://bugzilla.kernel.org/show_bug.cgi?id=14354
2914
2915    for all the gory details.
2916
2917    Requested-by: Eric Sandeen <sandeen@redhat.com>
2918    Cc: Theodore Tso <tytso@mit.edu>
2919    Cc: Alexey Fisher <bug-track@fisher-privat.net>
2920    Cc: Maxim Levitsky <maximlevitsky@gmail.com>
2921    Cc: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
2922    Cc: Mathias Burén <mathias.buren@gmail.com>
2923    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2924commit d93a8f829fe1d2f3002f2c6ddb553d12db420412
2925Author: Linus Torvalds <torvalds@linux-foundation.org>
2926Date:   Sun Oct 11 15:57:57 2009 -0700
2927
2928    Revert "USB: Work around BIOS bugs by quiescing USB controllers earlier"
2929
2930    This reverts commit db8be50c4307dac2b37305fc59c8dc0f978d09ea, as per
2931
2932            http://bugzilla.kernel.org/show_bug.cgi?id=14374
2933            http://marc.info/?l=linux-kernel&m=125446885705223&w=4
2934
2935    We simply can't do the USB handoff at FIXUP_HEADER time, since it will
2936    often require us to have valid IO mappings etc.  But that in turn
2937    requires a whole different approach, not this trivial one-liner.
2938
2939    Maybe we could teach all the USB quirk handoff handlers to only do the
2940    quirk if the device has all its registers set up (since if it isn't
2941    initialized, it's unlikely to be active), but regardless that will need
2942    a whole lot more code than just saying "let's do it really early".
2943
2944    The proper fix is almost certainly to just leave the legacy IOMMU
2945    mappings active until after all devices have been initialized.
2946
2947    Reported-by: Nick Piggin <npiggin@suse.de>
2948    Cc: David Woodhouse <David.Woodhouse@intel.com>
2949    Cc: Greg Kroah-Hartman <gregkh@suse.de>
2950    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2951commit e207e143e2fb6a2790b1ce3687c8aedc3ddc357b
2952Author: Linus Torvalds <torvalds@linux-foundation.org>
2953Date:   Wed Sep 30 07:48:37 2009 -0700
2954
2955    Revert "x86, mce: do not compile mcelog message on AMD"
2956
2957    This reverts commit 22223c9b417be5fd0ab2cf9ad17eb7bd1e19f7b9, as
2958    requested by Andi Kleen:
2959
2960      "Obviously kernels compiled with AMD support can still run on non AMD
2961       systems, so messages like this can never be removed at compile time."
2962
2963    Requsted-by: Andi Kleen <andi@firstfloor.org>
2964    Cc: Borislav Petkov <borislav.petkov@amd.com>
2965    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2966commit 4a21b8cb3550f19f838f7c48345fbbf6a0e8536b
2967Author: Linus Torvalds <torvalds@linux-foundation.org>
2968Date:   Thu Jul 16 09:14:23 2009 -0700
2969
2970    Revert "ppp: Fix throttling bugs"
2971
2972    This reverts commit a6540f731d506d9e82444cf0020e716613d4c46c, as
2973    requested by Alan:
2974
2975      "... as it was wrong, the pty code is now fixed and the fact this
2976       isn't reverted is breaking pptp setups."
2977
2978    Requested-by: Alan Cox <alan@lxorguk.ukuu.org.uk>
2979    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2980commit 81e4e1ba7ed4a1fdcf0e2ee944f1575010471464
2981Author: Linus Torvalds <torvalds@linux-foundation.org>
2982Date:   Sat Jul 11 11:22:34 2009 -0700
2983
2984    Revert "fuse: Fix build error" as unnecessary
2985
2986    This reverts commit 097041e576ee3a50d92dd643ee8ca65bf6a62e21.
2987
2988    Trond had a better fix, which is the parent of this one ("Fix compile
2989    error due to congestion_wait() changes")
2990
2991    Requested-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2992    Acked-by: Larry Finger <Larry.Finger@lwfinger.net>
2993    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2994commit 6c96895e999f5c44a95a5cad6a6e32e7bd6e28b6
2995Author: Linus Torvalds <torvalds@linux-foundation.org>
2996Date:   Wed Jul 8 09:20:11 2009 -0700
2997
2998    Revert "fb: Initialize fb_info mutexes in framebuffer_alloc()"
2999
3000    This reverts commit 4148df9b0f38bdd362dd91d52076926c11cbe5a9.
3001
3002    Let's hope that the mm_lock initialization is now correct with all
3003    drivers, following Krzysztof's patches.
3004
3005    Requested-by: Krzysztof Helt <krzysztof.h1@poczta.fm>
3006    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
3007commit 236e946b53ffd5e2f5d7e6abebbe72a9f0826d15
3008Author: Linus Torvalds <torvalds@linux-foundation.org>
3009Date:   Wed Jun 24 16:23:03 2009 -0700
3010
3011    Revert "PCI: use ACPI _CRS data by default"
3012
3013    This reverts commit 9e9f46c44e487af0a82eb61b624553e2f7118f5b.
3014
3015    Quoting from the commit message:
3016
3017     "At this point, it seems to solve more problems than it causes, so let's
3018      try using it by default.  It's an easy revert if it ends up causing
3019      trouble."
3020
3021    And guess what? The _CRS code causes trouble.
3022
3023    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
3024commit 752a4787511bf7515f99609ff4ae52341b5bfcde
3025Author: Linus Torvalds <torvalds@linux-foundation.org>
3026Date:   Mon Jun 22 11:24:43 2009 -0700
3027
3028    Revert "char: moxa, prevent opening unavailable ports"
3029
3030    This reverts commit a90b037583d5f1ae3e54e9c687c79df82d1d34a4, which
3031    already got fixed as commit f0e8527726b9e56649b9eafde3bc0fbc4dd2dd47:
3032    the same patch (trivial differences) got applied twice.
3033
3034    Requested-by: Alan Cox <alan@lxorguk.ukuu.org.uk>
3035    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
3036commit 27704a16c9e0fb4c6b04344c7c4c40ac16148ec0
3037Author: Linus Torvalds <torvalds@linux-foundation.org>
3038Date:   Fri Jun 5 10:21:52 2009 -0700
3039
3040    Revert "drm: don't associate _DRM_DRIVER maps with a master"
3041
3042    This reverts commit 6c51d1cfa0a370b48a157163340190cf5fd2346b, which
3043    apparently causes DRI initialization failures on Radeons.
3044
3045    Reported-by: Markus Trippelsdorf <markus@trippelsdorf.de>
3046    Requested-by: Dave Airlie <airlied@gmail.com>
3047    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
3048commit 0f181328287db30671e9997329cff71395d4af8b
3049Author: Linus Torvalds <torvalds@linux-foundation.org>
3050Date:   Wed May 13 08:29:12 2009 -0700
3051
3052    Revert "Ignore madvise(MADV_WILLNEED) for hugetlbfs-backed regions"
3053
3054    This reverts commit a425a638c858fd10370b573bde81df3ba500e271.
3055
3056    Now that the previous commit removed the "readpage" actor for hugetlb
3057    files, read-ahead will no longer mess up the mapping, and there's no
3058    longer any reason to treat hugetlbfs mappings specially.
3059
3060    Tested-and-acked-by: Mel Gorman <mel@csn.ul.ie>
3061    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
3062commit 3d4f16348b77efbf81b7fa186a18a0eb815b6b84
3063Author: Linus Torvalds <torvalds@linux-foundation.org>
3064Date:   Mon Apr 27 12:00:27 2009 -0700
3065
3066    Revert "linux.conf.au 2009: Tuz"
3067
3068    This reverts commit 8032b526d1a3bd91ad633dd3a3b5fdbc47ad54f1.
3069
3070    Hey, it was only meant to be a single release.  Now they can all die as
3071    far as I'm concerned.
3072
3073    [ Just kidding.  They're cute and cuddly.
3074
3075      Except when they have horrible nasty facial diseases.  Oh, and I guess
3076      they're not actually that cuddly even when disease-free. ]
3077
3078    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
3079commit b0cbc861a3c05e634520b049b5cc27ad6febb51f
3080Author: Linus Torvalds <torvalds@linux-foundation.org>
3081Date:   Sat Apr 11 12:45:20 2009 -0700
3082
3083    Revert "ACPI battery: fix async boot oops"
3084
3085    This reverts commit 5d38258ec026921a7b266f4047ebeaa75db358e5, since the
3086    underlying problem got fixed properly in the previous commit ("async:
3087    Fix module loading async-work regression").
3088
3089    Cc: Arkadiusz Miskiewicz <a.miskiewicz@gmail.com>
3090    Cc: Vegard Nossum <vegard.nossum@gmail.com>
3091    Cc: Len Brown <len.brown@intel.com>
3092    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
3093commit 2450cf51a1bdba7037e91b1bcc494b01c58aaf66
3094Author: Linus Torvalds <torvalds@linux-foundation.org>
3095Date:   Mon Mar 2 16:23:33 2009 -0800
3096
3097    Revert "menu: fix embedded menu snafu"
3098
3099    This reverts commit 155b25bcc28631a5b5230191aa3f56c40dfffa3f, which was
3100    totally wrong - the "embedded" options still exists (very much so) even
3101    on non-embedded platforms.
3102
3103    It's just that we don't bother with actually asking about them when
3104    we're not embedded, we just take their default values (which is usually
3105    'y' - the options add features that may not be worth it in a constrained
3106    environment).
3107
3108    Noticed-by: Andrew Morton <akpm@linux-foundation.org>
3109    Cc: Randy Dunlap <randy.dunlap@oracle.com>
3110    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
3111commit 50246dd41ccbcb47beb06d6c1d9355f6b7137a11
3112Author: Linus Torvalds <torvalds@linux-foundation.org>
3113Date:   Fri Jan 16 08:14:51 2009 -0800
3114
3115    Revert "PCI PM: Register power state of devices during initialization"
3116
3117    This reverts commit 98e6e286d7b01deb7453b717aa38ebb69d6cefc0, as Yinghai
3118    Lu reports that it breaks kexec with at least the e1000 and e1000e
3119    drivers.  The reason is that the shutdown sequence puts the hardware
3120    into D3 sleep, and the commit causes us to claim that it then is in D0
3121    (running) state just because we don't understand the PM capabilities.
3122
3123    Which then later makes "pci_set_power_state()" not do anything, and the
3124    device never wakes up properly and just returns 0xff to everything.
3125
3126    Reported-by: Yinghai Lu <yinghai@kernel.org>
3127    Acked-by: From: Rafael J. Wysocki <rjw@sisk.pl>
3128    Cc: Jesse Barnes <jesse.barnes@intel.com>
3129    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
3130commit b5db0e38653bfada34a92f360b4111566ede3842
3131Author: Linus Torvalds <torvalds@linux-foundation.org>
3132Date:   Thu Jan 15 15:32:12 2009 -0800
3133
3134    Revert "x86 PAT: remove CPA WARN_ON for zero pte"
3135
3136    This reverts commit 58dab916dfb57328d50deb0aa9b3fc92efa248ff, which
3137    makes my Nehalem come to a nasty crawling almost-halt.  It looks like it
3138    turns off caching of regular kernel RAM, with the understandable
3139    slowdown of a few orders of magnitude as a result.
3140
3141    Acked-by: Ingo Molnar <mingo@elte.hu>
3142    Cc: Yinghai Lu <yinghai@kernel.org>
3143    Cc: Peter Anvin <hpa@zytor.com>
3144    Cc: Venkatesh Pallipadi <venkatesh.pallipadi@intel.com>
3145    Cc: Suresh Siddha <suresh.b.siddha@intel.com>
3146    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
3147commit ca7e716c7833aeaeb8fedd6d004c5f5d5e14d325
3148Author: Linus Torvalds <torvalds@linux-foundation.org>
3149Date:   Sun Dec 14 15:46:01 2008 -0800
3150
3151    Revert "sched_clock: prevent scd->clock from moving backwards"
3152
3153    This reverts commit 5b7dba4ff834259a5623e03a565748704a8fe449, which
3154    caused a regression in hibernate, reported by and bisected by Fabio
3155    Comolli.
3156
3157    This revert fixes
3158
3159     http://bugzilla.kernel.org/show_bug.cgi?id=12155
3160     http://bugzilla.kernel.org/show_bug.cgi?id=12149
3161
3162    Bisected-by: Fabio Comolli <fabio.comolli@gmail.com>
3163    Requested-by: Rafael J. Wysocki <rjw@sisk.pl>
3164    Acked-by: Dave Kleikamp <shaggy@linux.vnet.ibm.com>
3165    Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
3166    Cc: Ingo Molnar <mingo@elte.hu>
3167    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
3168commit 6c34bc2976b30dc8b56392c020e25bae1f363cab
3169Author: Linus Torvalds <torvalds@linux-foundation.org>
3170Date:   Wed Dec 10 09:26:17 2008 -0800
3171
3172    Revert "radeonfb: accelerate imageblit and other improvements"
3173
3174    This reverts commit b1ee26bab14886350ba12a5c10cbc0696ac679bf, along with
3175    the "fixes" for it that all just caused problems:
3176
3177     - c4c6fa9891f3d1bcaae4f39fb751d5302965b566 "radeonfb: fix problem with
3178       color expansion & alignment"
3179
3180     - f3179748a157c21d44d929fd3779421ebfbeaa93 "radeonfb: Disable new color
3181       expand acceleration unless explicitely enabled"
3182
3183    because even when disabled, it breaks for people. See
3184
3185            http://bugzilla.kernel.org/show_bug.cgi?id=12191
3186
3187    for the latest example.
3188
3189    Acked-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
3190    Acked-by: David S. Miller <davem@davemloft.net>
3191    Cc: Krzysztof Halasa <khc@pm.waw.pl>
3192    Cc: James Cloos <cloos@jhcloos.com>
3193    Cc: "Rafael J. Wysocki" <rjw@sisk.pl>
3194    Cc: Krzysztof Helt <krzysztof.h1@poczta.fm>
3195    Cc: Jean-Luc Coulon <jean.luc.coulon@gmail.com>
3196    Cc: Andrew Morton <akpm@linux-foundation.org>
3197    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
3198commit f10a3a32ae2edbe94920ce6827f4c2550eb6ed48
3199Author: Linus Torvalds <torvalds@linux-foundation.org>
3200Date:   Fri Dec 5 13:30:03 2008 -0800
3201
3202    Revert "ACPI: battery: Convert discharge energy rate to current properly"
3203
3204    This reverts commit 558073dd56707864f09d563b64e7c37c021e89d2, along with
3205    the failed try to fix the regression it caused ("ACPI: Fix ACPI battery
3206    regression introduced by commit 558073"), which just made things worse.
3207
3208    Commit aaad077638be1a25871bcae5e43952d6b63abfca (that failed "Fix ACPI
3209    battery regression") got the voltage conversion confused, and fixed the
3210    problem with Rafael's battery monitor apparently just by mistake.
3211
3212    So revert them both, getting us back to the 2.6.27 state in this, and
3213    let's revisit it when people understand what's going on.
3214
3215    Noted-by: Paul Martin <pm@debian.org>
3216    Requested-by: Rafael J. Wysocki <rjw@sisk.pl>
3217    Cc: Len Brown <len.brown@intel.com>
3218    Cc: Alexey Starikovskiy <astarikovskiy@suse.de>
3219    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
3220commit 33b07db9f38fe73b3895f8d4db8fdee03e3afec3
3221Author: Linus Torvalds <torvalds@linux-foundation.org>
3222Date:   Mon Dec 1 07:55:14 2008 -0800
3223
3224    Revert "of_platform_driver noise on sparce"
3225
3226    This reverts commit e669dae6141ff97d3c7566207f5de3b487dcf837, since it
3227    is incomplete, and clashes with fuller patches and the sparc 32/64
3228    unification effort.
3229
3230    Requested-by: David Miller <davem@davemloft.net>
3231    Acked-by: Al Viro <viro@ZenIV.linux.org.uk>
3232    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
3233commit 4403b406d4369a275d483ece6ddee0088cc0d592
3234Author: Linus Torvalds <torvalds@linux-foundation.org>
3235Date:   Sat Oct 25 19:53:38 2008 -0700
3236
3237    Revert "Call init_workqueues before pre smp initcalls."
3238
3239    This reverts commit a802dd0eb5fc97a50cf1abb1f788a8f6cc5db635 by moving
3240    the call to init_workqueues() back where it belongs - after SMP has been
3241    initialized.
3242
3243    It also moves stop_machine_init() - which needs workqueues - to a later
3244    phase using a core_initcall() instead of early_initcall().  That should
3245    satisfy all ordering requirements, and was apparently the reason why
3246    init_workqueues() was moved to be too early.
3247
3248    Cc: Heiko Carstens <heiko.carstens@de.ibm.com>
3249    Cc: Rusty Russell <rusty@rustcorp.com.au>
3250    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
3251commit 93f78da405685a756beeaeae4b5e41fcec39eab3
3252Author: Linus Torvalds <torvalds@linux-foundation.org>
3253Date:   Tue Oct 14 12:12:02 2008 -0700
3254
3255    Revert "vt: fix background color on line feed"
3256
3257    This reverts commit c9e587abfdec2c2aaa55fab83bcb4972e2f84f9b, and the
3258    subsequent commits that fixed it up:
3259
3260     - afa9b649 "fbcon: prevent cursor disappearance after switching to 512
3261       character font"
3262
3263     - d850a2fa "vt/fbcon: fix background color on line feed"
3264
3265     - 7fe3915a "vt/fbcon: update scrl_erase_char after 256/512-glyph font
3266       switch"
3267
3268    by request of Alan Cox. Quoth Alan:
3269      "Unfortunately it's wrong and its been causing breakages because
3270       various apps like ncurses expect our previous (and correct)
3271       behaviour."
3272
3273    Alexander sent out a similar patch.
3274
3275    Requested-by: Alan Cox <alan@lxorguk.ukuu.org.uk>
3276    Tested-by: Jan Engelhardt <jengelh@medozas.de>
3277    Cc: Alexander V. Lukyanov <lav@netis.ru>
3278    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
3279commit e09e6e2b6a5daf653794926ab50a784b14b6de53
3280Author: Linus Torvalds <torvalds@linux-foundation.org>
3281Date:   Tue Oct 7 07:54:34 2008 -0700
3282
3283    Revert "V4L/DVB (8904): cx88: add missing unlock_kernel"
3284
3285    This reverts commit 135aedc38e812b922aa56096f36a3d72ffbcf2fb, as
3286    requested by Hans Verkuil.
3287
3288    It was a patch for 2.6.28 where the BKL was pushed down from v4l core to
3289    the drivers, not for 2.6.27!
3290
3291    Requested-by: Hans Verkuil <hverkuil@xs4all.nl>
3292    Cc: Mauro Carvalho Chehab <mchehab@infradead.org>
3293    Signed-of-by: Linus Torvalds <torvalds@linux-foundation.org>
3294commit e95926d05d028a6bf0ab60b21b484c3d622fdcd1
3295Author: Linus Torvalds <torvalds@linux-foundation.org>
3296Date:   Tue Sep 16 19:32:30 2008 -0700
3297
3298    Revert "b43/b43legacy: add RFKILL_STATE_HARD_BLOCKED support"
3299
3300    This reverts commit bc19d6e0b74ef03a3baf035412c95192b54dfc6f, which as
3301    Larry Finger reports causes the radio LED on his system to no longer
3302    respond to rfkill switch events.
3303
3304    Reported-by: Larry Finger <Larry.Finger@lwfinger.net>
3305    Requested-by: John W. Linville <linville@tuxdriver.com>
3306    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
3307commit bf7394ccc13fe291d9258f01113b4c61214ddeae
3308Author: Linus Torvalds <torvalds@linux-foundation.org>
3309Date:   Fri Sep 5 12:38:09 2008 -0700
3310
3311    Revert "mac80211: Use IWEVASSOCREQIE instead of IWEVCUSTOM"
3312
3313    This reverts commit 087d833e5a9f67ba933cb32eaf5a2279c1a5b47c, which was
3314    reported to break wireless at least in some combinations with 32bit user
3315    space and a 64bit kernel.  Alex Williamnson bisected it to this commit.
3316
3317    Reported-and-bisected-by: Alex Williamson <alex.williamson@hp.com>
3318    Acked-by: John W. Linville <linville@tuxdriver.com>
3319    Cc: David Miller <davem@davemloft.net>
3320    Cc: Jouni Malinen <jouni.malinen@atheros.com>
3321    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
3322commit b4609472116bb806a95e98d04767189406c74c70
3323Author: Linus Torvalds <torvalds@linux-foundation.org>
3324Date:   Fri Aug 29 14:38:03 2008 -0700
3325
3326    Revert "x86: fix HPET regression in 2.6.26 versus 2.6.25, check hpet against BAR, v3"
3327
3328    This reverts commit a2bd7274b47124d2fc4dfdb8c0591f545ba749dd.
3329
3330    It wasn't really right to begin with (there's a better fix for the
3331    problem with e820 reservations clashing with PCI BAR's pending), but it
3332    also actually causes more regressions, so it should be reverted even
3333    before the better fix is finalized.
3334
3335    Rafael reports that this commit broke AHCI detection, and thus causes
3336    the kernel to not boot on his quad core test box.
3337
3338    Reported-and-bisected-by: Rafael J. Wysocki <rjw@sisk.pl>
3339    Cc: Yinghai Lu <yhlu.kernel@gmail.com>
3340    Cc: David Witbrodt <dawitbro@sbcglobal.net>
3341    Cc: Ingo Molnar <mingo@elte.hu>
3342    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
3343commit 8560c650f340565b720fd57d1f9c99ab216d99d0
3344Author: Linus Torvalds <torvalds@linux-foundation.org>
3345Date:   Wed Aug 27 13:35:31 2008 -0700
3346
3347    Revert "pktcdvd: push BKL down into driver"
3348
3349    This reverts commit 5b6155ee70e9c4d2ad7e6f514c8eee06e2711c3a, because
3350    the block device ioctl's really aren't ready for it.
3351
3352    In particular, the "struct file *" and the "struct inode *" arguments do
3353    not necessarily match, which means that the unlocked version of the
3354    ioctl (that only gets a "struct file *") isn't actually able to handle
3355    the cases it needs to handle.
3356
3357    This fixes bugzilla
3358
3359            http://bugzilla.kernel.org/show_bug.cgi?id=11401
3360
3361    Reported-and-bisected-by: Laurent Riffard <laurent.riffard@free.fr>
3362    Acked-by: Peter Osterlund <petero2@telia.com>
3363    Cc: Alan Cox <alan@redhat.com>
3364    Cc: Andrew Morton <akpm@linux-foundation.org>
3365    Cc: Jens Axboe <jens.axboe@oracle.com>
3366    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
3367commit f607e3a03c90e8c050cb0c12ec9967c2925cc812
3368Author: Linus Torvalds <torvalds@linux-foundation.org>
3369Date:   Tue Aug 19 13:34:59 2008 -0700
3370
3371    Revert "[CPUFREQ][2/2] preregister support for powernow-k8"
3372
3373    This reverts commit 34ae7f35a21694aa5cb8829dc5142c39d73d6ba0, which has
3374    been reported to cause a number of problems.  During suspend and resume,
3375    it apparently causes a crash in a CPU hotplug notifier to happen,
3376    although the exact details are sketchy because of the inability to get
3377    good traces during the suspend sequence.
3378
3379    See buzilla entries
3380
3381            http://bugzilla.kernel.org/show_bug.cgi?id=11296
3382            http://bugzilla.kernel.org/show_bug.cgi?id=11339
3383
3384    for more examples and details.
3385
3386    [ Mark: "Revert the patch for now.  I'm still looking into getting a
3387      reliable reproduction and I do not have a fix at this time." ]
3388
3389    Requested-by: Rafael J. Wysocki <rjw@sisk.pl>
3390    Acked-by: Mark Langsdorf <mark.langsdorf@amd.com>
3391    Acked-by: Dave Jones <davej@redhat.com>
3392    Signed-off-by: Linus Torvalds <torvalds@inux-foundation.org>
3393commit 3838f59fc2ea9821f3ea13adb555bfc6ea43c74c
3394Author: Linus Torvalds <torvalds@linux-foundation.org>
3395Date:   Mon Aug 11 10:29:11 2008 -0700
3396
3397    Revert "fbcon: bgcolor fix"
3398
3399    This reverts commit 2d04a4a72d7e1519b4838f24bdd4b5d0f3f426dc, which made
3400    it impossible to make the softcursor use the highlight colors.
3401
3402    Yes, the fourth bit should be "blinking", but since we cannot reasonably
3403    blink in fbcon, highlighting it with a bright background is preferable.
3404
3405    Reported-by: Pavel Machek <pavel@suse.cz>
3406    Cc: Stefano Stabellini <stefano.stabellini@eu.citrix.com>
3407    Cc: Krzysztof Helt <krzysztof.h1@poczta.fm>
3408    Cc: Antonino A. Daplas <adaplas@pol.net>
3409    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
3410commit 685d87f7ccc649ab92b55e18e507a65d0e694eb9
3411Author: Linus Torvalds <torvalds@linux-foundation.org>
3412Date:   Wed Aug 6 19:24:47 2008 -0700
3413
3414    Revert "pcm_native.c: remove unused label"
3415
3416    This reverts commit 680db0136e0778a0d7e025af7572c6a8d82279e2.  The label
3417    is actually used, but hidden behind CONFIG_SND_DEBUG and the horrible
3418    snd_assert() macro.
3419
3420    That macro could probably be improved to be along the lines of
3421
3422            #define snd_assert(expr, args...) do { if ((void)(expr),0) { args; } } while (0)
3423
3424    or similar to make sure that we always both evaluate 'expr' and parse
3425    'args', but while gcc should optimize it all away, I'm too lazy to
3426    really verify that.  So I'll just admit defeat and will continue to live
3427    with the annoying warning.
3428
3429    Noted-by: Robert P. J. Day <rpjday@crashcourse.ca>
3430    Signed-off-by: Linus "Grr.." Torvalds
3431commit d6606683a5e3dac35cb979c7195f54ed827567bd
3432Author: Linus Torvalds <torvalds@linux-foundation.org>
3433Date:   Wed Aug 6 12:04:54 2008 -0700
3434
3435    Revert duplicate "mm/hugetlb.c must #include <asm/io.h>"
3436
3437    This reverts commit 7cb93181629c613ee2b8f4ffe3446f8003074842, since we
3438    did that patch twice, and the problem was already fixed earlier by
3439    78a34ae29bf1c9df62a5bd0f0798b6c62a54d520.
3440
3441    Reported-by: Andi Kleen <andi@firstfloor.org>
3442    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
3443commit 1a3f7d98e5f50f21ce6fb1406a35531d9596c5c6
3444Author: Linus Torvalds <torvalds@linux-foundation.org>
3445Date:   Mon Aug 4 16:50:38 2008 -0700
3446
3447    Revert "UFS: add const to parser token table"
3448
3449    This reverts commit f9247273cb69ba101877e946d2d83044409cc8c5 (and
3450    fb2e405fc1fc8b20d9c78eaa1c7fd5a297efde43 - "fix fs/nfs/nfsroot.c
3451    compilation" - that fixed a missed conversion).
3452
3453    The changes cause problems for at least the sparc build.  Let's re-do
3454    them when the exact issues are resolved.
3455
3456    Requested-by: Andrew Morton <akpm@linux-foundation.org>
3457    Requested-by: Steven Whitehouse <swhiteho@redhat.com>
3458    Cc: David Miller <davem@davemloft.net>
3459    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
3460commit fca082c9f1e11ec07efa8d2f9f13688521253f36
3461Author: Linus Torvalds <torvalds@linux-foundation.org>
3462Date:   Mon Aug 4 16:36:20 2008 -0700
3463
3464    Revert "[SCSI] extend the last_sector_bug flag to cover more sectors"
3465
3466    This reverts commit 2b142900784c6e38c8d39fa57d5f95ef08e735d8, since it
3467    seems to break some other USB storage devices (at least a JMicron USB to
3468    ATA bridge).  As such, while it apparently fixes some cardreaders, it
3469    would need to be made conditional on the exact reader it fixes in order
3470    to avoid causing regressions.
3471
3472    Cc: Alan Jenkins <alan-jenkins@tuffmail.co.uk>
3473    Cc: James Bottomley <James.Bottomley@HansenPartnership.com>
3474    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
3475commit 09ca8adbe9f724a7e96f512c0039c4c4a1c5dcc0
3476Author: Linus Torvalds <torvalds@linux-foundation.org>
3477Date:   Sun Jul 6 10:27:25 2008 -0700
3478
3479    Revert "USB: don't explicitly reenable root-hub status interrupts"
3480
3481    This reverts commit e872154921a6b5256a3c412dd69158ac0b135176.
3482
3483    Andrey Borzenkov reports that it resulted in a totally hung machine for
3484    him when loading the OHCI driver.  Extensive netconsole capture with
3485    SysRq output shows that modprobe gets stuck in ohci_hub_status_data()
3486    when probing and enabling the OHCI controller, see for example
3487
3488            http://lkml.org/lkml/2008/7/5/236
3489
3490    for an analysis.
3491
3492    The problem appears to be an interrupt flood triggered by the commit
3493    that gets reverted, and Andrey confirmed that the revert makes things
3494    work for him again.
3495
3496    Reported-and-tested-by: Andrey Borzenkov <arvidjaar@mail.ru>
3497    Acked-by: Alan Stern <stern@rowland.harvard.edu>
3498    Acked-by: David Brownell <david-b@pacbell.net>
3499    Cc: Greg Kroah-Hartman <gregkh@suse.de>
3500    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
3501commit 00b41ec2611dc98f87f30753ee00a53db648d662
3502Author: Linus Torvalds <torvalds@linux-foundation.org>
3503Date:   Sat May 10 20:43:22 2008 -0700
3504
3505    Revert "semaphore: fix"
3506
3507    This reverts commit bf726eab3711cf192405d21688a4b21e07b6188a, as it has
3508    been reported to cause a regression with processes stuck in __down(),
3509    apparently because some missing wakeup.
3510
3511    Quoth Sven Wegener:
3512     "I'm currently investigating a regression that has showed up with my
3513      last git pull yesterday.  Bisecting the commits showed bf726e
3514      "semaphore: fix" to be the culprit, reverting it fixed the issue.
3515
3516      Symptoms: During heavy filesystem usage (e.g.  a kernel compile) I get
3517      several compiler processes in uninterruptible sleep, blocking all i/o
3518      on the filesystem.  System is an Intel Core 2 Quad running a 64bit
3519      kernel and userspace.  Filesystem is xfs on top of lvm.  See below for
3520      the output of sysrq-w."
3521
3522    See
3523
3524            http://lkml.org/lkml/2008/5/10/45
3525
3526    for full report.
3527
3528    In the meantime, we can just fix the BKL performance regression by
3529    reverting back to the good old BKL spinlock implementation instead,
3530    since any sleeping lock will generally perform badly, especially if it
3531    tries to be fair.
3532
3533    Reported-by: Sven Wegener <sven.wegener@stealer.net>
3534    Cc: Andrew Morton <akpm@linux-foundation.org>
3535    Cc: Ingo Molnar <mingo@elte.hu>
3536    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
3537commit 8d539108560ec121d59eee05160236488266221c
3538Author: Linus Torvalds <torvalds@linux-foundation.org>
3539Date:   Thu May 8 18:41:48 2008 -0700
3540
3541    Revert "PCI: remove default PCI expansion ROM memory allocation"
3542
3543    This reverts commit 9f8daccaa05c14e5643bdd4faf5aed9cc8e6f11e, which was
3544    reported to break X startup (xf86-video-ati-6.8.0). See
3545
3546            http://bugs.freedesktop.org/show_bug.cgi?id=15523
3547
3548    for details.
3549
3550    Reported-by: Laurence Withers <l@lwithers.me.uk>
3551    Cc: Gary Hade <garyhade@us.ibm.com>
3552    Cc: Greg KH <greg@kroah.com>
3553    Cc: Jan Beulich <jbeulich@novell.com>
3554    Cc: "Jun'ichi Nomura" <j-nomura@ce.jp.nec.com>
3555    Cc: Andrew Morton <akpm@linux-foundation.org>
3556    Cc: Ingo Molnar <mingo@elte.hu>
3557    Cc: Thomas Gleixner <tglx@linutronix.de>
3558    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
3559commit c0a18111e571138747a98af18b3a2124df56a0d1
3560Author: Linus Torvalds <torvalds@linux-foundation.org>
3561Date:   Tue May 6 17:09:27 2008 -0700
3562
3563    Revert "uml: fix gcc problem"
3564
3565    This reverts commit 22eecde2f9034764a3fd095eecfa3adfb8ec9a98.  Uli
3566    reports that it breaks UML on x86-64 with the Fedora 8 gcc (gcc 4.1.2),
3567    causing a crash on startup. See
3568
3569            http://marc.info/?l=linux-kernel&m=121011722806093&w=2
3570
3571    for a trace.
3572
3573    Reported-by: Ulrich Drepper <drepper@redhat.com>
3574    Cc: Ingo Molnar <mingo@elte.hu>
3575    Cc: Thomas Gleixner <tglx@linutronix.de>
3576    Cc: Andrew Morton <akpm@linux-foundation.org>
3577    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
3578commit 099c736a470c8080a166e7a089f1e48e15f9947c
3579Author: Linus Torvalds <torvalds@linux-foundation.org>
3580Date:   Mon Apr 7 13:20:08 2008 -0700
3581
3582    Revert "smc91x: fix build breakage from the SMC_GET_MAC_ADDR API upgrade"
3583
3584    This reverts commit 9e6db60825ef7e7999abc610ce256ba768e58162, which was
3585    merged without the API it needed, causing build breakage.
3586
3587    Reported-by: Bryan Wu <cooloney@kernel.org>
3588    Acked-by: Jeff Garzik <jgarzik@redhat.com>
3589    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
3590commit 797de7bdb253624c16144f40b72ec65d63cdcca2
3591Author: Linus Torvalds <torvalds@linux-foundation.org>
3592Date:   Sat Apr 5 12:14:13 2008 -0700
3593
3594    Revert "ACPI: Ignore _BQC object when registering backlight device"
3595
3596    This reverts commit 7c0ea45be4f114d85ee35caeead8e1660699c46f which
3597    caused a regression with the backlight being set to off when a laptop
3598    doesn't have a _BQC entry to query the actual backlight value.  The code
3599    blindly then falls back on a value of 0.
3600
3601    See
3602            http://bugzilla.kernel.org/show_bug.cgi?id=10387
3603            http://lkml.org/lkml/2008/4/2/366
3604
3605    for details.
3606
3607    Bisected-and-reported-by: Andrey Borzenkov <arvidjaar@mail.ru>
3608    Cc: Zhao Yakui <yakui.zhao@intel.com>
3609    Cc: Zhang Rui <rui.zhang@intel.com>
3610    Cc: Len Brown <len.brown@intel.com>
3611    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
3612commit e72e9c23ee025a4c063ca112ba0a6059f9ecc9b7
3613Author: Linus Torvalds <torvalds@linux-foundation.org>
3614Date:   Thu Mar 27 20:56:33 2008 -0700
3615
3616    Revert "SLUB: remove useless masking of GFP_ZERO"
3617
3618    This reverts commit 3811dbf67162bd08412f1b0e02e554f353e93bdb.
3619
3620    The masking was not at all useless, and it was sensible.  We handle
3621    GFP_ZERO in the caller, and passing it down to any page allocator logic
3622    is buggy and wrong.
3623
3624    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
3625commit 12c22d6ef299ccf0955e5756eb57d90d7577ac68
3626Author: Linus Torvalds <torvalds@linux-foundation.org>
3627Date:   Wed Mar 26 11:22:40 2008 -0700
3628
3629    Revert "PCI: remove transparent bridge sizing"
3630
3631    This reverts commit 8fa5913d54f3b1e09948e6a0db34da887e05ff1f, which
3632    caused various interesting problems for people, including wrong resource
3633    allocations.  See for example bugzilla entry "2.6.25-rc2: ohci1394
3634    problem (MMIO broken)" at
3635
3636            http://bugzilla.kernel.org/show_bug.cgi?id=10080
3637
3638    And Gary Hade says:
3639
3640     "The same change had also exposed an issue reported by Paul Martin that
3641      has been causing an Oops while hotplugging ThinkPads to a ThinkPad
3642      Dock II.  See
3643
3644            http://lkml.org/lkml/2008/2/19/405
3645            http://bugzilla.kernel.org/show_bug.cgi?id=9961
3646
3647      I have a fix for the ThinkPad docking Oops but if the issue being
3648      discussed here is caused by the transparent bridge sizing removal
3649      change I totally agree that it should be reverted."
3650
3651      The transparent bridge sizing removal change was motivated by
3652      insufficient PCI memory resource for a transparent bridge window that
3653      was being created as a result of expansion ROM(s) being included in
3654      the transparent bridge sizing calculations.
3655
3656      A later "PCI: Remove default PCI expansion ROM memory allocation"
3657      change ( re: http://lkml.org/lkml/2007/12/11/361 ) removes the
3658      expansion ROM(s) from the transparent bridge sizing calculations which
3659      actually resolves the original issue in a different manner.  So, even
3660      if the "PCI: remove transparent bridge sizing" is not problematic it
3661      is no longer needed anyway."
3662
3663    Identified-by: Ivan Kokshaysky <ink@jurassic.park.msu.ru>
3664    Tested-by: Thomas Meyer <thomas@m3y3r.de>
3665    Acked-by: Gary Hade <garyhade@us.ibm.com>
3666    Acked-by: Ingo Molnar <mingo@elte.hu>
3667    Cc: Stefan Richter <stefanr@s5r6.in-berlin.de>
3668    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
3669commit a345b4ba2086bacc63884e5d72268415a97bcbff
3670Author: Linus Torvalds <torvalds@woody.linux-foundation.org>
3671Date:   Mon Mar 3 10:02:44 2008 -0800
3672
3673    Revert "x86: fix pmd_bad and pud_bad to support huge pages"
3674
3675    This reverts commit cded932b75ab0a5f9181ee3da34a0a488d1a14fd.
3676
3677    Arjan bisected down a boot-time hang to this, saying:
3678      ".. it prevents the kernel to finish booting on my (Penryn based)
3679       laptop.  The boot stops right after freeing the init memory."
3680
3681    and while it's not clear exactly what triggers it, at this stage we're
3682    better off just reverting it while Ingo tries to figure out what went
3683    wrong.
3684
3685    Requested-by: Arjan van de Ven <arjan@linux.intel.com>
3686    Cc: Hans Rosenfeld <hans.rosenfeld@amd.com>
3687    Cc: Nish Aravamudan <nish.aravamudan@gmail.com>
3688    Acked-by: Ingo Molnar <mingo@elte.hu>
3689    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
3690commit 00e962c5408b9f2d0bebd2308673fe982cb9a5fe
3691Author: Linus Torvalds <torvalds@woody.linux-foundation.org>
3692Date:   Tue Feb 19 09:08:49 2008 -0800
3693
3694    Revert "SLUB: Alternate fast paths using cmpxchg_local"
3695
3696    This reverts commit 1f84260c8ce3b1ce26d4c1d6dedc2f33a3a29c0c, which is
3697    suspected to be the reason for some very occasional and hard-to-trigger
3698    crashes that usually look related to memory allocation (mostly reported
3699    in networking, but since that's generally the most common source of
3700    shortlived allocations - and allocations in interrupt contexts - that in
3701    itself is not a big clue).
3702
3703    See for example
3704            http://bugzilla.kernel.org/show_bug.cgi?id=9973
3705            http://lkml.org/lkml/2008/2/19/278
3706    etc.
3707
3708    One promising suspicion for what the root cause of bug is (which also
3709    explains why it's so hard to trigger in practice) came from Eric
3710    Dumazet:
3711
3712       "I wonder how SLUB_FASTPATH is supposed to work, since it is affected
3713        by a classical ABA problem of lockless algo.
3714
3715        cmpxchg_local(&c->freelist, object, object[c->offset]) can succeed,
3716        while an interrupt came (on this cpu), and several allocations were
3717        done, and one free was performed at the end of this interruption, so
3718        'object' was recycled.
3719
3720        c->freelist can then contain the previous value (object), but
3721        object[c->offset] was changed by IRQ.
3722
3723        We then put back in freelist an already allocated object."
3724
3725    but another reason for the revert is simply that everybody agrees that
3726    this code was the main suspect just by virtue of the pattern of oopses.
3727
3728    Cc: Torsten Kaiser <just.for.lkml@googlemail.com>
3729    Cc: Christoph Lameter <clameter@sgi.com>
3730    Cc: Mathieu Desnoyers <mathieu.desnoyers@polymtl.ca>
3731    Cc: Pekka Enberg <penberg@cs.helsinki.fi>
3732    Cc: Ingo Molnar <mingo@elte.hu>
3733    Cc: Eric Dumazet <dada1@cosmosbay.com>
3734    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
3735commit dbcc2ec60fe15448e5dfe4153f7347ac82b4da73
3736Author: Linus Torvalds <torvalds@woody.linux-foundation.org>
3737Date:   Thu Jan 24 13:35:10 2008 -0800
3738
3739    Revert "mac80211: warn when receiving frames with unaligned data"
3740
3741    This reverts commit 81100eb80add328c4d2a377326f15aa0e7236398 for the
3742    release, to avoid the unnecessary warning noise that is only really
3743    relevant to wireless driver developers.
3744
3745    The warning will probably go right back in after I cut the release, but
3746    at least we won't unnecessarily worry users.
3747
3748    Acked-by: John W. Linville <linville@tuxdriver.com>
3749    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
3750commit c23f72cae9523d29ff94eec8f30ccbdaf234b20e
3751Author: Linus Torvalds <torvalds@woody.linux-foundation.org>
3752Date:   Mon Jan 14 21:21:29 2008 -0800
3753
3754    Revert "writeback: introduce writeback_control.more_io to indicate more io"
3755
3756    This reverts commit 2e6883bdf49abd0e7f0d9b6297fc3be7ebb2250b, as
3757    requested by Fengguang Wu.  It's not quite fully baked yet, and while
3758    there are patches around to fix the problems it caused, they should get
3759    more testing.  Says Fengguang: "I'll resend them both for -mm later on,
3760    in a more complete patchset".
3761
3762    See
3763
3764            http://bugzilla.kernel.org/show_bug.cgi?id=9738
3765
3766    for some of this discussion.
3767
3768    Requested-by: Fengguang Wu <wfg@mail.ustc.edu.cn>
3769    Cc: Andrew Morton <akpm@linux-foundation.org>
3770    Cc: Peter Zijlstra <peterz@infradead.org>
3771    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
3772commit d238998fbfa49f30b02f0a5de5294ca53c58348c
3773Author: Linus Torvalds <torvalds@woody.linux-foundation.org>
3774Date:   Tue Jan 8 11:46:37 2008 -0800
3775
3776    Revert "hda_intel suspend latency: shorten codec read"
3777
3778    This reverts commit 57a04513cb35086d54bcb2cb92e6627fc8fa0fae.
3779
3780    Harald Dunkel reports that it broke sound for him:
3781      "Alsa stopped working for me.  I still can access /dev/dsp, change the
3782       volume and so on, but the speakers are quiet."
3783
3784    Reverting it fixed things for him.
3785
3786    Reported-and-tested-by: Harald Dunkel <harald.dunkel@t-online.de>
3787    Acked-by: Takashi Iwai <tiwai@suse.de>
3788    Acked-by: Ingo Molnar <mingo@elte.hu>
3789    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
3790commit 7b3d9545f9ac8b31528dd2d6d8ec8d19922917b8
3791Author: Linus Torvalds <torvalds@woody.linux-foundation.org>
3792Date:   Sun Jan 6 10:17:12 2008 -0800
3793
3794    Revert "scsi: revert "[SCSI] Get rid of scsi_cmnd->done""
3795
3796    This reverts commit ac40532ef0b8649e6f7f83859ea0de1c4ed08a19, which gets
3797    us back the original cleanup of 6f5391c283d7fdcf24bf40786ea79061919d1e1d.
3798
3799    It turns out that the bug that was triggered by that commit was
3800    apparently not actually triggered by that commit at all, and just the
3801    testing conditions had changed enough to make it appear to be due to it.
3802
3803    The real problem seems to have been found by Peter Osterlund:
3804
3805      "pktcdvd sets it [block device size] when opening the /dev/pktcdvd
3806       device, but when the drive is later opened as /dev/scd0, there is
3807       nothing that sets it back.  (Btw, 40944 is possible if the disk is a
3808       CDRW that was formatted with "cdrwtool -m 10236".)
3809
3810       The problem is that pktcdvd opens the cd device in non-blocking mode
3811       when pktsetup is run, and doesn't close it again until pktsetup -d is
3812       run.  The effect is that if you meanwhile open the cd device,
3813       blkdev.c:do_open() doesn't call bd_set_size() because
3814       bdev->bd_openers is non-zero."
3815
3816    In particular, to repeat the bug (regardless of whether commit
3817    6f5391c283d7fdcf24bf40786ea79061919d1e1d is applied or not):
3818
3819      " 1. Start with an empty drive.
3820        2. pktsetup 0 /dev/scd0
3821        3. Insert a CD containing an isofs filesystem.
3822        4. mount /dev/pktcdvd/0 /mnt/tmp
3823        5. umount /mnt/tmp
3824        6. Press the eject button.
3825        7. Insert a DVD containing a non-writable filesystem.
3826        8. mount /dev/scd0 /mnt/tmp
3827        9. find /mnt/tmp -type f -print0 | xargs -0 sha1sum >/dev/null
3828        10. If the DVD contains data beyond the physical size of a CD, you
3829            get I/O errors in the terminal, and dmesg reports lots of
3830            "attempt to access beyond end of device" errors."
3831
3832    which in turn is because the nested open after the media change won't
3833    cause the size to be set properly (because the original open still holds
3834    the block device, and we only do the bd_set_size() when we don't have
3835    other people holding the device open).
3836
3837    The proper fix for that is probably to just do something like
3838
3839            bdev->bd_inode->i_size = (loff_t)get_capacity(disk)<<9;
3840
3841    in fs/block_dev.c:do_open() even for the cases where we're not the
3842    original opener (but *not* call bd_set_size(), since that will also
3843    change the block size of the device).
3844
3845    Cc: Peter Osterlund <petero2@telia.com>
3846    Cc: James Bottomley <James.Bottomley@HansenPartnership.com>
3847    Cc: Matthew Wilcox <matthew@wil.cx>
3848    Cc: Ingo Molnar <mingo@elte.hu>
3849    Cc: Andrew Morton <akpm@linux-foundation.org>
3850    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
3851commit 00684418707c7a1e36ebdedc4b30fbba5d5860b1
3852Author: Linus Torvalds <torvalds@woody.linux-foundation.org>
3853Date:   Tue Dec 25 20:16:16 2007 -0800
3854
3855    Revert "x86: fix show cpuinfo cpu number always zero"
3856
3857    This reverts commit fbdcf18df73758b2e187ab94678b30cd5f6ff9f9.
3858
3859    As pointed out by Yanmin Zhang, the problem was already fixed
3860    differently (and correctly), and rather than fix anything, it actually
3861    causes us to create a sub-optimal sched-domains hierarchy (not setting
3862    up the domain belonging to the core) when CONFIG_X86_HT=y.
3863
3864    Requested-by: Yanmin Zhang <yanmin_zhang@linux.intel.com>
3865    Acked-by: Ingo Molnar <mingo@elte.hu>
3866    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
3867commit 4af75653031c6d454b4ace47c1536f0d2e727e3e
3868Author: Linus Torvalds <torvalds@woody.linux-foundation.org>
3869Date:   Tue Dec 11 19:28:36 2007 -0800
3870
3871    Revert "make bnx2x select ZLIB_INFLATE"
3872
3873    This reverts commit 70eba18b5664f90d7620905e005b89388e5fd94b, as per
3874    Jeff Garzik:
3875
3876       "That was meant for 2.6.25, and actually (due to patching) applied to
3877        a completely unrelated 2.6.24 net driver."
3878
3879    Noted-by: Roland Dreier <rdreier@cisco.com>
3880    Requested-by: Jeff Garzik <jgarzik@pobox.com>
3881    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
3882commit af1bff4f1d117f99ad8a88e6532baff475fb44a5
3883Author: Linus Torvalds <torvalds@woody.linux-foundation.org>
3884Date:   Mon Dec 10 07:40:54 2007 -0800
3885
3886    Revert "PCI: fix IDE legacy mode resources"
3887
3888    This reverts commit fd6e732186ab522c812ab19c2c5e5befb8ec8115, which
3889    helped up things on MIPS, but was wrong for everything else.  As Ralf
3890    Baechle puts it:
3891
3892      "It seems the whole MIPS resource managment is complicated enough (out
3893       of necessity) that only a few people actually grok it.  Ioports being
3894       actually memory mapped on MIPS only makes the confusion worse, sigh."
3895
3896    Requested-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
3897    Acked-by: Alan Cox <alan@redhat.com>
3898    Acked-by: Ralf Baechle <ralf@linux-mips.org>
3899    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
3900commit 279e1dab949d33737557babfe9f74e0b74fbe39a
3901Author: Linus Torvalds <torvalds@woody.linux-foundation.org>
3902Date:   Thu Nov 15 08:44:36 2007 -0800
3903
3904    Revert "skge: fix ram buffer size calculation"
3905
3906    This reverts commit 7fb7ac241162dc51ec0f7644d4a97b2855213c32.
3907
3908    Heikki Orsila reports that it causes a regression:
3909
3910      "Doing
3911
3912            nc host port < /dev/zero
3913
3914       on a sending machine (not skge) to an skge machine that is receiving:
3915
3916            nc -l -p port >/dev/null
3917
3918       with ~60 MiB/s speed, causes the interface go malfunct. A slow
3919       transfer doesn't cause a problem."
3920
3921    See
3922
3923            http://bugzilla.kernel.org/show_bug.cgi?id=9321
3924
3925    for some more information.
3926
3927    There is a workaround (also reported by Heikki):
3928
3929      "After some fiddling, I noticed that not changing the register write
3930       order on patch:
3931
3932       +       skge_write32(hw, RB_ADDR(q, RB_END), end);
3933               skge_write32(hw, RB_ADDR(q, RB_WP), start);
3934               skge_write32(hw, RB_ADDR(q, RB_RP), start);
3935       -       skge_write32(hw, RB_ADDR(q, RB_END), end);
3936
3937       fixes the visible effect..  Possibly not the root cause of the
3938       problem, but changing the order back fixes networking here."
3939
3940    but that has yet to be ack'ed or tested more widely, so the whole
3941    problem-causing commit gets reverted until this is resolved properly.
3942
3943    Bisected-and-requested-by: Heikki Orsila <shdl@zakalwe.fi>
3944    Cc: Stephen Hemminger <shemminger@linux-foundation.org>
3945    Cc: Jeff Garzik <jeff@garzik.org>
3946    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
3947commit 0b832a4b93932103d73c0c3f35ef1153e288327b
3948Author: Linus Torvalds <torvalds@woody.linux-foundation.org>
3949Date:   Tue Nov 13 08:07:31 2007 -0800
3950
3951    Revert "ext2/ext3/ext4: add block bitmap validation"
3952
3953    This reverts commit 7c9e69faa28027913ee059c285a5ea8382e24b5d, fixing up
3954    conflicts in fs/ext4/balloc.c manually.
3955
3956    The cost of doing the bitmap validation on each lookup - even when the
3957    bitmap is cached - is absolutely prohibitive.  We could, and probably
3958    should, do it only when adding the bitmap to the buffer cache.  However,
3959    right now we are better off just reverting it.
3960
3961    Peter Zijlstra measured the cost of this extra validation as a 85%
3962    decrease in cached iozone, and while I had a patch that took it down to
3963    just 17% by not being _quite_ so stupid in the validation, it was still
3964    a big slowdown that could have been avoided by just doing it right.
3965
3966    Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
3967    Cc: Andrew Morton <akpm@linux-foundation.org>
3968    Cc: Aneesh Kumar <aneesh.kumar@linux.vnet.ibm.com>
3969    Cc: Andreas Dilger <adilger@clusterfs.com>
3970    Cc: Mingming Cao <cmm@us.ibm.com>
3971    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
3972commit 44048d700bcbfaf4bcca6e2e0a73d89d01ec0878
3973Author: Linus Torvalds <torvalds@woody.linux-foundation.org>
3974Date:   Mon Nov 12 14:14:44 2007 -0800
3975
3976    Revert "Bias the placement of kernel pages at lower PFNs"
3977
3978    This reverts commit 5adc5be7cd1bcef6bb64f5255d2a33f20a3cf5be.
3979
3980    Alexey Dobriyan reports that it causes huge slowdowns under some loads,
3981    in his case a "mkfs.ext2" on a 30G partition.  With the placement bias,
3982    the mkfs took over four minutes, with it reverted it's back to about ten
3983    seconds for Alexey.
3984
3985    Reported-and-tested-by: Alexey Dobriyan <adobriyan@gmail.com>
3986    Cc: Mel Gorman <mel@csn.ul.ie>
3987    Cc: Andrew Morton <akpm@linux-foundation.org>
3988    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
3989commit 3902beb48d369d5e19f66acc2f857865ddc9b3bf
3990Author: Linus Torvalds <torvalds@woody.linux-foundation.org>
3991Date:   Sat Nov 10 14:22:46 2007 -0800
3992
3993    Revert "[ARM] 4642/2: netX: default config for netx based boards"
3994
3995    This reverts commit f33bac8dd4573428b94c67149c5607be489092d1, which was
3996    totally bogus.
3997
3998    The arm/configs/netx_defconfig file already existed - in the right
3999    place. Namely under "arch".
4000
4001    Noticed-by: Paul Mundt <lethal@linux-sh.org>
4002    Acked-by: Robert Schwebel <r.schwebel@pengutronix.de>
4003    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
4004commit 6a22c57b8d2a62dea7280a6b2ac807a539ef0716
4005Author: Linus Torvalds <torvalds@woody.linux-foundation.org>
4006Date:   Mon Oct 29 11:36:04 2007 -0700
4007
4008    Revert "x86_64: allocate sparsemem memmap above 4G"
4009
4010    This reverts commit 2e1c49db4c640b35df13889b86b9d62215ade4b6.
4011
4012    First off, testing in Fedora has shown it to cause boot failures,
4013    bisected down by Martin Ebourne, and reported by Dave Jobes.  So the
4014    commit will likely be reverted in the 2.6.23 stable kernels.
4015
4016    Secondly, in the 2.6.24 model, x86-64 has now grown support for
4017    SPARSEMEM_VMEMMAP, which disables the relevant code anyway, so while the
4018    bug is not visible any more, it's become invisible due to the code just
4019    being irrelevant and no longer enabled on the only architecture that
4020    this ever affected.
4021
4022    Reported-by: Dave Jones <davej@redhat.com>
4023    Tested-by: Martin Ebourne <fedora@ebourne.me.uk>
4024    Cc: Zou Nan hai <nanhai.zou@intel.com>
4025    Cc: Suresh Siddha <suresh.b.siddha@intel.com>
4026    Cc: Andrew Morton <akpm@linux-foundation.org>
4027    Acked-by: Andy Whitcroft <apw@shadowen.org>
4028    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
4029commit e8b8c977734193adedf2b0f607d6252c78e86394
4030Author: Linus Torvalds <torvalds@woody.linux-foundation.org>
4031Date:   Fri Oct 19 21:25:45 2007 -0700
4032
4033    Revert "kconfig: tristate choices with mixed tristate and boolean values"
4034
4035    This reverts commit a5bf3d891a6a0fb5aa122792d965e3774108b923.
4036
4037    David Brownell notes that this causes a regression visible in the
4038    drivers/usb/gadget Kconfig file:
4039
4040      "That Kconfig hasn't changed (other than adding new drivers), and it's
4041       worked that way for several years now ...  so the issue seems to be
4042       changes in menuconfig/kconfig/etc semantics.
4043
4044       The issue is that when USB_GADGET=m, it's no longer possible to
4045       configure peripheral controller drivers as modules ...  the
4046       controller drivers can now only be configured for static linkage.
4047
4048       It should be making a choice of one of the controller drivers which
4049       could work on the target system, and allow that driver to be linked
4050       either as a module (ok iff USB_GADGET=m) or statically."
4051
4052    Reverting this commit resolves the problem, and also fixes a second
4053    problem that David noticed: various dependent options couldn't be enabled.
4054
4055    Tested-and-reported-by: David Brownell <david-b@pacbell.net>
4056    Cc: Randy Dunlap <rdunlap@xenotime.net>
4057    Cc: Jan Beulich <jbeulich@novell.com>,
4058    Cc: Andrew Morton <akpm@linux-foundation.org>,
4059    Cc: Sam Ravnborg <sam@ravnborg.org>,
4060    Cc: Roman Zippel <zippel@linux-m68k.org>
4061    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
4062commit ff0ce6845bc18292e80ea40d11c3d3a539a3fc5e
4063Author: Linus Torvalds <torvalds@woody.linux-foundation.org>
4064Date:   Wed Sep 26 15:52:17 2007 -0700
4065
4066    Revert "[PATCH] x86-64: fix x86_64-mm-sched-clock-share"
4067
4068    This reverts commit 184c44d2049c4db7ef6ec65794546954da2c6a0e.
4069
4070    As noted by Dave Jones:
4071       "Linus, please revert the above cset.  It doesn't seem to be
4072        necessary (it was added to fix a miscompile in 'make allnoconfig'
4073        which doesn't seem to be repeatable with it reverted) and actively
4074       breaks the ARM SA1100 framebuffer driver."
4075
4076    Requested-by: Dave Jones <davej@redhat.com>
4077    Cc: Russell King <rmk+lkml@arm.linux.org.uk>
4078    Cc: Andrew Morton <akpm@linux-foundation.org>
4079    Cc: Andi Kleen <ak@suse.de>
4080    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
4081commit f7f847b01571e86044dc77e03d92f43699652f8d
4082Author: Linus Torvalds <torvalds@woody.linux-foundation.org>
4083Date:   Wed Sep 26 15:21:33 2007 -0700
4084
4085    Revert "x86-64: Disable local APIC timer use on AMD systems with C1E"
4086
4087    This reverts commit e66485d747505e9d960b864fc6c37f8b2afafaf0, since
4088    Rafael Wysocki noticed that the change only works for his in -mm, not in
4089    mainline (and that both "noapictimer" _and_ "apicmaintimer" are broken
4090    on his hardware, but that's apparently not a regression, just a symptom
4091    of the same issue that causes the automatic apic timer disable to not
4092    work).
4093
4094    It turns out that it really doesn't work correctly on x86-64, since
4095    x86-64 doesn't use the generic clock events for timers yet.
4096
4097    Thanks to Rafal for testing, and here's the ugly details on x86-64 as
4098    per Thomas:
4099
4100      "I just looked into the code and the logic vs.  noapictimer on SMP is
4101       completely broken.
4102
4103       On i386 the noapictimer option not only disables the local APIC
4104       timer, it also registers the CPUs for broadcasting via IPI on SMP
4105       systems.
4106
4107       The x86-64 code uses the broadcast only when the local apic timer is
4108       active, i.e.  "noapictimer" is not on the command line.  This defeats
4109       the whole purpose of "noapictimer".  It should be there to make boxen
4110       work, where the local APIC timer actually has a hardware problem,
4111       e.g.  the nx6325.
4112
4113       The current implementation of x86_64 only fixes the ACPI c-states
4114       related problem where the APIC timer stops in C3(2), nothing else.
4115
4116       On nx6325 and other AMD X2 equipped systems which have the C1E
4117       enabled we run into the following:
4118
4119       PIT keeps jiffies (and the system) running, but the local APIC timer
4120       interrupts can get out of sync due to this C1E effect.
4121
4122       I don't think this is a critical problem, but it is wrong
4123       nevertheless.
4124
4125       I think it's safe to revert the C1E patch and postpone the fix to the
4126       clock events conversion."
4127
4128    On further reflection, Thomas noted:
4129
4130       "It's even worse than I thought on the first check:
4131
4132        "noapictimer" on the command line of an SMP box prevents _ONLY_ the
4133        boot CPU apic timer from being used.  But the secondary CPU is still
4134        unconditionally setting up the APIC timer and uses the non
4135        calibrated variable calibration_result, which is of course 0, to
4136        setup the APIC timer.  Wreckage guaranteed."
4137
4138    so we'll just have to wait for the x86 merge to hopefully fix this up
4139    for x86-64.
4140
4141    Tested-and-requested-by: Rafael J. Wysocki <rjw@sisk.pl>
4142    Acked-by: Thomas Gleixner <tglx@linutronix.de>
4143    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
4144commit da8f153e51290e7438ba7da66234a864e5d3e1c1
4145Author: Linus Torvalds <torvalds@woody.linux-foundation.org>
4146Date:   Fri Sep 21 12:09:41 2007 -0700
4147
4148    Revert "x86_64: Quicklist support for x86_64"
4149
4150    This reverts commit 34feb2c83beb3bdf13535a36770f7e50b47ef299.
4151
4152    Suresh Siddha points out that this one breaks the fundamental
4153    requirement that you cannot free page table pages before the TLB caches
4154    are flushed.  The quicklists do not give the same kinds of guarantees
4155    that the mmu_gather structure does, at least not in NUMA configurations.
4156
4157    Requested-by: Suresh Siddha <suresh.b.siddha@intel.com>
4158    Acked-by: Andi Kleen <ak@suse.de>
4159    Cc: Andrew Morton <akpm@linux-foundation.org>
4160    Cc: Christoph Lameter <clameter@sgi.com>
4161    Cc: Asit Mallick <asit.k.mallick@intel.com>
4162    Cc: Tony Luck <tony.luck@intel.com>
4163    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
4164commit 8eb891fc809b2300137bcd247025628c06c95a63
4165Author: Linus Torvalds <torvalds@woody.linux-foundation.org>
4166Date:   Mon Aug 20 23:38:44 2007 -0700
4167
4168    Revert "USB: EHCI cpufreq fix"
4169
4170    This reverts commit 196705c9bbc03540429b0f7cf9ee35c2f928a534.  It was
4171    reported to cause a regression by Daniel Exner, and Arjan van de Ven
4172    points out that we actually already have infrastructure in place for
4173    setting limits on acceptable DMA latency that would be the much more
4174    correct fix for the problem with some Broadcom EHCI controllers.
4175
4176    Fixed up trivial conflicts due to the changes to support big-endian host
4177    controller descriptors in drivers/usb/host/{ehci-sched.c,ehci.h}.
4178
4179    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
4180commit 88ffc3505988196ef5cfdc0278ad89025c2a7b1a
4181Author: Linus Torvalds <torvalds@woody.linux-foundation.org>
4182Date:   Thu Aug 9 08:10:16 2007 -0700
4183
4184    Revert "genirq: temporary fix for level-triggered IRQ resend"
4185
4186    This reverts commit 0fc4969b866671dfe39b1a9119d0fdc7ea0f63e5.  It was
4187    always meant to be temporary, but it's generating more useless noise
4188    than anything else, and we probably should never have done it in the
4189    generic kernel (only had the people involved test it on their own).
4190
4191    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
4192commit 602033ed5907a59ce86f709082a35be047743a86
4193Author: Linus Torvalds <torvalds@woody.linux-foundation.org>
4194Date:   Thu Jul 26 12:07:21 2007 -0700
4195
4196    Revert most of "x86: Fix alternatives and kprobes to remap write-protected kernel text"
4197
4198    This reverts most of commit 19d36ccdc34f5ed444f8a6af0cbfdb6790eb1177.
4199
4200    The way to DEBUG_RODATA interactions with KPROBES and CPU hotplug is to
4201    just not mark the text as being write-protected in the first place.
4202    Both of those facilities depend on rewriting instructions.
4203
4204    Having "helpful" debug facilities that just cause more problem is not
4205    being helpful.  It just adds complexity and bugs. Not worth it.
4206
4207    Reported-by: Rafael J. Wysocki <rjw@sisk.pl>
4208    Cc: Andi Kleen <ak@suse.de>
4209    Cc: Andrew Morton <akpm@linux-foundation.org>
4210    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
4211commit 2008220879af095d00ca27eb168a55c8595fbc0b
4212Author: Linus Torvalds <torvalds@woody.linux-foundation.org>
4213Date:   Fri Jul 20 13:28:54 2007 -0700
4214
4215    Revert "sys_time() speedup"
4216
4217    This basically reverts commit 4e44f3497d41db4c3b9051c61410dee8ae4fb49c,
4218    while waiting for it to be re-done more completely.  There are cases of
4219    people mixing "time()" with higher-resolution time sources, and we need
4220    to take the nanosecond offsets into account.
4221
4222    Ingo has a patch that does that, but it's still under some discussion.
4223    In the meantime, just revert back to the old simple situation of just
4224    doing the whole exact timesource calculations.
4225
4226    But rather than using do_gettimeofday(), use the internal nanosecond
4227    resolution getnstimeofday(), which at least avoids one unnecessary
4228    conversion (since we really don't care about whether the fractional
4229    seconds are nanoseconds or microseconds - we'll just throw them away).
4230
4231    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
4232commit 1985026d32e69ed6dac3ba0ef8ff10366f060ed3
4233Author: Linus Torvalds <torvalds@woody.linux-foundation.org>
4234Date:   Tue Jul 17 15:57:42 2007 -0700
4235
4236    Revert drivers/ide/ide.c scsi_cmd_ioctl() usage changes
4237
4238    The old IDE driver is not ready to take generic SCSI commands, even if
4239    it uses them for some specific issues (ie the tray open/close ioctls for
4240    IDE CD-ROM's). Pointed out by Bartlomiej.
4241
4242    I'm sure we'll have it fixed properly soon enough, but for now we should
4243    not allow it to cause problems.
4244
4245    Cc: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>
4246    Cc: Jens Axboe <jens.axboe@oracle.com>
4247    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
4248commit 2e27afb300b56d83bb03fbfa68852b9c1e2920c6
4249Author: Linus Torvalds <torvalds@woody.linux-foundation.org>
4250Date:   Mon Jul 16 14:31:08 2007 -0700
4251
4252    Revert "[NET]: Fix races in net_rx_action vs netpoll."
4253
4254    This reverts commit 29578624e354f56143d92510fff33a8b2aaa2c03.
4255
4256    Ingo Molnar reports complete breakage with his e1000 card (no
4257    networking, card reports transmit timeouts), and bisected it down to
4258    this commit.  Let's figure out what went wrong, but not keep breaking
4259    machines until we do.
4260
4261    Cc: Ingo Molnar <mingo@elte.hu>
4262    Cc: Olaf Kirch <olaf.kirch@oracle.com>
4263    Cc: David Miller <davem@davemloft.net>
4264    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
4265commit 8d9107e8c50e1c4ff43c91c8841805833f3ecfb9
4266Author: Linus Torvalds <torvalds@woody.linux-foundation.org>
4267Date:   Fri Jul 13 16:53:18 2007 -0700
4268
4269    Revert "SELinux: use SECINITSID_NETMSG instead of SECINITSID_UNLABELED for NetLabel"
4270
4271    This reverts commit 9faf65fb6ee2b4e08325ba2d69e5ccf0c46453d0.
4272
4273    It bit people like Michal Piotrowski:
4274
4275      "My system is too secure, I can not login :)"
4276
4277    because it changed how CONFIG_NETLABEL worked, and broke older SElinux
4278    policies.
4279
4280    As a result, quoth James Morris:
4281
4282      "Can you please revert this patch?
4283
4284       We thought it only affected people running MLS, but it will affect others.
4285
4286       Sorry for the hassle."
4287
4288    Cc: James Morris <jmorris@namei.org>
4289    Cc: Stephen Smalley <sds@tycho.nsa.gov>
4290    Cc: Michal Piotrowski <michal.k.k.piotrowski@gmail.com>
4291    Cc: Paul Moore <paul.moore@hp.com>
4292    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
4293commit 8b69ad0e690eb5f38c23087247a12e5fde1baeff
4294Author: Linus Torvalds <torvalds@woody.linux-foundation.org>
4295Date:   Fri Jul 13 10:43:52 2007 -0700
4296
4297    Revert "[CPUFREQ] powernow-k8: clarify number of cores."
4298
4299    This reverts commit 904f7a3f042b5c6aa9e53ce83f2c9de5e33170ff.
4300
4301    As noted by Peter Anvin:
4302
4303      "It causes build failures on i386.
4304
4305       Yet another case of unnecessary divergence between i386 and x86-64
4306       I'm afraid..."
4307
4308    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
4309commit 773208946a132fb733ba273ee8562814f828cc28
4310Author: Linus Torvalds <torvalds@woody.linux-foundation.org>
4311Date:   Thu Jul 12 17:06:50 2007 -0700
4312
4313    Revert "USB: fix gregkh-usb-usb-use-menuconfig-objects"
4314
4315    This reverts commit acb11c8b8020f1f1b2545152020675ef32d09a58.
4316
4317    It was broken.  We most certainly *do* want the default to be the old
4318    behaviour (and the common case!), instead of breaking everybodys
4319    configuration and making 99% of all people have to override the default.
4320
4321    What were you guys thinking?
4322
4323    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
4324commit de081fa517fed81b0369f2e90ca87c30182879c8
4325Author: Linus Torvalds <torvalds@woody.linux-foundation.org>
4326Date:   Thu Jul 12 16:40:08 2007 -0700
4327
4328    Revert "[BNX2]: Seems to not need net/tcp.h"
4329
4330    This reverts commit 963bd949b12158d9b5380b718b31c4b33372ed73.  The
4331    driver _does_ need the networking header files;
4332
4333              CC [M]  drivers/net/bnx2.o
4334            drivers/net/bnx2.c: In function 'bnx2_start_xmit':
4335            drivers/net/bnx2.c:5177: warning: implicit declaration of function 'tcp_optlen'
4336            drivers/net/bnx2.c:5181: error: invalid application of 'sizeof' to incomplete type 'struct ipv6hdr'
4337            drivers/net/bnx2.c:5202: error: invalid application of 'sizeof' to incomplete type 'struct tcphdr'
4338            drivers/net/bnx2.c:5207: warning: implicit declaration of function 'tcp_hdr'
4339            drivers/net/bnx2.c:5207: error: invalid type argument of '->'
4340            make[2]: *** [drivers/net/bnx2.o] Error 1
4341            make[1]: *** [drivers/net] Error 2
4342            make: *** [drivers] Error 2
4343
4344    Cc: Ilpo Jävinen <ilpo.jarvinen@helsinki.fi>
4345    Cc: David Miller <davem@davemloft.net>
4346    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
4347commit 1a93fa86bf6eb0d8bed84ef18c8b35be389edee5
4348Author: Linus Torvalds <torvalds@woody.linux-foundation.org>
4349Date:   Thu May 24 08:50:39 2007 -0700
4350
4351    Revert "HOWTO: bug report addition" (or "HOWTO: mention bughunting")
4352
4353    This reverts commit 722385f75efd82d9f480f0765a1e97a4d83cac0d (or commit
4354    3f27100872b21e4cc70d07b96eeb3611b30bce63, it's your choice ;), since the
4355    same patch to Documentation/HOWTO got added twice because it just kept
4356    applying cleanly.
4357
4358    Noted by Qi Yong.
4359
4360    Cc: Qi Yong <qiyong@fc-cn.com>
4361    Acked-by: Diego Calleja <diegocg@gmail.com>
4362    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
4363commit efa5bf1dd2cf3cdee0bfe97cfd76ff2296179ae4
4364Author: Linus Torvalds <torvalds@woody.linux-foundation.org>
4365Date:   Mon May 21 18:41:17 2007 -0700
4366
4367    Revert "kbuild: make better section mismatch reports on i386, arm and mips"
4368
4369    This reverts commit f892b7d480eec809a5dfbd6e65742b3f3155e50e, which
4370    totally broke the build on x86 with CONFIG_RELOCATABLE (which, as far as
4371    I can tell, is the only case where it should even matter!) due to a
4372    SIGSEGV in modpost.
4373
4374    Cc: Sam Ravnborg <sam@ravnborg.org>
4375    Cc: Atsushi Nemoto <anemo@mba.ocn.ne.jp>
4376    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
4377commit b46522394dde7341a5253658096f354b65cbd90d
4378Author: Linus Torvalds <torvalds@woody.linux-foundation.org>
4379Date:   Thu May 17 20:18:11 2007 -0700
4380
4381    Revert "[PATCH] x86: Drop cc-options call for all options supported in gcc 3.2+"
4382
4383    This reverts commit c8fdd247255a3a027cd9f66dcf93e6847d1d2f85.
4384
4385    It turns out the kernel was correct, and the gcc complaint was a gcc
4386    bug.  The preferred stack boundary is expressed not in bytes, but in the
4387    the log2() of the preferred boundary, so "-mpreferred-stack-boundary=2"
4388    is in fact exactly what we want, but a gcc that is compiled for x86-64
4389    will consider it an error (because the 64-bit calling sequence says that
4390    the stack should be 16-byte aligned) even if we are then using "-m32" to
4391    generate 32-bit code.
4392
4393    Noted-by: Mikulas Patocka <mikulas@artax.karlin.mff.cuni.cz>
4394    Cc: Jan Hubicka <jh@suse.cz>
4395    Acked-by: Andi Kleen <ak@suse.de>
4396    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
4397commit faa8b6c3c2e1454175609167a25ae525d075f045
4398Author: Linus Torvalds <torvalds@woody.linux-foundation.org>
4399Date:   Mon May 14 15:24:24 2007 -0700
4400
4401    Revert "ipmi: add new IPMI nmi watchdog handling"
4402
4403    This reverts commit f64da958dfc83335de1d2bef9d3868f30feb4e53.
4404
4405    Andi Kleen is unhappy with the changes, and they really do not seem
4406    worth it.  IPMI could use DIE_NMI_IPI instead of the new callback, even
4407    though that ends up having its own set of problems too, mainly because
4408    the IPMI code cannot really know the NMI was from IPMI or not.
4409
4410    Manually fix up conflicts in arch/x86_64/kernel/traps.c and
4411    drivers/char/ipmi/ipmi_watchdog.c.
4412
4413    Cc: Andi Kleen <ak@suse.de>
4414    Cc: Mathieu Desnoyers <mathieu.desnoyers@polymtl.ca>
4415    Cc: Corey Minyard <minyard@acm.org>
4416    Cc: Andrew Morton <akpm@linux-foundation.org>
4417    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
4418commit 44ce6294d07555c3d313757105fd44b78208407f
4419Author: Linus Torvalds <torvalds@woody.linux-foundation.org>
4420Date:   Wed May 9 18:51:36 2007 -0700
4421
4422    Revert "md: improve partition detection in md array"
4423
4424    This reverts commit 5b479c91da90eef605f851508744bfe8269591a0.
4425
4426    Quoth Neil Brown:
4427
4428      "It causes an oops when auto-detecting raid arrays, and it doesn't
4429       seem easy to fix.
4430
4431       The array may not be 'open' when do_md_run is called, so
4432       bdev->bd_disk might be NULL, so bd_set_size can oops.
4433
4434       This whole approach of opening an md device before it has been
4435       assembled just seems to get more and more painful.  I think I'm going
4436       to have to come up with something clever to provide both backward
4437       comparability with usage expectation, and sane integration into the
4438       rest of the kernel."
4439
4440    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
4441commit 01e73be3c8f254ef19d787f9b6757468175267eb
4442Author: Linus Torvalds <torvalds@woody.linux-foundation.org>
4443Date:   Tue May 8 20:12:30 2007 -0700
4444
4445    Revert "fbdev: ignore VESA modes if framebuffer is disabled"
4446
4447    This reverts commit 464bdd33e9baad9806c7adbd8dfc37081a55f27e.
4448
4449    Peter Anvin correctly points out that VESA modes have nothing to do with
4450    frame buffers per se - they are often just regular extended text modes.
4451    Disabling them just because we don't have frame buffer support is very
4452    wrong.
4453
4454    Cc: H. Peter Anvin <hpa@zytor.com>
4455    Cc: Antonino A. Daplas <adaplas@gmail.com>,
4456    Cc: Andrew Morton <akpm@linux-foundation.org>
4457    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
4458commit e3ebadd95cb621e2c7436f3d3646447ac9d5c16d
4459Author: Linus Torvalds <torvalds@woody.linux-foundation.org>
4460Date:   Mon May 7 08:44:24 2007 -0700
4461
4462    Revert "[PATCH] x86: __pa and __pa_symbol address space separation"
4463
4464    This was broken.  It adds complexity, for no good reason.  Rather than
4465    separate __pa() and __pa_symbol(), we should deprecate __pa_symbol(),
4466    and preferably __pa() too - and just use "virt_to_phys()" instead, which
4467    is more readable and has nicer semantics.
4468
4469    However, right now, just undo the separation, and make __pa_symbol() be
4470    the exact same as __pa().  That fixes the bugs this patch introduced,
4471    and we can do the fairly obvious cleanups later.
4472
4473    Do the new __phys_addr() function (which is now the actual workhorse for
4474    the unified __pa()/__pa_symbol()) as a real external function, that way
4475    all the potential issues with compile/link-time optimizations of
4476    constant symbol addresses go away, and we can also, if we choose to, add
4477    more sanity-checking of the argument.
4478
4479    Cc: Eric W. Biederman <ebiederm@xmission.com>
4480    Cc: Vivek Goyal <vgoyal@in.ibm.com>
4481    Cc: Andi Kleen <ak@suse.de>
4482    Cc: Andrew Morton <akpm@linux-foundation.org>
4483    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
4484commit 46fcc86dd71d70211e965102fb69414c90381880
4485Author: Linus Torvalds <torvalds@woody.linux-foundation.org>
4486Date:   Thu Apr 19 18:21:01 2007 -0700
4487
4488    Revert "e1000: fix NAPI performance on 4-port adapters"
4489
4490    This reverts commit 60cba200f11b6f90f35634c5cd608773ae3721b7.  It's been
4491    linked to lockups of the e1000 hardware, see for example
4492
4493            https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=229603
4494
4495    but it's likely that the commit itself is not really introducing the
4496    bug, but just allowing an unrelated problem to rear its ugly head (ie
4497    one current working theory is that the code exposes us to a hardware
4498    race condition by decreasing the amount of time we spend in each NAPI
4499    poll cycle).
4500
4501    We'll revert it until root cause is known.  Intel has a repeatable
4502    reproduction on two different machines and bus traces of the hardware
4503    doing something bad.
4504
4505    Acked-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
4506    Cc: Jeff Garzik <jeff@garzik.org>
4507    Cc: David S. Miller <davem@davemloft.net>
4508    Cc: Greg KH <gregkh@suse.de>
4509    Cc: Dave Jones <davej@redhat.com>
4510    Cc: Auke Kok <auke-jan.h.kok@intel.com>
4511    Cc: Andrew Morton <akpm@linux-foundation.org>
4512    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
4513commit 296d93cd0205433489b0689533426ce0a8cf2dec
4514Author: Linus Torvalds <torvalds@woody.linux-foundation.org>
4515Date:   Fri Mar 23 08:03:47 2007 -0700
4516
4517    Revert "ACPI: Only use IPI on known broken machines (AMD, Dothan/BaniasPentium M)"
4518
4519    This reverts commit 25496caec111481161e7f06bbfa12a533c43cc6f, which
4520    broke bootup on at least Ingo's ThinkPad T60.  Need to figure out
4521    exactly what is wrong before we can re-do the logic.
4522
4523    Requested-by: Ingo Molnar <mingo@elte.hu>
4524    Acked-by: Thomas Gleixner <tglx@linutronix.de>
4525    Cc: Thomas Renninger <trenn@suse.de>
4526    Cc: Len Brown <len.brown@intel.com>
4527    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
4528commit be521466feb3bb1cd89de82a2b1d080e9ebd3cb6
4529Author: Linus Torvalds <torvalds@woody.linux-foundation.org>
4530Date:   Sat Mar 10 14:22:07 2007 -0800
4531
4532    Revert "USB: pxa2xx_udc: fix hardcoded irq number"
4533
4534    This reverts commit d2487cb4257dafb686f682285854fe7f02ca29d8.
4535
4536    Russell King points out that it's obviously bogus, and I have to agree.
4537    Not only does "irq" not even exist in that scope, but we obviously need
4538    to free the irq that we actually requested, and that's IRQ_USB.
4539
4540    Reported-by: Russell King <rmk@arm.linux.org.uk>
4541    Cc: Andrew Morton <akpm@linux-foundation.org>
4542    Cc: Greg KH <greg@kroah.com>
4543    Cc: David Brownell <david-b@pacbell.net>,
4544    Cc: Milan Svoboda <msvoboda@ra.rockwell.com>
4545    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
4546commit 38f3323037de22bb0089d08be27be01196e7148b
4547Author: Linus Torvalds <torvalds@woody.linux-foundation.org>
4548Date:   Tue Mar 6 19:38:01 2007 -0800
4549
4550    Revert "[PATCH] LOG2: Alter get_order() so that it can make use of ilog2() on a constant"
4551
4552    This reverts commit 39d61db0edb34d60b83c5e0d62d0e906578cc707.
4553
4554    The commit was buggy in multiple ways:
4555     - the conversion to ilog2() was incorrect to begin with
4556     - it tested the wrong #defines, so on all architectures but FRV you'd
4557       never see the bug except for constant arguments.
4558     - the new "get_order()" macro used its arguments multiple times, and
4559       didn't even parenthesize them properly
4560     - despite the comments, it was not true that you could use it for
4561       constant initializers, since not all architectures even use the
4562       generic page.h header file.
4563
4564    All of the problems are individually fixable, but it all boils down to:
4565    better just revert it, and re-do it from scratch.
4566
4567    Cc: David Howells <dhowells@redhat.com>
4568    Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
4569    Cc: Paul Mackerras <paulus@samba.org>
4570    Cc: Andrew Morton <akpm@osdl.org>
4571    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
4572commit 221dee285ee38099b82437531bcae9fa9cb64cc4
4573Author: Linus Torvalds <torvalds@woody.linux-foundation.org>
4574Date:   Mon Feb 26 14:55:48 2007 -0800
4575
4576    Revert "[CPUFREQ] constify cpufreq_driver where possible."
4577
4578    This reverts commit aeeddc1435c37fa3fc844f31d39c185b08de4158, which was
4579    half-baked and broken.  It just resulted in compile errors, since
4580    cpufreq_register_driver() still changes the 'driver_data' by setting
4581    bits in the flags field.  So claiming it is 'const' _really_ doesn't
4582    work.
4583
4584    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
4585commit ea3d5226f52ef30f52aa0a04f47f5919c7facacf
4586Author: Linus Torvalds <torvalds@woody.linux-foundation.org>
4587Date:   Mon Feb 26 09:21:46 2007 -0800
4588
4589    Revert "[PATCH] i386: add idle notifier"
4590
4591    This reverts commit 2ff2d3d74705d34ab71b21f54634fcf50d57bdd5.
4592
4593    Uwe Bugla reports that he cannot mount a floppy drive any more, and Jiri
4594    Slaby bisected it down to this commit.
4595
4596    Benjamin LaHaise also points out that this is a big hot-path, and that
4597    interrupt delivery while idle is very common and should not go through
4598    all these expensive gyrations.
4599
4600    Fix up conflicts in arch/i386/kernel/apic.c and arch/i386/kernel/irq.c
4601    due to other unrelated irq changes.
4602
4603    Cc: Stephane Eranian <eranian@hpl.hp.com>
4604    Cc: Andi Kleen <ak@suse.de>
4605    Cc: Andrew Morton <akpm@osdl.org>
4606    Cc: Uwe Bugla <uwe.bugla@gmx.de>
4607    Cc: Jiri Slaby <jirislaby@gmail.com>
4608    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
4609commit b5bf28cde894b3bb3bd25c13a7647020562f9ea0
4610Author: Linus Torvalds <torvalds@woody.linux-foundation.org>
4611Date:   Wed Feb 21 11:21:44 2007 -0800
4612
4613    Revert "e1000: fix shared interrupt warning message"
4614
4615    This reverts commit d2ed16356ff4fb9de23fbc5e5d582ce580390106.
4616
4617    As Thomas Gleixner reports:
4618      "e1000 is not working anymore. ifup fails permanentely.
4619        ADDRCONF(NETDEV_UP): eth0: link is not ready
4620       nothing else"
4621
4622    The broken commit was identified with "git bisect".
4623
4624    Auke Kok says:
4625      "I think we need to drop this now.  The report that says that this
4626       *fixes* something might have been on regular interrupts only.  I
4627       currently suspect that it breaks all MSI interrupts, which would make
4628       sense if I look a the code.  Very bad indeed."
4629
4630    Cc: Jesse Brandeburg <jesse.brandeburg@intel.com>
4631    Acked-by: Auke Kok <auke-jan.h.kok@intel.com>
4632    Cc: Andrew Morton <akpm@linux-foundation.org>
4633    Cc: Jeff Garzik <jeff@garzik.org>
4634    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
4635commit 552ce544edfbe9bce79952a8c0f8d65b7f2d16bb
4636Author: Linus Torvalds <torvalds@woody.linux-foundation.org>
4637Date:   Tue Feb 13 12:08:18 2007 -0800
4638
4639    Revert "[PATCH] Fix d_path for lazy unmounts"
4640
4641    This reverts commit eb3dfb0cb1f4a44e2d0553f89514ce9f2a9fcaf1.
4642
4643    It causes some strange Gnome problem with dbus-daemon getting stuck, so
4644    we'll revert it until that problem is understood.
4645
4646    Reported by both walt and Greg KH, who both independently git-bisected
4647    the problem to this commit.
4648
4649    Andreas is looking at it.
4650
4651    Reported-by: walt <wa1ter@myrealbox.com>
4652    Reported-by: Greg KH <greg@kroah.com>
4653    Acked-by: Andreas Gruenbacher <agruen@suse.de>
4654    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
4655commit 435f8a605d3b56bb96212f4d70b62ecbd0629340
4656Author: Linus Torvalds <torvalds@woody.linux-foundation.org>
4657Date:   Fri Feb 2 08:07:42 2007 -0800
4658
4659    Revert "[PATCH] fix typo in geode_configre()@cyrix.c"
4660
4661    This reverts commit e4f0ae0ea63caceff37a13f281a72652b7ea71ba.
4662
4663    It's not wrong, but it's not right either, and everybody seems to agree
4664    that the right fix is probably to do the ccr3 write after the ccr4 one
4665    (and that we also should clean it up a bit).  And after that we need to
4666    really validate that all the bits that we write to ccr4 actually do
4667    work.
4668
4669    The old 2.6.19 code was insane, and basically didn't change ccr4 at all
4670    (even though it certainly looks like it was the *intent* to do so).  So
4671    let's revert the change that may fix things, just because it's not what
4672    was actually ever tested when the code was written, even if it _was_ the
4673    intent.
4674
4675    There's a discussion on http://lkml.org/lkml/2007/1/9/63 that was
4676    started by the patch that now gets reverted, and that discussion may
4677    well contain the proper long-term fix.
4678
4679    Suggested-by: Adrian Bunk <bunk@stusta.de>
4680    Acked-by: Andrew Morton <akpm@linux-foundation.org>
4681    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
4682commit 6fd6b17c6d9713f56b5f20903ec3e00fa6cc435e
4683Author: Linus Torvalds <torvalds@woody.linux-foundation.org>
4684Date:   Wed Jan 31 16:43:36 2007 -0800
4685
4686    Revert "[PATCH] mm: micro optimise zone_watermark_ok"
4687
4688    This reverts commit e80ee884ae0e3794ef2b65a18a767d502ad712ee.
4689
4690    Pawel Sikora had a boot-time oops due to it - because the sign change
4691    invalidates the following comparisons, since 'free_pages' can be
4692    negative.
4693
4694    The micro-optimization just isn't worth it.
4695
4696    Bisected-by: Pawel Sikora <pluto@agmk.net>
4697    Acked-by: Andrew Morton <akpm@osdl.org>
4698    Cc: Nick Piggin <nickpiggin@yahoo.com.au>
4699    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
4700commit bcdddfb66cc998252d34758ce4109cedc0d24a5c
4701Author: Linus Torvalds <torvalds@woody.linux-foundation.org>
4702Date:   Tue Jan 30 14:11:12 2007 -0800
4703
4704    Revert "net: ifb error path loop fix"
4705
4706    This reverts commit 0c0b3ae68ec93b1db5c637d294647d1cca0df763.
4707
4708    Quoth David:
4709
4710      "Jeff, please revert
4711
4712       It's wrong.  We had a lengthy analysis of this piece of code
4713       several months ago, and it is correct.
4714
4715       Consider, if we run the loop and we get an error
4716       the following happens:
4717
4718       1) attempt of ifb_init_one(i) fails, therefore we should
4719          not try to "ifb_free_one()" on "i" since it failed
4720       2) the loop iteration first increments "i", then it
4721          check for error
4722
4723       Therefore we must decrement "i" twice before the first
4724       free during the cleanup.  One to "undo" the for() loop
4725       increment, and one to "skip" the ifb_init_one() case which
4726       failed."
4727
4728    Reported-by: David Miller <davem@davemloft.net>
4729    Acked-by: Jeff Garzik <jgarzik@pobox.com>
4730    Cc: Andrew Morton <akpm@osdl.org>
4731    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
4732commit 444f378b237a0f728f5c4aba752c08d13c209344
4733Author: Linus Torvalds <torvalds@woody.linux-foundation.org>
4734Date:   Tue Jan 30 13:35:18 2007 -0800
4735
4736    Revert "[PATCH] namespaces: fix exit race by splitting exit"
4737
4738    This reverts commit 7a238fcba0629b6f2edbcd37458bae56fcf36be5 in
4739    preparation for a better and simpler fix proposed by Eric Biederman
4740    (and fixed up by Serge Hallyn)
4741
4742    Acked-by: Serge E. Hallyn <serue@us.ibm.com>
4743    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
4744commit 6d3154cc1143f62c3b80d9929caeaec6db8cb451
4745Author: Linus Torvalds <torvalds@woody.linux-foundation.org>
4746Date:   Mon Jan 22 08:53:24 2007 -0800
4747
4748    Revert "[PATCH] Fix up mmap_kmem"
4749
4750    This reverts commit 99a10a60ba9bedcf5d70ef81414d3e03816afa3f.
4751
4752    As per Hugh Dickins:
4753
4754      "Nadia Derbey has reported that mmap of /dev/kmem no longer works with
4755       the kernel virtual address as offset, and Franck has confirmed that
4756       his patch came from a misunderstanding of what an offset means to
4757       /dev/kmem - whereas his patch description seems to say that he was
4758       correcting the offset on a few plaforms, there was no such problem to
4759       correct, and his patch was in fact changing its API on all platforms."
4760
4761    Suggested-by: Hugh Dickins <hugh@veritas.com>
4762    Cc: Franck Bui-Huu <fbuihuu@gmail.com>
4763    Cc: Nadia Derbey <Nadia.Derbey@bull.net>
4764    Cc: Andi Kleen <ak@suse.de>
4765    Cc: Arjan van de Ven <arjan@infradead.org>
4766    Cc: Andrew Morton <akpm@osdl.org>
4767    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
4768commit fea5f1e19611d94fbf3905875a427c4cb959cd06
4769Author: Linus Torvalds <torvalds@woody.osdl.org>
4770Date:   Mon Jan 8 15:04:46 2007 -0800
4771
4772    Revert "[PATCH] x86-64: Try multiple timer variants in check_timer"
4773
4774    This reverts commit b026872601976f666bae77b609dc490d1834bf77, which has
4775    been linked to several problem reports with IO-APIC and the timer.
4776    Machines either don't boot because the timer doesn't happen, or we get
4777    double timer interrupts because we end up double-routing the timer irq
4778    through multiple interfaces.
4779
4780    See for example
4781
4782            http://lkml.org/lkml/2006/12/16/101
4783            http://lkml.org/lkml/2007/1/3/9
4784            http://bugzilla.kernel.org/show_bug.cgi?id=7789
4785
4786    about some of the discussion.
4787
4788    Patches to fix this cleanup exist (and have been confirmed to work fine
4789    at least for some of the affected cases) and we'll revisit it for
4790    2.6.21, but this late in the -rc series we're better off just reverting
4791    the incomplete commit that caused the problems.
4792
4793    Suggested-by: Adrian Bunk <bunk@stusta.de>
4794    Cc: Eric W. Biederman <ebiederm@xmission.com>
4795    Cc: Yinghai Lu <yinghai.lu@amd.com>
4796    Cc: Andrew Morton <akpm@osdl.org>
4797    Cc: Andi Kleen <ak@suse.de>
4798    Signed-off-by: Linus Torvalds <torvalds@osdl.org>
4799commit 90cb28e8f76e57751ffe14abd09c2d53a6aea7c8
4800Author: Linus Torvalds <torvalds@woody.osdl.org>
4801Date:   Sat Jan 6 13:28:21 2007 -0800
4802
4803    Revert "[PATCH] binfmt_elf: randomize PIE binaries (2nd try)"
4804
4805    This reverts commit 59287c0913cc9a6c75712a775f6c1c1ef418ef3b.
4806
4807    Hugh Dickins reports that it causes random failures on x86 with SuSE
4808    10.2, and points out
4809
4810      "Isn't that randomization, anywhere from 0x10000 to ELF_ET_DYN_BASE,
4811       sure to place the ET_DYN from time to time just where the comment
4812       says it's trying to avoid? I assume that somehow results in the error
4813       reported."
4814
4815    (where the comment in question is the existing comment in the source
4816    code about mmap/brk clashes).
4817
4818    Suggested-by: Hugh Dickins <hugh@veritas.com>
4819    Acked-by: Marcus Meissner <meissner@suse.de>
4820    Cc: Andrew Morton <akpm@osdl.org>
4821    Cc: Andi Kleen <ak@suse.de>
4822    Cc: Ingo Molnar <mingo@elte.hu>
4823    Cc: Dave Jones <davej@codemonkey.org.uk>
4824    Cc: Arjan van de Ven <arjan@linux.intel.com>
4825    Signed-off-by: Linus Torvalds <torvalds@osdl.org>
4826commit 36f696cd7ffc82c48bf5826bbb4e005370fb9c6a
4827Author: Linus Torvalds <torvalds@woody.osdl.org>
4828Date:   Mon Jan 1 10:55:45 2007 -0800
4829
4830    Revert "[PATCH] x86_64: fix boot hang caused by CALGARY_IOMMU_ENABLED_BY_DEFAULT"
4831
4832    This reverts commit a9622f6219ce58faba1417743bf3078501eb3434.  Now that
4833    the Calgary code apparently detects itself properly, it's not needed any
4834    more.
4835
4836    Signed-off-by: Linus Torvalds <torvalds@osdl.org>
4837commit b8e6ec865fd1d8838b6ce9516977b65e9f08f876
4838Author: Linus Torvalds <torvalds@woody.osdl.org>
4839Date:   Sun Nov 26 16:27:17 2006 -0800
4840
4841    Revert "[PATCH] Enforce "unsigned long flags;" when spinlocking"
4842
4843    This reverts commit ee3ce191e8eaa4cc15c51a28b34143b36404c4f5, since it
4844    broke on at least ARM, MIPS and PA-RISC due to complicated header file
4845    dependencies.
4846
4847    Conflicts in include/linux/spinlock.h (due to the "nested" variety
4848    fixes) fixed up by hand.
4849
4850    Cc: Alexey Dobriyan <adobriyan@gmail.com>
4851    Cc: Ralf Baechle <ralf@linux-mips.org>
4852    Cc: Kyle McMartin <kyle@parisc-linux.org>
4853    Cc: Russell King <rmk+lkml@arm.linux.org.uk>
4854    Signed-off-by: Linus Torvalds <torvalds@osdl.org>
4855commit b976fe19acc565e5137e6f12af7b6633a23e6b7c
4856Author: Linus Torvalds <torvalds@evo.osdl.org>
4857Date:   Fri Nov 17 19:31:09 2006 -0800
4858
4859    Revert "ACPI: created a dedicated workqueue for notify() execution"
4860
4861    This reverts commit 37605a6900f6b4d886d995751fcfeef88c4e462c.
4862
4863    Again.
4864
4865    This same bug has now been introduced twice: it was done earlier by
4866    commit b8d35192c55fb055792ff0641408eaaec7c88988, only to be reverted
4867    last time in commit 72945b2b90a5554975b8f72673ab7139d232a121.
4868
4869    We must NOT try to queue up notify handlers to another thread than the
4870    normal ACPI execution thread, because the notifications on some systems
4871    seem to just keep on accumulating until we run out of memory and/or
4872    threads.
4873
4874    Keeping events within the one deferred execution thread automatically
4875    throttles the events properly.
4876
4877    At least the Compaq N620c will lock up completely on the first thermal
4878    event without this patch reverted.
4879
4880    Cc: David Brownell <david-b@pacbell.net>
4881    Cc: Len Brown <len.brown@intel.com>
4882    Cc: Alexey Starikovskiy <alexey.y.starikovskiy@linux.intel.com>
4883    Signed-off-by: Linus Torvalds <torvalds@osdl.org>
4884commit 9a3a04ac386f44175b6a4142eaeab3d4170a57f3
4885Author: Linus Torvalds <torvalds@woody.osdl.org>
4886Date:   Tue Nov 14 15:20:51 2006 -0800
4887
4888    Revert "[PATCH] fix Data Acess error in dup_fd"
4889
4890    This reverts commit 0130b0b32ee53dc7add773fcea984f6a26ef1da3.
4891
4892    Sergey Vlasov points out (and Vadim Lobanov concurs) that the bug it was
4893    supposed to fix must be some unrelated memory corruption, and the "fix"
4894    actually causes more problems:
4895
4896      "However, the new code does not look safe in all cases.  If some other
4897       task has opened more files while dup_fd() released oldf->file_lock, the
4898       new code will update open_files to the new larger value.  But newf was
4899       allocated with the old smaller value of open_files, therefore subsequent
4900       accesses to newf may try to write into unallocated memory."
4901
4902    so revert it.
4903
4904    Cc: Sharyathi Nagesh <sharyath@in.ibm.com>
4905    Cc: Sergey Vlasov <vsu@altlinux.ru>
4906    Cc: Vadim Lobanov <vlobanov@speakeasy.net>
4907    Cc: Andrew Morton <akpm@osdl.org>
4908    Signed-off-by: Linus Torvalds <torvalds@osdl.org>
4909commit de8e7c12430a73654ae3cedbc45428d56c6b777b
4910Author: Linus Torvalds <torvalds@g5.osdl.org>
4911Date:   Wed Nov 8 10:09:28 2006 -0800
4912
4913    Revert "[PATCH] i386: Add MMCFG resources to i386 too"
4914
4915    This reverts commit de09bddb9d6f96785be470c832b881e6d72d589f.  It tried
4916    to reserve the MMCONFIG mmio memory ranges, but since the MMCONFIG
4917    information is broken and often bogus (which is why we don't dare use it
4918    most of the time _anyway_), it does more harm than good.
4919
4920    Cc: Jeff Chua <jeff.chua.linux@gmail.com>
4921    Cc: Adrian Bunk <bunk@stusta.de>
4922    Cc: Andi Kleen <ak@suse.de>
4923    Signed-off-by: Linus Torvalds <torvalds@osdl.org>
4924commit 80491eb90c750fcd7d13830062f27ae9b7cc5f75
4925Author: Linus Torvalds <torvalds@g5.osdl.org>
4926Date:   Sat Nov 4 09:55:00 2006 -0800
4927
4928    Revert unintentional "volatile" changes in ipc/msg.c
4929
4930    Commit 5a06a363ef48444186f18095ae1b932dddbbfa89 ("[PATCH] ipc/msg.c:
4931    clean up coding style") breaks fakeroot on Alpha (variously hangs or
4932    oopses), according to a report by Falk Hueffner.
4933
4934    The fact that the code seems to rely on compiler access ordering through
4935    the use of "volatile" is a pretty certain sign that the code has locking
4936    problems, and we should fix those properly and then remove the whole
4937    "volatile" entirely.
4938
4939    But in the meantime, the movement of "volatile" was unintentional, and
4940    should be reverted.
4941
4942    Cc: Falk Hueffner <falk@debian.org>
4943    Cc: Andrew Morton <akpm@osdl.org>
4944    Acked-by: Ingo Molnar <mingo@elte.hu>
4945    Signed-off-by: Linus Torvalds <torvalds@osdl.org>
4946commit 209ad53bc19667a128d9c51beba873a5c62bff6e
4947Author: Linus Torvalds <torvalds@g5.osdl.org>
4948Date:   Sun Oct 29 17:31:49 2006 -0800
4949
4950    Revert "r8169: mac address change support"
4951
4952    This reverts commit a2b98a697fa4e7564f78905b83db122824916cf9.
4953
4954    As per Guennadi Liakhovetski, the mac address change support code breaks
4955    some normal uses (_without_ any address changes), and until it's all
4956    sorted out, we're better off without it.
4957
4958    Says Francois:
4959
4960      "Go revert it.
4961
4962       Despite what I claimed, I can not find a third-party confirmation by
4963       email that it works elsewhere.
4964
4965       It would probably be enough to remove the call to
4966       __rtl8169_set_mac_addr() in rtl8169_hw_start() though."
4967
4968    See also
4969
4970            http://bugzilla.kernel.org/show_bug.cgi?id=6032
4971
4972    Cc: Guennadi Liakhovetski <g.liakhovetski@gmx.de>
4973    Acked-by: Francois Romieu <romieu@fr.zoreil.com>
4974    Signed-off-by: Linus Torvalds <torvalds@osdl.org>
4975commit 0c0e4668e0e65dd1404e8cf066d147235f95561d
4976Author: Linus Torvalds <torvalds@g5.osdl.org>
4977Date:   Mon Oct 23 14:25:30 2006 -0700
4978
4979    Revert unintentional and bogus change to drivers/pci/quirks.c
4980
4981    In commit 4e8a5201506423e0241202de1349422af4260296 ("[PKT_SCHED] netem:
4982    Orphan SKB when adding to queue.") Davem mistakenly also included a
4983    temporary diff in his tree that disabled the pci_fixup_video VGA quirk,
4984    which broke sparc64.
4985
4986    This reverts that part of the commit. Sayeth Davem:
4987
4988      "Greg KH has a patch coming to you soon which will move that VGA code
4989       back into x86/x86_64/IA64 specific areas and will fix the sparc64
4990       problem properly."
4991
4992    Special thanks to Claudio Martins <ctpm@ist.utl.pt> for noticing the
4993    error in the first place.
4994
4995    Cc: Claudio Martins <ctpm@ist.utl.pt>
4996    Cc: David Miller <davem@davemloft.net>
4997    Signed-off-by: Linus Torvalds <torvalds@osdl.org>
4998commit b2ef7858db6394b758818358a43c7dd5f232bbcc
4999Author: Linus Torvalds <torvalds@g5.osdl.org>
5000Date:   Fri Oct 20 10:40:48 2006 -0700
5001
5002    Revert "[mv643xx] Add pci device table for auto module loading."
5003
5004    This reverts commit 4596c75c23dde2623cbeec69357d5eb13d28387e as
5005    requested by Olaf Hering. It causes compile errors, and says Olaf:
5006
5007      "This change is also wrong, the autoloading works perfect with 2.6.18,
5008       no need to add random PCI ids.
5009
5010       See commit a0245f7ad5214cb00131d7cd176446e067c913dc, platform devices
5011       have now a modalias entry in sysfs.  The network card is not a PCI
5012       device."
5013
5014    Signed-off-by: Linus Torvalds <torvalds@osdl.org>
5015commit e24bb60e11e3fe9858b71874a4ac59333adbc4fc
5016Author: Linus Torvalds <torvalds@g5.osdl.org>
5017Date:   Sun Oct 8 15:38:22 2006 -0700
5018
5019    Revert "[POWERPC] Don't get PCI IRQ from OF for devices with no IRQ"
5020
5021    This reverts commit 41550c5128150175197257b6ceab2cd50dea7b51.
5022
5023    Quoth Ben Herrenschmidt:
5024     "Please revert this one for now.  It seems to break G5s :( Looks like
5025      PCI cells inside Apple IO ASICs don't have a PCI_INTERRUPT_LINE set.
5026      I need to figure out a better fix."
5027
5028    Signed-off-by: Linus Torvalds <torvalds@osdl.org>
5029commit 79e453d49bd49ba1b576f89310cc565c9e4ca379
5030Author: Linus Torvalds <torvalds@g5.osdl.org>
5031Date:   Tue Sep 19 08:15:22 2006 -0700
5032
5033    Revert mmiocfg heuristics and blacklist changes
5034
5035    This reverts commits 11012d419cfc0e0f78ca356aca03674217910124 and
5036    40dd2d20f220eda1cd0da8ea3f0f9db8971ba237, which allowed us to use the
5037    MMIO accesses for PCI config cycles even without the area being marked
5038    reserved in the e820 memory tables.
5039
5040    Those changes were needed for EFI-environment Intel macs, but broke some
5041    newer Intel 965 boards, so for now it's better to revert to our old
5042    2.6.17 behaviour and at least avoid introducing any new breakage.
5043
5044    Andi Kleen has a set of patches that work with both EFI and the broken
5045    Intel 965 boards, which will be applied once they get wider testing.
5046
5047    Cc: Arjan van de Ven <arjan@infradead.org>
5048    Cc: Edgar Hucek <hostmaster@ed-soft.at>
5049    Cc: Andi Kleen <ak@suse.de>
5050    Signed-off-by: Linus Torvalds <torvalds@osdl.org>
5051commit 2f1b9250516aa8744496abbbbb8a6066cb2b56db
5052Author: Linus Torvalds <torvalds@evo.osdl.org>
5053Date:   Wed Jul 12 21:04:16 2006 -0700
5054
5055    Revert "[PATCH] pcmcia: Make ide_cs work with the memory space of CF-Cards if IO space is not available"
5056
5057    This reverts commit 5040cb8b7e61b7a03e8837920b9eb2c839bb1947.
5058
5059    It breaks previously working ide-cs PIO configurations, causing problems
5060    like
5061
5062            ide2: I/O resource 0xF883200E-0xF883200E not free.
5063            ide2: ports already in use, skipping probe
5064
5065    rather than a working kernel.
5066
5067    Cc: Thomas Kleffel <tk@maintech.de>
5068    Cc: Dominik Brodowski <linux@dominikbrodowski.net>
5069    Signed-off-by: Linus Torvalds <torvalds@osdl.org>
5070commit 953969ddf5b049361ed1e8471cc43dc4134d2a6f
5071Author: Linus Torvalds <torvalds@g5.osdl.org>
5072Date:   Sun Jul 9 08:47:46 2006 -0700
5073
5074    Revert "ACPI: dock driver"
5075
5076    This reverts commit a5e1b94008f2a96abf4a0c0371a55a56b320c13e.
5077
5078    Adrian Bunk points out that it has build errors, and apparently no
5079    maintenance. Throw it out.
5080
5081    Signed-off-by: Linus Torvalds <torvalds@osdl.org>
5082commit d38b69689c349f35502b92e20dafb30c62d49d63
5083Author: Linus Torvalds <torvalds@g5.osdl.org>
5084Date:   Mon Jun 26 16:59:26 2006 -0700
5085
5086    Revert "kbuild: fix make -rR breakage"
5087
5088    This reverts commit e5c44fd88c146755da6941d047de4d97651404a9.
5089
5090    Thanks to Daniel Ritz and Michal Piotrowski for noticing the problem.
5091
5092    Daniel says:
5093
5094      "[The] reason is a recent change that made modules always shows as
5095       module.mod.  it breaks modprobe and probably many scripts..besides
5096       lsmod looking horrible
5097
5098       stuff like this in modprobe.conf:
5099            install pcmcia_core /sbin/modprobe --ignore-install pcmcia_core; /sbin/modprobe pcmcia
5100       makes modprobe fork/exec endlessly calling itself...until oom
5101       interrupts it"
5102
5103    Signed-off-by: Linus Torvalds <torvalds@osdl.org>
5104commit 09c0dc68625c06f5b1e786aad0d5369b592179e6
5105Author: Linus Torvalds <torvalds@g5.osdl.org>
5106Date:   Mon Jun 26 11:55:42 2006 -0700
5107
5108    Revert "[PATCH] kthread: update loop.c to use kthread"
5109
5110    This reverts commit c7b2eff059fcc2d1b7085ee3d84b79fd657a537b.
5111
5112    Hugh Dickins explains:
5113
5114     "It seems too little tested: "losetup -d /dev/loop0" fails with
5115      EINVAL because nothing sets lo_thread; but even when you patch
5116      loop_thread() to set lo->lo_thread = current, it can't survive
5117      more than a few dozen iterations of the loop below (with a tmpfs
5118      mounted on /tst):
5119
5120            j=0
5121            cp /dev/zero /tst
5122            while :
5123            do
5124                let j=j+1
5125                echo "Doing pass $j"
5126                losetup /dev/loop0 /tst/zero
5127                mkfs -t ext2 -b 1024 /dev/loop0 >/dev/null 2>&1
5128                mount -t ext2 /dev/loop0 /mnt
5129                umount /mnt
5130                losetup -d /dev/loop0
5131            done
5132
5133      it collapses with failed ioctl then BUG_ON(!bio).
5134
5135      I think the original lo_done completion was more subtle and safe
5136      than the kthread conversion has allowed for."
5137
5138    Signed-off-by: Linus Torvalds <torvalds@osdl.org>
5139commit 3448097fccdce4ea8f0fcad4f37f502a8cd72e68
5140Author: Linus Torvalds <torvalds@g5.osdl.org>
5141Date:   Sun Jun 25 18:41:00 2006 -0700
5142
5143    Revert "swsusp special saveable pages support" commits
5144
5145    This reverts commits
5146
5147      3e3318dee0878d42ed62a19c292a2ac284135db3 [PATCH] swsusp: x86_64 mark special saveable/unsaveable pages
5148      b6370d96e09944c6e3ae8d5743ca8a8ab1f79f6c [PATCH] swsusp: i386 mark special saveable/unsaveable pages
5149      ce4ab0012b32c1a4a1d6e934aeb73bf3151c48d9 [PATCH] swsusp: add architecture special saveable pages support
5150
5151    because not only do they apparently cause page faults on x86, the
5152    infrastructure doesn't compile on powerpc.
5153
5154    Signed-off-by: Linus Torvalds <torvalds@osdl.org>
5155commit 83626b01275d0228516b4d97da008328fc37c934
5156Author: Linus Torvalds <torvalds@g5.osdl.org>
5157Date:   Sat Jun 24 17:47:09 2006 -0700
5158
5159    Revert "[PATCH] usb: drivers/usb/core/devio.c dereferences a userspace pointer"
5160
5161    This reverts commit 786dc1d3d7333f269e17d742886eac2188a2d9cc.
5162
5163    As Al so eloquently points out, the patch is crap. The old code was fine,
5164    the new code was bogus.
5165
5166    It never dereferenced a user pointer, the "->" operator was to an array
5167    member, which gives the _address_ of the member (in user space), not an
5168    actual dereference at all.
5169
5170    Signed-off-by: Linus Torvalds <torvalds@osdl.org>
5171commit 9a8fca0499c611ab37b5c0d4481ca09d3f6e8101
5172Author: Linus Torvalds <torvalds@g5.osdl.org>
5173Date:   Tue May 30 20:32:15 2006 -0700
5174
5175    Revert "[PATCH] i386/x86_64: Force pci=noacpi on HP XW9300"
5176
5177    This reverts commit 5491d0f3e206beb95eeb506510d62a1dab462df1.
5178
5179    As per Andi:
5180
5181      "After some discussion with people who have the affected system it
5182       seems best to revert for 2.6.17.  It broke a common BIOS workaround
5183       and PCI-X still doesn't work.  Alternative is for people to change
5184       the BIOS which seems to be better right now."
5185
5186    Signed-off-by: Linus Torvalds <torvalds@osdl.org>
5187commit f1adad78dd2fc8edaa513e0bde92b4c64340245c
5188Author: Linus Torvalds <torvalds@g5.osdl.org>
5189Date:   Sun May 21 18:54:09 2006 -0700
5190
5191    Revert "[PATCH] sched: fix interactive task starvation"
5192
5193    This reverts commit 5ce74abe788a26698876e66b9c9ce7e7acc25413 (and its
5194    dependent commit 8a5bc075b8d8cf7a87b3f08fad2fba0f5d13295e), because of
5195    audio underruns.
5196
5197    Reported by Rene Herman <rene.herman@keyaccess.nl>, who also pinpointed
5198    the exact cause of the underruns:
5199
5200      "Audio underruns galore, with only ogg123 and firefox (browsing the
5201       GIT tree online is also a nice trigger by the way).
5202
5203       If I back it out, everything is fine for me again."
5204
5205    Cc: Rene Herman <rene.herman@keyaccess.nl>
5206    Cc: Mike Galbraith <efault@gmx.de>
5207    Acked-by: Con Kolivas <kernel@kolivas.org>
5208    Signed-off-by: Linus Torvalds <torvalds@osdl.org>
5209commit 9be2f7c38e0bd64e8a0f74ea68df1e73e2ddfcc3
5210Author: Linus Torvalds <torvalds@g5.osdl.org>
5211Date:   Sat May 13 08:01:23 2006 -0700
5212
5213    Revert "[PATCH] i386: export: memory more than 4G through /proc/iomem"
5214
5215    This reverts commit 10dbe196a8da6b3196881269c6639c0ec11c36cb.
5216
5217    The resource struct is still 32-bit, so trying to save a 64-bit memory
5218    size there obviously won't work.
5219
5220    When we merge the 64-bit resource series, we can re-enable this.
5221
5222    Thanks to Sachin Sant and Maneesh Soni for debugging
5223
5224    Cc: Maneesh Soni <maneesh@in.ibm.com>
5225    Cc: Sachin Sant <sachinp@in.ibm.com>
5226    Cc: Russell King <rmk+lkml@arm.linux.org.uk>
5227    Cc: Sharyathi Nagesh <sharyath@in.ibm.com>
5228    Cc: Arjan van de Ven <arjan@infradead.org>
5229    Cc: Vivek Goyal <vgoyal@in.ibm.com>
5230    Signed-off-by: Linus Torvalds <torvalds@osdl.org>
5231commit 1a2acc9e9214699a99389e323e6686e9e0e2ca67
5232Author: Linus Torvalds <torvalds@g5.osdl.org>
5233Date:   Fri May 12 12:08:46 2006 -0700
5234
5235    Revert "[BLOCK] Fix oops on removal of SD/MMC card"
5236
5237    This reverts commit 56cf6504fc1c0c221b82cebc16a444b684140fb7.
5238
5239    Both Erik Mouw and Andrew Vasquez independently pinpointed this commit
5240    as causing problems, where the slab cache for a driver is never released
5241    (most obviously causing problems when immediately re-loading that
5242    driver, resulting in a "kmem_cache_create: duplicate cache <xyz>"
5243    message, but it can also cause other trouble).
5244
5245    James Bottomley dug into it, and reports:
5246
5247      "OK, here's the scoop.  The problem patch adds a get of driverfs_dev in
5248       add_disk(), but doesn't put it again until disk_release() (which occurs
5249       on final put_disk() of the gendisk).
5250
5251       However, in SCSI, the driverfs_dev is the sdev_gendev.  That means
5252       there's a reference held on sdev_gendev  until final disk put.
5253       Unfortunately, we use the driver model driver_remove to trigger
5254       del_gendisk (which removes the gendisk from visibility and decrements
5255       the refcount), so we've introduced an unbreakable deadlock in the
5256       reference counting with this.
5257
5258       I suggest simply reversing this patch at the moment.  If Russell and
5259       Jens can tell me what they're trying to do I'll see if there's another
5260       way to do it."
5261
5262    so hereby the patch gets reverted, waiting for a better fix.
5263
5264    Cc: Jens Axboe <axboe@suse.de>
5265    Cc: Russell King <rmk@arm.linux.org.uk>
5266    Cc: James Bottomley <James.Bottomley@SteelEye.com>
5267    Cc: Erik Mouw <erik@harddisk-recovery.com>
5268    Cc: Andrew Vasquez <andrew.vasquez@qlogic.com>
5269    Signed-off-by: Linus Torvalds <torvalds@osdl.org>
5270commit 601e7f024edbea8018de34c83a7398623214e636
5271Author: Linus Torvalds <torvalds@g5.osdl.org>
5272Date:   Mon May 8 13:38:42 2006 -0700
5273
5274    Revert "kbuild: fix modpost segfault for 64bit mipsel kernel"
5275
5276    This reverts commit c8d8b837ebe4b4f11e1b0c4a2bdc358c697692ed, which
5277    caused problems for the x86 build. Quoth Sam:
5278
5279      "It was discussed on mips list but apparently the fix was bogus.  I
5280       will not have time to look into it so mips can carry this local fix
5281       until we get a proper fix in mainline."
5282
5283    Signed-off-by: Linus Torvalds <torvalds@osdl.org>
5284commit e0f4ab8a1741193891f096aa63df9ac8672af54c
5285Author: Linus Torvalds <torvalds@g5.osdl.org>
5286Date:   Tue Mar 21 14:51:37 2006 -0800
5287
5288    Revert "V4L/DVB (3543): Fix Makefile to adapt to bt8xx/ conversion"
5289
5290    This reverts commit 08f1d0b99f4e2203935d86640a7fec5c233b777c
5291
5292    The "bt8xx/ conversion" for drivers/video/ hasn't actually percolated
5293    all the way to this tree, so the Makefile change escaped too soon.
5294
5295    Build breakage noticed by Jeff Garzik <jeff@garzik.org>
5296
5297    Signed-off-by: Linus Torvalds <torvalds@osdl.org>
5298commit cbf0ec6ee001ab89471397ac84105b2e3b2fd986
5299Author: Linus Torvalds <torvalds@g5.osdl.org>
5300Date:   Tue Mar 14 08:01:47 2006 -0800
5301
5302    Revert "[PATCH] x86-64: Fix up handling of non canonical user RIPs"
5303
5304    This reverts commit c33d4568aca9028a22857f94f5e0850012b6444b.
5305
5306    Andrew Clayton and Hugh Dickins report that it's broken for them and
5307    causes strange page table and slab corruption, and spontaneous reboots.
5308
5309    Let's get it right next time.
5310
5311    Cc: Andrew Clayton <andrew@rootshell.co.uk>
5312    Cc: Hugh Dickins <hugh@veritas.com>
5313    Cc: Andi Kleen <ak@suse.de>
5314    Signed-off-by: Linus Torvalds <torvalds@osdl.org>
5315commit 637029c6cb5efcbaa3d5831af4c1972bdd629779
5316Author: Linus Torvalds <torvalds@g5.osdl.org>
5317Date:   Mon Feb 27 20:41:56 2006 -0800
5318
5319    Revert "[PATCH] x86_64: Only do the clustered systems have unsynchronized TSC assumption on IBM systems"
5320
5321    This reverts commit 13a229abc25640813f1480c0478dfc6bdbc1c19e.
5322
5323    Quoth Andi:
5324      "After some consideration and feedback from various people it turns
5325       out this wasn't that good an idea.  It has some problems and needs
5326       more work.  Since it was only an optimization anyways it's best to
5327       just back it out again for now."
5328
5329    Signed-off-by: Linus Torvalds <torvalds@osdl.org>
5330commit 15508d22d00277a1f2a1022dce38f2772c810d32
5331Author: Linus Torvalds <torvalds@g5.osdl.org>
5332Date:   Thu Feb 9 08:00:14 2006 -0800
5333
5334    Revert "[PATCH] kconfig: detect if -lintl is needed when linking conf,mconf"
5335
5336    This reverts commit 5e375bc7d586e0df971734a5a5f1f080ffd89b68.
5337
5338    Kyle McMartin steps on his soap-box:
5339
5340      "Sigh.  Can everyone please stop assuming gcc can output to /dev/null?
5341       On several platforms, ld tries to lseek in the output file, and fails
5342       if it can't."
5343
5344    Signed-off-by: Linus Torvalds <torvalds@osdl.org>
5345commit cef5076987dd545ac74f4efcf1c962be8eac34b0
5346Author: Linus Torvalds <torvalds@g5.osdl.org>
5347Date:   Sun Feb 5 10:51:57 2006 -0800
5348
5349    Revert "[PATCH] x86_64: Fix the node cpumask of a cpu going down"
5350
5351    This reverts commit 10f4dc8b27ac42f930ac55adb8c521264dc997f8.
5352
5353    Quoth Andi Kleen:
5354      "Kiran decided that it makes the problem worse than it was before.
5355       Fixing it fully requires more work which is too much for 2.6.16.  So
5356       please revert that commit for now."
5357
5358    Signed-off-by: Linus Torvalds <torvalds@osdl.org>
5359commit 392c14beaca2ee85a98d0c6b453501be67423a20
5360Author: Linus Torvalds <torvalds@g5.osdl.org>
5361Date:   Thu Dec 29 13:01:54 2005 -0800
5362
5363    Revert radeon AGP aperture offset changes
5364
5365    This reverts the series of commits
5366
5367            67dbb4ea33731415fe09c62149a34f472719ac1d
5368            281ab031a8c9e5b593142eb4ec59a87faae8676a
5369            47807ce381acc34a7ffee2b42e35e96c0f322e52
5370
5371    that changed the GART VM start offset.  It fixed some machines, but
5372    seems to continually interact badly with some X versions.
5373
5374    Quoth Ben Herrenschmidt:
5375
5376      "So I think at this point, the best is that we keep the old bogus code
5377       that at least is consistent with the bug in the server. I'm working on a
5378       big patch to X that reworks the memory map stuff completely and fixes
5379       those issues on the server side, I'll do a DRM patch matching this X fix
5380       as well so that the memory map is only ever set in one place and with
5381       what I hope is a correct algorithm..."
5382
5383    Signed-off-by: Linus Torvalds <torvalds@osdl.org>
5384commit 49d7bc64283970ee83d2c954d04ba00d04e5943d
5385Author: Linus Torvalds <torvalds@g5.osdl.org>
5386Date:   Mon Dec 12 11:25:04 2005 -0800
5387
5388    Revert revert of "[SCSI] fix usb storage oops"
5389
5390    This reverts commit 1b0997f561bf46689cc6e0903f342e9bf2506bf1, which in
5391    turn reverted 34ea80ec6a02ad02e6b9c75c478c18e5880d6713 (which is thus
5392    re-instated).
5393
5394    Quoth James Bottomley:
5395
5396      "All it's doing is deferring the device_put() from the
5397       scsi_put_command() to after the scsi_run_queue(), which doesn't fix
5398       the sleep while atomic problem of the device release method.  In both
5399       cases we still get the semaphore in atomic context problem which is
5400       caused by scsi_reap_target() doing a device_del(), which I assumed
5401       (wrongly) was valid from atomic context."
5402
5403    who also promised to fix scsi_reap_target().
5404
5405    Signed-off-by: Linus Torvalds <torvalds@osdl.org>
5406commit 1b0997f561bf46689cc6e0903f342e9bf2506bf1
5407Author: Linus Torvalds <torvalds@g5.osdl.org>
5408Date:   Fri Dec 2 15:54:37 2005 -0800
5409
5410    Revert "[SCSI] fix usb storage oops"
5411
5412    This reverts commit 34ea80ec6a02ad02e6b9c75c478c18e5880d6713.
5413
5414    It does a put_device() from softirq context, which is bad since it gets
5415    a semaphore for reading.
5416
5417    Signed-off-by: Linus Torvalds <torvalds@osdl.org>
5418commit 346f7dbb17cb7d84317b4410df2e8f7bf2e57f44
5419Author: Linus Torvalds <torvalds@g5.osdl.org>
5420Date:   Wed Nov 30 10:22:30 2005 -0800
5421
5422    Revert "[PATCH] pci_ids.h: remove duplicate entries"
5423
5424    This reverts commit c9d6073fb3cda856132dd544d537679f9715436c.
5425
5426    It was totally bogus.
5427
5428    Signed-off-by: Linus Torvalds <torvalds@osdl.org>
5429commit f747307ed1defcdfd37a3ef84c48e3138691cd26
5430Author: Linus Torvalds <torvalds@g5.osdl.org>
5431Date:   Tue Nov 29 14:21:57 2005 -0800
5432
5433    Revert "[PATCH] drivers/message/fusion/mptbase.c: make code static"
5434
5435    This reverts commit 252ac865535e1ea9cc2d28be83f477d8d8b961a2.
5436
5437    It impacts the LSI customers using the mptstm target mode drivers
5438    (source tar-ball at
5439
5440      ftp://ftp.lsil.com/HostAdapterDrivers/linux/Fusion-MPT/mptstm-1.00.13-src.tar.gz
5441
5442    for those who care).
5443
5444    Signed-off-by: Linus Torvalds <torvalds@osdl.org>
5445commit 2d0ebb36038c0626cde662a3b06da9787cfb68c3
5446Author: Linus Torvalds <torvalds@g5.osdl.org>
5447Date:   Wed Nov 23 08:44:05 2005 -0800
5448
5449    Revert "[NET]: Shut up warnings in net/core/flow.c"
5450
5451    This reverts commit af2b4079ab154bd12e8c12b02db5f31b31babe63
5452
5453    Changing the #define to an inline function breaks on non-SMP builds,
5454    since wuite a few places in the kernel do not implement the ipi handler
5455    when compiling for UP.
5456
5457    Signed-off-by: Linus Torvalds <torvalds@osdl.org>
5458commit 0fde7f591860e48dd210144ff24c061da03bfa39
5459Author: Linus Torvalds <torvalds@g5.osdl.org>
5460Date:   Sat Nov 12 10:14:02 2005 -0800
5461
5462    Revert "[PATCH] fbcon: Add rl (Roman Large) font"
5463
5464    This reverts 998e6d51162707685336ff99c029c8911b270d32 commit.
5465commit 1e4c85f97fe26fbd70da12148b3992c0e00361fd
5466Author: Linus Torvalds <torvalds@g5.osdl.org>
5467Date:   Mon Oct 31 19:16:17 2005 -0800
5468
5469    Revert "i386: move apic init in init_IRQs"
5470
5471    Commit f2b36db692b7ff6972320ad9839ae656a3b0ee3e causes a bootup hang on
5472    at least one machine.  Revert for now until we understand why.  The old
5473    code may be ugly, but it works.
5474
5475    Signed-off-by: Linus Torvalds <torvalds@osdl.org>
5476commit 79b95a454bb5c1d9b7287d1016a70885ba3f346c
5477Author: Linus Torvalds <torvalds@g5.osdl.org>
5478Date:   Thu Oct 27 16:28:39 2005 -0700
5479
5480    Revert "x86-64: Avoid unnecessary double bouncing for swiotlb"
5481
5482    Commit id 6142891a0c0209c91aa4a98f725de0d6e2ed4918
5483
5484    Andi Kleen reports that it seems to break things for some people,
5485    and since it's purely a small optimization, revert it for now.
5486
5487    Signed-off-by: Linus Torvalds <torvalds@osdl.org>
5488commit a362f463a6d316d14daed0f817e151835ce97ff7
5489Author: Linus Torvalds <torvalds@g5.osdl.org>
5490Date:   Thu Oct 27 09:07:33 2005 -0700
5491
5492    Revert "remove false BUG_ON() from run_posix_cpu_timers()"
5493
5494    This reverts commit 3de463c7d9d58f8cf3395268230cb20a4c15bffa.
5495
5496    Roland has another patch that allows us to leave the BUG_ON() in place
5497    by just making sure that the condition it tests for really is always
5498    true.
5499
5500    That goes in next.
5501
5502    Signed-off-by: Linus Torvalds <torvalds@osdl.org>
5503commit 9465bee863bc4c6cf1566c12d6f92a8133e3da5c
5504Author: Linus Torvalds <torvalds@g5.osdl.org>
5505Date:   Fri Oct 21 15:36:00 2005 -0700
5506
5507    Revert "Fix cpu timers exit deadlock and races"
5508
5509    Revert commit e03d13e985d48ac4885382c9e3b1510c78bd047f, to be replaced
5510    by a much nicer fix from Roland.
5511commit 6e3254c4e2927c117044a02acf5f5b56e1373053
5512Author: Linus Torvalds <torvalds@g5.osdl.org>
5513Date:   Fri Sep 30 12:38:27 2005 -0700
5514
5515    Revert "x86-64: Reverse order of bootmem lists"
5516
5517    As requested by Thomas Gleixner <tglx@linutronix.de>:
5518
5519      "5d3d0f7704ed0bc7eaca0501eeae3e5da1ea6c87 breaks a couple of ARM
5520       boards, which depend on the historical bootmem allocation order.
5521       There is a cleaner solution around to remove the pgdat list
5522       completely, but this is a topic for post 2.6.14
5523
5524       Andi signalled ACK already."
5525
5526    Signed-off-by: Linus Torvalds <torvalds@osdl.org>
5527commit 4a8342d233a39ee582e9f7260e12d2f5fd194a05
5528Author: Linus Torvalds <torvalds@g5.osdl.org>
5529Date:   Thu Sep 29 15:18:21 2005 -0700
5530
5531    Revert task flag re-ordering, add comments
5532
5533    Roland points out that the flags end up having non-obvious dependencies
5534    elsewhere, so revert aa55a08687059aa169d10a313c41f238c2070488 and add
5535    some comments about why things are as they are.
5536
5537    We'll just have to fix up the broken comparisons. Roland has a patch.
5538
5539    Signed-off-by: Linus Torvalds <torvalds@osdl.org>
5540commit c231c7db30faf93419fc22d680f74d816bea70e2
5541Author: Linus Torvalds <torvalds@g5.osdl.org>
5542Date:   Wed Aug 17 13:07:28 2005 -0700
5543
5544    Revert unnecessary zlib_inflate/inftrees.c fix
5545
5546    It turns out that empty distance code tables are not an error, and that
5547    a compressed block with only literals can validly have an empty table
5548    and should not be flagged as a data error.
5549
5550    Some old versions of gzip had problems with this case, but it does not
5551    affect the zlib code in the kernel.
5552
5553    Analysis and explanations thanks to Sergey Vlasov <vsu@altlinux.ru>
5554
5555    Signed-off-by: Linus Torvalds <torvalds@osdl.org>
5556commit 2ba84684e8cf6f980e4e95a2300f53a505eb794e
5557Author: Linus Torvalds <torvalds@g5.osdl.org>
5558Date:   Sun Aug 14 18:21:30 2005 -0700
5559
5560    Revert PCIBIOS_MIN_IO changes for 2.6.13
5561
5562    This reverts commits
5563
5564      71db63acff69618b3d9d3114bd061938150e146b
5565            [PATCH] increase PCIBIOS_MIN_IO on x86
5566
5567    and
5568
5569      0b2bfb4e7ff61f286676867c3508569bea6fbf7a
5570        ACPI: increase PCIBIOS_MIN_IO on x86
5571
5572    since Lukas Sandströ<lukass@etek.chalmers.se> reports that this breaks
5573    his on-board nvidia audio.
5574
5575    We should re-visit this later. For now we revert the change
5576
5577    Signed-off-by: Linus Torvalds <torvalds@osdl.org>
5578commit b4b08e581fac8e0ba9ae348bdc13246c9798c99e
5579Author: Linus Torvalds <torvalds@g5.osdl.org>
5580Date:   Sun Aug 14 15:43:39 2005 -0700
5581
5582    Revert "dc395x: Fix support for highmem"
5583
5584    It introduces a repeatable oops in the driver, which is a bigger problem
5585    than the patch tries to solve. From the original description:
5586
5587        Author: Jamie Lenehan <lenehan@twibble.org>
5588        Date:   Thu Mar 3 14:41:40 2005 +0200
5589
5590        [PATCH] dc395x: Fix support for highmem
5591
5592        From: Guennadi Liakhovetski <g.liakhovetski@gmx.de>
5593
5594        Removes the page_to_virt and maps sg lists dynamically.
5595        This makes the driver work with highmem pages.
5596
5597        Signed-off-by: Guennadi Liakhovetski <g.liakhovetski@gmx.de>
5598        Signed-off-by: Jamie Lenehan <lenehan@twibble.org>
5599        Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
5600
5601    Signed-off-by: Guennadi Liakhovetski <g.liakhovetski@gmx.de>
5602    Signed-off-by: Linus Torvalds <torvalds@osdl.org>
5603commit dc836b5b6fcde95f750a4790d8200fabaf563dc9
5604Author: Linus Torvalds <torvalds@g5.osdl.org>
5605Date:   Mon Aug 8 18:46:09 2005 -0700
5606
5607    Revert "[PATCH] PCI: restore BAR values..."
5608
5609    Revert commit fec59a711eef002d4ef9eb8de09dd0a26986eb77, which is
5610    breaking sparc64 that doesn't have a working pci_update_resource.
5611
5612    We'll re-do this after 2.6.13 when we'll do it all properly.
5613commit 697a2d63a3844caaa2b6565ab7f3d69086af94d4
5614Author: Linus Torvalds <torvalds@evo.osdl.org>
5615Date:   Mon Aug 1 12:37:54 2005 -0700
5616
5617    Revert ACPI interrupt resume changes
5618
5619    If there are devices that use interrupts over a suspend event, ACPI must
5620    restore the PCI interrupt links on resume.  Anything else breaks any
5621    device that hasn't been converted to the new (dubious) PM rules.
5622
5623    Drivers that need the irq free/re-aquire sequence can be done one by one
5624    independently of this one.
5625commit 889371f61fd5bb914d0331268f12432590cf7e85
5626Author: Linus Torvalds <torvalds@g5.osdl.org>
5627Date:   Sat Jul 30 13:41:56 2005 -0700
5628
5629    Revert "yenta free_irq on suspend"
5630
5631    ACPI is wrong.  Devices should not release their IRQ's on suspend and
5632    re-aquire them on resume.  ACPI should just re-init the IRQ controller
5633    instead of breaking most drivers very subtly.
5634
5635    Breakage reported by Hugh Dickins <hugh@veritas.com>
5636
5637    Undo: d8c4b4195c7d664baf296818bf756775149232d3
5638
5639    Signed-off-by: Linus Torvalds <torvalds@osdl.org>
5640commit 2ac6608c41f8c45371ea9dddae7f99bc2c15d5cf
5641Author: Linus Torvalds <torvalds@g5.osdl.org>
5642Date:   Thu Jul 28 10:34:47 2005 -0700
5643
5644    Revert broken "statement with no effect" warning fix
5645
5646    It may shut up gcc, but it also incorrectly changes the semantics of the
5647    smp_call_function() helpers.
5648
5649    You can fix the warning other ways if you are interested (create another
5650    inline function that takes no arguments and returns zero), but
5651    preferably gcc just shouldn't complain about unused return values from
5652    statement expressions in the first place.