Userland should be able to trust the pid and uid of the sender of a
signal if the si_code is SI_TKILL.
Unfortunately, the kernel has historically allowed sigqueueinfo() to
send any si_code at all (as long as it was negative - to distinguish it
from kernel-generated signals like SIGILL etc), so it could spoof a
SI_TKILL with incorrect siginfo values.
Happily, it looks like glibc has always set si_code to the appropriate
SI_QUEUE, so there are probably no actual user code that ever uses
anything but the appropriate SI_QUEUE flag.
So just tighten the check for si_code (we used to allow any negative
value), and add a (one-time) warning in case there are binaries out
there that might depend on using other si_code values.
A virtualized display device is usually viewed with the vncviewer
application, either by 'xm vnc domU' or with vncviewer localhost:port.
vncviewer and the RFB protocol provides absolute coordinates to the
virtual display. These coordinates are either passed through to a PV
guest or converted to relative coordinates for a HVM guest.
A PV guest receives these coordinates and passes them to the kernels
evdev driver. There it can be picked up by applications such as the
xorg-input drivers. Using absolute coordinates avoids issues such as
guest mouse pointer not tracking host mouse pointer due to wrong mouse
acceleration settings in the guests X display.
Advertise either absolute or relative coordinates to the input system
and the evdev driver, depending on what dom0 provides. The xorg-input
driver prefers relative coordinates even if a devices provides both.
If a device doesn't support power management (pm_cap == 0) but it is
acpi_pci_power_manageable() because there is a _PS0 method declared for
it and _EJ0 is also declared for the slot then nobody is going to set
current_state = PCI_D0 for this device. This is what I think it is
happening:
pci_enable_device
|
__pci_enable_device_flags
/* here we do not set current_state because !pm_cap */
|
do_pci_enable_device
|
pci_set_power_state
|
__pci_start_power_transition
|
pci_platform_power_transition
/* platform_pci_power_manageable() calls acpi_pci_power_manageable that
* returns true */
|
platform_pci_set_power_state
/* acpi_pci_set_power_state gets called and does nothing because the
* acpi device has _EJ0, see the comment "If the ACPI device has _EJ0,
* ignore the device" */
at this point if we refer to the commit message that introduced the
comment above (10b3dcae0f275e2546e55303d64ddbb58cec7599), it is up to
the hotplug driver to set the state to D0.
However AFAICT the pci hotplug driver never does, in fact
drivers/pci/hotplug/acpiphp_glue.c:register_slot sets the slot flags to
(SLOT_ENABLED | SLOT_POWEREDON) but it does not set the pci device
current state to PCI_D0.
So my proposed fix is also to set current_state = PCI_D0 in
register_slot.
Comments are very welcome.
The oom killer naturally defers killing anything if it finds an eligible
task that is already exiting and has yet to detach its ->mm. This avoids
unnecessarily killing tasks when one is already in the exit path and may
free enough memory that the oom killer is no longer needed. This is
detected by PF_EXITING since threads that have already detached its ->mm
are no longer considered at all.
The problem with always deferring when a thread is PF_EXITING, however, is
that it may never actually exit when being traced, specifically if another
task is tracing it with PTRACE_O_TRACEEXIT. The oom killer does not want
to defer in this case since there is no guarantee that thread will ever
exit without intervention.
This patch will now only defer the oom killer when a thread is PF_EXITING
and no ptracer has stopped its progress in the exit path. It also ensures
that a child is sacrificed for the chosen parent only if it has a
different ->mm as the comment implies: this ensures that the thread group
leader is always targeted appropriately.
We shouldn't defer oom killing if a thread has already detached its ->mm
and still has TIF_MEMDIE set. Memory needs to be freed, so find kill
other threads that pin the same ->mm or find another task to kill.
This patch prevents unnecessary oom kills or kernel panics by reverting
two commits:
495789a5 (oom: make oom_score to per-process value) cef1d352 (oom: multi threaded process coredump don't make deadlock)
First, 495789a5 (oom: make oom_score to per-process value) ignores the
fact that all threads in a thread group do not necessarily exit at the
same time.
It is imperative that select_bad_process() detect threads that are in the
exit path, specifically those with PF_EXITING set, to prevent needlessly
killing additional tasks. If a process is oom killed and the thread group
leader exits, select_bad_process() cannot detect the other threads that
are PF_EXITING by iterating over only processes. Thus, it currently
chooses another task unnecessarily for oom kill or panics the machine when
nothing else is eligible.
By iterating over threads instead, it is possible to detect threads that
are exiting and nominate them for oom kill so they get access to memory
reserves.
Second, cef1d352 (oom: multi threaded process coredump don't make
deadlock) erroneously avoids making the oom killer a no-op when an
eligible thread other than current isfound to be exiting. We want to
detect this situation so that we may allow that exiting thread time to
exit and free its memory; if it is able to exit on its own, that should
free memory so current is no loner oom. If it is not able to exit on its
own, the oom killer will nominate it for oom kill which, in this case,
only means it will get access to memory reserves.
Without this change, it is easy for the oom killer to unnecessarily target
tasks when all threads of a victim don't exit before the thread group
leader or, in the worst case, panic the machine.
This patch reverts 5a03b051 ("thp: use compaction in kswapd for GFP_ATOMIC
order > 0") due to reports stating that kswapd CPU usage was higher and
IRQs were being disabled more frequently. This was reported at
http://www.spinics.net/linux/fedora/alsa-user/msg09885.html.
Without this patch applied, CPU usage by kswapd hovers around the 20% mark
according to the tester (Arthur Marsh:
http://www.spinics.net/linux/fedora/alsa-user/msg09899.html). With this
patch applied, it's around 2%.
The problem is not related to THP which specifies __GFP_NO_KSWAPD but is
triggered by high-order allocations hitting the low watermark for their
order and waking kswapd on kernels with CONFIG_COMPACTION set. The most
common trigger for this is network cards configured for jumbo frames but
it's also possible it'll be triggered by fork-heavy workloads (order-1)
and some wireless cards which depend on order-1 allocations.
The symptoms for the user will be high CPU usage by kswapd in low-memory
situations which could be confused with another writeback problem. While
a patch like 5a03b051 may be reintroduced in the future, this patch plays
it safe for now and reverts it.
[mel@csn.ul.ie: Beefed up the changelog] Signed-off-by: Andrea Arcangeli <aarcange@redhat.com> Signed-off-by: Mel Gorman <mel@csn.ul.ie> Reported-by: Arthur Marsh <arthur.marsh@internode.on.net> Tested-by: Arthur Marsh <arthur.marsh@internode.on.net> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
If an administrator tries to swapon a file backed by NFS, the inode mutex is
taken (as it is for any swapfile) but later identified to be a bad swapfile
due to the lack of bmap and tries to cleanup. During cleanup, an attempt is
made to close the file but with inode->i_mutex still held. Closing an NFS
file syncs it which tries to acquire the inode mutex leading to deadlock. If
lockdep is enabled the following appears on the console;
=============================================
[ INFO: possible recursive locking detected ]
2.6.38-rc8-autobuild #1
---------------------------------------------
swapon/2192 is trying to acquire lock:
(&sb->s_type->i_mutex_key#13){+.+.+.}, at: vfs_fsync_range+0x47/0x7c
but task is already holding lock:
(&sb->s_type->i_mutex_key#13){+.+.+.}, at: sys_swapon+0x28d/0xae7
other info that might help us debug this:
1 lock held by swapon/2192:
#0: (&sb->s_type->i_mutex_key#13){+.+.+.}, at: sys_swapon+0x28d/0xae7
This patch releases the mutex if its held before calling filep_close()
so swapon fails as expected without deadlock when the swapfile is backed
by NFS. If accepted for 2.6.39, it should also be considered a -stable
candidate for 2.6.38 and 2.6.37.
Up to 2.6.22, you could use remap_file_pages(2) on a tmpfs file or a
shared mapping of /dev/zero or a shared anonymous mapping. In 2.6.23 we
disabled it by default, but set VM_CAN_NONLINEAR to enable it on safe
mappings. We made sure to set it in shmem_mmap() for tmpfs files, but
missed it in shmem_zero_setup() for the others. Fix that at last.
list_del() leaves poison in the prev and next pointers. The next
list_empty() will compare those poisons, and say the list isn't empty.
Any list operations that assume the node is on a list because of such a
check will be fooled into dereferencing poison. One needs to INIT the
node after the del, and fortunately there's already a wrapper for that -
list_del_init().
Some of the dels are followed by deallocations, so can be ignored, and one
can be merged with an add to make a move. Apart from that, I erred on the
side of caution in making nodes list_empty()-queriable.
Signed-off-by: Phil Carmody <ext-phil.2.carmody@nokia.com> Reviewed-by: Paul Menage <menage@google.com> Cc: Li Zefan <lizf@cn.fujitsu.com> Acked-by: Kirill A. Shutemov <kirill@shutemov.name> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
The test program below will hang because io_getevents() uses
add_wait_queue_exclusive(), which means the wake_up() in io_destroy() only
wakes up one of the threads. Fix this by using wake_up_all() in the aio
code paths where we want to make sure no one gets stuck.
| Mount-cache hash table entries: 512
| slab error in verify_redzone_free(): cache `idr_layer_cache': memory outside object was overwritten
| Backtrace:
| [<c0227088>] (dump_backtrace+0x0/0x110) from [<c0431afc>] (dump_stack+0x18/0x1c)
| [<c0431ae4>] (dump_stack+0x0/0x1c) from [<c0293304>] (__slab_error+0x28/0x30)
| [<c02932dc>] (__slab_error+0x0/0x30) from [<c0293a74>] (cache_free_debugcheck+0x1c0/0x2b8)
| [<c02938b4>] (cache_free_debugcheck+0x0/0x2b8) from [<c0293f78>] (kmem_cache_free+0x3c/0xc0)
| [<c0293f3c>] (kmem_cache_free+0x0/0xc0) from [<c032b1c8>] (ida_get_new_above+0x19c/0x1c0)
| [<c032b02c>] (ida_get_new_above+0x0/0x1c0) from [<c02af7ec>] (alloc_vfsmnt+0x54/0x144)
| [<c02af798>] (alloc_vfsmnt+0x0/0x144) from [<c0299830>] (vfs_kern_mount+0x30/0xec)
| [<c0299800>] (vfs_kern_mount+0x0/0xec) from [<c0299908>] (kern_mount_data+0x1c/0x20)
| [<c02998ec>] (kern_mount_data+0x0/0x20) from [<c02146c4>] (sysfs_init+0x68/0xc8)
| [<c021465c>] (sysfs_init+0x0/0xc8) from [<c02137d4>] (mnt_init+0x90/0x1b0)
| [<c0213744>] (mnt_init+0x0/0x1b0) from [<c0213388>] (vfs_caches_init+0x100/0x140)
| [<c0213288>] (vfs_caches_init+0x0/0x140) from [<c0208c0c>] (start_kernel+0x2e8/0x368)
| [<c0208924>] (start_kernel+0x0/0x368) from [<c0208034>] (__enable_mmu+0x0/0x2c)
| c0113268: redzone 1:0xd84156c5c032b3ac, redzone 2:0xd84156c5635688c0.
| slab error in cache_alloc_debugcheck_after(): cache `idr_layer_cache': double free, or memory outside object was overwritten
| ...
| c011307c: redzone 1:0x9f91102ffffffff, redzone 2:0x9f911029d74e35b
| slab: Internal list corruption detected in cache 'idr_layer_cache'(24), slabp c0113000(16). Hexdump:
|
| 000: 20 4f 10 c0 20 4f 10 c0 7c 00 00 00 7c 30 11 c0
| 010: 10 00 00 00 10 00 00 00 00 00 c9 17 fe ff ff ff
| 020: fe ff ff ff fe ff ff ff fe ff ff ff fe ff ff ff
| 030: fe ff ff ff fe ff ff ff fe ff ff ff fe ff ff ff
| 040: fe ff ff ff fe ff ff ff fe ff ff ff fe ff ff ff
| 050: fe ff ff ff fe ff ff ff fe ff ff ff 11 00 00 00
| 060: 12 00 00 00 13 00 00 00 14 00 00 00 15 00 00 00
| 070: 16 00 00 00 17 00 00 00 c0 88 56 63
| kernel BUG at /home/rmk/git/linux-2.6-rmk/mm/slab.c:2928!
Reference: https://lkml.org/lkml/2011/2/7/238 Reported-and-analyzed-by: Russell King <rmk@arm.linux.org.uk> Signed-off-by: Pekka Enberg <penberg@kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
This structure was accidentally defined such that its layout can
differ between 32-bit and 64-bit processes. Add compat structure
definitions and an ioctl wrapper function.
Signed-off-by: Ben Hutchings <bhutchings@solarflare.com> Acked-by: Alexander Duyck <alexander.h.duyck@intel.com> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
ASoC audio for mini2440 platform in current kenrel doesn't work.
First problem is samsung_asoc_dma device is missing in initialization.
Next problem is with codec. Codec is initialized but never probed
because no platform_device exist for codec driver. It leads to errors
during codec binding to asoc dai. Next problem was platform data which
was passed from board to asoc main driver but not passed to codec when
called codec_soc_probe().
Following patch should fix issues. But not sure if in correct way.
Please review.
Signed-off-by: Marek Belisko <marek.belisko@open-nandra.com> Acked-by: Liam Girdwood <lrg@ti.com> Signed-off-by: Mark Brown <broonie@opensource.wolfsonmicro.com> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Similar to commit 7e59e097c09b82760bb0fe08b0fa2b704d76c3f4, this patch
avoids unnecessary volume control indices for more
Realtek auto-parsers, e g the ALC66x family, on the "Surround" and "Side"
controls.
These indices cause these volume controls to be ignored by PulseAudio and
vmaster and should be removed whenever possible.
Reported-by: Jan Losinski <losinski@wh2.tu-dresden.de> Signed-off-by: David Henningsson <david.henningsson@canonical.com> Signed-off-by: Takashi Iwai <tiwai@suse.de> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
An integer overflow occurs in the calculation of RHlinear when the
relative humidity is greater than around 30%. The consequence is a subtle
(but noticeable) error in the resulting humidity measurement.
Signed-off-by: Vivien Didelot <vivien.didelot@savoirfairelinux.com> Signed-off-by: Jean Delvare <khali@linux-fr.org> Cc: Jonathan Cameron <jic23@cam.ac.uk> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
The MUSB driver doesn't see its platform device on DM644x EVM board anymore
since commit 73b089b052a69020b953312a624a6e1eb5b81fab (usb: musb: split davinci
to its own platform_driver) because the new probe is called as subsys_initcall()
now, and the device is registered later than that by the board code. Move the
registration to davinci_evm_init() -- it's safe to do so because the MUSB core
device still gets initialized as fs_initcall() -- which is late enough for the
I2C GPIO expander (which controls VBUS) to be initialized.
Signed-off-by: Sergei Shtylyov <sshtylyov@ru.mvista.com> Acked-by: Felipe Balbi <balbi@ti.com> Tested-by: Sekhar Nori <nsekhar@ti.com> Signed-off-by: Kevin Hilman <khilman@ti.com> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
The latest binutils (2.21.0.20110302/Ubuntu) breaks the build
yet another time, under CONFIG_XEN=y due to a .size directive that
refers to a slightly differently named (hence, to the now very
strict and unforgiving assembler, non-existent) symbol.
[ mingo:
This unnecessary build breakage caused by new binutils
version 2.21 gets escallated back several kernel releases spanning
several years of Linux history, affecting over 130,000 upstream
kernel commits (!), on CONFIG_XEN=y 64-bit kernels (i.e. essentially
affecting all major Linux distro kernel configs).
Git annotate tells us that this slight debug symbol code mismatch
bug has been introduced in 2008 in commit 3d75e1b8:
Human reviewers almost never catch such small mismatches, and binutils
never even warned about it either.
This new binutils version thus breaks the Xen build on all upstream kernels
since v2.6.27, out of the blue.
This makes a straightforward Git bisection of all 64-bit Xen-enabled kernels
impossible on such binutils, for a bisection window of over hundred
thousand historic commits. (!)
This is a major fail on the side of binutils and binutils needs to turn
this show-stopper build failure into a warning ASAP. ]
Signed-off-by: Alexander van Heukelum <heukelum@fastmail.fm> Cc: Jeremy Fitzhardinge <jeremy@goop.org> Cc: Jan Beulich <jbeulich@novell.com> Cc: H.J. Lu <hjl.tools@gmail.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: "H. Peter Anvin" <hpa@zytor.com> Cc: Kees Cook <kees.cook@canonical.com>
LKML-Reference: <1299877178-26063-1-git-send-email-heukelum@fastmail.fm> Signed-off-by: Ingo Molnar <mingo@elte.hu> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
During redetection of a SDIO card, a request for a new card RCA
was submitted to the card, but was then overwritten by the old RCA.
This caused the card to be deselected instead of selected when using
the incorrect RCA. This bug's been present since the "oldcard"
handling was introduced in 2.6.32.
When the multi input quirk is set, there is a new input device
created for every feature report. Since the idea is to present
features per hid device, not per input device, revert back to
the original report loop and change the feature_mapping() callback
to not take the input device as argument.
Signed-off-by: Henrik Rydberg <rydberg@euromail.se> Tested-by: Benjamin Tissoires <benjmain.tissoires@gmail.com> Signed-off-by: Jiri Kosina <jkosina@suse.cz> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
This device does not tolerate delayed opening and goes into a coma if
we try to that. Ubuntu even has a crutch for udev that opened the device
upon seeing it for the first time, but it did not work if we happened to
boot with the device attached, since by the time userspace got around
opening the device it was too late. Let's start the device immediately
to deal with this issue.
According to intel CPU manual, every time PGD entry is changed in i386 PAE
mode, we need do a full TLB flush. Current code follows this and there is
comment for this too in the code.
But current code misses the multi-threaded case. A changed page table
might be used by several CPUs, every such CPU should flush TLB. Usually
this isn't a problem, because we prepopulate all PGD entries at process
fork. But when the process does munmap and follows new mmap, this issue
will be triggered.
When it happens, some CPUs keep doing page faults:
Mike Galbraith reported finding a lockup ("perma-spin bug") where the
cpumask passed to smp_call_function_many was cleared by other cpu(s)
while a cpu was preparing its call_data block, resulting in no cpu to
clear the last ref and unlock the block.
Having cpus clear their bit asynchronously could be useful on a mask of
cpus that might have a translation context, or cpus that need a push to
complete an rcu window.
Instead of adding a BUG_ON and requiring yet another cpumask copy, just
detect the race and handle it.
Note: arch_send_call_function_ipi_mask must still handle an empty
cpumask because the data block is globally visible before the that arch
callback is made. And (obviously) there are no guarantees to which cpus
are notified if the mask is changed during the call; only cpus that were
online and had their mask bit set during the whole call are guaranteed
to be called.
Reported-by: Mike Galbraith <efault@gmx.de> Reported-by: Jan Beulich <JBeulich@novell.com> Acked-by: Jan Beulich <jbeulich@novell.com> Signed-off-by: Milton Miller <miltonm@bga.com> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Paul McKenney's review pointed out two problems with the barriers in the
2.6.38 update to the smp call function many code.
First, a barrier that would force the func and info members of data to
be visible before their consumption in the interrupt handler was
missing. This can be solved by adding a smp_wmb between setting the
func and info members and setting setting the cpumask; this will pair
with the existing and required smp_rmb ordering the cpumask read before
the read of refs. This placement avoids the need a second smp_rmb in
the interrupt handler which would be executed on each of the N cpus
executing the call request. (I was thinking this barrier was present
but was not).
Second, the previous write to refs (establishing the zero that we the
interrupt handler was testing from all cpus) was performed by a third
party cpu. This would invoke transitivity which, as a recient or
concurrent addition to memory-barriers.txt now explicitly states, would
require a full smp_mb().
However, we know the cpumask will only be set by one cpu (the data
owner) and any preivous iteration of the mask would have cleared by the
reading cpu. By redundantly writing refs to 0 on the owning cpu before
the smp_wmb, the write to refs will follow the same path as the writes
that set the cpumask, which in turn allows us to keep the barrier in the
interrupt handler a smp_rmb instead of promoting it to a smp_mb (which
will be be executed by N cpus for each of the possible M elements on the
list).
I moved and expanded the comment about our (ab)use of the rcu list
primitives for the concurrent walk earlier into this function. I
considered moving the first two paragraphs to the queue list head and
lock, but felt it would have been too disconected from the code.
Cc: Paul McKinney <paulmck@linux.vnet.ibm.com> Signed-off-by: Milton Miller <miltonm@bga.com> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Peter pointed out there was nothing preventing the list_del_rcu in
smp_call_function_interrupt from running before the list_add_rcu in
smp_call_function_many.
Fix this by not setting refs until we have gotten the lock for the list.
Take advantage of the wmb in list_add_rcu to save an explicit additional
one.
I tried to force this race with a udelay before the lock & list_add and
by mixing all 64 online cpus with just 3 random cpus in the mask, but
was unsuccessful. Still, inspection shows a valid race, and the fix is
a extension of the existing protection window in the current code.
Reported-by: Peter Zijlstra <peterz@infradead.org> Signed-off-by: Milton Miller <miltonm@bga.com> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
When ext3_dx_add_entry() has to split an index node, it has to ensure that
name_len of dx_node's fake_dirent is also zero, because otherwise e2fsck
won't recognise it as an intermediate htree node and consider the htree to
be corrupted.
Signed-off-by: Eric Sandeen <sandeen@redhat.com> Signed-off-by: Jan Kara <jack@suse.cz> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Some versions of grep don't treat '\s' properly. When building perf on such
systems and using a kernel tarball the perf version is unable to be determined
from the main kernel Makefile and the user is left with a version of '..'.
Replacing the use of '\s' with '[[:space:]]', which should work in all grep
versions, gives a usable version number.
Events on POWER7 can roll back if a speculative event doesn't
eventually complete. Unfortunately in some rare cases they will
raise a performance monitor exception. We need to catch this to
ensure we reset the PMC. In all cases the PMC will be 256 or less
cycles from overflow.
Signed-off-by: Anton Blanchard <anton@samba.org> Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <20110309143842.6c22845e@kryten> Signed-off-by: Ingo Molnar <mingo@elte.hu> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
RPC task RPC_TASK_QUEUED bit is set must be checked before trying to wake up
task rpc_killall_tasks() because task->tk_waitqueue can not be set (equal to
NULL).
Also, as Trond Myklebust mentioned, such approach (instead of checking
tk_waitqueue to NULL) allows us to "optimise away the call to
rpc_wake_up_queued_task() altogether for those
tasks that aren't queued".
Here is an example of dereferencing of tk_waitqueue equal to NULL:
This fixes a race in which the task->tk_callback() puts the rpc_task
to sleep, setting a new callback. Under certain circumstances, the current
code may end up executing the task->tk_action before it gets round to the
callback.
The use of blk_execute_rq_nowait() implies __blk_put_request() is needed
in stpg_endio() rather than blk_put_request() -- blk_finish_request() is
called with queue lock already held.
Signed-off-by: Joseph Gruher <joseph.r.gruher@intel.com> Signed-off-by: Ilgu Hong <ilgu.hong@promise.com> Signed-off-by: Mike Snitzer <snitzer@redhat.com> Signed-off-by: James Bottomley <James.Bottomley@suse.de> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
the target infrastructure fails to send the correct conventional size
to READ_CAPACITY that force a retry with READ_CAPACITY_16, which reads
the capacity for devices > 2TB. Fix by adding the correct return to
trigger RC(16).
Reported-by: Ben Jarvis <bjarvismn@gmail.com> Signed-off-by: Signed-off-by: Nicholas A. Bellinger <nab@linux-iscsi.org> Signed-off-by: James Bottomley <James.Bottomley@suse.de> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
loopback_pos_update() can be called in the timer callback, thus the lock
held should be irq-safe. Otherwise you'll get AB/BA deadlock together
with substream->self_group.lock.
The user-supplied index into the adapters array needs to be checked, or
an out-of-bounds kernel pointer could be accessed and used, leading to
potentially exploitable memory corruption.
Commit 280c73d ("PCI: centralize the capabilities code in
pci-sysfs.c") changed the initialisation of the "rom" and "vpd"
attributes, and made the failure path for the "vpd" attribute
incorrect. We must free the new attribute structure (attr), but
instead we currently free dev->vpd->attr. That will normally be NULL,
resulting in a memory leak, but it might be a stale pointer, resulting
in a double-free.
Some broken BIOSes on ICH4 chipset report an ACPI region which is in
conflict with legacy IDE ports when ACPI is disabled. Even though the
regions overlap, IDE ports are working correctly (we cannot find out
the decoding rules on chipsets).
So the only problem is the reported region itself, if we don't reserve
the region in the quirk everything works as expected.
This patch avoids reserving any quirk regions below PCIBIOS_MIN_IO
which is 0x1000. Some regions might be (and are by a fast google
query) below this border, but the only difference is that they won't
be reserved anymore. They should still work though the same as before.
The conflicts look like (1f.0 is bridge, 1f.1 is IDE ctrl):
pci 0000:00:1f.1: address space collision: [io 0x0170-0x0177] conflicts with 0000:00:1f.0 [io 0x0100-0x017f]
At 0x0100 a 128 bytes long ACPI region is reported in the quirk for
ICH4. ata_piix then fails to find disks because the IDE legacy ports
are zeroed:
ata_piix 0000:00:1f.1: device not available (can't reserve [io 0x0000-0x0007])
Per ICH4 and ICH6 specs, ACPI and GPIO regions are valid iff ACPI_EN
and GPIO_EN bits are set to 1. Add checks for these bits into the
quirks prior to the region creation.
If BIOS doesn't allocate resources for the SR-IOV BARs, zero the Flash
BAR and program the SR-IOV BARs to use the old Flash Memory Space.
Please refer to Intel 82576 Gigabit Ethernet Controller Datasheet
section 7.9.2.14.2 for details.
http://download.intel.com/design/network/datashts/82576_Datasheet.pdf
Signed-off-by: Yu Zhao <yu.zhao@intel.com> Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
This quirk was added before SR-IOV was in production and now all machines that
originally had this issue alreayd have bios updates to correct the issue. The
quirk itself is no longer needed and in fact causes bugs if run. Remove it.
As reported on http://ubuntuforums.org/showthread.php?t=1594007 the
PKB-1700 needs same special handling as WKB-2000. This change is
originally based on patch posted by user asmoore82 on the Ubuntu
forums.
The magic trackpad and mouse both report touch orientation in opposite
direction to the bcm5974 driver and what is written in
Documents/input/multi-touch-protocol.txt. This patch reverts the
direction, so that all in-kernel devices with this feature behave the
same way.
Since no known application has been utilizing this information yet, it
seems appropriate also for stable.
Cc: Michael Poole <mdpoole@troilus.org> Signed-off-by: Henrik Rydberg <rydberg@euromail.se> Acked-by: Chase Douglas <chase.douglas@canonical.com> Signed-off-by: Jiri Kosina <jkosina@suse.cz> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Currently some special handling for the unusual case like dual-ADCs
or a single-input-src is done in the tree-parse time in
set_capture_mixer(). But this setup could be overwritten by static
init verbs.
This patch moves the initialization into the init phase so that
such input-src setup won't be lost.
When the mux for digital mic is different from the mux for other mics,
the current auto-parser doesn't handle them in a right way but provides
only one mic. This patch fixes the issue.
Without this change, a volume control named "Surround" or "Side" would
get an unnecessary index, causing it to be ignored by the vmaster and
PulseAudio.
Use the Mult and bMaxBurst values from the endpoint companion
descriptor to calculate the max length of an isoc transfer.
Add USB_SS_MULT macro to access Mult field of bmAttributes, at
Sarah's suggestion.
This patch should be queued for the 2.6.36 and 2.6.37 stable trees, since
those were the first kernels to have isochronous support for SuperSpeed
devices.
Signed-off-by: Paul Zimmerman <paulz@synopsys.com> Signed-off-by: Sarah Sharp <sarah.a.sharp@linux.intel.com> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
When an endpoint stalls, we need to update the xHCI host's internal
dequeue pointer to move it past the stalled transfer. This includes
updating the cycle bit (TRB ownership bit) if we have moved the dequeue
pointer past a link TRB with the toggle cycle bit set.
When we're trying to find the new dequeue segment, find_trb_seg() is
supposed to keep track of whether we've passed any link TRBs with the
toggle cycle bit set. However, this while loop's body
Will never get executed if the ring only contains one segment.
find_trb_seg() will return immediately, without updating the new cycle
bit. Since find_trb_seg() has no idea where in the segment the TD that
stalled was, make the caller, xhci_find_new_dequeue_state(), check for
this special case and update the cycle bit accordingly.
This patch should be queued to kernels all the way back to 2.6.31.
When an endpoint stalls, the xHCI driver must move the endpoint ring's
dequeue pointer past the stalled transfer. To do that, the driver issues
a Set TR Dequeue Pointer command, which will complete some time later.
Takashi was having issues with USB 1.1 audio devices that stalled, and his
analysis of the code was that the old code would not update the xHCI
driver's ring dequeue pointer after the command completes. However, the
dequeue pointer is set in xhci_find_new_dequeue_state(), just before the
set command is issued to the hardware.
Setting the dequeue pointer before the Set TR Dequeue Pointer command
completes is a dangerous thing to do, since the xHCI hardware can fail the
command. Instead, store the new dequeue pointer in the xhci_virt_ep
structure, and update the ring's dequeue pointer when the Set TR dequeue
pointer command completes.
While we're at it, make sure we can't queue another Set TR Dequeue Command
while the first one is still being processed. This just won't work with
the internal xHCI state code. I'm still not sure if this is the right
thing to do, since we might have a case where a driver queues multiple
URBs to a control ring, one of the URBs Stalls, and then the driver tries
to cancel the second URB. There may be a race condition there where the
xHCI driver might try to issue multiple Set TR Dequeue Pointer commands,
but I would have to think very hard about how the Stop Endpoint and
cancellation code works. Keep the fix simple until when/if we run into
that case.
This patch should be queued to kernels all the way back to 2.6.31.
The hcd->state variable is a disaster. It's not clearly owned by
either usbcore or the host controller drivers, and they both change it
from time to time, potentially stepping on each other's toes. It's
not protected by any locks. And there's no mechanism to prevent it
from going through an invalid transition.
This patch (as1451) takes a first step toward fixing these problems.
As it turns out, usbcore uses hcd->state for essentially only two
things: checking whether the controller's root hub is running and
checking whether the controller has died. Therefore the patch adds
two new atomic bitflags to the hcd structure, to store these pieces of
information. The new flags are used only by usbcore, and a private
spinlock prevents invalid combinations (a dead controller's root hub
cannot be running).
The patch does not change the places where usbcore sets hcd->state,
since HCDs may depend on them. Furthermore, there is one place in
usb_hcd_irq() where usbcore still must use hcd->state: An HCD's
interrupt handler can implicitly indicate that the controller died by
setting hcd->state to HC_STATE_HALT. Nevertheless, the new code is a
big improvement over the current code.
The patch makes one other change. The hcd_bus_suspend() and
hcd_bus_resume() routines now check first whether the host controller
has died; if it has then they return immediately without calling the
HCD's bus_suspend or bus_resume methods.
This fixes the major problem reported in Bugzilla #29902: The system
fails to suspend after a host controller dies during system resume.
Signed-off-by: Alan Stern <stern@rowland.harvard.edu> Tested-by: Alex Terekhov <a.terekhov@gmail.com> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
If a device plug/unplug is detected on an ATI SB700 USB controller in D3,
it appears to set the port status register but not the controller status
register. As a result we'll fail to detect the plug event. Check the port
status register on resume as well in order to catch this case.
Signed-off-by: Matthew Garrett <mjg@redhat.com> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
The document says:
|2.1 Problem description
| When at least two USB devices are simultaneously running, it is observed that
| sometimes the INT corresponding to one of the USB devices stops occurring. This may
| be observed sometimes with USB-to-serial or USB-to-network devices.
| The problem is not noticed when only USB mass storage devices are running.
|2.2 Implication
| This issue is because of the clearing of the respective Done Map bit on reading the ATL
| PTD Done Map register when an INT is generated by another PTD completion, but is not
| found set on that read access. In this situation, the respective Done Map bit will remain
| reset and no further INT will be asserted so the data transfer corresponding to that USB
| device will stop.
|2.3 Workaround
| An SOF INT can be used instead of an ATL INT with polling on Done bits. A time-out can
| be implemented and if a certain Done bit is never set, verification of the PTD completion
| can be done by reading PTD contents (valid bit).
| This is a proven workaround implemented in software.
Russell King run into this with an USB-to-serial converter. This patch
implements his suggestion to enable the high frequent SOF interrupt only
at the time we have ATL packages queued. It goes even one step further
and enables the SOF interrupt only if we have more than one ATL packet
queued at the same time.
Tested-by: Russell King <rmk+kernel@arm.linux.org.uk> Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
The driver for the RTL8187L chips returns IEEE80211_TX_STAT_ACK for all
packets, even if the maximum number of retries was exhausted. In addition
it fails to setup max_rates in the ieee80211_hw struct, This behavior
may be responsible for the problems noted in Bug 14168. As the bug is very
old, testers have not been found, and I do not have the case where the
indicated signal is less than -70 dBm.
Signed-off-by: Larry Finger <Larry.Finger@lwfinger.net> Acked-by: Hin-Tak Leung <htl10@users.sourceforge.net> Signed-off-by: John W. Linville <linville@tuxdriver.com> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
We need to protect not only the dmm_map list, but the individual
map_obj's, otherwise, we might be building the scatter-gather list with
garbage. So, use the existing proc_lock for that.
I observed race conditions which caused kernel panics while running
stress tests, also, Tuomas Kulve found it happening quite often in
Gumstix Over. This patch fixes those.
Cc: Tuomas Kulve <tuomas@kulve.fi> Signed-off-by: Felipe Contreras <felipe.contreras@nokia.com> Signed-off-by: Omar Ramirez Luna <omar.ramirez@ti.com> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
I picked up a new DAK-780EX(professional digitl reverb/mix system),
which use CH341T chipset to communication with computer on 3/2011
and the CH341T's vendor code is 1a86
Looking up the CH341T's vendor and product id's I see:
1a86 QinHeng Electronics
5523 CH341 in serial mode, usb to serial port converter
CH341T,CH341 are the products of the same company, maybe
have some common hardware, and I test the ch341.c works
well with CH341T
On https://bugs.launchpad.net/ubuntu/+source/linux/+bug/636091, one of
the cases reported is a big timeout on option_send_setup, which causes
some side effects as tty_lock is held. Looks like some of ZTE MF626
devices also don't like the RTS/DTR setting in option_send_setup, like
with 4G XS Stick W14. The reporter confirms which this it solves the
long freezes in his system.
Signed-off-by: Herton Ronaldo Krzesinski <herton.krzesinski@canonical.com> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
When a driver doesn't know how much data a device is going to send,
the buffer size should be at least as big as the endpoint's maxpacket
value. The serial drivers don't follow this rule; many of them
request only 256-byte bulk-in buffers. As a result, they suffer
overflow errors if a high-speed device wants to send a lot of data,
because high-speed bulk endpoints are required to have a maxpacket
size of 512.
This patch (as1450) fixes the problem by using the driver's
bulk_in_size value as a minimum, always allocating buffers no smaller
than the endpoint's maxpacket size.
In 8250.c original ns16550 autoconfig code, we change the divisor latch when
we goto to high speed mode, we're assuming the previous speed is legacy. This
some times is not true.
For example in a system with both CONFIG_SERIAL_8250 and
CONFIG_SERIAL_8250_PNP set, in this case, the code (autoconfig) will be called
twice, one in serial8250_init/probe() and the other is from
serial_pnp_probe. When serial_pnp_probe calls the autoconfig for NS16550A,
it's already in high speed mode, change the divisor latch (quot << 3) in this
case will make the UART console garbled.
For any reason if the NS16550A was not work in high speed mode (e.g. we hold
NS16550A from going to high speed mode in autoconfig_16550a()), now we are
resume from suspend, we should also set the uartclk to the correct
value. Otherwise it is still the old 1843200 and that will bring issues.
The hardware rx filter flag triggered by FIF_PROMISC_IN_BSS is overly broad
and covers even frames with PHY errors. When this flag is enabled, this message
shows up frequently during scanning or hardware resets:
ath: Could not stop RX, we could be confusing the DMA engine when we start RX up
Since promiscuous mode is usually not particularly useful, yet enabled by
default by bridging (either used normally in 4-addr mode, or with hacks
for various virtualization software), we should sacrifice it for better
reliability during normal operation.
This patch leaves it enabled if there are active monitor mode interfaces, since
it's very useful for debugging.
Signed-off-by: Felix Fietkau <nbd@openwrt.org> Signed-off-by: John W. Linville <linville@tuxdriver.com> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
There are few places where we are checking for macversion and revsions
before RTC is powered ON. However we are reading the macversion and
revisions only after RTC is powered ON and so both macversion and
revisions are actully zero and this leads to incorrect srev checks
Incorrect srev checks can cause registers to be configured wrongly and can
cause unexpected behavior. Fixing this seems to address the ASPM issue that
we have observed. The laptop becomes very slow and hangs mostly with ASPM L1
enabled without this fix.
fix this by reading the macversion and revisisons even before we start
using them. There is no reason why should we delay reading this info
until RTC is powered on as this is just a register information.
Signed-off-by: Senthil Balasubramanian <senthilkumar@atheros.com> Signed-off-by: John W. Linville <linville@tuxdriver.com> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
We need to read and backup AR_WA register value permanently and reading
this after the chip is awakened results in this register being zeroed out.
This seems to fix the ASPM with L1 enabled issue that we have observed.
The laptop becomes very slow and hangs mostly with ASPM L1 enabled without
this fix.
Signed-off-by: Senthil Balasubramanian <senthilkumar@atheros.com> Signed-off-by: John W. Linville <linville@tuxdriver.com> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
[Digression: what is upowerd doing reading those power hungry files?]
Reported-by: Paul Menzel <paulepanter@users.sourceforge.net> Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Signed-off-by: Dave Airlie <airlied@redhat.com> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
I stumbled over this magic bit in the gen3 INSTPM:
Bit11 Interrupt-Based AGPBUSY# Enable:
‘0’ = Pending GMCH interrupts will not cause AGPBUSY# assertion.
‘1’ = Pending GMCH interrupts will cause AGPBUSY# assertion and hence
can cause the CPU to exit C3. There is no suppression of cacheable
writes.
Note that in either case in C3 the interrupts are not lost. They will be
forwarded to the ICH when the GMCH is out of C3.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Tested-by: Daniel Vetter <daniel.vetter@ffwll.ch> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
In tomoyo_check_open_permission() since 2.6.36, TOMOYO was by error
recalculating already calculated pathname when checking allow_rewrite
permission. As a result, memory will leak whenever a file is opened for writing
without O_APPEND flag. Also, performance will degrade because TOMOYO is
calculating pathname regardless of profile configuration.
This patch fixes the leak and performance degrade.
Intel Archiecture Software Developer's Manual section 7.1.3 specifies that a
core serializing instruction such as "cpuid" should be executed on _each_ core
before the new instruction is made visible.
Failure to do so can lead to unspecified behavior (Intel XMC erratas include
General Protection Fault in the list), so we should avoid this at all cost.
This problem can affect modified code executed by interrupt handlers after
interrupt are re-enabled at the end of stop_machine, because no core serializing
instruction is executed between the code modification and the moment interrupts
are reenabled.
Because stop_machine_text_poke performs the text modification from the first CPU
decrementing stop_machine_first, modified code executed in thread context is
also affected by this problem. To explain why, we have to split the CPUs in two
categories: the CPU that initiates the text modification (calls text_poke_smp)
and all the others. The scheduler, executed on all other CPUs after
stop_machine, issues an "iret" core serializing instruction, and therefore
handles core serialization for all these CPUs. However, the text modification
initiator can continue its execution on the same thread and access the modified
text without any scheduler call. Given that the CPU that initiates the code
modification is not guaranteed to be the one actually performing the code
modification, it falls into the XMC errata.
Q: Isn't this executed from an IPI handler, which will return with IRET (a
serializing instruction) anyway?
A: No, now stop_machine uses per-cpu workqueue, so that handler will be
executed from worker threads. There is no iret anymore.
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
LKML-Reference: <20110303160137.GB1590@Krystal> Reviewed-by: Masami Hiramatsu <masami.hiramatsu.pt@hitachi.com> Cc: Arjan van de Ven <arjan@infradead.org> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Steven Rostedt <rostedt@goodmis.org> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Andi Kleen <andi@firstfloor.org> Cc: Frederic Weisbecker <fweisbec@gmail.com> Signed-off-by: H. Peter Anvin <hpa@linux.intel.com> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
A userland read of more than PAGE_SIZE bytes from /dev/zero results in
(a) not all of the bytes returned being zero, and
(b) memory corruption due to zeroing of bytes beyond the user buffer.
This is caused by improper constraints on the assembly __clear_user function.
The constrints don't indicate to the compiler that the pointer argument is
modified. Since the function is inline, this results in double-incrementing
of the pointer when __clear_user() is invoked through a multi-page read() of
/dev/zero.
Signed-off-by: Steven J. Magnani <steve@digidescorp.com> Acked-by: Michal Simek <monstr@monstr.eu> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Commit 7f74f8f28a2bd9db9404f7d364e2097a0c42cc12
(x86 quirk: Fix polarity for IRQ0 pin2 override on SB800
systems) introduced a regression. It removed some SB600 specific
code to determine the revision ID without adapting a
corresponding revision ID check for SB600.
When processing a SIDR REQ, the ib_cm allocates a new cm_id. The
refcount of the cm_id is initialized to 1. However, cm_process_work
will decrement the refcount after invoking all callbacks. The result
is that the cm_id will end up with refcount set to 0 by the end of the
sidr req handler.
If a user tries to destroy the cm_id, the destruction will proceed,
under the incorrect assumption that no other threads are referencing
the cm_id. This can lead to a crash when the cm callback thread tries
to access the cm_id.
This problem was noticed as part of a larger investigation with kernel
crashes in the rdma_cm when running on a real time OS.
Signed-off-by: Sean Hefty <sean.hefty@intel.com> Acked-by: Doug Ledford <dledford@redhat.com> Signed-off-by: Roland Dreier <roland@purestorage.com> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
It turns out that while a maximum of 8 partitions may be what people
"should" have had, you can actually fit up to 18 entries(*) in a sector.
And some people clearly were taking advantage of that, like Michael
Cree, who had ten partitions on one of his OSF disks.
(*) The OSF partition data starts at byte offset 64 in the first sector,
and the array of 16-byte partition entries start at offset 148 in
the on-disk partition structure.
They were able to reproduce the crash multiple times with the
following details:
Crash seems to always happen on the:
mutex_unlock(&conn_id->handler_mutex);
as conn_id looks to have been freed during this code path.
An examination of the code shows that a race exists in the request
handlers. When a new connection request is received, the rdma_cm
allocates a new connection identifier. This identifier has a single
reference count on it. If a user calls rdma_destroy_id() from another
thread after receiving a callback, rdma_destroy_id will proceed to
destroy the id and free the associated memory. However, the request
handlers may still be in the process of running. When control returns
to the request handlers, they can attempt to access the newly created
identifiers.
Fix this by holding a reference on the newly created rdma_cm_id until
the request handler is through accessing it.
Signed-off-by: Sean Hefty <sean.hefty@intel.com> Acked-by: Doug Ledford <dledford@redhat.com> Signed-off-by: Roland Dreier <roland@purestorage.com> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
ata_eh_analyze_serror() suppresses hotplug notifications if LPM is
being used because LPM generates spurious hotplug events. It compared
whether link->lpm_policy was different from ATA_LPM_MAX_POWER to
determine whether LPM is enabled; however, this is incorrect as for
drivers which don't implement LPM, lpm_policy is always
ATA_LPM_UNKNOWN. This disabled hotplug detection for all drivers
which don't implement LPM.
Fix it by comparing whether lpm_policy is greater than
ATA_LPM_MAX_POWER.
Emit warning when "mem=nopentium" is specified on any arch other
than x86_32 (the only that arch supports it).
Signed-off-by: Kamal Mostafa <kamal@canonical.com> BugLink: http://bugs.launchpad.net/bugs/553464 Cc: Yinghai Lu <yinghai@kernel.org> Cc: Len Brown <len.brown@intel.com> Cc: Rafael J. Wysocki <rjw@sisk.pl>
LKML-Reference: <1296783486-23033-2-git-send-email-kamal@canonical.com> Signed-off-by: Ingo Molnar <mingo@elte.hu> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Avoid removing all of memory and panicing when "mem={invalid}"
is specified, e.g. mem=blahblah, mem=0, or mem=nopentium (on
platforms other than x86_32).
Signed-off-by: Kamal Mostafa <kamal@canonical.com> BugLink: http://bugs.launchpad.net/bugs/553464 Cc: Yinghai Lu <yinghai@kernel.org> Cc: Len Brown <len.brown@intel.com> Cc: Rafael J. Wysocki <rjw@sisk.pl>
LKML-Reference: <1296783486-23033-1-git-send-email-kamal@canonical.com> Signed-off-by: Ingo Molnar <mingo@elte.hu> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
When the fuction graph tracer starts, it needs to make a special
stack for each task to save the real return values of the tasks.
All running tasks have this stack created, as well as any new
tasks.
On CPU hot plug, the new idle task will allocate a stack as well
when init_idle() is called. The problem is that cpu hotplug does
not create a new idle_task. Instead it uses the idle task that
existed when the cpu went down.
ftrace_graph_init_task() will add a new ret_stack to the task
that is given to it. Because a clone will make the task
have a stack of its parent it does not check if the task's
ret_stack is already NULL or not. When the CPU hotplug code
starts a CPU up again, it will allocate a new stack even
though one already existed for it.
The solution is to treat the idle_task specially. In fact, the
function_graph code already does, just not at init_idle().
Instead of using the ftrace_graph_init_task() for the idle task,
which that function expects the task to be a clone, have a
separate ftrace_graph_init_idle_task(). Also, we will create a
per_cpu ret_stack that is used by the idle task. When we call
ftrace_graph_init_idle_task() it will check if the idle task's
ret_stack is NULL, if it is, then it will assign it the per_cpu
ret_stack.
Reported-by: Benjamin Herrenschmidt <benh@kernel.crashing.org> Suggested-by: Peter Zijlstra <a.p.zijlstra@chello.nl> Signed-off-by: Steven Rostedt <rostedt@goodmis.org> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
The new vfs locking scheme introduced in 2.6.38 breaks NFS sillyrename
because the latter relies on being able to determine the parent
directory of the dentry in the ->iput() callback in order to send the
appropriate unlink rpc call.
Looking at the code that cares about races with dput(), there doesn't
seem to be anything that specifically uses d_parent as a test for
whether or not there is a race:
- __d_lookup_rcu(), __d_lookup() all test for d_hashed() after d_parent
- shrink_dcache_for_umount() is safe since nothing else can rearrange
the dentries in that super block.
- have_submount(), select_parent() and d_genocide() can test for a
deletion if we set the DCACHE_DISCONNECTED flag when the dentry
is removed from the parent's d_subdirs list.
projects / karo-tx-linux.git / log