In fsl_rio_dbell_handler() the code currently simply acknowledges the QFI
queue full interrupt, but does nothing to resolve the queue full
condition. Instead, it jumps to the end of the isr. When a queue full
condition occurs, the isr is then re-entered immediately and continually,
forever.
The fix is to just fall through and read out current doorbell entries.
Signed-off-by: Thomas Taranowski <tom@baringforge.com> Cc: Alexandre Bounine <alexandre.bounine@idt.com> Cc: Kumar Gala <galak@kernel.crashing.org> Cc: Matt Porter <mporter@kernel.crashing.org> Cc: Li Yang <leoli@freescale.com> Cc: Thomas Moll <thomas.moll@sysgo.com> Cc: Micha Nelissen <micha@neli.hopto.org> Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org> Cc: Grant Likely <grant.likely@secretlab.ca> 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>
In 2.6.37 I was running into oopses with repeated module
loads & unloads. I tracked this down to:
fb1813f4 ext4: use dedicated slab caches for group_info structures
(this was in addition to the features advert unload problem)
The kstrdup & subsequent kfree of the cache name was causing
a double free. In slub, at least, if I read it right it allocates
& frees the name itself, slab seems to do something different...
so in slub I think we were leaking -our- cachep->name, and double
freeing the one allocated by slub.
After getting lost in slab/slub/slob a bit, I just looked at other
sized-caches that get allocated. jbd2, biovec, sgpool all do it
more or less the way jbd2 does. Below patch follows the jbd2
method of dynamically allocating a cache at mount time from
a list of static names.
(This might also possibly fix a race creating the caches with
parallel mounts running).
[Folded in a fix from Dan Carpenter which fixed an off-by-one error in
the original patch]
This fixes a corruption problem with the multi-block
writepages submittal change for ext4, from commit bd2d0210cf22f2bd0cef72eb97cf94fc7d31d8cc ("ext4: use bio
layer instead of buffer layer in mpage_da_submit_io").
(Note that this corruption is not present in 2.6.37 on
ext4, because the corruption was detected after the
feature was merged in 2.6.37-rc1, and so it was turned
off by adding a non-default mount option,
mblk_io_submit. With this commit, which hopefully
fixes the last of the bugs with this feature, we'll be
able to turn on this performance feature by default in
2.6.38, and remove the mblk_io_submit option.)
The ext4 code path to bundle multiple pages for
writeback in ext4_bio_write_page() had a bug: we should
be clearing buffer head dirty flags *before* we submit
the bio, not in the completion routine.
The patch below was tested on 2.6.37 under KVM with the
postgresql script which was submitted by Jon Nelson as
documented in commit 1449032be1.
Without the patch, I'd hit the corruption problem about
50-70% of the time. With the patch, I executed the
script > 100 times with no corruption seen.
I also fixed a bug to make sure ext4_end_bio() doesn't
dereference the bio after the bio_put() call.
Reported-by: Jon Nelson <jnelson@jamponi.net> Reported-by: Matthias Bayer <jackdachef@gmail.com> Signed-off-by: Curt Wohlgemuth <curtw@google.com> Signed-off-by: "Theodore Ts'o" <tytso@mit.edu> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Ext4 features interface was not properly unregistered which led to
problems while unloading/reloading ext4 module. This commit fixes that by
adding proper kobject unregistration code into ext4_exit_fs() as well as
fail-path of ext4_init_fs()
Commit 40389687 moved a call to ext4_forget() out of
ext4_free_branches and let ext4_free_blocks() handle calling
bforget(). But that change unfortunately did not replace the call to
ext4_forget() with brelse(), which was needed to drop the in-use count
of the indirect block's buffer head, which lead to a memory leak when
deleting files that used indirect blocks. Fix this.
When ext4_trim_fs() is called to trim a part of a single group, the
logic will wrongly set last block of the interval to 'len' instead
of 'first_block + len'. Thus a shorter interval is possibly trimmed.
Fix it.
CC: Lukas Czerner <lczerner@redhat.com> Signed-off-by: Jan Kara <jack@suse.cz> Signed-off-by: "Theodore Ts'o" <tytso@mit.edu> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
When wb_writeback() is called in WB_SYNC_ALL mode, work->nr_to_write is
usually set to LONG_MAX. The logic in wb_writeback() then calls
__writeback_inodes_sb() with nr_to_write == MAX_WRITEBACK_PAGES and we
easily end up with non-positive nr_to_write after the function returns, if
the inode has more than MAX_WRITEBACK_PAGES dirty pages at the moment.
When nr_to_write is <= 0 wb_writeback() decides we need another round of
writeback but this is wrong in some cases! For example when a single
large file is continuously dirtied, we would never finish syncing it
because each pass would be able to write MAX_WRITEBACK_PAGES and inode
dirty timestamp never gets updated (as inode is never completely clean).
Thus __writeback_inodes_sb() would write the redirtied inode again and
again.
Fix the issue by setting nr_to_write to LONG_MAX in WB_SYNC_ALL mode. We
do not need nr_to_write in WB_SYNC_ALL mode anyway since
write_cache_pages() does livelock avoidance using page tagging in
WB_SYNC_ALL mode.
This makes wb_writeback() call __writeback_inodes_sb() only once on
WB_SYNC_ALL. The latter function won't livelock because it works on
- a finite set of files by doing queue_io() once at the beginning
- a finite set of pages by PAGECACHE_TAG_TOWRITE page tagging
After this patch, program from http://lkml.org/lkml/2010/10/24/154 is no
longer able to stall sync forever.
[fengguang.wu@intel.com: fix locking comment] Signed-off-by: Jan Kara <jack@suse.cz> Signed-off-by: Wu Fengguang <fengguang.wu@intel.com> Cc: Johannes Weiner <hannes@cmpxchg.org> Cc: Dave Chinner <david@fromorbit.com> Cc: Christoph Hellwig <hch@lst.de> Cc: Jan Engelhardt <jengelh@medozas.de> Cc: Jens Axboe <axboe@kernel.dk> 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>
Background writeback is easily livelockable in a loop in wb_writeback() by
a process continuously re-dirtying pages (or continuously appending to a
file). This is in fact intended as the target of background writeback is
to write dirty pages it can find as long as we are over
dirty_background_threshold.
But the above behavior gets inconvenient at times because no other work
queued in the flusher thread's queue gets processed. In particular, since
e.g. sync(1) relies on flusher thread to do all the IO for it, sync(1)
can hang forever waiting for flusher thread to do the work.
Generally, when a flusher thread has some work queued, someone submitted
the work to achieve a goal more specific than what background writeback
does. Moreover by working on the specific work, we also reduce amount of
dirty pages which is exactly the target of background writeout. So it
makes sense to give specific work a priority over a generic page cleaning.
Thus we interrupt background writeback if there is some other work to do.
We return to the background writeback after completing all the queued
work.
This may delay the writeback of expired inodes for a while, however the
expired inodes will eventually be flushed to disk as long as the other
works won't livelock.
[fengguang.wu@intel.com: update comment] Signed-off-by: Jan Kara <jack@suse.cz> Signed-off-by: Wu Fengguang <fengguang.wu@intel.com> Cc: Johannes Weiner <hannes@cmpxchg.org> Cc: Dave Chinner <david@fromorbit.com> Cc: Christoph Hellwig <hch@lst.de> Cc: Jan Engelhardt <jengelh@medozas.de> Cc: Jens Axboe <axboe@kernel.dk> 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>
Check whether background writeback is needed after finishing each work.
When bdi flusher thread finishes doing some work check whether any kind of
background writeback needs to be done (either because
dirty_background_ratio is exceeded or because we need to start flushing
old inodes). If so, just do background write back.
This way, bdi_start_background_writeback() just needs to wake up the
flusher thread. It will do background writeback as soon as there is no
other work.
This is a preparatory patch for the next patch which stops background
writeback as soon as there is other work to do.
Signed-off-by: Jan Kara <jack@suse.cz> Signed-off-by: Wu Fengguang <fengguang.wu@intel.com> Cc: Johannes Weiner <hannes@cmpxchg.org> Cc: Dave Chinner <david@fromorbit.com> Cc: Christoph Hellwig <hch@lst.de> Cc: Jan Engelhardt <jengelh@medozas.de> Cc: Jens Axboe <axboe@kernel.dk> 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>
move_native_irq() masks and unmasks the interrupt line
unconditionally, but the interrupt line might be masked due to a
threaded oneshot handler in progress. Unmasking the line in that case
can lead to interrupt storms. Observed on PREEMPT_RT.
2.6.37 added an unmap_and_move_huge_page() for memory failure recovery,
but its anon_vma handling was still based around the 2.6.35 conventions.
Update it to use page_lock_anon_vma, get_anon_vma, page_unlock_anon_vma,
drop_anon_vma in the same way as we're now changing unmap_and_move().
I don't particularly like to propose this for stable when I've not seen
its problems in practice nor tested the solution: but it's clearly out of
synch at present.
Increased usage of page migration in mmotm reveals that the anon_vma
locking in unmap_and_move() has been deficient since 2.6.36 (or even
earlier). Review at the time of f18194275c39835cb84563500995e0d503a32d9a
("mm: fix hang on anon_vma->root->lock") missed the issue here: the
anon_vma to which we get a reference may already have been freed back to
its slab (it is in use when we check page_mapped, but that can change),
and so its anon_vma->root may be switched at any moment by reuse in
anon_vma_prepare.
Perhaps we could fix that with a get_anon_vma_unless_zero(), but let's
not: just rely on page_lock_anon_vma() to do all the hard thinking for us,
then we don't need any rcu read locking over here.
In removing the rcu_unlock label: since PageAnon is a bit in
page->mapping, it's impossible for a !page->mapping page to be anon; but
insert VM_BUG_ON in case the implementation ever changes.
migrate_pages() -> unmap_and_move() only calls rcu_read_lock() for
anonymous pages, as introduced by git commit 989f89c57e6361e7d16fbd9572b5da7d313b073d ("fix rcu_read_lock() in page
migraton"). The point of the RCU protection there is part of getting a
stable reference to anon_vma and is only held for anon pages as file pages
are locked which is sufficient protection against freeing.
However, while a file page's mapping is being migrated, the radix tree is
double checked to ensure it is the expected page. This uses
radix_tree_deref_slot() -> rcu_dereference() without the RCU lock held
triggering the following warning.
This patch introduces radix_tree_deref_slot_protected() which calls
rcu_dereference_protected(). Users of it must pass in the
mapping->tree_lock that is protecting this dereference. Holding the tree
lock protects against parallel updaters of the radix tree meaning that
rcu_dereference_protected is allowable.
[akpm@linux-foundation.org: remove unneeded casts] Signed-off-by: Mel Gorman <mel@csn.ul.ie> Cc: Minchan Kim <minchan.kim@gmail.com> Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com> Cc: Milton Miller <miltonm@bga.com> Cc: Nick Piggin <nickpiggin@yahoo.com.au> Cc: Wu Fengguang <fengguang.wu@intel.com> 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>
Its reason is the wrong way of accounting hd_struct->in_flight. When a bio is
merged into a request belongs to different partition by ELEVATOR_FRONT_MERGE.
The detailed root cause is as follows.
Assuming that there are two partition, sda1 and sda2.
1. A request for sda2 is in request_queue. Hence sda1's hd_struct->in_flight
is 0 and sda2's one is 1.
2. A bio belongs to sda1 is issued and is merged into the request mentioned on
step1 by ELEVATOR_BACK_MERGE. The first sector of the request is changed
from sda2 region to sda1 region. However the two partition's
hd_struct->in_flight are not changed.
The patch fixes the problem by caching the partition lookup
inside the request structure, hence making sure that the increment
and decrement will always happen on the same partition struct. This
also speeds up IO with accounting enabled, since it cuts down on
the number of lookups we have to do.
Also add a refcount to struct hd_struct to keep the partition in
memory as long as users exist. We use kref_test_and_get() to ensure
we don't add a reference to a partition which is going away.
Add kref_test_and_get() function, which atomically add a reference only if
refcount is not zero. This prevent to add a reference to an object that is
already being removed.
rtc-cmos was setting suspend/resume hooks at the device_driver level.
However, the platform bus code (drivers/base/platform.c) only looks for
resume hooks at the dev_pm_ops level, or within the platform_driver.
Switch rtc_cmos to use dev_pm_ops so that suspend/resume code is executed
again.
Paul said:
: The user visible symptom in our (XO laptop) case was that rtcwake would
: fail to wake the laptop. The RTC alarm would expire, but the wakeup
: wasn't unmasked.
:
: As for severity, the impact may have been reduced because if I recall
: correctly, the bug only affected platforms with CONFIG_PNP disabled.
Signed-off-by: Paul Fox <pgf@laptop.org> Signed-off-by: Daniel Drake <dsd@laptop.org> Acked-by: Rafael J. Wysocki <rjw@sisk.pl> 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>
Commit 34aacb2920 ("procfs: Use generic_file_llseek in /proc/kcore") broke
seeking on /proc/kcore. This changes it back to use default_llseek in
order to restore the original behavior.
The problem with generic_file_llseek is that it only allows seeks up to
inode->i_sb->s_maxbytes, which is 2GB-1 on procfs, where the memory file
offset values in the /proc/kcore PT_LOAD segments may exceed or start
beyond that offset value.
A similar revert was made for /proc/vmcore.
Signed-off-by: Dave Anderson <anderson@redhat.com> Acked-by: Frederic Weisbecker <fweisbec@gmail.com> 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>
Re-initializing the wait object in rdma_init()/rdma_fini() causes a
timing window which can lead to a deadlock during close. Once this
deadlock hits, all RDMA activity over the T4 device will be stuck.
There's no need to re-init the wait object, so remove it.
Signed-off-by: Steve Wise <swise@opengridcomputing.com> Signed-off-by: Roland Dreier <rolandd@cisco.com> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
On older gcc (3.3) dynamic debug fails to compile:
include/net/inet_connection_sock.h: In function `inet_csk_reset_xmit_timer':
include/net/inet_connection_sock.h:236: error: duplicate label declaration `do_printk'
include/net/inet_connection_sock.h:219: error: this is a previous declaration
include/net/inet_connection_sock.h:236: error: duplicate label declaration `out'
include/net/inet_connection_sock.h:219: error: this is a previous declaration
include/net/inet_connection_sock.h:236: error: duplicate label `do_printk'
include/net/inet_connection_sock.h:236: error: duplicate label `out'
Fix, by reverting the usage of JUMP_LABEL() in dynamic debug for now.
Reported-by: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp> Tested-by: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp> Signed-off-by: Jason Baron <jbaron@redhat.com> Signed-off-by: Steven Rostedt <rostedt@goodmis.org> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
There is no reason to be freeing the delegation cred in the rcu callback,
and doing so is resulting in a lockdep complaint that rpc_credcache_lock
is being called from both softirq and non-softirq contexts.
Commit c0204fd2b8fe047b18b67e07e1bf2a03691240cd (NFS: Clean up
nfs4_proc_create()) broke NFSv3 exclusive open by removing the code
that passes the O_EXCL flag down to nfs3_proc_create(). This patch
reverts that offending hunk from the original commit.
Reported-by: Nick Bowler <nbowler@elliptictech.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com> Tested-by: Nick Bowler <nbowler@elliptictech.com> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
vm_map_ram() is not available on NOMMU platforms, and causes trouble
on incoherrent architectures such as ARM when we access the page data
through both the direct and the virtual mapping.
The alternative is to use the direct mapping to access page data
for the case when we are not crossing a page boundary, but to copy
the data into a linear scratch buffer when we are accessing data
that spans page boundaries.
When we disable the WM8994 FLL code path sharing means that we end up
writing out a configuration. Currently this is the currently active
input and output frequency (which causes snd_soc_update_bits() to
suppress actual writes both immediately and in the common case where
we reenable the same configuration later) but we allow machine drivers
to pass through a source of zero. Since the register values written
are one less than the source constants this causes corruption of other
bitfields in the register.
Fix this by using the most recently configured FLL source when none is
provided.
Signed-off-by: Mark Brown <broonie@opensource.wolfsonmicro.com> Acked-by: Liam Girdwood <lrg@slimlogic.co.uk> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
The commit 53d7d69d8ffdfa60c5b66cc2e9ee0774aaaef5c0
ALSA: hdmi - support infoframe for DisplayPort
dropped the initialization of CA field accidentally.
This resulted in only two-channel LPCM mode on Nvidia machines.
If a timer interrupt was delayed too much, hrtimer_forward_now() will
forward the timer expiry more than once. When this happens, the
additional number of elapsed ALSA timer ticks must be passed to
snd_timer_interrupt() to prevent the ALSA timer from falling behind.
This mostly fixes MIDI slowdown problems on highly-loaded systems with
badly behaved interrupt handlers.
Signed-off-by: Clemens Ladisch <clemens@ladisch.de> Reported-and-tested-by: Arthur Marsh <arthur.marsh@internode.on.net> Signed-off-by: Takashi Iwai <tiwai@suse.de> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
The Conexant codec driver adds the jack arrays in init callback which
may be called also in each PM resume. This results in the addition of
new jack element at each time.
The fix is to check whether the requested jack is already present in
the array.
This patch fixes the non-compiling AC97C driver for AVR32 architecture by
include mach/hardware.h only for AT91 architecture. The AVR32 architecture does
not supply the hardware.h include file.
Fix playback/capture channels patch to change supported playback
channels of au8830 to 1,2,4 and capture channels to 1,2.
This prevent oops when oss emulation use SNDCTL_DSP_CHANNELS to
set 3 Channels
The dynamic PCM restriction based on ELD information may lead to the
problem in some cases, e.g. when the receiver is turned off. Then it
may send a TV HDMI default such as channels = 2. Since it's still
plugged, the driver doesn't know whether it's the right configuration
for future use. Now, when an app opens the device at this moment,
then turn on the receiver, the app still sends channels=2.
The right solution is to implement some kind of notification and
automatic re-open mechanism. But, this is a goal far ahead.
This patch provides a workaround for such a case by providing a new
module option static_hdmi_pcm for snd-hda-codec-hdmi module. When
this is set to true, the driver doesn't change PCM parameters per
ELD information. For users who need the static configuration like
the scenario above, set this to true.
The parameter can be changed dynamically via sysfs, too.
When multiple headphone pins are defined without line-out pins, the
driver takes them as primary outputs. But it forgot to set line_out_type
to HP by assuming there is some rest of HP pins. This results in some
mis-handling of these pins for Realtek codec parser. It takes as if
these are pure line-out jacks.
It seems that ix2505v driver ignores a i2c error in ix2505v_read_status_reg.
This looks like a typing error using (ret = 1) instead of correct (ret == 1).
gcc 4.5+ doesn't properly evaluate some inlined expressions.
A previous patch were proposed by Andrew Morton using noinline.
However, the entire inlined function is bogus, so let's just
remove it and be happy.
After a module loads you will have loaded the world roaming regulatory
domain or a custom regulatory domain. Further regulatory hints are
welcomed and should be respected unless the regulatory hint is coming
from a country IE as the IEEE spec allows for a country IE to be a subset
of what is allowed by the local regulatory agencies.
So disable all channels that do not fit a regulatory domain sent
from a unless the hint is from a country IE and the country IE had
no information about the band we are currently processing.
This fixes a few regulatory issues, for example for drivers that depend
on CRDA and had no 5 GHz freqencies allowed were not properly disabling
5 GHz at all, furthermore it also allows users to restrict devices
further as was intended.
If you recieve a country IE upon association we will also disable the
channels that are not allowed if the country IE had at least one
channel on the respective band we are procesing.
This was the original intention behind this design but it was
completely overlooked...
Cc: David Quan <david.quan@atheros.com> Cc: Jouni Malinen <jouni.malinen@atheros.com>
cc: Easwar Krishnan <easwar.krishnan@atheros.com> Signed-off-by: Luis R. Rodriguez <lrodriguez@atheros.com> Signed-off-by: John W. Linville <linville@tuxdriver.com> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
We should be enabling country IE hints for WIPHY_FLAG_STRICT_REGULATORY
even if we haven't yet recieved regulatory domain hint for the driver
if it needed one. Without this Country IEs are not passed on to drivers
that have set WIPHY_FLAG_STRICT_REGULATORY, today this is just all
Atheros chipset drivers: ath5k, ath9k, ar9170, carl9170.
This was part of the original design, however it was completely
overlooked...
Cc: Easwar Krishnan <easwar.krishnan@atheros.com> Signed-off-by: Luis R. Rodriguez <lrodriguez@atheros.com> Signed-off-by: John W. Linville <linville@tuxdriver.com> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Cc: Easwar Krishnan <easwar.krishnan@atheros.com> Signed-off-by: Luis R. Rodriguez <lrodriguez@atheros.com> Signed-off-by: John W. Linville <linville@tuxdriver.com> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Dean noticed that 'err' wasn't being set when the "goto err_dma"
statement is executed in the following hunk from the commit. It's value
will be zero as a result of a successful call to e1000_init_hw_struct().
This patch changes the error condition to be correctly propagated.
Signed-off-by: Dean Nelson <dnelson@redhat.com> Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com> Tested-by: Emil Tantilov <emil.s.tantilov@intel.com> Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
There was a configuration page timing out during the initial port
enable at driver load time. The port enable would fail, and this would
result in the driver unloading itself, meanwhile the driver was accessing
freed memory in another context resulting in the panic. The fix is to
prevent access to freed memory once the driver had issued the diag reset
which woke up the sleeping port enable process. The routine
_base_reset_handler was reorganized so the last sleeping process woken up was
the port_enable.
The ioc->hba_queue_depth is not properly resized when the controller
firmware reports that it supports more outstanding IO than what can be fit
inside the reply descriptor pool depth. This is reproduced by setting the
controller global credits larger than 30,000. The bug results in an
incorrect sizing of the queues. The fix is to resize the queue_size by
dividing queue_diff by two.
False timeout after hard resets, there were two issues which leads
to timeout.
(1) Panic because of invalid memory access in the broadcast asyn
event processing routine due to a race between accessing the scsi command
pointer from broadcast asyn event processing thread and completing
the same scsi command from the interrupt context.
(2) Broadcast asyn event notifcations are not handled due to events
ignored while the broadcast asyn event is activity being processed
from the event process kernel thread.
In addition, changed the ABRT_TASK_SET to ABORT_TASK in the
broadcast async event processing routine. This is less disruptive to other
request that generate Broadcast Asyn Primitives besides target
reset. e.g clear reservations, microcode download,and mode select.
The "internal device reset complete" event is not supported
for older firmware prior to MPI Rev K We added
a check in the driver so the "internal device reset" event is
ignored for older firmware. When ignored, the tm_busy flag doesn't
get set nor cleared. Without this fix, IO queues would be froozen
indefinetly after the "internal device reset" event, as the "complete" event
never sent to clear the flag.
When zoning end devices, the driver is not sending device
removal handshake alogrithm to firmware. This results in controller
firmware not sending sas topology add events the next time the device is
added. The fix is the driver should be doing the device removal handshake
even though the PHYSTATUS_VACANT bit is set in the PhyStatus of the
event data. The current design is avoiding the handshake when the
VACANT bit is set in the phy status.
Issue:
IR shutdown(sending) and IR shutdown(complete) messages not
listed in /var/log/messages when driver is removed.
The driver needs to issue a MPI2_RAID_ACTION_SYSTEM_SHUTDOWN_INITIATED
request when the driver is unloaded so the IR metadata journal is updated.
If this request is not sent, then the volume would need a "check
consistency" issued on the next bootup if the volume was roamed from one
initiator to another. The current driver supports this feature only when the
system is rebooted, however this also need to be supported if the driver is
unloaded
Fix:
To fix this issue, the driver is going
to need to call the _scsih_ir_shutdown prior to reporting
the volumes missing from the OS, hence the device handles
are still present.
libsas makes use of scsi_schedule_eh() but forgets to clear the
host_eh_scheduled flag in its error handling routine. Because of this,
the error handler thread never gets to sleep; it's constantly awake and
trying to run the error routine leading to console spew and inability to
run anything else (at least on a UP system). The fix is to clear the
flag as we splice the work queue.
Signed-off-by: James Bottomley <James.Bottomley@suse.de> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Our current handling of medium error assumes that data is returned up
to the bad sector. This assumption holds good for all disk devices,
all DIF arrays and most ordinary arrays. However, an LSI array engine
was recently discovered which reports a medium error without returning
any data. This means that when we report good data up to the medium
error, we've reported junk originally in the buffer as good. Worse,
if the read consists of requested data plus a readahead, and the error
occurs in readahead, we'll just strip off the readahead and report
junk up to userspace as good data with no error.
The fix for this is to have the error position computation take into
account the amount of data returned by the driver using the scsi
residual data. Unfortunately, not every driver fills in this data,
but for those who don't, it's set to zero, which means we'll think a
full set of data was transferred and the behaviour will be identical
to the prior behaviour of the code (believe the buffer up to the error
sector). All modern drivers seem to set the residual, so that should
fix up the LSI failure/corruption case.
Reported-by: Douglas Gilbert <dgilbert@interlog.com> Signed-off-by: James Bottomley <James.Bottomley@suse.de> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Fix the check for ap interupts to look for facility bits 2 and 65.
Make sure that we only register interrupts for aps, if the machine
has ap interrupt support.
This patch is relevant only for the 2.6.37 stable series.
Signed-off-by: Felix Beck <felix.beck@de.ibm.com> Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
If QIOASSIST is enabled for a qdio device the SIGA instruction requires
a modified function code. This function code modifier was missing for
SIGA-R and SIGA-S which can lead to a kernel panic caused by an
operand exception.
Signed-off-by: Jan Glauber <jang@linux.vnet.ibm.com> Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
The bit 6 & 7 of AR_WA (0x4004) should be enabled only
for the chips that are supporting L0s functionality
while resuming back from S3/S4.
Enabling these bits for AR9280 is causing system hang
within a few S3/S4-resume cycles.
Cc: Jack Lee <jlee@atheros.com> Signed-off-by: Rajkumar Manoharan <rmanoharan@atheros.com> Signed-off-by: John W. Linville <linville@tuxdriver.com> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
AR9003's PAPRD was enabled prematurely, and is causing some
large discrepancies on throughput and network connectivity.
For example downlink (RX) throughput against an AR9280 AP
can vary widlely from 43-73 Mbit/s while disabling this
gets AR9382 (2x2) up to around 93 Mbit/s in a 2.4 GHz HT20 setup.
Cc: Paul Shaw <paul.shaw@atheros.com> Signed-off-by: Luis R. Rodriguez <lrodriguez@atheros.com> Signed-off-by: John W. Linville <linville@tuxdriver.com> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
The power detector adc offset calibration has to be done
on 4 minutes interval (longcal * pa_skip_count). But the commit
"ath9k_hw: fix a noise floor calibration related race condition"
makes the PA calibration executed more frequently beased on
nfcal_pending value. Running PAOffset calibration lesser than
longcal interval doesn't help anything and the worse part is that
it causes NF load timeouts and RX deaf conditions.
In a very noisy environment, where the distance b/w AP & station
is ~10 meter and running a downlink udp traffic with frequent
background scan causes "Timeout while waiting for nf to load:
AR_PHY_AGC_CONTROL=0x40d1a" and moves the chip into deaf state.
This issue was originaly reported in Android platform where
the network-manager application does bgscan more frequently
on AR9271 chips. (AR9285 family usb device).
Signed-off-by: Vasanthakumar Thiagarajan <vasanth@atheros.com> Signed-off-by: Rajkumar Manoharan <rmanoharan@atheros.com> Signed-off-by: John W. Linville <linville@tuxdriver.com> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
ath9k supports its own set of virtual wiphys, and it uses
the mac80211 idle notifications to know when a device needs
to be idle or not. We recently changed ath9k to force idle
on driver stop() and on resume but forgot to take into account
ath9k's own virtual wiphy idle states. These are used internally
by ath9k to check if the device's radio should be powered down
on each idle call. Without this change its possible that the
device could have been forced off but the virtual wiphy idle
was left on.
Cc: Paul Stewart <pstew@google.com> Cc: Amod Bodas <amod.bodas@atheros.com> Signed-off-by: Luis R. Rodriguez <lrodriguez@atheros.com> Signed-off-by: John W. Linville <linville@tuxdriver.com> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
mac80211 will notify drivers when to go idle and ath9k
assumed that it would get further notifications for idle
states after a device stop() config call but as per agreed
semantics the idle state of the radio is left up to driver
after mac80211 issues the stop() callback. The driver is
resposnbile for ensuring the device remains idle after
that even between suspend / resume calls.
This fixes suspend/resume when you issue suspend and resume
twice on ath9k when ath9k_stop() was already called. We need
to put the radio to full sleep in order for resume to work
correctly.
What might seem fishy is we are turning the radio off
after resume. The reason why we do this is because we know
we should not have anything enabled after a mac80211 tells
us to stop(), if we resume and never get a start() we won't
get another stop() by mac80211 so to be safe always bring
the 802.11 device with the radio disabled after resume,
this ensures that if we suspend we already have the radio
disabled and only a start() will ever trigger it on.
Cc: Paul Stewart <pstew@google.com> Cc: Amod Bodas <amod.bodas@atheros.com> Signed-off-by: Luis R. Rodriguez <lrodriguez@atheros.com> Signed-off-by: John W. Linville <linville@tuxdriver.com> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Re-enable the interrupt when it occurs to see all transitions.
Signed-off-by: Don Fry <donald.h.fry@intel.com> Signed-off-by: Wey-Yi Guy <wey-yi.w.guy@intel.com> Signed-off-by: John W. Linville <linville@tuxdriver.com> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Since commit 6cd0b1cb872b3bf9fc5de4536404206ab74bafdd "iwlagn: fix
hw-rfkill while the interface is down", we enable interrupts when
device is not ready to receive them. However hardware, when it is in
some inconsistent state, can generate other than rfkill interrupts
and crash the system. I can reproduce crash with "kernel BUG at
drivers/net/wireless/iwlwifi/iwl-agn.c:1010!" message, when forcing
firmware restarts.
To fix only enable rfkill interrupt when down device and after probe.
I checked patch on laptop with 5100 device, rfkill change is still
passed to user space when device is down.
Signed-off-by: Stanislaw Gruszka <sgruszka@redhat.com> Acked-by: Wey-Yi Guy <wey-yi.w.guy@intel.com> Signed-off-by: John W. Linville <linville@tuxdriver.com> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
The change 'mac80211: Fix BUG in pskb_expand_head when transmitting shared skbs'
added a check for copying the skb if it's shared, however the tx info variable
still points at the cb of the old skb
Signed-off-by: Felix Fietkau <nbd@openwrt.org> Acked-by: Helmut Schaa <helmut.schaa@googlemail.com> Signed-off-by: John W. Linville <linville@tuxdriver.com> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Some drivers (e.g. ath9k) do not always disable beacons when they're
supposed to. When an interface is changed using the change_interface op,
the mode specific sdata part is in an undefined state and trying to
get a beacon at this point can produce weird crashes.
To fix this, add a check for ieee80211_sdata_running before using
anything from the sdata.
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>
When the buffer size is set to zero in the block ack parameter set
field, we should use the maximum supported number of subframes. The
existing code was bogus and was doing some unnecessary calculations
that lead to wrong values.
Thanks Johannes for helping me figure this one out.
Cc: Johannes Berg <johannes@sipsolutions.net> Signed-off-by: Luciano Coelho <coelho@ti.com> Reviewed-by: Johannes Berg <johannes@sipsolutions.net> Signed-off-by: John W. Linville <linville@tuxdriver.com>
removed the synchronization against RCU and thus
opened a race window where we can use a key for
TX while it is already freed. Put a synchronisation
into the right place to close that window.
Reported-by: Jussi Kivilinna <jussi.kivilinna@mbnet.fi> Signed-off-by: Johannes Berg <johannes.berg@intel.com> Signed-off-by: John W. Linville <linville@tuxdriver.com>
Under memory pressure, the mac80211 mesh code
may helpfully print a message that it failed
to clone a mesh frame and then will proceed
to crash trying to use it anyway. Fix that.
Avoid the reference whenever the frame copy is unsuccessful
regardless of the debug message being suppressed or printed.
Signed-off-by: Milton Miller <miltonm@bga.com> Signed-off-by: John W. Linville <linville@tuxdriver.com> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
PCI PM suspend callback took down the interface and resume brought
it back up. In the mac80211 context this is done in subsequent calls.
Rework implementation so that suspend only stores config, and sets
PCI power state. The resume return to full power state (D0), restores
the config, and brings hardware back up. Full bringup is done by
subsequent mac80211 calls.
The hv_netvsc gets RNDIS_STATUS_MEDIA_CONNECT event after the VM
is live migrated. Adding call to netif_notify_peers() for this event
to send GARP (Gratuitous ARP) to notify network peers. Otherwise,
the VM's network connection may stop after a live migration.
This patch should also be applied to stable kernel 2.6.32 and later.
Removed the call to netif_stop_queue() in netvsc_probe() as
the queue is not initialized at that point and further call
to it after queue initialization is really not necessary.
This change was prompted after an upstream change went into
2.6.37 (netif_tx_stop_queue) that now checks if netif_stop_queue
is called before register with netdev is done.
This will eliminate the warning message to the log when hv_netvsc
driver starts up.
The ni_labpc driver module only requests a shared IRQ for PCI devices,
requesting a non-shared IRQ for non-PCI devices.
As this module is also used by the ni_labpc_cs module for certain
National Instruments PCMCIA cards, it also needs to request a shared IRQ
for PCMCIA devices, otherwise you get a IRQ mismatch with the CardBus
controller.
Signed-off-by: Ian Abbott <abbotti@mev.co.uk> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
This patch adds several new device ids to the r8712u staging driver.
The new ids were retrieved from latest vendor driver (v2.6.6.0.20101111)
downloadable from www.realtek.com.tw
Signed-off-by: Axel Koellhofer <AxelKoellhofer@web.de> Signed-off-by: Larry Finger <Larry.Finger@lwfinger.net> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
In zram_read() and zram_write() we were not incrementing the
index number and thus were reading/writing values from/to
incorrect sectors on zram disk, resulting in data corruption.
projects / karo-tx-linux.git / log