cpuidle->enter() callbacks should not call into tracing because RCU
has already been disabled. Instead of doing the broadcast thing
itself, simply advertise to the cpuidle core that those states stop
the timer.
The local variable 'cpumask_t mask' is in the stack memory, and its address
is assigned to 'desc->affinity' in 'irq_set_affinity_hint()'.
But the memory area where this variable is located is at risk of being
modified.
During LTP testing, the following error was generated:
Fix it by using 'cpumask_of(cpu)' to get the cpumask.
Signed-off-by: Hao Si <si.hao@zte.com.cn> Signed-off-by: Lin Chen <chen.lin5@zte.com.cn> Signed-off-by: Yi Wang <wang.yi59@zte.com.cn> Signed-off-by: Li Yang <leoyang.li@nxp.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
This is because during the quota enable ioctl we lock first the mutex
qgroup_ioctl_lock and then start a transaction, and starting a transaction
acquires a fs freeze semaphore (at the VFS level). However, every other
code path, except for the quota disable ioctl path, we do the opposite:
we start a transaction and then lock the mutex.
So fix this by making the quota enable and disable paths to start the
transaction without having the mutex locked, and then, after starting the
transaction, lock the mutex and check if some other task already enabled
or disabled the quotas, bailing with success if that was the case.
Signed-off-by: Filipe Manana <fdmanana@suse.com> Reviewed-by: David Sterba <dsterba@suse.com> Signed-off-by: David Sterba <dsterba@suse.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
When adding or removing a qgroup relation we are doing a GFP_KERNEL
allocation which is not safe because we are holding a transaction
handle open and that can make us deadlock if the allocator needs to
recurse into the filesystem. So just surround those calls with a
nofs context.
Signed-off-by: Filipe Manana <fdmanana@suse.com> Reviewed-by: David Sterba <dsterba@suse.com> Signed-off-by: David Sterba <dsterba@suse.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
Robin Murphy pointed out that if the arm-smmu driver probes before
the qcom_scm driver, we may call qcom_scm_qsmmu500_wait_safe_toggle()
before the __scm is initialized.
Now, getting this to happen is a bit contrived, as in my efforts it
required enabling asynchronous probing for both drivers, moving the
firmware dts node to the end of the dtsi file, as well as forcing a
long delay in the qcom_scm_probe function.
With those tweaks we ran into the following crash:
[ 2.631040] arm-smmu 15000000.iommu: Stage-1: 48-bit VA -> 48-bit IPA
[ 2.633372] Unable to handle kernel NULL pointer dereference at virtual address 0000000000000000
...
[ 2.633402] [0000000000000000] user address but active_mm is swapper
[ 2.633409] Internal error: Oops: 96000005 [#1] PREEMPT SMP
[ 2.633415] Modules linked in:
[ 2.633427] CPU: 5 PID: 117 Comm: kworker/u16:2 Tainted: G W 5.10.0-rc1-mainline-00025-g272a618fc36-dirty #3971
[ 2.633430] Hardware name: Thundercomm Dragonboard 845c (DT)
[ 2.633448] Workqueue: events_unbound async_run_entry_fn
[ 2.633456] pstate: 80c00005 (Nzcv daif +PAN +UAO -TCO BTYPE=--)
[ 2.633465] pc : qcom_scm_qsmmu500_wait_safe_toggle+0x78/0xb0
[ 2.633473] lr : qcom_smmu500_reset+0x58/0x78
[ 2.633476] sp : ffffffc0105a3b60
...
[ 2.633567] Call trace:
[ 2.633572] qcom_scm_qsmmu500_wait_safe_toggle+0x78/0xb0
[ 2.633576] qcom_smmu500_reset+0x58/0x78
[ 2.633581] arm_smmu_device_reset+0x194/0x270
[ 2.633585] arm_smmu_device_probe+0xc94/0xeb8
[ 2.633592] platform_drv_probe+0x58/0xa8
[ 2.633597] really_probe+0xec/0x398
[ 2.633601] driver_probe_device+0x5c/0xb8
[ 2.633606] __driver_attach_async_helper+0x64/0x88
[ 2.633610] async_run_entry_fn+0x4c/0x118
[ 2.633617] process_one_work+0x20c/0x4b0
[ 2.633621] worker_thread+0x48/0x460
[ 2.633628] kthread+0x14c/0x158
[ 2.633634] ret_from_fork+0x10/0x18
[ 2.633642] Code: a9034fa0d0007f7329107fa091342273 (f9400020)
To avoid this, this patch adds a check on qcom_scm_is_available() in
the qcom_smmu_impl_init() function, returning -EPROBE_DEFER if its
not ready.
This allows the driver to try to probe again later after qcom_scm has
finished probing.
Reported-by: Robin Murphy <robin.murphy@arm.com> Signed-off-by: John Stultz <john.stultz@linaro.org> Reviewed-by: Robin Murphy <robin.murphy@arm.com> Cc: Robin Murphy <robin.murphy@arm.com> Cc: Will Deacon <will@kernel.org> Cc: Andy Gross <agross@kernel.org> Cc: Maulik Shah <mkshah@codeaurora.org> Cc: Bjorn Andersson <bjorn.andersson@linaro.org> Cc: Saravana Kannan <saravanak@google.com> Cc: Marc Zyngier <maz@kernel.org> Cc: Lina Iyer <ilina@codeaurora.org> Cc: iommu@lists.linux-foundation.org Cc: linux-arm-msm <linux-arm-msm@vger.kernel.org> Link: https://lore.kernel.org/r/20201112220520.48159-1-john.stultz@linaro.org Signed-off-by: Will Deacon <will@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
Given the case that bootloader(such as UEFI)'s FSPI driver might not
handle all interrupts before loading kernel, those legacy interrupts
would assert immidiately once kernel's FSPI driver enable them. Further,
if it was FSPI_INTR_IPCMDDONE, the irq handler nxp_fspi_irq_handler()
would call complete(&f->c) to notify others. However, f->c might not be
initialized yet at that time, then cause kernel panic.
Of cause, we should fix this issue within bootloader. But it would be
better to have this pacth to make dirver more robust (by clearing all
interrupt status bits before enabling interrupts).
The Exynos DRM uses Common Clock Framework thus it cannot be built on
platforms without it (e.g. compile test on MIPS with RALINK and
SOC_RT305X):
/usr/bin/mips-linux-gnu-ld: drivers/gpu/drm/exynos/exynos_mixer.o: in function `mixer_bind':
exynos_mixer.c:(.text+0x958): undefined reference to `clk_set_parent'
Reported-by: kernel test robot <lkp@intel.com> Signed-off-by: Krzysztof Kozlowski <krzk@kernel.org> Signed-off-by: Inki Dae <inki.dae@samsung.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
On systems without HW-based collections (i.e. anything except GIC-500),
we rely on firmware to perform the ITS save/restore. This doesn't
really work, as although FW can properly save everything, it cannot
fully restore the state of the command queue (the read-side is reset
to the head of the queue). This results in the ITS consuming previously
processed commands, potentially corrupting the state.
Instead, let's always save the ITS state on suspend, disabling it in the
process, and restore the full state on resume. This saves us from broken
FW as long as it doesn't enable the ITS by itself (for which we can't do
anything).
This amounts to simply dropping the ITS_FLAGS_SAVE_SUSPEND_STATE.
Sometimes it takes longer than 5 seconds (watchdog timeout) to complete
failover, migration, and other resets. In stead of scheduling another
timeout reset, we wait for the current one to complete.
Suggested-by: Brian King <brking@linux.vnet.ibm.com> Signed-off-by: Lijun Pan <ljp@linux.ibm.com> Reviewed-by: Dany Madden <drt@linux.ibm.com> Signed-off-by: Jakub Kicinski <kuba@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
The following errors are noticed during boot on a QCS404 board:
[ 2.926647] qcom_icc_rpm_smd_send mas 6 error -6
[ 2.934573] qcom_icc_rpm_smd_send mas 8 error -6
These errors show when we try to configure the GPU and display nodes.
Since these particular nodes aren't supported on RPM and are purely
local, we should just change their mas_rpm_id to -1 to avoid any
requests being sent for these master IDs.
Some nodes are incorrectly marked as RPM-controlled (they have RPM
master and slave ids assigned), but are actually controlled by the
application CPU instead. The RPM complains when we send requests for
resources that it can't control. Let's fix this by replacing the IDs,
with the default "-1" in which case no requests are sent.
If someone plays with the UFS clk scaling devfreq governor through sysfs,
ufshcd_devfreq_scale may be called even when HBA is not runtime ACTIVE.
This can lead to unexpected error. We cannot just protect it by calling
pm_runtime_get_sync() because that may cause a race condition since HBA
runtime suspend ops need to suspend clk scaling. To fix this call
pm_runtime_get_noresume() and check HBA's runtime status. Only proceed if
HBA is runtime ACTIVE, otherwise just bail.
WB-related sysfs entries can be accessed even when an UFS device does not
support the feature. The descriptors which are not supported by the UFS
device may be wrongly reported when they are accessed from their
corrsponding sysfs entries. Fix it by adding a sanity check of parameter
offset against the actual decriptor length.
Link: https://lore.kernel.org/r/1603346348-14149-1-git-send-email-cang@codeaurora.org Reviewed-by: Asutosh Das <asutoshd@codeaurora.org> Acked-by: Daejun Park <daejun7.park@samsung.com> Signed-off-by: Can Guo <cang@codeaurora.org> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
To start stack unwinding (SP, PC and BLINK) are needed. When the
explicit execution context (pt_regs etc) is not available, unwinder
assumes the task is sleeping (in __switch_to()) and fetches SP and BLINK
from kernel mode stack.
But this assumption is not true, specially in a SMP system, when top
runs on 1 core, there may be active running processes on all cores.
So when unwinding non courrent tasks, ensure they are NOT running.
And while at it, handle the self unwinding case explicitly.
This came out of investigation of a customer reported hang with
rcutorture+top
The scripts/dtc/checks.c requires that the node have empty "dma-ranges"
property must have the same "#address-cells" and "#size-cells" values as
the parent node. Otherwise, the following warnings is reported:
arch/arm64/boot/dts/broadcom/stingray/stingray-usb.dtsi:7.3-14: Warning \
(dma_ranges_format): /usb:dma-ranges: empty "dma-ranges" property but \
its #address-cells (1) differs from / (2)
arch/arm64/boot/dts/broadcom/stingray/stingray-usb.dtsi:7.3-14: Warning \
(dma_ranges_format): /usb:dma-ranges: empty "dma-ranges" property but \
its #size-cells (1) differs from / (2)
Arnd Bergmann figured out why it's necessary:
Also note that the #address-cells=<1> means that any device under
this bus is assumed to only support 32-bit addressing, and DMA will
have to go through a slow swiotlb in the absence of an IOMMU.
During CSA, we briefly nullify the phy context, in __iwl_mvm_unassign_vif_chanctx.
In case we have a FW assert right after it, it remains NULL though.
We end up running into endless loop due to mac80211 trying repeatedly to
move us to ASSOC state, and we keep returning -EINVAL. Later down the road
we hit a kernel panic.
On some platforms, the preset values aren't correct and then we may
get a completion timeout in the firmware. Change the LTR configuration
to avoid that. The firmware will do some more complex reinit of this
later, but for the boot process we use ~250usec.
Currently, our max tpt is limited to max HT A-MPDU for LB,
and max VHT A-MPDU for HB. Configure HE exponent value correctly to
achieve HE max A-MPDU, both on LB and HB.
Recently introduced async probe on mmc devices can shuffle block IDs.
Pin them to fixed values to ease booting in environments where UUIDs
are not practical. Use newly introduced aliases for mmcblk devices from [1].
When we read device memory, we lock a spinlock, write the address we
want to read from the device and then spin in a loop reading the data
in 32-bit quantities from another register.
As the description makes clear, this is rather inefficient, incurring
a PCIe bus transaction for every read. In a typical device today, we
want to read 786k SMEM if it crashes, leading to 192k register reads.
Occasionally, we've seen the whole loop take over 20 seconds and then
triggering the soft lockup detector.
Clearly, it is unreasonable to spin here for such extended periods of
time.
To fix this, break the loop down into an outer and an inner loop, and
break out of the inner loop if more than half a second elapsed. To
avoid too much overhead, check for that only every 128 reads, though
there's no particular reason for that number. Then, unlock and relock
to obtain NIC access again, reprogram the start address and continue.
This will keep (interrupt) latencies on the CPU down to a reasonable
time.
The 7211a0 has a tca_drv_sel bit in the USB SETUP register that
should never be enabled. This feature is only used if there is a
USB Type-C PHY, and the 7211 does not have one. If the bit is
enabled, the VBUS signal will never be asserted. In the 7211a0,
the bit was incorrectly defaulted to on so the driver had to clear
the bit. In the 7211c0 the state was inverted so the driver should
no longer clear the bit. This hasn't been a problem because all
current 7211 boards don't use the VBUS signal, but there are some
future customer boards that may use it.
On success, mmap should return the begin address of newly mapped area,
but patch "mm: mmap: merge vma after call_mmap() if possible" set
vm_start of newly merged vma to return value addr. Users of mmap will
get wrong address if vma is merged after call_mmap(). We fix this by
moving the assignment to addr before merging vma.
We have a driver which changes vm_flags, and this bug is found by our
testcases.
Fixes: d70cec898324 ("mm: mmap: merge vma after call_mmap() if possible") Signed-off-by: Liu Zixian <liuzixian4@huawei.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Reviewed-by: Jason Gunthorpe <jgg@nvidia.com> Reviewed-by: David Hildenbrand <david@redhat.com> Cc: Miaohe Lin <linmiaohe@huawei.com> Cc: Hongxiang Lou <louhongxiang@huawei.com> Cc: Hu Shiyuan <hushiyuan@huawei.com> Cc: Matthew Wilcox <willy@infradead.org> Link: https://lkml.kernel.org/r/20201203085350.22624-1-liuzixian4@huawei.com Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
The commit 3966c3feca3f ("x86/perf/amd: Remove need to check "running"
bit in NMI handler") introduced this. It seems x86_pmu_stop can be
called recursively (like when it losts some samples) like below:
While commit 35d1ce6bec13 ("perf/x86/intel/ds: Fix x86_pmu_stop
warning for large PEBS") fixed it for the normal cases, there's
another path to call x86_pmu_stop() recursively when a PEBS error was
detected (like two or more counters overflowed at the same time).
Like in the Kan's previous fix, we can skip the interrupt accounting
for large PEBS, so check the iregs which is set for PMI only.
Fixes: 3966c3feca3f ("x86/perf/amd: Remove need to check "running" bit in NMI handler") Reported-by: John Sperbeck <jsperbeck@google.com> Suggested-by: Peter Zijlstra <peterz@infradead.org> Signed-off-by: Namhyung Kim <namhyung@kernel.org> Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org> Link: https://lkml.kernel.org/r/20201126110922.317681-1-namhyung@kernel.org Signed-off-by: Sasha Levin <sashal@kernel.org>
The NO_160 flag specifies if the device doesn't have 160 MHz support,
but we errorneously assumed the opposite. If the flag was set, we
were considering that 160 MHz was supported, but it's actually the
opposite. Fix it by inverting the bits, i.e. NO_160 is 0x1 and 160
is 0x0.
drivers/vdpa/mlx5/ uses vhost_iotlb*() interfaces, so select
VHOST_IOTLB to make them be built.
However, if VHOST_IOTLB is the only VHOST symbol that is
set/enabled, the object file still won't be built because
drivers/Makefile won't descend into drivers/vhost/ to build it,
so make drivers/Makefile build the needed binary whenever
VHOST_IOTLB is set, like it does for VHOST_RING.
Fixes: 29064bfdabd5 ("vdpa/mlx5: Add support library for mlx5 VDPA implementation") Fixes: aff90770e54c ("vdpa/mlx5: Fix dependency on MLX5_CORE") Reported-by: kernel test robot <lkp@intel.com> Signed-off-by: Randy Dunlap <rdunlap@infradead.org> Cc: Eli Cohen <eli@mellanox.com> Cc: Parav Pandit <parav@mellanox.com> Cc: "Michael S. Tsirkin" <mst@redhat.com> Cc: Jason Wang <jasowang@redhat.com> Cc: virtualization@lists.linux-foundation.org Cc: Saeed Mahameed <saeedm@nvidia.com> Cc: Leon Romanovsky <leonro@nvidia.com> Cc: netdev@vger.kernel.org Link: https://lore.kernel.org/r/20201128213905.27409-1-rdunlap@infradead.org Signed-off-by: Michael S. Tsirkin <mst@redhat.com> Acked-by: Jason Wang <jasowang@redhat.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
Return -ENOMEM from the error handling case instead of 0.
Link: https://lore.kernel.org/r/20201127030206.104616-1-jingxiangfeng@huawei.com Fixes: 436ad9413353 ("scsi: storvsc: Allow only one remove lun work item to be issued per lun") Reviewed-by: Michael Kelley <mikelley@microsoft.com> Signed-off-by: Jing Xiangfeng <jingxiangfeng@huawei.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
my_tramp[12]? are declared as global functions in C, but they are not
marked global in the inline assembly definition. This mismatch confuses
Clang's Control-Flow Integrity checking. Fix the definitions by adding
.globl.
Link: https://lkml.kernel.org/r/20201113183414.1446671-1-samitolvanen@google.com Fixes: 9d907f1ae80b8 ("ftrace/samples: Add a sample module that implements modify_ftrace_direct()") Reviewed-by: Kees Cook <keescook@chromium.org> Signed-off-by: Sami Tolvanen <samitolvanen@google.com> Signed-off-by: Steven Rostedt (VMware) <rostedt@goodmis.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
In the error handling in c_can_power_up(), there are two bugs:
1) c_can_pm_runtime_get_sync() will increase usage counter if device is not
empty. Forgetting to call c_can_pm_runtime_put_sync() will result in a
reference leak here.
2) c_can_reset_ram() operation will set start bit when enable is true. We
should clear it in the error handling.
We fix it by adding c_can_pm_runtime_put_sync() for 1), and
c_can_reset_ram(enable is false) for 2) in the error handling.
Losing arbitration is normal in a CAN-bus network, it means that a higher
priority frame is being send and the pending message will be retried later.
Hence most driver only increment arbitration_lost, but the sun4i driver also
incremeants tx_error, causing errors to be reported on a normal functioning
CAN-bus. So stop counting them as errors.
Fixes: 0738eff14d81 ("can: Allwinner A10/A20 CAN Controller support - Kernel module") Signed-off-by: Jeroen Hofstee <jhofstee@victronenergy.com> Link: https://lore.kernel.org/r/20201127095941.21609-1-jhofstee@victronenergy.com
[mkl: split into two seperate patches] Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de> Signed-off-by: Sasha Levin <sashal@kernel.org>
Losing arbitration is normal in a CAN-bus network, it means that a higher
priority frame is being send and the pending message will be retried later.
Hence most driver only increment arbitration_lost, but the sja1000 driver also
incremeants tx_error, causing errors to be reported on a normal functioning
CAN-bus. So stop counting them as errors.
Fixes: 8935f57e68c4 ("can: sja1000: fix network statistics update") Signed-off-by: Jeroen Hofstee <jhofstee@victronenergy.com> Link: https://lore.kernel.org/r/20201127095941.21609-1-jhofstee@victronenergy.com
[mkl: split into two seperate patches] Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de> Signed-off-by: Sasha Levin <sashal@kernel.org>
The clocks mcan_class->cclk and mcan_class->hclk are not prepared by any call
during tcan4x5x_can_probe(), so remove erroneous clk_disable_unprepare() on
them.
The probe routine acquires the reset GPIO using GPIOD_OUT_LOW. Directly
afterwards it calls acx565akm_detect(), which sets the GPIO value to
HIGH. If the bootloader initialized the GPIO to HIGH before the probe
routine was called, there is only a very short time period of a few
instructions where the reset signal is LOW. Exact time depends on
compiler optimizations, kernel configuration and alignment of the stars,
but I expect it to be always way less than 10us. There are no public
datasheets for the panel, but acx565akm_power_on() has a comment with
timings and reset period should be at least 10us. So this potentially
brings the panel into a half-reset state.
The result is, that panel may not work after boot and can get into a
working state by re-enabling it (e.g. by blanking + unblanking), since
that does a clean reset cycle. This bug has recently been hit by Ivaylo
Dimitrov, but there are some older reports which are probably the same
bug. At least Tony Lindgren, Peter Ujfalusi and Jarkko Nikula have
experienced it in 2017 describing the blank/unblank procedure as
possible workaround.
Note, that the bug really goes back in time. It has originally been
introduced in the predecessor of the omapfb driver in commit 3c45d05be382
("OMAPDSS: acx565akm panel: handle gpios in panel driver") in 2012.
That driver eventually got replaced by a newer one, which had the bug
from the beginning in commit 84192742d9c2 ("OMAPDSS: Add Sony ACX565AKM
panel driver") and still exists in fbdev world. That driver has later
been copied to omapdrm and then was used as a basis for this driver.
Last but not least the omapdrm specific driver has been removed in
commit 45f16c82db7e ("drm/omap: displays: Remove unused panel drivers").
Reported-by: Jarkko Nikula <jarkko.nikula@bitmer.com> Reported-by: Peter Ujfalusi <peter.ujfalusi@ti.com> Reported-by: Tony Lindgren <tony@atomide.com> Reported-by: Aaro Koskinen <aaro.koskinen@iki.fi> Reported-by: Ivaylo Dimitrov <ivo.g.dimitrov.75@gmail.com> Cc: Merlijn Wajer <merlijn@wizzup.org> Cc: Laurent Pinchart <laurent.pinchart@ideasonboard.com> Cc: Tomi Valkeinen <tomi.valkeinen@ti.com> Fixes: 1c8fc3f0c5d2 ("drm/panel: Add driver for the Sony ACX565AKM panel") Signed-off-by: Sebastian Reichel <sebastian.reichel@collabora.com> Tested-by: Ivaylo Dimitrov <ivo.g.dimitrov.75@gmail.com> Tested-by: Aaro Koskinen <aaro.koskinen@iki.fi> Tested-by: Jarkko Nikula <jarkko.nikula@bitmer.com> Signed-off-by: Sam Ravnborg <sam@ravnborg.org> Link: https://patchwork.freedesktop.org/patch/msgid/20201127200429.129868-1-sebastian.reichel@collabora.com Signed-off-by: Sasha Levin <sashal@kernel.org>
In the Rockchip DRM LVDS component driver, the endpoint id provided to
drm_of_find_panel_or_bridge is grabbed from the endpoint's reg property.
However, the property may be missing in the case of a single endpoint.
Initialize the endpoint_id variable to 0 to avoid using an
uninitialized variable in that case.
send_login() does not check for the result of ibmvnic_send_crq() of the
login request. This results in the driver needlessly retrying the login
10 times even when CRQ is no longer active. Check the return code and
give up in case of errors in sending the CRQ.
The only time we want to retry is if we get a PARITALSUCCESS response
from the partner.
Fixes: 032c5e82847a2 ("Driver for IBM System i/p VNIC protocol") Signed-off-by: Dany Madden <drt@linux.ibm.com> Signed-off-by: Sukadev Bhattiprolu <sukadev@linux.ibm.com> Signed-off-by: Jakub Kicinski <kuba@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
If after ibmvnic sends a LOGIN it gets a FAILOVER, it is possible that
the worker thread will start reset process and free the login response
buffer before it gets a (now stale) LOGIN_RSP. The ibmvnic tasklet will
then try to access the login response buffer and crash.
Have ibmvnic track pending logins and discard any stale login responses.
Fixes: 032c5e82847a ("Driver for IBM System i/p VNIC protocol") Signed-off-by: Sukadev Bhattiprolu <sukadev@linux.ibm.com> Signed-off-by: Jakub Kicinski <kuba@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
If auto-priority failover is enabled, the backing device needs time
to settle if hard resetting fails for any reason. Add a delay of 60
seconds before retrying the hard-reset.
Fixes: 2770a7984db5 ("ibmvnic: Introduce hard reset recovery") Signed-off-by: Sukadev Bhattiprolu <sukadev@linux.ibm.com> Signed-off-by: Jakub Kicinski <kuba@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
When ibmvnic fails to reset, it breaks out of the reset loop and frees
all of the remaining resets from the workqueue. Doing so prevents the
adapter from recovering if no reset is scheduled after that. Instead,
have the driver continue to process resets on the workqueue.
Inconsistent login with the vnicserver is causing the device to be
removed. This does not give the device a chance to recover from error
state. This patch schedules a FATAL reset instead to bring the adapter
up.
Fixes: 032c5e82847a2 ("Driver for IBM System i/p VNIC protocol") Signed-off-by: Dany Madden <drt@linux.ibm.com> Signed-off-by: Lijun Pan <ljp@linux.ibm.com> Signed-off-by: Jakub Kicinski <kuba@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
In the error path of ip_vs_control_net_init(), remove_proc_entry() needs
to be called to remove the added proc entry, otherwise a memory leak
will occur.
Also, add some '#ifdef CONFIG_PROC_FS' because proc_create_net* return NULL
when PROC is not used.
Fixes: b17fc9963f83 ("IPVS: netns, ip_vs_stats and its procfs") Fixes: 61b1ab4583e2 ("IPVS: netns, add basic init per netns.") Reported-by: Hulk Robot <hulkci@huawei.com> Signed-off-by: Wang Hai <wanghai38@huawei.com> Acked-by: Julian Anastasov <ja@ssi.bg> Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
When a packet is fragmented by batman-adv, the original batman-adv header
is not modified. Only a new fragmentation is inserted between the original
one and the ethernet header. The code must therefore make sure that it has
a writable region of this size in the skbuff head.
But it is not useful to always reallocate the skbuff by this size even when
there would be more than enough headroom still in the skb. The reallocation
is just to costly during in this codepath.
Fixes: ee75ed88879a ("batman-adv: Fragment and send skbs larger than mtu") Signed-off-by: Sven Eckelmann <sven@narfation.org> Signed-off-by: Simon Wunderlich <sw@simonwunderlich.de> Signed-off-by: Sasha Levin <sashal@kernel.org>
The batadv net_device is trying to propagate the needed_headroom and
needed_tailroom from the lower devices. This is needed to avoid cost
intensive reallocations using pskb_expand_head during the transmission.
But the fragmentation code split the skb's without adding extra room at the
end/beginning of the various fragments. This reduced the performance of
transmissions over complex scenarios (batadv on vxlan on wireguard) because
the lower devices had to perform the reallocations at least once.
Fixes: ee75ed88879a ("batman-adv: Fragment and send skbs larger than mtu") Signed-off-by: Sven Eckelmann <sven@narfation.org> Signed-off-by: Simon Wunderlich <sw@simonwunderlich.de> Signed-off-by: Sasha Levin <sashal@kernel.org>
If a batman-adv packets has to be fragmented, then the original batman-adv
packet header is not stripped away. Instead, only a new header is added in
front of the packet after it was split.
This size must be considered to avoid cost intensive reallocations during
the transmission through the various device layers.
Fixes: 7bca68c7844b ("batman-adv: Add lower layer needed_(head|tail)room to own ones") Reported-by: Linus Lüssing <linus.luessing@c0d3.blue> Signed-off-by: Sven Eckelmann <sven@narfation.org> Signed-off-by: Simon Wunderlich <sw@simonwunderlich.de> Signed-off-by: Sasha Levin <sashal@kernel.org>
Pinned pages are not properly accounted particularly when
mapping error occurs on IOTLB update. Clean up dangling
pinned pages for the error path.
The memory usage for bookkeeping pinned pages is reverted
to what it was before: only one single free page is needed.
This helps reduce the host memory demand for VM with a large
amount of memory, or in the situation where host is running
short of free memory.
Fixes: 4c8cf31885f6 ("vhost: introduce vDPA-based backend") Signed-off-by: Si-Wei Liu <si-wei.liu@oracle.com> Link: https://lore.kernel.org/r/1604618793-4681-1-git-send-email-si-wei.liu@oracle.com Signed-off-by: Michael S. Tsirkin <mst@redhat.com> Acked-by: Jason Wang <jasowang@redhat.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
Commit 642e450b6b59 ("xsk: Do not discard packet when NETDEV_TX_BUSY")
addressed the problem that packets were discarded from the Tx AF_XDP
ring, when the driver returned NETDEV_TX_BUSY. Part of the fix was
bumping the skbuff reference count, so that the buffer would not be
freed by dev_direct_xmit(). A reference count larger than one means
that the skbuff is "shared", which is not the case.
If the "shared" skbuff is sent to the generic XDP receive path,
netif_receive_generic_xdp(), and pskb_expand_head() is entered the
BUG_ON(skb_shared(skb)) will trigger.
This patch adds a variant to dev_direct_xmit(), __dev_direct_xmit(),
where a user can select the skbuff free policy. This allows AF_XDP to
avoid bumping the reference count, but still keep the NETDEV_TX_BUSY
behavior.
Fixes: 642e450b6b59 ("xsk: Do not discard packet when NETDEV_TX_BUSY") Reported-by: Yonghong Song <yhs@fb.com> Signed-off-by: Björn Töpel <bjorn.topel@intel.com> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Link: https://lore.kernel.org/bpf/20201123175600.146255-1-bjorn.topel@gmail.com Signed-off-by: Sasha Levin <sashal@kernel.org>
Fix to check the write(2) failure including partial write
correctly and try to rollback the partial write, because
if there is no BOOTCONFIG_MAGIC string, we can not remove it.
Link: https://lkml.kernel.org/r/160576521135.320071.3883101436675969998.stgit@devnote2 Fixes: 85c46b78da58 ("bootconfig: Add bootconfig magic word for indicating bootconfig explicitly") Suggested-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Masami Hiramatsu <mhiramat@kernel.org> Signed-off-by: Steven Rostedt (VMware) <rostedt@goodmis.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
Variant 1 needs the exact same quirk as variant 2, so relax the DMI check
for the existing quirk a bit so that it matches both variant 1 and 2
(note the other variants will still not match).
Variant 2 already has an existing quirk (which now also matches variant 1)
Variant 3 uses a cx2072x codec, so is not applicable here.
Variant 4 almost works with the defaults, but it also needs a quirk to
fix jack-detection, add a new quirk for this.
Variant 5 does use a RT5640 codec (based on old dmesg output), but was
otherwise not tested, keep using the defaults for this variant.
Fixes: ec8e8418ff7d ("ASoC: Intel: bytcr_rt5640: Add quirks for various devices") Signed-off-by: Hans de Goede <hdegoede@redhat.com> Acked-by: Pierre-Louis Bossart <pierre-louis.bossart@linux.intel.com> Link: https://lore.kernel.org/r/20201118121515.11441-1-hdegoede@redhat.com Signed-off-by: Mark Brown <broonie@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
Since the kprobe handlers have protection that prohibits other handlers from
executing in other contexts (like if an NMI comes in while processing a
kprobe, and executes the same kprobe, it will get fail with a "busy"
return). Lockdep is unaware of this protection. Use lockdep's nesting api to
differentiate between locks taken in INT3 context and other context to
suppress the false warnings.
While I was doing zram testing, I found sometimes decompression failed
since the compression buffer was corrupted. With investigation, I found
below commit calls cond_resched unconditionally so it could make a
problem in atomic context if the task is reschedule.
BUG: sleeping function called from invalid context at mm/vmalloc.c:108
in_atomic(): 1, irqs_disabled(): 0, non_block: 0, pid: 946, name: memhog
3 locks held by memhog/946:
#0: ffff9d01d4b193e8 (&mm->mmap_lock#2){++++}-{4:4}, at: __mm_populate+0x103/0x160
#1: ffffffffa3d53de0 (fs_reclaim){+.+.}-{0:0}, at: __alloc_pages_slowpath.constprop.0+0xa98/0x1160
#2: ffff9d01d56b8110 (&zspage->lock){.+.+}-{3:3}, at: zs_map_object+0x8e/0x1f0
CPU: 0 PID: 946 Comm: memhog Not tainted 5.9.3-00011-gc5bfc0287345-dirty #316
Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS 1.13.0-1 04/01/2014
Call Trace:
unmap_kernel_range_noflush+0x2eb/0x350
unmap_kernel_range+0x14/0x30
zs_unmap_object+0xd5/0xe0
zram_bvec_rw.isra.0+0x38c/0x8e0
zram_rw_page+0x90/0x101
bdev_write_page+0x92/0xe0
__swap_writepage+0x94/0x4a0
pageout+0xe3/0x3a0
shrink_page_list+0xb94/0xd60
shrink_inactive_list+0x158/0x460
We can fix this by removing the ZSMALLOC_PGTABLE_MAPPING feature (which
contains the offending calling code) from zsmalloc.
Even though this option showed some amount improvement(e.g., 30%) in
some arm32 platforms, it has been headache to maintain since it have
abused APIs[1](e.g., unmap_kernel_range in atomic context).
Since we are approaching to deprecate 32bit machines and already made
the config option available for only builtin build since v5.8, lastly it
has been not default option in zsmalloc, it's time to drop the option
for better maintenance.
Clang's integrated assembler produces the warning for assembly files:
warning: DWARF2 only supports one section per compilation unit
If -Wa,-gdwarf-* is unspecified, then debug info is not emitted for
assembly sources (it is still emitted for C sources). This will be
re-enabled for newer DWARF versions in a follow up patch.
Enables defconfig+CONFIG_DEBUG_INFO to build cleanly with
LLVM=1 LLVM_IAS=1 for x86_64 and arm64.
The 64-bit signed bounds should not affect 32-bit signed bounds unless the
verifier knows that upper 32-bits are either all 1s or all 0s. For example the
register with smin_value==1 doesn't mean that s32_min_value is also equal to 1,
since smax_value could be larger than 32-bit subregister can hold.
The verifier refines the smax/s32_max return value from certain helpers in
do_refine_retval_range(). Teach the verifier to recognize that smin/s32_min
value is also bounded. When both smin and smax bounds fit into 32-bit
subregister the verifier can propagate those bounds.
This reverts commit 4179b00c04d1 ("geneve: pull IP header before ECN decapsulation").
Eric says: "network header should have been pulled already before
hitting geneve_rx()". Let's revert the syzbot fix since it's causing
more harm than good, and revisit.
Userspace might match on prefix bytes of header fields if they are on
the byte boundary, this requires that the mask is adjusted accordingly.
Use NFT_OFFLOAD_MATCH_EXACT() for meta since prefix byte matching is not
allowed for this type of selector.
The bitwise expression might be optimized out by userspace, hence the
kernel needs to infer the prefix from the number of payload bytes to
match on. This patch adds nft_payload_offload_mask() to calculate the
bitmask to match on the prefix.
There are reports wrt lockdep splat in nftables, e.g.:
------------[ cut here ]------------
WARNING: CPU: 2 PID: 31416 at net/netfilter/nf_tables_api.c:622
lockdep_nfnl_nft_mutex_not_held+0x28/0x38 [nf_tables]
...
These are caused by an earlier, unrelated bug such as a n ABBA deadlock
in a different subsystem.
In such an event, lockdep is disabled and lockdep_is_held returns true
unconditionally. This then causes the WARN() in nf_tables.
Make the WARN conditional on lockdep still active to avoid this.
Building on arch/s390/ results in this build error:
cc1: some warnings being treated as errors
../drivers/md/dm-writecache.c: In function 'persistent_memory_claim':
../drivers/md/dm-writecache.c:323:1: error: no return statement in function returning non-void [-Werror=return-type]
Fix this by replacing the BUG() with an -EOPNOTSUPP return.
Fixes: 48debafe4f2f ("dm: add writecache target") Reported-by: Randy Dunlap <rdunlap@infradead.org> Signed-off-by: Mike Snitzer <snitzer@redhat.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
During cci_isr() errors read from register fields belonging to
i2c master1 are currently assigned to the status field belonging to
i2c master0. This patch corrects this error, and always assigns
master1 errors to the status field of master1.
This code does not ensure that the whole buffer is initialized and none
of the callers check for errors so potentially none of the buffer is
initialized. Add a memset to eliminate this bug.
Fixes: e3037485c68e ("rtw88: new Realtek 802.11ac driver") Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com> Signed-off-by: Kalle Valo <kvalo@codeaurora.org> Link: https://lore.kernel.org/r/X8ilOfVz3pf0T5ec@mwanda Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
GFS2's freeze/thaw mechanism uses a special freeze glock to control its
operation. It does this with a sync glock operation (glops.c) called
freeze_go_sync. When the freeze glock is demoted (glock's do_xmote) the
glops function causes the file system to be frozen. This is intended. However,
GFS2's mount and unmount processes also hold the freeze glock to prevent other
processes, perhaps on different cluster nodes, from mounting the frozen file
system in read-write mode.
Before this patch, there was no check in freeze_go_sync for whether a freeze
in intended or whether the glock demote was caused by a normal unmount.
So it was trying to freeze the file system it's trying to unmount, which
ends up in a deadlock.
This patch adds an additional check to freeze_go_sync so that demotes of the
freeze glock are ignored if they come from the unmount process.
Fixes: 20b329129009 ("gfs2: Fix regression in freeze_go_sync") Signed-off-by: Bob Peterson <rpeterso@redhat.com> Signed-off-by: Andreas Gruenbacher <agruenba@redhat.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Commit 0e539ca1bbbe ("gfs2: Fix NULL pointer dereference in gfs2_rgrp_dump")
introduced additional locking in gfs2_rgrp_go_dump, which is also used for
dumping resource group glocks via debugfs. However, on that code path, the
glock spin lock is already taken in dump_glock, and taking it again in
gfs2_glock2rgrp leads to deadlock. This can be reproduced with:
Fix to return a negative error code from the error handling case
instead of 0 in function wm_adsp_load(), as done elsewhere in this
function.
Fixes: 170b1e123f38 ("ASoC: wm_adsp: Add support for new Halo core DSPs") Reported-by: Hulk Robot <hulkci@huawei.com> Signed-off-by: Luo Meng <luomeng12@huawei.com> Acked-by: Richard Fitzgerald <rf@opensource.cirrus.com> Link: https://lore.kernel.org/r/20201123133839.4073787-1-luomeng12@huawei.com Signed-off-by: Mark Brown <broonie@kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
In the commit fdeba99b1e58
("tipc: fix use-after-free in tipc_bcast_get_mode"), we're trying
to make sure the tipc_net_finalize_work work item finished if it
enqueued. But calling flush_scheduled_work() is not just affecting
above work item but either any scheduled work. This has turned out
to be overkill and caused to deadlock as syzbot reported:
======================================================
WARNING: possible circular locking dependency detected
5.9.0-rc2-next-20200828-syzkaller #0 Not tainted
------------------------------------------------------
kworker/u4:6/349 is trying to acquire lock: ffff8880aa063d38 ((wq_completion)events){+.+.}-{0:0}, at: flush_workqueue+0xe1/0x13e0 kernel/workqueue.c:2777
but task is already holding lock: ffffffff8a879430 (pernet_ops_rwsem){++++}-{3:3}, at: cleanup_net+0x9b/0xb10 net/core/net_namespace.c:565
v1:
To fix the original issue, we replace above calling by introducing
a bit flag. When a namespace cleaned-up, bit flag is set to zero and:
- tipc_net_finalize functionial just does return immediately.
- tipc_net_finalize_work does not enqueue into the scheduled work queue.
v2:
Use cancel_work_sync() helper to make sure ONLY the
tipc_net_finalize_work() stopped before releasing bcbase object.
Reported-by: syzbot+d5aa7e0385f6a5d0f4fd@syzkaller.appspotmail.com Fixes: fdeba99b1e58 ("tipc: fix use-after-free in tipc_bcast_get_mode") Acked-by: Jon Maloy <jmaloy@redhat.com> Signed-off-by: Hoang Huu Le <hoang.h.le@dektech.com.au> Signed-off-by: Jakub Kicinski <kuba@kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
If gfs2 tries to mount a (corrupt) file system that has no resource
groups it still tries to set preferences on the first one, which causes
a kernel null pointer dereference. This patch adds a check to function
gfs2_ri_update so this condition is detected and reported back as an
error.
Reported-by: syzbot+e3f23ce40269a4c9053a@syzkaller.appspotmail.com Signed-off-by: Bob Peterson <rpeterso@redhat.com> Signed-off-by: Andreas Gruenbacher <agruenba@redhat.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
To detect potential bugs in CAN protocol implementations (double removal of
receiver entries) a WARN() statement has been used if no matching list item was
found for removal.
The fault injection issued by syzkaller was able to create a situation where
the closing of a socket runs simultaneously to the notifier call chain for
removing the CAN network device in use.
This case is very unlikely in real life but it doesn't break anything.
Therefore we just replace the WARN() statement with pr_warn() to preserve the
notification for the CAN protocol development.
Reported-by: syzbot+381d06e0c8eaacb8706f@syzkaller.appspotmail.com Reported-by: syzbot+d0ddd88c9a7432f041e6@syzkaller.appspotmail.com Reported-by: syzbot+76d62d3b8162883c7d11@syzkaller.appspotmail.com Signed-off-by: Oliver Hartkopp <socketcan@hartkopp.net> Link: https://lore.kernel.org/r/20201126192140.14350-1-socketcan@hartkopp.net Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Lilith >_> and Claudio Bozzato of Cisco Talos security team reported
that collect_syscall() improperly casts the syscall registers to 64-bit
values leaking the uninitialized last 24 bytes on 32-bit platforms, that
are visible in /proc/self/syscall.
The cause is that info->data.args are u64 while syscall_get_arguments()
uses longs, as hinted by the bogus pointer cast in the function.
Let's just proceed like the other call places, by retrieving the
registers into an array of longs before assigning them to the caller's
array. This was successfully tested on x86_64, i386 and ppc32.
Commit 10befea91b61 ("mm: memcg/slab: use a single set of kmem_caches
for all allocations") introduced a regression into the handling of the
obj_cgroup_charge() return value. If a non-zero value is returned
(indicating of exceeding one of memory.max limits), the allocation
should fail, instead of falling back to non-accounted mode.
To make the code more readable, move memcg_slab_pre_alloc_hook() and
memcg_slab_post_alloc_hook() calling conditions into bodies of these
hooks.
Fixes: 10befea91b61 ("mm: memcg/slab: use a single set of kmem_caches for all allocations") Signed-off-by: Roman Gushchin <guro@fb.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Reviewed-by: Shakeel Butt <shakeelb@google.com> Cc: Johannes Weiner <hannes@cmpxchg.org> Cc: Michal Hocko <mhocko@kernel.org> Cc: <stable@vger.kernel.org> Link: https://lkml.kernel.org/r/20201127161828.GD840171@carbon.dhcp.thefacebook.com Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
According to the AMD IOMMU spec, the commit 73db2fc595f3
("iommu/amd: Increase interrupt remapping table limit to 512 entries")
also requires the interrupt table length (IntTabLen) to be set to 9
(power of 2) in the device table mapping entry (DTE).
Fixes: 73db2fc595f3 ("iommu/amd: Increase interrupt remapping table limit to 512 entries") Reported-by: Jerry Snitselaar <jsnitsel@redhat.com> Signed-off-by: Suravee Suthikulpanit <suravee.suthikulpanit@amd.com> Reviewed-by: Jerry Snitselaar <jsnitsel@redhat.com> Link: https://lore.kernel.org/r/20201207091920.3052-1-suravee.suthikulpanit@amd.com Signed-off-by: Will Deacon <will@kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Adrian Moreno was ruuning a kubernetes 1.19 + containerd/docker workload
using hugetlbfs. In this environment the issue is reproduced by:
- Start a simple pod that uses the recently added HugePages medium
feature (pod yaml attached)
- Start a DPDK app. It doesn't need to run successfully (as in transfer
packets) nor interact with real hardware. It seems just initializing
the EAL layer (which handles hugepage reservation and locking) is
enough to trigger the issue
- Delete the Pod (or let it "Complete").
This would result in a kworker thread going into a tight loop (top output):
We are spinning in the do-while loop in hugetlb_cgroup_css_offline.
Worse yet, we are holding the master cgroup lock (cgroup_mutex) while
infinitely spinning. Little else can be done on the system as the
cgroup_mutex can not be acquired.
Do note that the issue can be reproduced by simply offlining a hugetlb
cgroup containing pages with reservation counts.
The loop in hugetlb_cgroup_css_offline is moving page counts from the
cgroup being offlined to the parent cgroup. This is done for each
hstate, and is repeated until hugetlb_cgroup_have_usage returns false.
The routine moving counts (hugetlb_cgroup_move_parent) is only moving
'usage' counts. The routine hugetlb_cgroup_have_usage is checking for
both 'usage' and 'reservation' counts. Discussion about what to do with
reservation counts when reparenting was discussed here:
The decision was made to leave a zombie cgroup for with reservation
counts. Unfortunately, the code checking reservation counts was
incorrectly added to hugetlb_cgroup_have_usage.
To fix the issue, simply remove the check for reservation counts. While
fixing this issue, a related bug in hugetlb_cgroup_css_offline was
noticed. The hstate index is not reinitialized each time through the
do-while loop. Fix this as well.
Fixes: 1adc4d419aa2 ("hugetlb_cgroup: add interface for charge/uncharge hugetlb reservations") Reported-by: Adrian Moreno <amorenoz@redhat.com> Signed-off-by: Mike Kravetz <mike.kravetz@oracle.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Tested-by: Adrian Moreno <amorenoz@redhat.com> Reviewed-by: Shakeel Butt <shakeelb@google.com> Cc: Mina Almasry <almasrymina@google.com> Cc: David Rientjes <rientjes@google.com> Cc: Greg Thelen <gthelen@google.com> Cc: Sandipan Das <sandipan@linux.ibm.com> Cc: Shuah Khan <shuah@kernel.org> Cc: <stable@vger.kernel.org> Link: https://lkml.kernel.org/r/20201203220242.158165-1-mike.kravetz@oracle.com Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
When investigating a slab cache bloat problem, significant amount of
negative dentry cache was seen, but confusingly they neither got shrunk
by reclaimer (the host has very tight memory) nor be shrunk by dropping
cache. The vmcore shows there are over 14M negative dentry objects on
lru, but tracing result shows they were even not scanned at all.
Further investigation shows the memcg's vfs shrinker_map bit is not set.
So the reclaimer or dropping cache just skip calling vfs shrinker. So
we have to reboot the hosts to get the memory back.
I didn't manage to come up with a reproducer in test environment, and
the problem can't be reproduced after rebooting. But it seems there is
race between shrinker map bit clear and reparenting by code inspection.
The hypothesis is elaborated as below.
The memcg hierarchy on our production environment looks like:
root
/ \
system user
The main workloads are running under user slice's children, and it
creates and removes memcg frequently. So reparenting happens very often
under user slice, but no task is under user slice directly.
So with the frequent reparenting and tight memory pressure, the below
hypothetical race condition may happen:
CPU A CPU B
reparent
dst->nr_items == 0
shrinker:
total_objects == 0
add src->nr_items to dst
set_bit
return SHRINK_EMPTY
clear_bit
child memcg offline
replace child's kmemcg_id with
parent's (in memcg_offline_kmem())
list_lru_del() between shrinker runs
see parent's kmemcg_id
dec dst->nr_items
reparent again
dst->nr_items may go negative
due to concurrent list_lru_del()
The second run of shrinker:
read nr_items without any
synchronization, so it may
see intermediate negative
nr_items then total_objects
may return 0 coincidently
keep the bit cleared
dst->nr_items != 0
skip set_bit
add scr->nr_item to dst
After this point dst->nr_item may never go zero, so reparenting will not
set shrinker_map bit anymore. And since there is no task under user
slice directly, so no new object will be added to its lru to set the
shrinker map bit either. That bit is kept cleared forever.
How does list_lru_del() race with reparenting? It is because reparenting
replaces children's kmemcg_id to parent's without protecting from
nlru->lock, so list_lru_del() may see parent's kmemcg_id but actually
deleting items from child's lru, but dec'ing parent's nr_items, so the
parent's nr_items may go negative as commit 2788cf0c401c ("memcg:
reparent list_lrus and free kmemcg_id on css offline") says.
Since it is impossible that dst->nr_items goes negative and
src->nr_items goes zero at the same time, so it seems we could set the
shrinker map bit iff src->nr_items != 0. We could synchronize
list_lru_count_one() and reparenting with nlru->lock, but it seems
checking src->nr_items in reparenting is the simplest and avoids lock
contention.
Fixes: fae91d6d8be5 ("mm/list_lru.c: set bit in memcg shrinker bitmap on first list_lru item appearance") Suggested-by: Roman Gushchin <guro@fb.com> Signed-off-by: Yang Shi <shy828301@gmail.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Reviewed-by: Roman Gushchin <guro@fb.com> Reviewed-by: Shakeel Butt <shakeelb@google.com> Acked-by: Kirill Tkhai <ktkhai@virtuozzo.com> Cc: Vladimir Davydov <vdavydov.dev@gmail.com> Cc: <stable@vger.kernel.org> [4.19] Link: https://lkml.kernel.org/r/20201202171749.264354-1-shy828301@gmail.com Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
'format_corename()' will splite 'core_pattern' on spaces when it is in
pipe mode, and take helper_argv[0] as the path to usermode executable.
It works fine in most cases.
However, if there is a space between '|' and '/file/path', such as
'| /usr/lib/systemd/systemd-coredump %P %u %g', then helper_argv[0] will
be parsed as '', and users will get a 'Core dump to | disabled'.
It is not friendly to users, as the pattern above was valid previously.
Fix this by ignoring the spaces between '|' and '/file/path'.
Fixes: 315c69261dd3 ("coredump: split pipe command whitespace before expanding template") Signed-off-by: Menglong Dong <dong.menglong@zte.com.cn> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Cc: Paul Wise <pabs3@bonedaddy.net> Cc: Jakub Wilk <jwilk@jwilk.net> [https://bugs.debian.org/924398] Cc: Neil Horman <nhorman@tuxdriver.com> Cc: <stable@vger.kernel.org> Link: https://lkml.kernel.org/r/5fb62870.1c69fb81.8ef5d.af76@mx.google.com Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Fixes sparse warnings:
drivers/md/dm.c:508:12: warning: context imbalance in 'dm_prepare_ioctl' - wrong count at exit
drivers/md/dm.c:543:13: warning: context imbalance in 'dm_unprepare_ioctl' - wrong count at exit
Fixes: 971888c46993f ("dm: hold DM table for duration of ioctl rather than use blkdev_get") Cc: stable@vger.kernel.org Signed-off-by: Mike Snitzer <snitzer@redhat.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
With virtio multiqueue, normally each queue IRQ is mapped to a CPU.
Commit 0d9f0a52c8b9f ("virtio_scsi: use virtio IRQ affinity") exposed
an existing shortcoming of the arch code by moving virtio_scsi to
the automatic IRQ affinity assignment.
The affinity is correctly computed in msi_desc but this is not applied
to the system IRQs.
It appears the affinity is correctly passed to rtas_setup_msi_irqs() but
lost at this point and never passed to irq_domain_alloc_descs()
(see commit 06ee6d571f0e ("genirq: Add affinity hint to irq allocation"))
because irq_create_mapping() doesn't take an affinity parameter.
Use the new irq_create_mapping_affinity() function, which allows to forward
the affinity setting from rtas_setup_msi_irqs() to irq_domain_alloc_descs().
With this change, the virtqueues are correctly dispatched between the CPUs
on pseries.
Fixes: e75eafb9b039 ("genirq/msi: Switch to new irq spreading infrastructure") Signed-off-by: Laurent Vivier <lvivier@redhat.com> Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Reviewed-by: Greg Kurz <groug@kaod.org> Acked-by: Michael Ellerman <mpe@ellerman.id.au> Cc: stable@vger.kernel.org Link: https://lore.kernel.org/r/20201126082852.1178497-3-lvivier@redhat.com Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
There is currently no way to convey the affinity of an interrupt
via irq_create_mapping(), which creates issues for devices that
expect that affinity to be managed by the kernel.
In order to sort this out, rename irq_create_mapping() to
irq_create_mapping_affinity() with an additional affinity parameter that
can be passed down to irq_domain_alloc_descs().
irq_create_mapping() is re-implemented as a wrapper around
irq_create_mapping_affinity().
No functional change.
Fixes: e75eafb9b039 ("genirq/msi: Switch to new irq spreading infrastructure") Signed-off-by: Laurent Vivier <lvivier@redhat.com> Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Reviewed-by: Greg Kurz <groug@kaod.org> Cc: Michael Ellerman <mpe@ellerman.id.au> Cc: stable@vger.kernel.org Link: https://lore.kernel.org/r/20201126082852.1178497-2-lvivier@redhat.com Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
__io_compat_recvmsg_copy_hdr() with REQ_F_BUFFER_SELECT reads out iov
len but never assigns it to iov/fast_iov, leaving sr->len with garbage.
Hopefully, following io_buffer_select() truncates it to the selected
buffer size, but the value is still may be under what was specified.
projects / thirdparty / kernel / stable.git / log
