| Age | Commit message (Collapse) | Author | Files | Lines |
|---|
|
Exclude active GPU pages from the purview of the background shrinker
(kswapd), as these cause uncontrollable GPU stalls. Given that the
shrinker is rerun until the freelists are satisfied, we should have
opportunity in subsequent passes to recover the pages once idle. If the
machine does run out of memory entirely, we have the forced idling in the
oom-notifier as a means of releasing all the pages we can before an oom
is prematurely executed.
Note that this relies upon an up-front retire_requests to keep the
inactive list in shape, which was added in a previous patch, mostly as
execlist ctx pinning band-aids.
Signed-off-by: Chris Wilson <[email protected]>
Reviewed-by: Damien Lespiau <[email protected]>
[danvet: Add note about retire_requests.]
Signed-off-by: Daniel Vetter <[email protected]>
|
|
With UMS gone, we no longer use it during suspend. And with the last
user removed from the shrinker, we can remove the dead code.
Signed-off-by: Chris Wilson <[email protected]>
Signed-off-by: Daniel Vetter <[email protected]>
|
|
We can forgo an evict-everything here as the shrinker operation itself
will unbind any vma as required. If we explicitly idle the GPU through a
switch to the default context, we not only create a request in an
illegal context (e.g. whilst shrinking during execbuf with a request
already allocated), but switching to the default context will not free
up the memory backing the active contexts - unless in the unlikely
situation that context had already been closed (and just kept arrive by
being the current context). The saving is near zero and the danger real.
To compensate for the loss of the forced retire, add a couple of
retire-requests to i915_gem_shirnk() - this should help free up any
transitive cache from the requests.
Note that the second retire_requests is for the benefit of the
hand-rolled execlist ctx active tracking: We need to manually kick
requests to get those unpinned again. Once that's fixed we can try to
remove this again.
Signed-off-by: Chris Wilson <[email protected]>
[danvet: Add summary of why we need a pile of retire_requests.]
Signed-off-by: Daniel Vetter <[email protected]>
|
|
Often it is very useful to know why we suddenly purge vast tracts of
memory and surprisingly up until now we didn't even have a tracepoint
for when we shrink our memory.
Note that there are slab_start/end tracepoints already, but those
don't cover the internal recursion when we directly call into our
shrinker code. Hence a separate tracepoint seems justified. Also note
that we don't really need a separate tracepoint for the actual amount
of pages freed since we already have an unbind tracpoint for that.
Signed-off-by: Chris Wilson <[email protected]>
[danvet: Add a note that there's also slab_start/end and why they're
insufficient.]
Signed-off-by: Daniel Vetter <[email protected]>
|
|
Add the item of i915_component.h in DocBook and add the DOC for
i915_component.h. Explain the struct i915_audio_component_ops and
struct i915_audio_component_audio_ops usage.
Signed-off-by: Libin Yang <[email protected]>
Signed-off-by: Daniel Vetter <[email protected]>
|
|
Add the kerneldoc for i915_audio_component in i915_component.h
Signed-off-by: Libin Yang <[email protected]>
Signed-off-by: Daniel Vetter <[email protected]>
|
|
drm-intel-next-queued
Pull in the i915/hda changes for N/CTS setting so I can apply the
follow-up documentation work for drm/i915.
Some conflicts because ofc we had to rework i915 while that N/CTS work
was going on. But not more than adjacent changes really.
Signed-off-by: Daniel Vetter <[email protected]>
|
|
As the shrinker_control now passes us unsigned long targets, update our
shrinker functions to match.
Signed-off-by: Chris Wilson <[email protected]>
Signed-off-by: Daniel Vetter <[email protected]>
|
|
I've botched this, so let's fix it.
Botched in
commit eb0b44adc08c0be01a027eb009e9cdadc31e65a2
Author: Daniel Vetter <[email protected]>
Date: Wed Mar 18 14:47:59 2015 +0100
drm/i915: kerneldoc for i915_gem_shrinker.c
v2: Be a good citizen^Wmaintainer and add the proper commit citation.
Noticed by Jani.
Reviewed-by: Jani Nikula <[email protected]>
Signed-off-by: Daniel Vetter <[email protected]>
|
|
The new amdgpu driver passes a user space pointer in a 64-bit structure
member, which is the correct way to do it, but it attempts to
directly cast it to a __user pointer in the kernel, which causes
a warning in three places:
drm/amd/amdgpu/amdgpu_cs.c: In function 'amdgpu_cs_parser_init':
drm/amd/amdgpu/amdgpu_cs.c:180:21: warning: cast to pointer from integer of different size [-Wint-to-pointer-cast]
chunk_array_user = (uint64_t __user *)(cs->in.chunks);
This changes all three to add an intermediate cast to 'unsigned long'
as other drivers do. This avoids the warning and works correctly on
both 32-bit and 64-bit architectures.
Fixes: e60b344f6c0eff ("drm/amdgpu: optimize amdgpu_parser_init")
Reviewed-by: Christian König <[email protected]>
Signed-off-by: Arnd Bergmann <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>
|
|
perf_regs.c does not get built on Powerpc as CONFIG_PERF_REGS is false.
So the weak definition for 'sample_regs_masks' doesn't get picked up.
Adding perf_regs.o to util/Build unconditionally, exposes a redefinition
error for 'perf_reg_value()' function (due to the static inline version
in util/perf_regs.h). So use #ifdef HAVE_PERF_REGS_SUPPORT' around that
function.
Signed-off-by: Sukadev Bhattiprolu <[email protected]>
Acked-by: Jiri Olsa <[email protected]>
Cc: Naveen N. Rao <[email protected]>
Cc: Stephane Eranian <[email protected]>
Cc: [email protected]
Link: http://lkml.kernel.org/r/[email protected]
Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>
|
|
git://git.kernel.org/pub/scm/linux/kernel/git/broonie/regmap
Pull regmap fixes from Mark Brown:
"A couple of fixes for the debugfs information on the register map,
fixing issues with very small reads potentially causing underflows and
wraparounds"
* tag 'regmap-fix-v4.3-rc4' of git://git.kernel.org/pub/scm/linux/kernel/git/broonie/regmap:
regmap: debugfs: Don't bother actually printing when calculating max length
regmap: debugfs: Ensure we don't underflow when printing access masks
|
|
git://git.kernel.org/pub/scm/linux/kernel/git/broonie/spi
Pull spi fixes from Mark Brown:
"A couple of very minor fixes, one for error handling in the Davinci
driver probe function and another making the Renesas sh-msiof DT
binding documentation correspond to what's actually implemented"
* tag 'spi-fix-v4.3-rc4' of git://git.kernel.org/pub/scm/linux/kernel/git/broonie/spi:
spi: sh-msiof: Match renesas,rx-fifo-size in DT bindings doc with driver
spi: davinci: fix handling platform_get_irq result
|
|
git://git.kernel.org/pub/scm/linux/kernel/git/broonie/regulator
Pull regulator fixes from Mark Brown:
"Two fixes here, one device specific fix for axp20x and a core fix for
cases where one regulator is supplying another which broke probe
deferral, substituting in a dummy regulator too aggressively"
* tag 'regulator-fix-v4.3-rc4' of git://git.kernel.org/pub/scm/linux/kernel/git/broonie/regulator:
regulator: core: Handle probe deferral from DT when resolving supplies
regulator: axp20x: Fix enable bit indexes for DCDC4 and DCDC5
|
|
If of_parse_display_timing() fails we are printing an error message and
jumping to the error path but we missed freeing "dt".
Signed-off-by: Sudip Mukherjee <[email protected]>
Signed-off-by: Tomi Valkeinen <[email protected]>
|
|
spi-linus
|
|
git://git.kernel.org/pub/scm/linux/kernel/git/cmetcalf/linux-tile
Pull strscpy fixes from Chris Metcalf :
"This patch series fixes up a couple of architecture issues where
strscpy wasn't configured correctly (missing on h8300, duplicating
local and asm-generic copies on powerpc and tile).
It also adds a use of zero_bytemask() to the final store for strscpy
to avoid writing uninitialized data to the destination. However, to
make this work we had to add support for zero_bytemask() to the two
architectures that didn't have it (alpha and tile), because they were
providing their own local copies, but didn't provide the
zero_bytemask() that was previously only required when building with
CONFIG_DCACHE_WORD_ACCESS"
[ Side note: there is still no actual users of strscpy except for the
one preexisting use in arch/tile that predates the generic version.
So this is all about fixing the infrastructure so that we eventually
can start using it. - Linus ]
* 'strscpy' of git://git.kernel.org/pub/scm/linux/kernel/git/cmetcalf/linux-tile:
strscpy: zero any trailing garbage bytes in the destination
word-at-a-time.h: support zero_bytemask() on alpha and tile
word-at-a-time.h: fix some Kbuild files
|
|
Pull MTD fixes from Brian Norris:
"A few MTD fixes:
- mxc_nand: a "refactoring only" change in 4.3-rc1 had some bad
pointer (array) arithmetic. Fix that
- sunxi_nand:
- Fix an old list manipulation / memory management bug in the device
release() code path
- Correct a few mistakes in OOB write support"
* tag 'for-linus-20151006' of git://git.infradead.org/linux-mtd:
mxc_nand: fix copy_spare
mtd: nand: sunxi: fix sunxi_nand_chips_cleanup()
mtd: nand: sunxi: fix OOB handling in ->write_xxx() functions
|
|
Pull NFS client bugfixes from Trond Myklebust:
"Highlights include:
Bugfixes:
- Fix a use-after-free bug in the RPC/RDMA client
- Fix a write performance regression
- Fix up page writeback accounting
- Don't try to reclaim unused state owners
- Fix a NFSv4 nograce recovery hang
- reset states to use open_stateid when returning delegation
voluntarily
- Fix a tracepoint NULL-pointer dereference"
* tag 'nfs-for-4.3-3' of git://git.linux-nfs.org/projects/trondmy/linux-nfs:
NFS: Fix a tracepoint NULL-pointer dereference
nfs4: reset states to use open_stateid when returning delegation voluntarily
NFSv4: Fix a nograce recovery hang
NFSv4.1: nfs4_opendata_check_deleg needs to handle NFS4_OPEN_CLAIM_DELEG_CUR_FH
NFSv4: Don't try to reclaim unused state owners
NFS: Fix a write performance regression
NFS: Fix up page writeback accounting
xprtrdma: disconnect and flush cqs before freeing buffers
|
|
This reverts commit 998ef75ddb5709bbea0bf1506cd2717348a3c647.
The commit itself does not appear to be buggy per se, but it is exposing
a bug in ext4 (and Ted thinks ext3 too, but we solved that by getting
rid of it). It's too late in the release cycle to really worry about
this, even if Dave Hansen has a patch that may actually fix the
underlying ext4 problem. We can (and should) revisit this for the next
release.
The problem is that moving the prefaulting later now exposes a special
case with partially successful writes that isn't handled correctly. And
the prefaulting likely isn't normally even that much of a performance
issue - it looks like at least one reason Dave saw this in his
performance tests is that he also ran them on Skylake that now supports
the new SMAP code, which makes the normally very cheap user space
prefaulting noticeably more expensive.
Bisected-and-acked-by: Ted Ts'o <[email protected]>
Analyzed-and-acked-by: Dave Hansen <[email protected]>
Cc: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
|
|
qxl_bo_unref calls drm_gem_object_unreference_unlocked which
locks dev->struct_mutex. However this lock could be already
locked if the call came from qxl_gem_object_free.
As we don't need to call qxl_bo_ref/qxl_bo_unref cause
qxl_release_list_add will hold a reference by itself avoid
to call them and the possible deadlock.
Signed-off-by: Frediano Ziglio <[email protected]>
Signed-off-by: Dave Airlie <[email protected]>
|
|
This avoid a dependency lock error.
According to https://lwn.net/Articles/548909/ users of WW mutex API
should avoid using different context.
When a buffer is reserved with qxl_bo_reserve a ww_mutex_lock without
context is used. However during qxl_draw_dirty_fb different locks
with specific context are used.
This is detected during a machine booting with a debug kernel with lock
dependency checking enabled.
Like many other function in this file to avoid this problem object
pinning is used. Once the object is pinned is not necessary to keep
the lock so it can be released avoiding the locking problem.
Signed-off-by: Frediano Ziglio <[email protected]>
Signed-off-by: Dave Airlie <[email protected]>
|
|
Commit "c0fe07a drm/qxl: rewrite framebuffer support" has a bug in the
dirty rectangle tracking: Instead of ignoring an empty dirty rectangle
when adding a new dirty region the dirty region gets extended to the
upper left corner. Fix it.
Cc: [email protected]
Signed-off-by: Gerd Hoffmann <[email protected]>
Signed-off-by: Dave Airlie <[email protected]>
|
|
Running xfstest generic/013 with the tracepoint nfs:nfs4_open_file
enabled produces a NULL-pointer dereference when calculating fileid and
filehandle of the opened file. Fix this by checking if state is NULL
before trying to use the inode pointer.
Reported-by: Olga Kornievskaia <[email protected]>
Signed-off-by: Anna Schumaker <[email protected]>
Signed-off-by: Trond Myklebust <[email protected]>
|
|
HDMI sinks are permitted to de-assert and re-assert the HPD signal to
indicate that their EDID has been updated, which may not involve a
change of video information.
An example of where such a situation can arise is when an AV receiver
is connected between the source and the display device. Events which
can cause the HPD to be deasserted include:
* turning on or switching to standby the AV receiver.
* turning on or switching to standby the display device.
Each of these can change the entire EDID data, or just a part of the
EDID data - it's up to the connected HDMI sink to do what they desire
here. For example
- with the AV receiver and display device both in standby, a source
connected to the AV receiver may provide its own EDID to the source.
- turning on the display device causes the display device's EDID to be
made available in an unmodified form to the source.
- subsequently turning on the AV receiver then provides a modified
version of the display device's EDID.
Moreover, HPD doesn't tell us whether something is actually listening
on the HDMI TDMS signals. The phy gives us a set of RXSENSE indications
which tell us whether there is a sink connected to the TMDS signals.
Currently, we use the HPD signal to enable or disable the HDMI block,
which is questionable when HPD is used in this manner. Using the
RXSENSE would be more appropriate, but there is some bad behaviour
which needs to be coped with. The iMX6 implementation lets the TMDS
signals float when the phy is "powered down", which cause spurious
interrupts. Rather than just using RXSENSE, use RXSENSE and HPD
becoming both active to signal the presence of a device, but loss
of RXSENSE to indicate that the device has been unplugged.
The side effect of this change is that a sink deasserting the HPD
signal to cause a re-read of the EDID data will not cause the bridge
to immediately disable the video signal.
Tested-by: Philipp Zabel <[email protected]>
Reviewed-by: Fabio Estevam <[email protected]>
Signed-off-by: Russell King <[email protected]>
|
|
When connected to HDMI sources, some DVI monitors de-assert their HPD
signal and TDMS loads for one seconds every four seconds when there is
no signal present on the connection.
Unfortunately, this behaviour is indistinguishable from a proper HDMI
setup with an AV receiver in the path to the display: the HDMI spec
requires us to detect HPD deassertions as short as 100ms, which indicate
that the EDID has changed.
Since it is possible to connect a DVI monitor to an AV receiver and then
to a HDMI source, merely working around this by detecting the lack of
HDMI vendor block in the EDID is insufficient - the AV receiver is at
liberty to modify the EDID as it sees fit, and it will place its own
parameters into the EDID including the HDMI vendor block.
DRM has support for forcing the state of a connector, which we should
implement to allow us to work around these broken DVI monitors - we can
tell DRM to force the connection state to indicate that there is always
a device connected to work around this problem. Although this requires
manual configuration, it is better than nothing at all.
When a forced connection state has been set, there is no point handling
our RXSENSE interrupts, so disable them in this circumstance.
Tested-by: Philipp Zabel <[email protected]>
Reviewed-by: Fabio Estevam <[email protected]>
Signed-off-by: Russell King <[email protected]>
|
|
Add support for interlaced video modes to the dw_hdmi bridge. This
mainly involves halving the vertical parameters to be programmed into
the bridge registers, and setting the interlace_allowed connector flag.
This brings working 1080i support. However, 480i and 576i fail to
work due to the lack of proper pixel repetition support, which is not
trivial to add due to the tabular PLL parameterisation. Hence, we
filter out these modes in our mode_valid() method.
Tested-by: Philipp Zabel <[email protected]>
Reviewed-by: Fabio Estevam <[email protected]>
Signed-off-by: Russell King <[email protected]>
|
|
The support for interlaced video modes seems to be broken; we don't use
anything other than the vtotal/htotal from the timing information to
define the various sync counters.
Freescale patches for interlaced video support contain an alternative
sync counter setup, which we include here. This setup produces the
hsync and vsync via the normal counter 2 and 3, but moves the display
enable signal from counter 5 to counter 6. Therefore, we need to
change the display controller setup as well.
The corresponding Freescale patches for this change are:
iMX6-HDMI-support-interlaced-display-mode.patch
IPU-fine-tuning-the-interlace-display-timing-for-CEA.patch
This produces a working interlace format output from the IPU.
Tested-by: Philipp Zabel <[email protected]>
Reviewed-by: Fabio Estevam <[email protected]>
Signed-off-by: Russell King <[email protected]>
|
|
Use a function to convert the sync pin to a bit mask for the DI_GENERAL
register, and move this out of the interlace/non-interlace path to the
common path.
Tested-by: Philipp Zabel <[email protected]>
Reviewed-by: Fabio Estevam <[email protected]>
Signed-off-by: Russell King <[email protected]>
|
|
It's possible that the destination can be shadowed in userspace
(as, for example, the perf buffers are now). So we should take
care not to leak data that could be inspected by userspace.
Signed-off-by: Chris Metcalf <[email protected]>
|
|
Both alpha and tile needed implementations of zero_bytemask.
The alpha version is untested.
Signed-off-by: Chris Metcalf <[email protected]>
|
|
arch/tile added word-at-a-time.h after the patch that added generic-y
entries; the generic-y entry is now stale.
arch/h8300 is newer than the generic-y patch for word-at-a-time.h,
and needs a generic-y entry.
arch/powerpc seems to have gotten a generic-y entry by mistake in
the first patch; this change removes it.
Signed-off-by: Chris Metcalf <[email protected]>
|
|
BUG: sleeping function called from invalid context at kernel/locking/rtmutex.c:917
in_atomic(): 0, irqs_disabled(): 128, pid: 342, name: perf
1 lock held by perf/342:
#0: (break_hook_lock){+.+...}, at: [<ffffffc0000851ac>] call_break_hook+0x34/0xd0
irq event stamp: 62224
hardirqs last enabled at (62223): [<ffffffc00010b7bc>] __call_rcu.constprop.59+0x104/0x270
hardirqs last disabled at (62224): [<ffffffc0000fbe20>] vprintk_emit+0x68/0x640
softirqs last enabled at (0): [<ffffffc000097928>] copy_process.part.8+0x428/0x17f8
softirqs last disabled at (0): [< (null)>] (null)
CPU: 0 PID: 342 Comm: perf Not tainted 4.1.6-rt5 #4
Hardware name: linux,dummy-virt (DT)
Call trace:
[<ffffffc000089968>] dump_backtrace+0x0/0x128
[<ffffffc000089ab0>] show_stack+0x20/0x30
[<ffffffc0007030d0>] dump_stack+0x7c/0xa0
[<ffffffc0000c878c>] ___might_sleep+0x174/0x260
[<ffffffc000708ac8>] __rt_spin_lock+0x28/0x40
[<ffffffc000708db0>] rt_read_lock+0x60/0x80
[<ffffffc0000851a8>] call_break_hook+0x30/0xd0
[<ffffffc000085a70>] brk_handler+0x30/0x98
[<ffffffc000082248>] do_debug_exception+0x50/0xb8
Exception stack(0xffffffc00514fe30 to 0xffffffc00514ff50)
fe20: 00000000 00000000 c1594680 0000007f
fe40: ffffffff ffffffff 92063940 0000007f 0550dcd8 ffffffc0 00000000 00000000
fe60: 0514fe70 ffffffc0 000be1f8 ffffffc0 0514feb0 ffffffc0 0008948c ffffffc0
fe80: 00000004 00000000 0514fed0 ffffffc0 ffffffff ffffffff 9282a948 0000007f
fea0: 00000000 00000000 9282b708 0000007f c1592820 0000007f 00083914 ffffffc0
fec0: 00000000 00000000 00000010 00000000 00000064 00000000 00000001 00000000
fee0: 005101e0 00000000 c1594680 0000007f c1594740 0000007f ffffffd8 ffffff80
ff00: 00000000 00000000 00000000 00000000 c1594770 0000007f c1594770 0000007f
ff20: 00665e10 00000000 7f7f7f7f 7f7f7f7f 01010101 01010101 00000000 00000000
ff40: 928e4cc0 0000007f 91ff11e8 0000007f
call_break_hook is called in atomic context (hard irq disabled), so replace
the sleepable lock to rcu lock, replace relevant list operations to rcu
version and call synchronize_rcu() in unregister_break_hook().
And, replace write lock to spinlock in {un}register_break_hook.
Signed-off-by: Yang Shi <[email protected]>
Signed-off-by: Will Deacon <[email protected]>
|
|
When booting a kernel without an initrd, the kernel reports that it
moves -1 bytes worth, having gone through the motions with initrd_start
equal to initrd_end:
Moving initrd from [4080000000-407fffffff] to [9fff49000-9fff48fff]
Prevent this by bailing out early when the initrd size is zero (i.e. we
have no initrd), avoiding the confusing message and other associated
work.
Fixes: 1570f0d7ab425c1e ("arm64: support initrd outside kernel linear map")
Cc: Mark Salter <[email protected]>
Signed-off-by: Mark Rutland <[email protected]>
Signed-off-by: Will Deacon <[email protected]>
|
|
These have not been tested very extensively and generally
seem to be problematic. Mark them as experimental for now.
bug:
https://bugs.freedesktop.org/show_bug.cgi?id=92270
Reviewed-by: Christian König <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>
Cc: [email protected]
|
|
So the problem this patch is trying to address is as follows:
CPU0 CPU1
context_switch(A, B)
ttwu(A)
LOCK A->pi_lock
A->on_cpu == 0
finish_task_switch(A)
prev_state = A->state <-.
WMB |
A->on_cpu = 0; |
UNLOCK rq0->lock |
| context_switch(C, A)
`-- A->state = TASK_DEAD
prev_state == TASK_DEAD
put_task_struct(A)
context_switch(A, C)
finish_task_switch(A)
A->state == TASK_DEAD
put_task_struct(A)
The argument being that the WMB will allow the load of A->state on CPU0
to cross over and observe CPU1's store of A->state, which will then
result in a double-drop and use-after-free.
Now the comment states (and this was true once upon a long time ago)
that we need to observe A->state while holding rq->lock because that
will order us against the wakeup; however the wakeup will not in fact
acquire (that) rq->lock; it takes A->pi_lock these days.
We can obviously fix this by upgrading the WMB to an MB, but that is
expensive, so we'd rather avoid that.
The alternative this patch takes is: smp_store_release(&A->on_cpu, 0),
which avoids the MB on some archs, but not important ones like ARM.
Reported-by: Oleg Nesterov <[email protected]>
Signed-off-by: Peter Zijlstra (Intel) <[email protected]>
Acked-by: Linus Torvalds <[email protected]>
Cc: <[email protected]> # v3.1+
Cc: Peter Zijlstra <[email protected]>
Cc: Thomas Gleixner <[email protected]>
Cc: [email protected]
Cc: [email protected]
Cc: [email protected]
Fixes: e4a52bcb9a18 ("sched: Remove rq->lock from the first half of ttwu()")
Link: http://lkml.kernel.org/r/[email protected]
Signed-off-by: Ingo Molnar <[email protected]>
|
|
Make sure we are not the root device before attempting to
read the pcie bridge registers to check the pcie gen speeed.
Fixes a crash when the device is passed through to a VM.
Reviewed-by: Christian König <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>
Cc: [email protected]
|
|
end_clone_bio() is a endio callback for clone bio and should check
and save the clone's bi_error for error reporting. However,
4246a0b63bd8 ("block: add a bi_error field to struct bio") changed
the function to check the original bio's bi_error, which is 0.
Without this fix, clone's error is ignored and reported to the
original request as success. Thus data corruption will be observed.
Fixes: 4246a0b63bd8 ("block: add a bi_error field to struct bio")
Signed-off-by: Jun'ichi Nomura <[email protected]>
Cc: Christoph Hellwig <[email protected]>
Signed-off-by: Mike Snitzer <[email protected]>
|
|
git://git.kernel.org/pub/scm/linux/kernel/git/xen/tip
Pull xen bug fixes from David Vrabel:
- Fix VM save performance regression with x86 PV guests
- Make kexec work in x86 PVHVM guests (if Xen has the soft-reset ABI)
- Other minor fixes.
* tag 'for-linus-4.3b-rc4-tag' of git://git.kernel.org/pub/scm/linux/kernel/git/xen/tip:
x86/xen/p2m: hint at the last populated P2M entry
x86/xen: Do not clip xen_e820_map to xen_e820_map_entries when sanitizing map
x86/xen: Support kexec/kdump in HVM guests by doing a soft reset
xen/x86: Don't try to write syscall-related MSRs for PV guests
xen: use correct type for HYPERVISOR_memory_op()
|
|
git://git.kernel.org/pub/scm/linux/kernel/git/s390/linux
Pull s390 fixes from Martin Schwidefsky:
"Three bug fixes and an update to the default configuration"
* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/s390/linux:
s390/defconfig: set SCSI_DH=y
s390/vtime: correct scaled cputime of partially idle CPUs
s390/boot/decompression: disable floating point in decompressor
s390/numa: use correct type for node_to_cpumask_map
|
|
The "fh_len" passed to ->fh_to_* is not guaranteed to be that same as
that returned by encode_fh - it may be larger.
With NFSv2, the filehandle is fixed length, so it may appear longer
than expected and be zero-padded.
So we must test that fh_len is at least some value, not exactly equal
to it.
Signed-off-by: NeilBrown <[email protected]>
Acked-by: David Sterba <[email protected]>
|
|
After reading one of chunk or tree root tree's root node from disk, if the
root node does not have EXTENT_BUFFER_UPTODATE flag set, we fail to release
the memory used by the root node. Fix this.
Signed-off-by: Chandan Rajendra <[email protected]>
|
|
Pull CIFS fixes from Steve French:
"Two fixes for problems pointed out by automated tools.
Thanks PaX/grsecurity team and Dan Carpenter (and the Smatch tool)"
* 'for-next' of git://git.samba.org/sfrench/cifs-2.6:
[CIFS] Update cifs version number
[SMB3] Do not fall back to SMBWriteX in set_file_size error cases
[SMB3] Missing null tcon check
|
|
With commit 633d6f17cd91ad5bf2370265946f716e42d388c6 (x86/xen: prepare
p2m list for memory hotplug) the P2M may be sized to accomdate a much
larger amount of memory than the domain currently has.
When saving a domain, the toolstack must scan all the P2M looking for
populated pages. This results in a performance regression due to the
unnecessary scanning.
Instead of reporting (via shared_info) the maximum possible size of
the P2M, hint at the last PFN which might be populated. This hint is
increased as new leaves are added to the P2M (in the expectation that
they will be used for populated entries).
Signed-off-by: David Vrabel <[email protected]>
Cc: <[email protected]> # 4.0+
|
|
git://git.kernel.org/pub/scm/linux/kernel/git/horms/renesas into fixes
Merge "Renesas ARM Based SoC Fixes for v4.3" from Simon Horman
* Add Add CPG/MSTP Clock Domain for sound on r8a779[01] SoCs.
This allows sound to work once again.
* tag 'renesas-fixes-for-v4.3' of git://git.kernel.org/pub/scm/linux/kernel/git/horms/renesas:
ARM: shmobile: r8a7791 dtsi: Add CPG/MSTP Clock Domain for sound
ARM: shmobile: r8a7790 dtsi: Add CPG/MSTP Clock Domain for sound
|
|
https://git.kernel.org/pub/scm/linux/kernel/git/mripard/linux into fixes
Merge "Allwinner fixes for 4.3" from Maxime Ripard:
Two patches, one that fixes one of the DT build, and the other raising the
voltage of the lowest OPP of the A20 to remain within the SoC operating
boundaries
* tag 'sunxi-fixes-for-4.3' of https://git.kernel.org/pub/scm/linux/kernel/git/mripard/linux:
ARM: dts: Fix Makefile target for sun4i-a10-itead-iteaduino-plus
ARM: dts: sunxi: Raise minimum CPU voltage for sun7i-a20 to meet SoC specifications
|
|
git://git.kernel.org/pub/scm/linux/kernel/git/kgene/linux-samsung into fixes
Merge "Samsung fixes for v4.3" from Kukjin Kim:
- fix invalid clock used for FIMD IOMMU
- fix thermal boot issue smdk5250-smdk5250
- fix S2R on exynos4412 trats2 boards
- fix LEDs on exynos5422-odroidxu3-common
- fix booting of all 8 cores on exynos542x
* tag 'samsung-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/kgene/linux-samsung:
ARM: dts: Fix wrong clock binding for sysmmu_fimd1_1 on exynos5420
ARM: dts: Fix bootup thermal issue on smdk5250
ARM: dts: add suspend opp to exynos4412
ARM: dts: Fix LEDs on exynos5422-odroidxu3
ARM: EXYNOS: reset Little cores when cpu is up
|
|
Whilst discussing possible ways to trigger an invalidate_range on a
userptr with an aliased GGTT mmapping (and so cause a struct_mutex
deadlock), the conclusion is that we can, and we must, prevent any
possible deadlock by avoiding taking the mutex at all during
invalidate_range. This has numerous advantages all of which stem from
avoid the sleeping function from inside the unknown context. In
particular, it simplifies the invalidate_range because we no longer
have to juggle the spinlock/mutex and can just hold the spinlock
for the entire walk. To compensate, we have to make get_pages a bit more
complicated in order to serialise with a pending cancel_userptr worker.
As we hold the struct_mutex, we have no choice but to return EAGAIN and
hope that the worker is then flushed before we retry after reacquiring
the struct_mutex.
The important caveat is that the invalidate_range itself is no longer
synchronous. There exists a small but definite period in time in which
the old PTE's page remain accessible via the GPU. Note however that the
physical pages themselves are not invalidated by the mmu_notifier, just
the CPU view of the address space. The impact should be limited to a
delay in pages being flushed, rather than a possibility of writing to
the wrong pages. The only race condition that this worsens is remapping
an userptr active on the GPU where fresh work may still reference the
old pages due to struct_mutex contention. Given that userspace is racing
with the GPU, it is fair to say that the results are undefined.
v2: Only queue (and importantly only take one refcnt) the worker once.
Signed-off-by: Chris Wilson <[email protected]>
Cc: Michał Winiarski <[email protected]>
Cc: Tvrtko Ursulin <[email protected]>
Reviewed-by: Tvrtko Ursulin <[email protected]>
Signed-off-by: Daniel Vetter <[email protected]>
|
|
Michał Winiarski found a really evil way to trigger a struct_mutex
deadlock with userptr. He found that if he allocated a userptr bo and
then GTT mmaped another bo, or even itself, at the same address as the
userptr using MAP_FIXED, he could then cause a deadlock any time we then
had to invalidate the GTT mmappings (so at will). Tvrtko then found by
repeatedly allocating GTT mmappings he could alias with an old userptr
mmap and also trigger the deadlock.
To counter act the deadlock, we make the observation that we only need
to take the struct_mutex if the object has any pages to revoke, and that
before userspace can alias with the userptr address space, it must have
invalidated the userptr->pages. Thus if we can check for those pages
outside of the struct_mutex, we can avoid the deadlock. To do so we
introduce a separate flag for userptr objects that we can inspect from
the mmu-notifier underneath its spinlock.
The patch makes one eye-catching change. That is the removal serial=0
after detecting a to-be-freed object inside the invalidate walker. I
felt setting serial=0 was a questionable pessimisation: it denies us the
chance to reuse the current iterator for the next loop (before it is
freed) and being explicit makes the reader question the validity of the
locking (since the object-free race could occur elsewhere). The
serialisation of the iterator is through the spinlock, if the object is
freed before the next loop then the notifier.serial will be incremented
and we start the walk from the beginning as we detect the invalid cache.
To try and tame the error paths and interactions with the userptr->active
flag, we have to do a fair amount of rearranging of get_pages_userptr().
v2: Grammar fixes
v3: Reorder set-active so that it is only set when obj->pages is set
(and so needs cancellation). Only the order of setting obj->pages and
the active-flag is crucial. Calling gup after invalidate-range begin
means the userptr sees the new set of backing storage (and so will not
need to invalidate its new pages), but we have to be careful not to set
the active-flag prior to successfully establishing obj->pages.
v4: Take the active->flag early so we know in the mmu-notifier when we
have to cancel a pending gup-worker.
v5: Rearrange the error path so that is not so convoluted
v6: Set pinned to 0 when negative before calling release_pages()
Reported-by: Michał Winiarski <[email protected]>
Testcase: igt/gem_userptr_blits/map-fixed*
Signed-off-by: Chris Wilson <[email protected]>
Cc: Michał Winiarski <[email protected]>
Cc: Tvrtko Ursulin <[email protected]>
Cc: [email protected]
Reviewed-by: Tvrtko Ursulin <[email protected]>
Signed-off-by: Daniel Vetter <[email protected]>
|
|
The userptr worker allows for a slight race condition where upon there
may two or more threads calling get_user_pages for the same object. When
we have the array of pages, then we serialise the update of the object.
However, the worker should only overwrite the obj->userptr.work pointer
if and only if it is the active one. Currently we clear it for a
secondary worker with the effect that we may rarely force a second
lookup.
v2: Rebase and rename a variable to avoid 80cols
v3: Mention v2
Signed-off-by: Chris Wilson <[email protected]>
Reviewed-by: Tvrtko Ursulin <[email protected]>
Signed-off-by: Daniel Vetter <[email protected]>
|