In tegra_i2c_fill_tx_fifo, once we have finished pushing all the bytes
to the I2C hardware controller, the interrupt might happen before we
have updated i2c_dev->msg_buf_remaining at the end of the function.
Then, in tegra_i2c_isr, we will call again tegra_i2c_fill_tx_fifo
triggering weird behaviour. This has been shown to happen under real
conditions.
Signed-off-by: Doug Anderson <dianders@chromium.org>
Tested-by: Vincent Palatin <vpalatin@chromium.org>
Acked-by: Rhyland Klein <rklein@nvidia.com>
Acked-by: Stephen Warren <swarren@nvidia.com>
Signed-off-by: Stephen Warren <swarren@nvidia.com>
Signed-off-by: Ben Dooks <ben-linux@fluff.org>
This patch was intended to be part of 7ca2d1a105a239e300b937e9c41a10a4bd08f569
"i2c: Tegra: Add DeviceTree support". However, an early version of that patch,
which was missing a chunk, was applied to next-i2c. This change is that
missing chunk.
Signed-off-by: John Bonesio <bones@secretlab.ca>
Signed-off-by: Stephen Warren <swarren@nvidia.com>
Signed-off-by: Ben Dooks <ben-linux@fluff.org>
Signed-off-by: Axel Lin <axel.lin@gmail.com>
Acked-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Signed-off-by: Ben Dooks <ben-linux@fluff.org>
The suspend/resume _noirq handlers were #ifdef'd out in the
!CONFIG_SUSPEND case, but were still assigned to the dev_pm_ops
struct. Fix by defining them to NULL in the !CONFIG_SUSPEND case.
Reported-by: Arnd Bergmann <arnd@arndb.de>
Acked-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Kevin Hilman <khilman@ti.com>
Rewrite the loop walking the id array during probe. The new code is
better adapted to a null-terminated array, and is also clearer and
more efficient than the original.
Signed-off-by: Jean Delvare <khali@linux-fr.org>
Cc: Axel Lin <axel.lin@gmail.com>
Cc: Guenter Roeck <guenter.roeck@ericsson.com>
Signed-off-by: Guenter Roeck <guenter.roeck@ericsson.com>
The chips supported by the max16065 driver should not be accessed using direct
i2ctools commands. Add warning to driver documentation to alert users.
Signed-off-by: Guenter Roeck <guenter.roeck@ericsson.com>
Acked-by: Jean Delvare <khali@linux-fr.org>
Current calculation is completely wrong. Add missing brackets to fix it.
Signed-off-by: Guenter Roeck <guenter.roeck@ericsson.com>
Acked-by: Jean Delvare <khali@linux-fr.org>
Cc: stable@kernel.org # 3.0+
This make sure we don't end up reusing the unlinked inode object.
The ideal way is to use inode i_generation. But i_generation is
not available in userspace always.
Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Some of the flags are OS/arch dependent we add a 9p
protocol value which maps to asm-generic/fcntl.h values in Linux
Based on the original patch from Venkateswararao Jujjuri <jvrao@linux.vnet.ibm.com>
Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
With msize equal to 512K (PAGE_SIZE * VIRTQUEUE_NUM), we hit multiple
crashes. This patch fix those.
Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Signed-off-by: Eric Van Hensbergen <ericvh@gmail.com>
We should only update attributes that we can change on stat2inode.
Also do file type initialization in v9fs_init_inode.
Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Signed-off-by: Eric Van Hensbergen <ericvh@gmail.com>
d_instantiate marks the dentry positive. So a parallel lookup and mkdir of
the directory can find dentry that doesn't have fid attached. This can result
in both the code path doing v9fs_fid_add which results in v9fs_dentry leak.
Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Signed-off-by: Eric Van Hensbergen <ericvh@gmail.com>
Sink type is always DP for DP bridges and EDID fetch on
DP bridges is always i2c over aux rather than plain i2c.
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
The conversion to per bus type registration functions means we don't need
to do module_get()s to hold the bus types in memory (their users will link
to them) so we removed all those calls. This left module_put() calls in
the cleanup paths which aren't needed and which cause unbalanced puts if
we ever try to unload anything.
Reported-by: Jonathan Cameron <jic23@cam.ac.uk>
Signed-off-by: Mark Brown <broonie@opensource.wolfsonmicro.com>
As it was decided not to export struct pt_regs to userspace, struct
sigcontext shouldn't be using it either. The pt_regs struct for OpenRISC
is kernel internal and the layout of the registers may change in the
future. The struct user_regs_struct is what is guaranteed to remain
stable, so struct sigcontext may use that instead.
This patch removes the usage of struct pt_regs in struct sigcontext and
makes according changes in signal.c to get the register layout right.
The usp field is removed from the sigcontext structure as this information
is already contained in the user_regs_struct.
Signed-off-by: Jonas Bonn <jonas@southpole.se>
Reviewed-by: Emilio Cota <cota@braap.org>
Prepared descriptors that are not submitted will not be freed. Add
prepared descriptor to a list to be able to release them upon
dmaengine_terminate_all().
Signed-off-by: Per Forlin <per.forlin@linaro.org>
Acked-by: Linus Walleij <linus.walleij@linaro.org>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
The client list may exist in two lists at the same time. This makes free
fail since the same desc is freed multiple times. Remove desc from
client list when adding it to the pending queue. Move free of client owned
descriptors from free_dma() to terminate_all().
Unable to handle kernel paging request at virtual address 00100104
pgd = dea8c000
[00100104] *pgd=1ea62831, *pte=00000000, *ppte=00000000
Internal error: Oops: 817 [#1] PREEMPT SMP
Modules linked in:
CPU: 0 Not tainted (3.1.0-rc3+ #58)
PC is at d40_free_chan_resources+0x64/0x330
Signed-off-by: Per Forlin <per.forlin@linaro.org>
Acked-by: Linus Walleij <linus.walleij@linaro.org>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
The link status code operates from a timer, and writes the index
register without first taking a lock. A well-placed interrupt
between writing the index register and reading the data register
could change the index register on us, which will return wrong data.
Add the necessary lock.
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
Add a missing array terminator to omap2430_usbhsotg_addrs[]. Without
this terminator, the omap_hwmod resource building code runs off the
end of the array, resulting in at least this error -- if not worse
behavior:
[ 0.578002] musb-omap2430: failed to claim resource 4
[ 0.583465] omap_device: musb-omap2430: build failed (-16)
[ 0.589294] Could not build omap_device for musb-omap2430 usb_otg_hs
This should have been part of commit
78183f3fdf ("omap_hwmod: use a null
structure record to terminate omap_hwmod_addr_space arrays") but was
evidently missed.
Signed-off-by: Paul Walmsley <paul@pwsan.com>
When ARCH_HAS_HOLES_MEMORYMODEL is selected, pfn_valid calls
memblock_is_memory to test validity of a pfn:
> memblock_is_memory(pfn << PAGE_SHIFT);
On LPAE systems this cuts off the top bits, as the shift occurs before
the value is promoted to a phys_addr_t.
This patch replaces the shift with a call to __pfn_to_phys (which casts
pfn to phys_addr_t before shifting), preventing the loss of significant
bits.
Signed-off-by: Mark Rutland <mark.rutland@arm.com>
Acked-by: Will Deacon <will.deacon@arm.com>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
* 'drm-fixes' of git://people.freedesktop.org/~airlied/linux:
drm/radeon/kms: make sure pci max read request size is valid on evergreen+ (v2)
drm/radeon/kms: set a default max_pixel_clock
* 'for-linus' of git://oss.sgi.com/xfs/xfs:
xfs: fix ->write_inode return values
xfs: fix xfs_mark_inode_dirty during umount
xfs: deprecate the nodelaylog mount option
The domain_flush_devices() function takes the domain->lock.
But this function is only called from update_domain() which
itself is already called unter the domain->lock. This causes
a deadlock situation when the dma-address-space of a domain
grows larger than 1GB.
Cc: stable@kernel.org
Signed-off-by: Joerg Roedel <joerg.roedel@amd.com>
The value is only set to true but never set back to false,
which causes to many completion-wait commands to be sent to
hardware. Fix it with this patch.
Cc: stable@kernel.org
Signed-off-by: Joerg Roedel <joerg.roedel@amd.com>
If the bios or OS sets the pci max read request size to 0 or an
invalid value (6,7), it can result in a hang or slowdown. Check
and set it to something sane if it's invalid.
Fixes:
https://bugzilla.kernel.org/show_bug.cgi?id=42162
v2: use pci reg defines from include/linux/pci_regs.h
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Cc: stable@kernel.org
Reviewed-by: Michel Dänzer <michel.daenzer@amd.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
We have hit a couple of customer bugs where they would like to
use those parameters to run an UP kernel - but both of those
options turn of important sources of interrupt information so
we end up not being able to boot. The correct way is to
pass in 'dom0_max_vcpus=1' on the Xen hypervisor line and
the kernel will patch itself to be a UP kernel.
Fixes bug: http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=637308
CC: stable@kernel.org
Acked-by: Ian Campbell <Ian.Campbell@eu.citrix.com>
Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
If vmalloc page_fault happens inside of interrupt handler with interrupts
disabled then on exit path from exception handler when there is no pending
interrupts, the following code (arch/x86/xen/xen-asm_32.S:112):
cmpw $0x0001, XEN_vcpu_info_pending(%eax)
sete XEN_vcpu_info_mask(%eax)
will enable interrupts even if they has been previously disabled according to
eflags from the bounce frame (arch/x86/xen/xen-asm_32.S:99)
testb $X86_EFLAGS_IF>>8, 8+1+ESP_OFFSET(%esp)
setz XEN_vcpu_info_mask(%eax)
Solution is in setting XEN_vcpu_info_mask only when it should be set
according to
cmpw $0x0001, XEN_vcpu_info_pending(%eax)
but not clearing it if there isn't any pending events.
Reproducer for bug is attached to RHBZ 707552
CC: stable@kernel.org
Signed-off-by: Igor Mammedov <imammedo@redhat.com>
Acked-by: Jeremy Fitzhardinge <jeremy@goop.org>
Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
Currently we always redirty an inode that was attempted to be written out
synchronously but has been cleaned by an AIL pushed internall, which is
rather bogus. Fix that by doing the i_update_core check early on and
return 0 for it. Also include async calls for it, as doing any work for
those is just as pointless. While we're at it also fix the sign for the
EIO return in case of a filesystem shutdown, and fix the completely
non-sensical locking around xfs_log_inode.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Signed-off-by: Alex Elder <aelder@sgi.com>
(cherry picked from commit 297db93bb74cf687510313eb235a7aec14d67e97)
Signed-off-by: Alex Elder <aelder@sgi.com>
Use the domain's maximum reservation to limit the amount of extra RAM
for the memory balloon. This reduces the size of the pages tables and
the amount of reserved low memory (which defaults to about 1/32 of the
total RAM).
On a system with 8 GiB of RAM with the domain limited to 1 GiB the
kernel reports:
Before:
Memory: 627792k/4472000k available
After:
Memory: 549740k/11132224k available
A increase of about 76 MiB (~1.5% of the unused 7 GiB). The reserved
low memory is also reduced from 253 MiB to 32 MiB. The total
additional usable RAM is 329 MiB.
For dom0, this requires at patch to Xen ('x86: use 'dom0_mem' to limit
the number of pages for dom0') (c/s 23790)
CC: stable@kernel.org
Signed-off-by: David Vrabel <david.vrabel@citrix.com>
Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
During umount we do not add a dirty inode to the lru and wait for it to
become clean first, but force writeback of data and metadata with
I_WILL_FREE set. Currently there is no way for XFS to detect that the
inode has been redirtied for metadata operations, as we skip the
mark_inode_dirty call during teardown. Fix this by setting i_update_core
nanually in that case, so that the inode gets flushed during inode reclaim.
Alternatively we could enable calling mark_inode_dirty for inodes in
I_WILL_FREE state, and let the VFS dirty tracking handle this. I decided
against this as we will get better I/O patterns from reclaim compared to
the synchronous writeout in write_inode_now, and always marking the inode
dirty in some way from xfs_mark_inode_dirty is a better safetly net in
either case.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Signed-off-by: Alex Elder <aelder@sgi.com>
(cherry picked from commit da6742a5a4cc844a9982fdd936ddb537c0747856)
Signed-off-by: Alex Elder <aelder@sgi.com>
This patch fixes the problem in sdhci-s3c host driver for Samsung Soc's.
During the card identification stage the mmc core driver enumerates for
the best bus width in combination with the highest available data rate.
It starts enumerating from the highest bus width (8) to lowest width (1).
In case of few MMC cards the 4-bit bus enumeration fails and tries
the 1-bit bus enumeration. When switched to 1-bit bus mode the host driver
has to clear the previous bus width setting and apply the new setting.
The current patch will clear the previous bus mode and apply the new
mode setting.
Signed-off-by: Girish K S <girish.shivananjappa@linaro.org>
Acked-by: Jaehoon Chung <jh80.chung@samsung.com>
Cc: <stable@kernel.org>
Signed-off-by: Chris Ball <cjb@laptop.org>
mmc_sd_init_uhs_card function sets the driver type, current limit
and bus speed mode on card as well as on host controller side.
Currently bus speed mode is set by sending CMD6 to card and
immediately setting the timing mode in host controller. But
then before initiating tuning sequence, it also tries to set
current limit by sending CMD6 to card which results in data
timeout errors in controller if bus speed mode is SDR50/SDR104 mode.
So basically bus speed mode should be set only after current limit
is set in the card and immediately after setting the bus speed mode,
tuning sequence should be initiated.
Signed-off-by: Subhash Jadavani <subhashj@codeaurora.org>
Reviewed-by: Arindam Nath <arindam.nath@amd.com>
Signed-off-by: Chris Ball <cjb@laptop.org>
This corrects a logic error that I introduced in
"mmc: sdhi: Add write16_hook"
Reported-by: Magnus Damm <magnus.damm@gmail.com>
Signed-off-by: Simon Horman <horms@verge.net.au>
Signed-off-by: Chris Ball <cjb@laptop.org>
The default multithread workqueue can cause the same work to be executed
concurrently on a different CPUs. This isn't really suitable for clock
gating as it might already gated the clock and gating it twice results both
host->clk_old and host->ios.clock to be set to 0.
To prevent this from happening we use system_nrt_wq instead.
Signed-off-by: Mika Westerberg <mika.westerberg@linux.intel.com>
Reviewed-by: Linus Walleij <linus.walleij@linaro.org>
Tested-by: Chris Ball <cjb@laptop.org>
Cc: <stable@kernel.org>
Signed-off-by: Chris Ball <cjb@laptop.org>
We have seen at least two different races when clock gating kicks in in a
middle of ios structure update.
First one happens when ios->clock is changed outside of aggressive clock
gating framework, for example via mmc_set_clock(). The race might happen
when we run following code:
mmc_set_ios():
...
if (ios->clock > 0)
mmc_set_ungated(host);
Now if gating kicks in right after the condition check we end up setting
host->clk_gated to false even though we have just gated the clock. Next
time a request is started we try to ungate and restore the clock in
mmc_host_clk_hold(). However since we have host->clk_gated set to false the
original clock is not restored.
This eventually will cause the host controller to hang since its clock is
disabled while we are trying to issue a request. For example on Intel
Medfield platform we see:
[ 13.818610] mmc2: Timeout waiting for hardware interrupt.
[ 13.818698] sdhci: =========== REGISTER DUMP (mmc2)===========
[ 13.818753] sdhci: Sys addr: 0x00000000 | Version: 0x00008901
[ 13.818804] sdhci: Blk size: 0x00000000 | Blk cnt: 0x00000000
[ 13.818853] sdhci: Argument: 0x00000000 | Trn mode: 0x00000000
[ 13.818903] sdhci: Present: 0x1fff0000 | Host ctl: 0x00000001
[ 13.818951] sdhci: Power: 0x0000000d | Blk gap: 0x00000000
[ 13.819000] sdhci: Wake-up: 0x00000000 | Clock: 0x00000000
[ 13.819049] sdhci: Timeout: 0x00000000 | Int stat: 0x00000000
[ 13.819098] sdhci: Int enab: 0x00ff00c3 | Sig enab: 0x00ff00c3
[ 13.819147] sdhci: AC12 err: 0x00000000 | Slot int: 0x00000000
[ 13.819196] sdhci: Caps: 0x6bee32b2 | Caps_1: 0x00000000
[ 13.819245] sdhci: Cmd: 0x00000000 | Max curr: 0x00000000
[ 13.819292] sdhci: Host ctl2: 0x00000000
[ 13.819331] sdhci: ADMA Err: 0x00000000 | ADMA Ptr: 0x00000000
[ 13.819377] sdhci: ===========================================
[ 13.919605] mmc2: Reset 0x2 never completed.
and it never recovers.
Second race might happen while running mmc_power_off():
static void mmc_power_off(struct mmc_host *host)
{
host->ios.clock = 0;
host->ios.vdd = 0;
[ clock gating kicks in here ]
/*
* Reset ocr mask to be the highest possible voltage supported for
* this mmc host. This value will be used at next power up.
*/
host->ocr = 1 << (fls(host->ocr_avail) - 1);
if (!mmc_host_is_spi(host)) {
host->ios.bus_mode = MMC_BUSMODE_OPENDRAIN;
host->ios.chip_select = MMC_CS_DONTCARE;
}
host->ios.power_mode = MMC_POWER_OFF;
host->ios.bus_width = MMC_BUS_WIDTH_1;
host->ios.timing = MMC_TIMING_LEGACY;
mmc_set_ios(host);
}
If the clock gating worker kicks in while we are only partially updated the
ios structure the host controller gets incomplete ios and might not work as
supposed. Again on Intel Medfield platform we get:
[ 4.185349] kernel BUG at drivers/mmc/host/sdhci.c:1155!
[ 4.185422] invalid opcode: 0000 [#1] PREEMPT SMP
[ 4.185509] Modules linked in:
[ 4.185565]
[ 4.185608] Pid: 4, comm: kworker/0:0 Not tainted 3.0.0+ #240 Intel Corporation Medfield/iCDKA
[ 4.185742] EIP: 0060:[<c136364e>] EFLAGS: 00010083 CPU: 0
[ 4.185827] EIP is at sdhci_set_power+0x3e/0xd0
[ 4.185891] EAX: f5ff98e0 EBX: f5ff98e0 ECX: 00000000 EDX: 00000001
[ 4.185970] ESI: f5ff977c EDI: f5ff9904 EBP: f644fe98 ESP: f644fe94
[ 4.186049] DS: 007b ES: 007b FS: 00d8 GS: 0000 SS: 0068
[ 4.186125] Process kworker/0:0 (pid: 4, ti=f644e000 task=f644c0e0 task.ti=f644e000)
[ 4.186219] Stack:
[ 4.186257] f5ff98e0 f644feb0 c1365173 00000282 f5ff9460 f5ff96e0 f5ff96e0 f644feec
[ 4.186418] c1355bd8 f644c0e0 c1499c3d f5ff96e0 f644fed4 00000006 f5ff96e0 00000286
[ 4.186579] f644fedc c107922b f644feec 00000286 f5ff9460 f5ff9700 f644ff10 c135839e
[ 4.186739] Call Trace:
[ 4.186802] [<c1365173>] sdhci_set_ios+0x1c3/0x340
[ 4.186883] [<c1355bd8>] mmc_gate_clock+0x68/0x120
[ 4.186963] [<c1499c3d>] ? _raw_spin_unlock_irqrestore+0x4d/0x60
[ 4.187052] [<c107922b>] ? trace_hardirqs_on+0xb/0x10
[ 4.187134] [<c135839e>] mmc_host_clk_gate_delayed+0xbe/0x130
[ 4.187219] [<c105ec09>] ? process_one_work+0xf9/0x5b0
[ 4.187300] [<c135841d>] mmc_host_clk_gate_work+0xd/0x10
[ 4.187379] [<c105ec82>] process_one_work+0x172/0x5b0
[ 4.187457] [<c105ec09>] ? process_one_work+0xf9/0x5b0
[ 4.187538] [<c1358410>] ? mmc_host_clk_gate_delayed+0x130/0x130
[ 4.187625] [<c105f3c8>] worker_thread+0x118/0x330
[ 4.187700] [<c1496cee>] ? preempt_schedule+0x2e/0x50
[ 4.187779] [<c105f2b0>] ? rescuer_thread+0x1f0/0x1f0
[ 4.187857] [<c1062cf4>] kthread+0x74/0x80
[ 4.187931] [<c1062c80>] ? __init_kthread_worker+0x60/0x60
[ 4.188015] [<c149acfa>] kernel_thread_helper+0x6/0xd
[ 4.188079] Code: 81 fa 00 00 04 00 0f 84 a7 00 00 00 7f 21 81 fa 80 00 00 00 0f 84 92 00 00 00 81 fa 00 00 0
[ 4.188780] EIP: [<c136364e>] sdhci_set_power+0x3e/0xd0 SS:ESP 0068:f644fe94
[ 4.188898] ---[ end trace a7b23eecc71777e4 ]---
This BUG() comes from the fact that ios.power_mode was still in previous
value (MMC_POWER_ON) and ios.vdd was set to zero.
We prevent these by inhibiting the clock gating while we update the ios
structure.
Both problems can be reproduced by simply running the device in a reboot
loop.
Signed-off-by: Mika Westerberg <mika.westerberg@linux.intel.com>
Reviewed-by: Linus Walleij <linus.walleij@linaro.org>
Tested-by: Chris Ball <cjb@laptop.org>
Cc: <stable@kernel.org>
Signed-off-by: Chris Ball <cjb@laptop.org>
As per suggestion by Linus Walleij:
> If you think the names of the functions are confusing then
> you may rename them, say like this:
>
> mmc_host_clk_ungate() -> mmc_host_clk_hold()
> mmc_host_clk_gate() -> mmc_host_clk_release()
>
> Which would make the usecases more clear
(This is CC'd to stable@ because the next two patches, which fix
observable races, depend on it.)
Signed-off-by: Mika Westerberg <mika.westerberg@linux.intel.com>
Reviewed-by: Linus Walleij <linus.walleij@linaro.org>
Cc: <stable@kernel.org>
Signed-off-by: Chris Ball <cjb@laptop.org>
* 'for-linus' of git://neil.brown.name/md:
md/raid5: fix a hang on device failure.
md: fix clearing of 'blocked' flag in the presence of bad blocks.
md/linear: avoid corrupting structure while waiting for rcu_free to complete.
md: use REQ_NOIDLE flag in md_super_write()
md: ensure changes to 'write-mostly' are reflected in metadata.
md: report failure if a 'set faulty' request doesn't.
* 'merge' of git://git.kernel.org/pub/scm/linux/kernel/git/benh/powerpc:
powerpc/p1023rds: Fix the error of bank-width of nor flash
powerpc/85xx: enable caam crypto driver by default
powerpc/85xx: enable the audio drivers in the defconfigs
The i_mutex lock and flush_completed_IO() added by commit 2581fdc810
in ext4_evict_inode() causes lockdep complaining about potential
deadlock in several places. In most/all of these LOCKDEP complaints
it looks like it's a false positive, since many of the potential
circular locking cases can't take place by the time the
ext4_evict_inode() is called; but since at the very least it may mask
real problems, we need to address this.
This change removes the flush_completed_IO() and i_mutex lock in
ext4_evict_inode(). Instead, we take a different approach to resolve
the software lockup that commit 2581fdc810 intends to fix. Rather
than having ext4-dio-unwritten thread wait for grabing the i_mutex
lock of an inode, we use mutex_trylock() instead, and simply requeue
the work item if we fail to grab the inode's i_mutex lock.
This should speed up work queue processing in general and also
prevents the following deadlock scenario: During page fault,
shrink_icache_memory is called that in turn evicts another inode B.
Inode B has some pending io_end work so it calls ext4_ioend_wait()
that waits for inode B's i_ioend_count to become zero. However, inode
B's ioend work was queued behind some of inode A's ioend work on the
same cpu's ext4-dio-unwritten workqueue. As the ext4-dio-unwritten
thread on that cpu is processing inode A's ioend work, it tries to
grab inode A's i_mutex lock. Since the i_mutex lock of inode A is
still hold before the page fault happened, we enter a deadlock.
Signed-off-by: Jiaying Zhang <jiayingz@google.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
We detected a serious issue with PERF_SAMPLE_READ and
timing information when events were being multiplexing.
Samples would have time_running > time_enabled. That
was easy to reproduce with a libpfm4 example (ran 3
times to cause multiplexing on Core 2):
$ syst_smpl -e uops_retired:freq=1 &
$ syst_smpl -e uops_retired:freq=1 &
$ syst_smpl -e uops_retired:freq=1 &
IIP:0x0000000040062d ... PERIOD:2355332948 ENA=40144625315 RUN=60014875184
syst_smpl: WARNING: time_running > time_enabled
63277537998 uops_retired:freq=1 , scaled
The bug was not present in kernel up to (and including) 3.0. It turns
out the bug was introduced by the following commit:
commit c479429591
events: Move lockless timer calculation into helper function
The parameters of the function got reversed yet the call sites
were not updated to reflect the change. That lead to time_running
and time_enabled being swapped. That had no effect when there was
no multiplexing because in that case time_running = time_enabled
but it would show up in any other scenario.
Signed-off-by: Stephane Eranian <eranian@google.com>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Link: http://lkml.kernel.org/r/20110829124112.GA4828@quad
Signed-off-by: Ingo Molnar <mingo@elte.hu>
