Checking for the lack of epitems refering to the epoll we want to insert into
is not enough; we might have an insertion of that epoll into another one that
has already collected the set of files to recheck for excessive reverse paths,
but hasn't gotten to creating/inserting the epitem for it.
However, any such insertion in progress can be detected - it will update the
generation count in our epoll when it's done looking through it for files
to check. That gets done under ->mtx of our epoll and that allows us to
detect that safely.
We are *not* holding epmutex here, so the generation count is not stable.
However, since both the update of ep->gen by loop check and (later)
insertion into ->f_ep_link are done with ep->mtx held, we are fine -
the sequence is
grab epmutex
bump loop_check_gen
...
grab tep->mtx // 1
tep->gen = loop_check_gen
...
drop tep->mtx // 2
...
grab tep->mtx // 3
...
insert into ->f_ep_link
...
drop tep->mtx // 4
bump loop_check_gen
drop epmutex
and if the fastpath check in another thread happens for that
eventpoll, it can come
* before (1) - in that case fastpath is just fine
* after (4) - we'll see non-empty ->f_ep_link, slow path
taken
* between (2) and (3) - loop_check_gen is stable,
with ->mtx providing barriers and we end up taking slow path.
Note that ->f_ep_link emptiness check is slightly racy - we are protected
against insertions into that list, but removals can happen right under us.
Not a problem - in the worst case we'll end up taking a slow path for
no good reason.
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Make sure to call sd_zbc_init_disk() when the sdkp->zoned field is known,
that is, once sd_read_block_characteristics() is executed in
sd_revalidate_disk(), so that host-aware disks also get initialized. To do
so, move sd_zbc_init_disk() call in sd_zbc_revalidate_zones() and make sure
to execute it for all zoned disks, including for host-aware disks used as
regular disks as these disk zoned model may be changed back to BLK_ZONED_HA
when partitions are deleted.
Link: https://lore.kernel.org/r/20200915073347.832424-3-damien.lemoal@wdc.com Fixes: 5795eb443060 ("scsi: sd_zbc: emulate ZONE_APPEND commands") Cc: <stable@vger.kernel.org> # v5.8+ Reported-by: Borislav Petkov <bp@alien8.de> Tested-by: Borislav Petkov <bp@suse.de> Reviewed-by: Johannes Thumshirn <johannes.thumshirn@wdc.com> Reviewed-by: Christoph Hellwig <hch@lst.de> Signed-off-by: Damien Le Moal <damien.lemoal@wdc.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
When CONFIG_BLK_DEV_ZONED is disabled, allow using host-aware ZBC disks as
regular disks. In this case, ensure that command completion is correctly
executed by changing sd_zbc_complete() to return good_bytes instead of 0
and causing a hang during device probe (endless retries).
When CONFIG_BLK_DEV_ZONED is enabled and a host-aware disk is detected to
have partitions, it will be used as a regular disk. In this case, make sure
to not do anything in sd_zbc_revalidate_zones() as that triggers warnings.
Since all these different cases result in subtle settings of the disk queue
zoned model, introduce the block layer helper function
blk_queue_set_zoned() to generically implement setting up the effective
zoned model according to the disk type, the presence of partitions on the
disk and CONFIG_BLK_DEV_ZONED configuration.
Link: https://lore.kernel.org/r/20200915073347.832424-2-damien.lemoal@wdc.com Fixes: b72053072c0b ("block: allow partitions on host aware zone devices") Cc: <stable@vger.kernel.org> Reported-by: Borislav Petkov <bp@alien8.de> Suggested-by: Christoph Hellwig <hch@infradead.org> Reviewed-by: Christoph Hellwig <hch@lst.de> Reviewed-by: Johannes Thumshirn <johannes.thumshirn@wdc.com> Signed-off-by: Damien Le Moal <damien.lemoal@wdc.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Current BDW virtual display port is initialized as PORT_B, so need
to use same port for VFIO EDID region, otherwise invalid EDID blob
pointer is assigned which caused kernel null pointer reference. We
might evaluate actual display hotplug for BDW to make this function
work as expected, anyway this is always required to be fixed first.
Reported-by: Alejandro Sior <aho@sior.be> Cc: Alejandro Sior <aho@sior.be> Fixes: 0178f4ce3c3b ("drm/i915/gvt: Enable vfio edid for all GVT supported platform") Reviewed-by: Hang Yuan <hang.yuan@intel.com> Signed-off-by: Zhenyu Wang <zhenyuw@linux.intel.com> Link: http://patchwork.freedesktop.org/patch/msgid/20200914030302.2775505-1-zhenyuw@linux.intel.com Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The introduced line event handling ABI in the commit
61f922db7221 ("gpio: userspace ABI for reading GPIO line events")
missed the fact that 64-bit kernel may serve for 32-bit applications.
In such case the very first check in the lineevent_read() will fail
due to alignment differences.
To workaround this introduce lineevent_get_size() helper which returns actual
size of the structure in user space.
Fixes: 61f922db7221 ("gpio: userspace ABI for reading GPIO line events") Suggested-by: Arnd Bergmann <arnd@arndb.de> Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com> Acked-by: Arnd Bergmann <arnd@arndb.de> Tested-by: Kent Gibson <warthog618@gmail.com> Signed-off-by: Bartosz Golaszewski <bgolaszewski@baylibre.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
Commit f227e3ec3b5c ("random32: update the net random state on interrupt
and activity") broke compilation and was temporarily fixed by Linus in 83bdc7275e62 ("random32: remove net_rand_state from the latent entropy
gcc plugin") by entirely moving net_rand_state out of the things handled
by the latent_entropy GCC plugin.
From what I understand when reading the plugin code, using the
__latent_entropy attribute on a declaration was the wrong part and
simply keeping the __latent_entropy attribute on the variable definition
was the correct fix.
The pipe splice code still used the old model of waiting for pipe IO by
using a non-specific "pipe_wait()" that waited for any pipe event to
happen, which depended on all pipe IO being entirely serialized by the
pipe lock. So by checking the state you were waiting for, and then
adding yourself to the wait queue before dropping the lock, you were
guaranteed to see all the wakeups.
Strictly speaking, the actual wakeups were not done under the lock, but
the pipe_wait() model still worked, because since the waiter held the
lock when checking whether it should sleep, it would always see the
current state, and the wakeup was always done after updating the state.
However, commit 0ddad21d3e99 ("pipe: use exclusive waits when reading or
writing") split the single wait-queue into two, and in the process also
made the "wait for event" code wait for _two_ wait queues, and that then
showed a race with the wakers that were not serialized by the pipe lock.
It's only splice that used that "pipe_wait()" model, so the problem
wasn't obvious, but Josef Bacik reports:
"I hit a hang with fstest btrfs/187, which does a btrfs send into
/dev/null. This works by creating a pipe, the write side is given to
the kernel to write into, and the read side is handed to a thread that
splices into a file, in this case /dev/null.
The box that was hung had the write side stuck here [pipe_write] and
the read side stuck here [splice_from_pipe_next -> pipe_wait].
[ more details about pipe_wait() scenario ]
The problem is we're doing the prepare_to_wait, which sets our state
each time, however we can be woken up either with reads or writes. In
the case above we race with the WRITER waking us up, and re-set our
state to INTERRUPTIBLE, and thus never break out of schedule"
Josef had a patch that avoided the issue in pipe_wait() by just making
it set the state only once, but the deeper problem is that pipe_wait()
depends on a level of synchonization by the pipe mutex that it really
shouldn't. And the whole "wait for any pipe state change" model really
isn't very good to begin with.
So rather than trying to work around things in pipe_wait(), remove that
legacy model of "wait for arbitrary pipe event" entirely, and actually
create functions that wait for the pipe actually being readable or
writable, and can do so without depending on the pipe lock serializing
everything.
Commit 387caf0b759a ("iommu/amd: Treat per-device exclusion
ranges as r/w unity-mapped regions") accidentally overwrites
the 'flags' field in IVMD (struct ivmd_header) when the I/O
virtualization memory definition is associated with the
exclusion range entry. This leads to the corrupted IVMD table
(incorrect checksum). The kdump kernel reports the invalid checksum:
ACPI BIOS Warning (bug): Incorrect checksum in table [IVRS] - 0x5C, should be 0x60 (20200717/tbprint-177)
AMD-Vi: [Firmware Bug]: IVRS invalid checksum
Fix the above-mentioned issue by modifying the 'struct unity_map_entry'
member instead of the IVMD header.
Cleanup: The *exclusion_range* functions are not used anymore, so
get rid of them.
Fixes: 387caf0b759a ("iommu/amd: Treat per-device exclusion ranges as r/w unity-mapped regions") Reported-and-tested-by: Baoquan He <bhe@redhat.com> Signed-off-by: Adrian Huang <ahuang12@lenovo.com> Cc: Jerry Snitselaar <jsnitsel@redhat.com> Link: https://lore.kernel.org/r/20200926102602.19177-1-adrianhuang0701@gmail.com Signed-off-by: Joerg Roedel <jroedel@suse.de> Signed-off-by: Sasha Levin <sashal@kernel.org>
When driver has been converted to the bitmap API the non-bitmap functions
started behaving differently on 32-bit BE architectures since the bytes in
two consequent unsigned longs are in different order in comparison to byte
array. Hence if the chip had had more than 32 lines the memset() call over
it would have not set up upper lines correctly.
Although it's currently a theoretical case (no supported chips of this type
has 32+ lines), it's better to provide a clean code to avoid people thinking
this is okay and potentially producing not fully working things.
Fixes: 35d13d94893f ("gpio: pca953x: convert to use bitmap API") Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com> Reviewed-by: Bartosz Golaszewski <bgolaszewski@baylibre.com> Link: https://lore.kernel.org/r/20200930142013.59247-2-andriy.shevchenko@linux.intel.com Signed-off-by: Linus Walleij <linus.walleij@linaro.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
check mtk_is_virt_gpio input parameter,
virtual gpio need to support eint mode.
add error handler for the ko case
to fix this boot fail:
pc : mtk_is_virt_gpio+0x20/0x38 [pinctrl_mtk_common_v2]
lr : mtk_gpio_get_direction+0x44/0xb0 [pinctrl_paris]
Fixes: edd546465002 ("pinctrl: mediatek: avoid virtual gpio trying to set reg") Signed-off-by: Hanks Chen <hanks.chen@mediatek.com> Acked-by: Sean Wang <sean.wang@kernel.org> Singed-off-by: Jie Yang <sin_jieyang@mediatek.com> Link: https://lore.kernel.org/r/1597922546-29633-1-git-send-email-hanks.chen@mediatek.com Signed-off-by: Linus Walleij <linus.walleij@linaro.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
autofs got broken in some configurations by commit 13c164b1a186
("autofs: switch to kernel_write") because there is now an extra LSM
permission check done by security_file_permission() in rw_verify_area().
autofs is one if the few places that really does want the much more
limited __kernel_write(), because the write is an internal kernel one
that shouldn't do any user permission checks (it also doesn't need the
file_start_write/file_end_write logic, since it's just a pipe).
There are a couple of other cases like that - accounting, core dumping,
and splice - but autofs stands out because it can be built as a module.
As a result, we need to export this internal __kernel_write() function
again.
We really don't want any other module to use this, but we don't have a
"EXPORT_SYMBOL_FOR_AUTOFS_ONLY()". But we can mark it GPL-only to at
least approximate that "internal use only" for licensing.
While in this area, make autofs pass in NULL for the file position
pointer, since it's always a pipe, and we now use a NULL file pointer
for streaming file descriptors (see file_ppos() and commit 438ab720c675:
"vfs: pass ppos=NULL to .read()/.write() of FMODE_STREAM files")
This effectively reverts commits 9db977522449 ("fs: unexport
__kernel_write") and 13c164b1a186 ("autofs: switch to kernel_write").
Fixes: 13c164b1a186 ("autofs: switch to kernel_write") Reported-by: Ondrej Mosnacek <omosnace@redhat.com> Acked-by: Christoph Hellwig <hch@lst.de> Acked-by: Acked-by: Ian Kent <raven@themaw.net> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
the expectation is that host tools are built with debug informations.
This however doesn't happen if the Makefile assigns a new value to the
HOST_EXTRACFLAGS instead of appending to it. So use += instead of := for
the first assignment.
Fixes: e3fd9b5384f3 ("scripts/dtc: consolidate include path options in Makefile") Signed-off-by: Uwe Kleine-König <u.kleine-koenig@pengutronix.de> Signed-off-by: Rob Herring <robh@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
Blk-mq should call commit_rqs once 'bd.last != true' and no more
request will come(so virtscsi can kick the virtqueue, e.g.). We already
do that in 'blk_mq_dispatch_rq_list/blk_mq_try_issue_list_directly' while
list not empty and 'queued > 0'. However, we can seen the same scene
once the last request in list call queue_rq and return error like
BLK_STS_IOERR which will not requeue the request, and lead that list
empty but need call commit_rqs too(Or the request for virtscsi will stay
timeout until other request kick virtqueue).
We found this problem by do fsstress test with offline/online virtscsi
device repeat quickly.
Due to a HW issue, in some scenarios the LAST bit might remain set.
This will cause an unexpected NACK after reading 16 bytes on the next
read.
Example: if user tries to read from a missing device, get a NACK,
then if the next command is a long read ( > 16 bytes),
the master will stop reading after 16 bytes.
To solve this, if a command fails, check if LAST bit is still
set. If it does, reset the module.
Fixes: 56a1485b102e (i2c: npcm7xx: Add Nuvoton NPCM I2C controller driver) Signed-off-by: Tali Perry <tali.perry1@gmail.com> Signed-off-by: Wolfram Sang <wsa@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
the i2c_ram structure is missing the sdmatmp field mentionned in
datasheet for MPC8272 at paragraph 36.5. With this field missing, the
hardware would write past the allocated memory done through
cpm_muram_alloc for the i2c_ram structure and land in memory allocated
for the buffers descriptors corrupting the cbd_bufaddr field. Since this
field is only set during setup(), the first i2c transaction would work
and the following would send data read from an arbitrary memory
location.
Fixes: 61045dbe9d8d ("i2c: Add support for I2C bus on Freescale CPM1/CPM2 controllers") Signed-off-by: Nicolas VINCENT <nicolas.vincent@vossloh.com> Acked-by: Jochen Friedrich <jochen@scram.de> Acked-by: Christophe Leroy <christophe.leroy@csgroup.eu> Signed-off-by: Wolfram Sang <wsa@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
GPIO_U is mapped to the least significant byte of input/output mask, and
the byte in "output" mask should be 0 because GPIO_U is input only. All
the other bits need to be 1 because GPIO_V/W/X support both input and
output modes.
Similarly, GPIO_Y/Z are mapped to the 2 least significant bytes, and the
according bits need to be 1 because GPIO_Y/Z support both input and
output modes.
Fixes: ab4a85534c3e ("gpio: aspeed: Add in ast2600 details to Aspeed driver") Signed-off-by: Tao Ren <rentao.bupt@gmail.com> Reviewed-by: Andrew Jeffery <andrew@aj.id.au> Signed-off-by: Bartosz Golaszewski <bgolaszewski@baylibre.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
Currently, the IRQ setup for the SGPIO driver enables all interrupts in
dual-edge trigger mode. Since the default handler is handle_bad_irq, any
state change on input GPIOs will trigger bad IRQ warnings.
This change applies sensible IRQ defaults: single-edge trigger, and all
IRQs disabled.
Signed-off-by: Jeremy Kerr <jk@codeconstruct.com.au> Fixes: 7db47faae79b ("gpio: aspeed: Add SGPIO driver") Reviewed-by: Joel Stanley <joel@jms.id.au> Acked-by: Andrew Jeffery <andrew@aj.id.au> Signed-off-by: Bartosz Golaszewski <bgolaszewski@baylibre.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
Currently, the aspeed-sgpio driver exposes up to 80 GPIO lines,
corresponding to the 80 status bits available in hardware. Each of these
lines can be configured as either an input or an output.
However, each of these GPIOs is actually an input *and* an output; we
actually have 80 inputs plus 80 outputs.
This change expands the maximum number of GPIOs to 160; the lower half
of this range are the input-only GPIOs, the upper half are the outputs.
We fix the GPIO directions to correspond to this mapping.
This also fixes a bug when setting GPIOs - we were reading from the
input register, making it impossible to set more than one output GPIO.
Signed-off-by: Jeremy Kerr <jk@codeconstruct.com.au> Fixes: 7db47faae79b ("gpio: aspeed: Add SGPIO driver") Reviewed-by: Joel Stanley <joel@jms.id.au> Reviewed-by: Andrew Jeffery <andrew@aj.id.au> Acked-by: Rob Herring <robh@kernel.org> Signed-off-by: Bartosz Golaszewski <bgolaszewski@baylibre.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
When pca953x_irq_pending returns false, the pending parameter won't
be set. But pca953x_irq_handler continues using this uninitialized
variable as pending irqs and will cause problem.
Fix the issue by initializing pending to 0.
Fixes: 064c73afe738 ("gpio: pca953x: Synchronize interrupt handler properly") Signed-off-by: Ye Li <ye.li@nxp.com> Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com> Signed-off-by: Bartosz Golaszewski <bgolaszewski@baylibre.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
if of_find_device_by_node() succeed, exynos_iommu_of_xlate() doesn't have
a corresponding put_device(). Thus add put_device() to fix the exception
handling for this function implementation.
Fixes: aa759fd376fb ("iommu/exynos: Add callback for initializing devices from device tree") Signed-off-by: Yu Kuai <yukuai3@huawei.com> Acked-by: Marek Szyprowski <m.szyprowski@samsung.com> Link: https://lore.kernel.org/r/20200918011335.909141-1-yukuai3@huawei.com Signed-off-by: Joerg Roedel <jroedel@suse.de> Signed-off-by: Sasha Levin <sashal@kernel.org>
transport_lookup_tmr_lun() uses "orig_fe_lun" member of struct se_cmd for
the lookup. Hence, update this field directly for the
TARGET_SCF_LOOKUP_LUN_FROM_TAG case.
Link: https://lore.kernel.org/r/1600300471-26135-1-git-send-email-sudhakar.panneerselvam@oracle.com Fixes: a36840d80027 ("target: Initialize LUN in transport_init_se_cmd()") Reported-by: Martin Wilck <mwilck@suse.com> Reviewed-by: Mike Christie <michael.christie@oracle.com> Signed-off-by: Sudhakar Panneerselvam <sudhakar.panneerselvam@oracle.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
The ChipID IO region has it's own clock, which is being disabled while
scanning for unused clocks. It turned out that some CPU hotplug, CPU idle
or even SOC firmware code depends on the reads from that area. Fix the
mysterious hang caused by entering deep CPU idle state by ignoring the
'chipid' clock during unused clocks scan, as there are no direct clients
for it which will keep it enabled.
Fixes: e062b571777f ("clk: exynos4: register clocks using common clock framework") Signed-off-by: Marek Szyprowski <m.szyprowski@samsung.com> Link: https://lore.kernel.org/r/20200922124046.10496-1-m.szyprowski@samsung.com Reviewed-by: Krzysztof Kozlowski <krzk@kernel.org> Acked-by: Sylwester Nawrocki <s.nawrocki@samsung.com> Signed-off-by: Stephen Boyd <sboyd@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
Andy reported that commit 6b41030fdc79 ("dmaengine: dmatest:
Restore default for channel") broke his scripts for the case
where "busy" channel is used for configuration with expectation
that run command would do nothing. Instead, behavior was
(unintentionally) changed to treat such case as under-configuration
and progress with defaults, i.e. run command would start a test
with default setting for channel (which would use all channels).
Restore original behavior with tracking status of channel setter
so we can distinguish between misconfigured and under-configured
cases in run command and act accordingly.
Fixes: 6b41030fdc79 ("dmaengine: dmatest: Restore default for channel") Reported-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com> Tested-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com> Signed-off-by: Vladimir Murzin <vladimir.murzin@arm.com> Tested-by: Peter Ujfalusi <peter.ujfalusi@ti.com> Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com> Link: https://lore.kernel.org/r/20200922115847.30100-1-andriy.shevchenko@linux.intel.com Signed-off-by: Vinod Koul <vkoul@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
Commit bff1cef5f23a ("clk: tegra: Don't enable already enabled PLLs")
added checks to avoid enabling PLLs that have already been enabled by
the bootloader. However, the PLL_E configuration inherited from the
bootloader isn't necessarily the one that is needed for the kernel.
This can cause SATA to fail like this:
[ 5.310270] phy phy-sata.6: phy poweron failed --> -110
[ 5.315604] tegra-ahci 70027000.sata: failed to power on AHCI controller: -110
[ 5.323022] tegra-ahci: probe of 70027000.sata failed with error -110
Fix this by always programming the PLL_E. This ensures that any mis-
configuration by the bootloader will be overwritten by the kernel.
After client is done with the COPY operation, it needs to invalidate
its pagecache (as it did no reading or writing of the data locally)
and it needs to invalidate it's attributes just like it would have
for a read on the source file and write on the destination file.
Once the linux server started giving out read delegations to
read+write opens, the destination file of the copy_file range
started having delegations and not doing syncup on close of the
file leading to xfstest failures for generic/430,431,432,433,565.
v2: changing cache_validity needs to be protected by the i_lock.
nfs_readdir_page_filler() iterates over entries in a directory, reusing
the same security label buffer, but does not reset the buffer's length.
This causes decode_attr_security_label() to return -ERANGE if an entry's
security label is longer than the previous one's. This error, in
nfs4_decode_dirent(), only gets passed up as -EAGAIN, which causes another
failed attempt to copy into the buffer. The second error is ignored and
the remaining entries do not show up in ls, specifically the getdents64()
syscall.
Reproduce by creating multiple files in NFS and giving one of the later
files a longer security label. ls will not see that file nor any that are
added afterwards, though they will exist on the backend.
In nfs_readdir_page_filler(), reset security label buffer length before
every reuse
Per the datasheet the i2c functions use MPP_Sel=0x1. They are documented
as using MPP_Sel=0x4 as well but mixing 0x1 and 0x4 is clearly wrong. On
the board tested 0x4 resulted in a non-functioning i2c bus so stick with
0x1 which works.
Fixes: d7ae8f8dee7f ("pinctrl: mvebu: pinctrl driver for 98DX3236 SoC") Signed-off-by: Chris Packham <chris.packham@alliedtelesis.co.nz> Reviewed-by: Andrew Lunn <andrew@lunn.ch> Link: https://lore.kernel.org/r/20200907211712.9697-2-chris.packham@alliedtelesis.co.nz Signed-off-by: Linus Walleij <linus.walleij@linaro.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
The raw interrupt status of GPIO maybe set before the interrupt is enabled,
which would trigger the interrupt event once enabled it from user side.
This is the case for edge interrupts only. Adding a clear operation when
setting interrupt type can avoid that.
There're a few considerations for the solution:
1) This issue is for edge interrupt only; The interrupts requested by users
are IRQ_TYPE_LEVEL_HIGH as default, so clearing interrupt when request
is useless.
2) The interrupt type can be set to edge when request and following up
with clearing it though, but the problem is still there once users set
the interrupt type to level trggier.
3) We can add a clear operation after each time of setting interrupt
enable bit, but it is redundant for level trigger interrupt.
Therefore, the solution is this patch seems the best for now.
Fixes: 9a3821c2bb47 ("gpio: Add GPIO driver for Spreadtrum SC9860 platform") Signed-off-by: Taiping Lai <taiping.lai@unisoc.com> Signed-off-by: Chunyan Zhang <chunyan.zhang@unisoc.com> Reviewed-by: Baolin Wang <baolin.wang7@gmail.com> Signed-off-by: Bartosz Golaszewski <bgolaszewski@baylibre.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
Actually, the problem mentioned in scripts/link-vmlinux.sh comments;
"In theory it's possible this results in even more stubs, but unlikely"
is happening here, and ends up with another kallsyms step required.
scripts/kallsyms.c already ignores various compiler stubs. Let's do
similar to make kallsysms for PowerPC always succeed in 2 steps.
The lldd may have made calls to delete a remote port or local port and
the delete is in progress when the cli then attempts to create a new
controller. Currently, this proceeds without error although it can't be
very successful.
Fix this by validating that both the host port and remote port are
present when a new controller is to be created.
Signed-off-by: James Smart <james.smart@broadcom.com> Reviewed-by: Himanshu Madhani <himanshu.madhani@oracle.com> Signed-off-by: Christoph Hellwig <hch@lst.de> Signed-off-by: Sasha Levin <sashal@kernel.org>
Currently, we use nvmeq->q_depth as the upper limit for a valid tag in
nvme_handle_cqe(), it is not correct. Because the available tag number
is recorded in tagset, which is not equal to nvmeq->q_depth.
The nvme driver registers interrupts for queues before initializing the
tagset, because it uses the number of successful request_irq() calls to
configure the tagset parameters. This allows a race condition with the
current tag validity check if the controller happens to produce an
interrupt with a corrupted CQE before the tagset is initialized.
Replace the driver's indirect tag check with the one already provided by
the block layer.
Signed-off-by: Xianting Tian <tian.xianting@h3c.com> Reviewed-by: Keith Busch <kbusch@kernel.org> Signed-off-by: Christoph Hellwig <hch@lst.de> Signed-off-by: Sasha Levin <sashal@kernel.org>
Some of the IS2 IP4_TCP_UDP keys are not correct, like L4_DPORT,
L4_SPORT and other L4 keys. This prevents offloaded tc-flower rules from
matching on src_port and dst_port for TCP and UDP packets.
Signed-off-by: Xiaoliang Yang <xiaoliang.yang_1@nxp.com> Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Sasha Levin <sashal@kernel.org>
The SPIE register contains counts for the TX FIFO so any time the irq
handler was invoked we would attempt to process the RX/TX fifos. Use the
SPIM value to mask the events so that we only process interrupts that
were expected.
This was a latent issue exposed by commit 3282a3da25bd ("powerpc/64:
Implement soft interrupt replay in C").
The commit eb1f00237aca ("lockdep,trace: Expose tracepoints"), started to
expose us for tracepoints. This lead to the following RCU splat on an ARM64
Qcom board.
[ 5.529634] WARNING: suspicious RCU usage
[ 5.537307] sdhci-pltfm: SDHCI platform and OF driver helper
[ 5.541092] 5.9.0-rc3 #86 Not tainted
[ 5.541098] -----------------------------
[ 5.541105] ../include/trace/events/lock.h:37 suspicious rcu_dereference_check() usage!
[ 5.541110]
[ 5.541110] other info that might help us debug this:
[ 5.541110]
[ 5.541116]
[ 5.541116] rcu_scheduler_active = 2, debug_locks = 1
[ 5.541122] RCU used illegally from extended quiescent state!
[ 5.541129] no locks held by swapper/0/0.
[ 5.541134]
[ 5.541134] stack backtrace:
[ 5.541143] CPU: 0 PID: 0 Comm: swapper/0 Not tainted 5.9.0-rc3 #86
[ 5.541149] Hardware name: Qualcomm Technologies, Inc. APQ 8016 SBC (DT)
[ 5.541157] Call trace:
[ 5.568185] sdhci_msm 7864900.sdhci: Got CD GPIO
[ 5.574186] dump_backtrace+0x0/0x1c8
[ 5.574206] show_stack+0x14/0x20
[ 5.574229] dump_stack+0xe8/0x154
[ 5.574250] lockdep_rcu_suspicious+0xd4/0xf8
[ 5.574269] lock_acquire+0x3f0/0x460
[ 5.574292] _raw_spin_lock_irqsave+0x80/0xb0
[ 5.574314] __pm_runtime_suspend+0x4c/0x188
[ 5.574341] psci_enter_domain_idle_state+0x40/0xa0
[ 5.574362] cpuidle_enter_state+0xc0/0x610
[ 5.646487] cpuidle_enter+0x38/0x50
[ 5.650651] call_cpuidle+0x18/0x40
[ 5.654467] do_idle+0x228/0x278
[ 5.657678] cpu_startup_entry+0x24/0x70
[ 5.661153] rest_init+0x1a4/0x278
[ 5.665061] arch_call_rest_init+0xc/0x14
[ 5.668272] start_kernel+0x508/0x540
Following the path in pm_runtime_put_sync_suspend() from
psci_enter_domain_idle_state(), it seems like we end up using the RCU.
Therefore, let's simply silence the splat by informing the RCU about it
with RCU_NONIDLE.
Note that, this is a temporary solution. Instead we should strive to avoid
using RCU_NONIDLE (and similar), but rather push rcu_idle_enter|exit()
further down, closer to the arch specific code. However, as the CPU PM
notifiers are also using the RCU, additional rework is needed.
Reported-by: Naresh Kamboju <naresh.kamboju@linaro.org> Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org> Acked-by: Paul E. McKenney <paulmck@kernel.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
These will naturally fail when attempted through SQPOLL, but either
with -EFAULT or -EBADF. Make it explicit that these are not workable
through SQPOLL and return -EINVAL, just like other ops that need to
use ->files.
It would seem none of the kernel continuous integration does this:
$ cd tools/io_uring
$ make
Otherwise it may have noticed:
cc -Wall -Wextra -g -D_GNU_SOURCE -c -o io_uring-bench.o
io_uring-bench.c
io_uring-bench.c:133:12: error: static declaration of ‘gettid’
follows non-static declaration
133 | static int gettid(void)
| ^~~~~~
In file included from /usr/include/unistd.h:1170,
from io_uring-bench.c:27:
/usr/include/x86_64-linux-gnu/bits/unistd_ext.h:34:16: note:
previous declaration of ‘gettid’ was here
34 | extern __pid_t gettid (void) __THROW;
| ^~~~~~
make: *** [<builtin>: io_uring-bench.o] Error 1
The problem on Ubuntu 20.04 (with lk 5.9.0-rc5) is that unistd.h
already defines gettid(). So prefix the local definition with
"lk_".
For 64bit CONFIG_BASE_SMALL=0 systems PID_MAX_LIMIT is set by default to 4194304. During boot the kernel sets a new value based on number of CPUs
but no lower than 32768. It is 1024 per CPU so with 128 CPUs the default
becomes 131072 which needs six digits.
This value can be increased during run time but must not exceed the
initial upper limit.
Systemd sometime after v241 sets it to the upper limit during boot. The
result is that when the pid exceeds five digits, the trace output is a
little hard to read because it is no longer properly padded (same like
on big iron with 98+ CPUs).
Radiotap header field 'Channel flags' has '2 GHz spectrum' set to
'true' for 6GHz packet.
Change it to 5GHz as there isn't a separate option available for 6GHz.
This patch sets skb->protocol before transmitting frames on the HDLC
device, so that a user listening on the HDLC device with an AF_PACKET
socket will see outgoing frames' sll_protocol field correctly set and
consistent with that of incoming frames.
1. Control frames in hdlc_cisco and hdlc_ppp
When these drivers send control frames, skb->protocol is not set.
This value should be set to htons(ETH_P_HDLC), because when receiving
control frames, their skb->protocol is set to htons(ETH_P_HDLC).
When receiving, hdlc_type_trans in hdlc.h is called, which then calls
cisco_type_trans or ppp_type_trans. The skb->protocol of control frames
is set to htons(ETH_P_HDLC) so that the control frames can be received
by hdlc_rcv in hdlc.c, which calls cisco_rx or ppp_rx to process the
control frames.
2. hdlc_fr
When this driver sends control frames, skb->protocol is set to internal
values used in this driver.
When this driver sends data frames (from upper stacked PVC devices),
skb->protocol is the same as that of the user data packet being sent on
the upper PVC device (for normal PVC devices), or is htons(ETH_P_802_3)
(for Ethernet-emulating PVC devices).
However, skb->protocol for both control frames and data frames should be
set to htons(ETH_P_HDLC), because when receiving, all frames received on
the HDLC device will have their skb->protocol set to htons(ETH_P_HDLC).
When receiving, hdlc_type_trans in hdlc.h is called, and because this
driver doesn't provide a type_trans function in struct hdlc_proto,
all frames will have their skb->protocol set to htons(ETH_P_HDLC).
The frames are then received by hdlc_rcv in hdlc.c, which calls fr_rx
to process the frames (control frames are consumed and data frames
are re-received on upper PVC devices).
Cc: Krzysztof Halasa <khc@pm.waw.pl> Signed-off-by: Xie He <xie.he.0141@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Sasha Levin <sashal@kernel.org>
This driver is a virtual driver stacked on top of Ethernet interfaces.
When this driver transmits data on the Ethernet device, the skb->protocol
setting is inconsistent with the Ethernet header prepended to the skb.
This causes a user listening on the Ethernet interface with an AF_PACKET
socket, to see different sll_protocol values for incoming and outgoing
frames, because incoming frames would have this value set by parsing the
Ethernet header.
This patch changes the skb->protocol value for outgoing Ethernet frames,
making it consistent with the Ethernet header prepended. This makes a
user listening on the Ethernet device with an AF_PACKET socket, to see
the same sll_protocol value for incoming and outgoing frames.
Cc: Martin Schiller <ms@dev.tdt.de> Signed-off-by: Xie He <xie.he.0141@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Sasha Levin <sashal@kernel.org>
Get and put the reference to the ctrl in the nvme_dev_open() and
nvme_dev_release() before and after module get/put for ctrl in char
device file operations.
Introduce char_dev relase function, get/put the controller and module
which allows us to fix the potential Oops which can be easily reproduced
with a passthru ctrl (although the problem also exists with pure user
access):
Reported-by: Eric Sandeen <esandeen@redhat.com> Signed-off-by: David Milburn <dmilburn@redhat.com> Tested-by: Eric Sandeen <sandeen@redhat.com> Signed-off-by: Christoph Hellwig <hch@lst.de> Signed-off-by: Sasha Levin <sashal@kernel.org>
Some WinCE devices face connectivity issues via the NDIS interface. They
fail to register, resulting in -110 timeout errors and failures during the
probe procedure.
In this kind of WinCE devices, the Windows-side ndis driver needs quite
more time to be loaded and configured, so that the linux rndis host queries
to them fail to be responded correctly on time.
More specifically, when INIT is called on the WinCE side - no other
requests can be served by the Client and this results in a failed QUERY
afterwards.
The increase of the waiting time on the side of the linux rndis host in
the command-response loop leaves the INIT process to complete and respond
to a QUERY, which comes afterwards. The WinCE devices with this special
"feature" in their ndis driver are satisfied by this fix.
Signed-off-by: Olympia Giannou <olympia.giannou@leica-geosystems.com> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Sasha Levin <sashal@kernel.org>
Increase Rx ring size to address issue where hardware is reaching
the receive work limit.
Before:
[ 102.223342] de2104x 0000:17:00.0 eth0: rx work limit reached
[ 102.245695] de2104x 0000:17:00.0 eth0: rx work limit reached
[ 102.251387] de2104x 0000:17:00.0 eth0: rx work limit reached
[ 102.267444] de2104x 0000:17:00.0 eth0: rx work limit reached
Signed-off-by: Lucy Yan <lucyyan@google.com> Reviewed-by: Moritz Fischer <mdf@kernel.org> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Sasha Levin <sashal@kernel.org>
The previous change "hv_netvsc: Switch the data path at the right time
during hibernation" adds the call of netvsc_vf_changed() upon
NETDEV_CHANGE, so it's necessary to avoid the duplicate call and message
when the VF is brought UP or DOWN.
Signed-off-by: Dexuan Cui <decui@microsoft.com> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Sasha Levin <sashal@kernel.org>
This patch causes a regression betwen Kernel 5.7 and 5.8 at wlcore:
with it applied, WiFi stops working, and the Kernel starts printing
this message every second:
PVC devices are virtual devices in this driver stacked on top of the
actual HDLC device. They are the devices normal users would use.
PVC devices have two types: normal PVC devices and Ethernet-emulating
PVC devices.
When transmitting data with PVC devices, the ndo_start_xmit function
will prepend a header of 4 or 10 bytes. Currently this driver requests
this headroom to be reserved for normal PVC devices by setting their
hard_header_len to 10. However, this does not work when these devices
are used with AF_PACKET/RAW sockets. Also, this driver does not request
this headroom for Ethernet-emulating PVC devices (but deals with this
problem by reallocating the skb when needed, which is not optimal).
This patch replaces hard_header_len with needed_headroom, and set
needed_headroom for Ethernet-emulating PVC devices, too. This makes
the driver to request headroom for all PVC devices in all cases.
Cc: Krzysztof Halasa <khc@pm.waw.pl> Cc: Martin Schiller <ms@dev.tdt.de> Signed-off-by: Xie He <xie.he.0141@gmail.com> Signed-off-by: Jakub Kicinski <kuba@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
Ubuntu mainline builds for ppc64le are failing with the below error (*):
CALL /home/kernel/COD/linux/scripts/atomic/check-atomics.sh
DESCEND bpf/resolve_btfids
resolve_btfids needs to be build as a host binary and it needs libbpf.
However, libbpf Makefile hardcodes an include path utilizing $(ARCH).
This results in mixing of cross-architecture headers resulting in a
build failure.
The specific header include path doesn't seem necessary for a libbpf
build. Hence, remove the same.
Using dev_kfree_skb for tx skbs breaks AQL. This worked until now only
by accident, because a mac80211 issue breaks AQL on drivers with firmware
rate control that report the rate via ieee80211_tx_status_ext as struct
rate_info.
Signed-off-by: Felix Fietkau <nbd@nbd.name> Acked-by: Toke Høiland-Jørgensen <toke@redhat.com> Signed-off-by: Kalle Valo <kvalo@codeaurora.org> Link: https://lore.kernel.org/r/20200812144943.91974-1-nbd@nbd.name Signed-off-by: Sasha Levin <sashal@kernel.org>
Fix the check for the mainline vboxsf code being used with the old
mount.vboxsf mount binary from the out-of-tree vboxsf version doing
a comparison between signed and unsigned data types.
Reported-by: kernel test robot <lkp@intel.com> Reported-by: Dan Carpenter <dan.carpenter@oracle.com> Signed-off-by: Hans de Goede <hdegoede@redhat.com> Signed-off-by: Al Viro <viro@zeniv.linux.org.uk> Signed-off-by: Sasha Levin <sashal@kernel.org>
When the timer counts to the upper limit, an overflow interrupt is
generated, and the count is reset with the value in the TIME_INI
register. But the software expects to start counting from 0 when
the count overflows, so it forces TIME_INI to 0 to solve the
potential interrupt storm problem.
Signed-off-by: Guo Ren <guoren@linux.alibaba.com> Tested-by: Xu Kai <xukai@nationalchip.com> Cc: Daniel Lezcano <daniel.lezcano@linaro.org> Cc: Thomas Gleixner <tglx@linutronix.de> Signed-off-by: Daniel Lezcano <daniel.lezcano@linaro.org> Link: https://lore.kernel.org/r/1597735877-71115-1-git-send-email-guoren@kernel.org Signed-off-by: Sasha Levin <sashal@kernel.org>
Since commit c330fb1ddc0a ("XEN uses irqdesc::irq_data_common::handler_data to store a per interrupt XEN data pointer which contains XEN specific information.")
Xen is using the chip_data pointer for storing IRQ specific data. When
running as a HVM domain this can result in problems for legacy IRQs, as
those might use chip_data for their own purposes.
Use a local array for this purpose in case of legacy IRQs, avoiding the
double use.
Cc: stable@vger.kernel.org Fixes: c330fb1ddc0a ("XEN uses irqdesc::irq_data_common::handler_data to store a per interrupt XEN data pointer which contains XEN specific information.") Signed-off-by: Juergen Gross <jgross@suse.com> Tested-by: Stefan Bader <stefan.bader@canonical.com> Reviewed-by: Boris Ostrovsky <boris.ostrovsky@oracle.com> Link: https://lore.kernel.org/r/20200930091614.13660-1-jgross@suse.com Signed-off-by: Juergen Gross <jgross@suse.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
A recent attempt to fix a ref count leak in
amdgpu_display_crtc_set_config() turned out to be doing too much and
"fixed" an intended decrease as if it were a leak. Undo that part to
restore the proper balance. This is the very nature of this function
to increase or decrease the power reference count depending on the
situation.
Consequences of this bug is that the power reference would
eventually get down to 0 while the display was still in use,
resulting in that display switching off unexpectedly.
Signed-off-by: Jean Delvare <jdelvare@suse.de> Fixes: e008fa6fb415 ("drm/amdgpu: fix ref count leak in amdgpu_display_crtc_set_config") Cc: stable@vger.kernel.org Cc: Navid Emamdoost <navid.emamdoost@gmail.com> Cc: Alex Deucher <alexander.deucher@amd.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
After commit 6827ca573c03 ("memstick: rtsx_usb_ms: Support runtime power
management"), removing module rtsx_usb_ms will be stuck.
The deadlock is caused by powering on and powering off at the same time,
the former one is when memstick_check() is flushed, and the later is called
by memstick_remove_host().
Soe let's skip allocating card to prevent this issue.
The temp buffer size variable for trace_find_next_entry() was incorrectly
being updated when the size did not change. The temp buffer size should only
be updated when it is reallocated.
This is mostly an issue when used with ftrace_dump(). That's because
ftrace_dump() can not allocate a new buffer, and instead uses a temporary
buffer with a fix size. But the variable that keeps track of that size is
incorrectly updated with each call, and it could fall into the path that
would try to reallocate the buffer and produce a warning.
Not only fix the updating of the temp buffer, but also do not free the temp
buffer before a new buffer is allocated (there's no reason to not continue
to use the current temp buffer if an allocation fails).
Cc: stable@vger.kernel.org Fixes: 8e99cf91b99bb ("tracing: Do not allocate buffer in trace_find_next_entry() in atomic") Reported-by: Anna-Maria Behnsen <anna-maria@linutronix.de> Signed-off-by: Steven Rostedt (VMware) <rostedt@goodmis.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The first thing that the ftrace function callback helper functions should do
is to check for recursion. Peter Zijlstra found that when
"rcu_is_watching()" had its notrace removed, it caused perf function tracing
to crash. This is because the call of rcu_is_watching() is tested before
function recursion is checked and and if it is traced, it will cause an
infinite recursion loop.
rcu_is_watching() should still stay notrace, but to prevent this should
never had crashed in the first place. The recursion prevention must be the
first thing done in callback functions.
Link: https://lore.kernel.org/r/20200929112541.GM2628@hirez.programming.kicks-ass.net Cc: stable@vger.kernel.org Cc: Paul McKenney <paulmck@kernel.org> Fixes: c68c0fa293417 ("ftrace: Have ftrace_ops_get_func() handle RCU and PER_CPU flags too") Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Reported-by: Peter Zijlstra (Intel) <peterz@infradead.org> Signed-off-by: Steven Rostedt (VMware) <rostedt@goodmis.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
By default, PCI drivers with runtime PM enabled will skip the calls
to suspend and resume on system PM. For this driver, we don't want
that, as we need to perform additional steps for system PM to work
properly on all systems. So instruct the PM core to not skip these
calls.
Fixes: a9c8088c7988 ("i2c: i801: Don't restore config registers on runtime PM") Reported-by: Volker Rümelin <volker.ruemelin@googlemail.com> Signed-off-by: Jean Delvare <jdelvare@suse.de> Cc: stable@vger.kernel.org Signed-off-by: Wolfram Sang <wsa@kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The kernel may fail to boot or devices may fail to come up when
initializing iscsi_tcp devices starting with Linux 5.8.
Commit a79af8a64d39 ("[SCSI] iscsi_tcp: use iscsi_conn_get_addr_param
libiscsi function") introduced getpeername() within the session spinlock.
Commit 1b66d253610c ("bpf: Add get{peer, sock}name attach types for
sock_addr") introduced BPF_CGROUP_RUN_SA_PROG_LOCK() within getpeername(),
which acquires a mutex and when used from iscsi_tcp devices can now lead to
"BUG: scheduling while atomic:" and subsequent damage.
Ensure that the spinlock is released before calling getpeername() or
getsockname(). sock_hold() and sock_put() are used to ensure that the
socket reference is preserved until after the getpeername() or
getsockname() complete.
BPLL clock must not be disabled because it is needed for proper DRAM
operation. This is normally handled by respective memory devfreq driver,
but when that driver is not yet probed or its probe has been deferred
the clock might get disabled what causes board hang. Fix this by calling
clk_prepare_enable() directly from the clock provider driver.
Cc: stable@vger.kernel.org Signed-off-by: Marek Szyprowski <m.szyprowski@samsung.com> Reviewed-by: Lukasz Luba <lukasz.luba@arm.com> Tested-by: Lukasz Luba <lukasz.luba@arm.com> Acked-by: Krzysztof Kozlowski <krzk@kernel.org> Link: https://lore.kernel.org/r/20200807133143.22748-1-m.szyprowski@samsung.com Fixes: 6e7674c3c6df ("memory: Add DMC driver for Exynos5422") Signed-off-by: Sylwester Nawrocki <s.nawrocki@samsung.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The original commit appears to have the logic reversed in
amd_fch_gpio_get_direction. Also confirmed by observing the value of
"direction" in the sys tree.
The offset of regmap is incorrect, j * 8 is move to the
wrong register.
for example:
asume i = 0, j = 1. we want to set KPY5 as interrupt
falling edge mode, regmap[0][1] should be TC3589x_GPIOIBE1 0xcd
but, regmap[i] + j * 8 = TC3589x_GPIOIBE0 + 8 ,point to 0xd4,
this is TC3589x_GPIOIE2 not TC3589x_GPIOIBE1.
The gpio-siox driver uses handle_nested_irq() to implement its
interrupt support. This is only capable of handling threaded irq
actions. For a hardirq action it triggers a NULL pointer oops.
(It calls action->thread_fn which is NULL then.)
Prevent registration of a hardirq action by setting
gpio_irq_chip::threaded to true.
Cc: u.kleine-koenig@pengutronix.de Fixes: be8c8facc707 ("gpio: new driver to work with a 8x12 siox") Cc: stable@vger.kernel.org Signed-off-by: Ahmad Fatoum <a.fatoum@pengutronix.de> Acked-by: Uwe Kleine-König <u.kleine-koenig@pengutronix.de> Signed-off-by: Bartosz Golaszewski <bgolaszewski@baylibre.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Commit 88b7381a939d ("USB: Select better matching USB drivers when
available") inadvertently broke usbip functionality. The commit in
question allows USB device drivers to be explicitly matched with
USB devices via the use of driver-provided identifier tables and
match functions, which is useful for a specialised device driver
to be chosen for a device that can also be handled by another,
more generic, device driver.
Prior, the USB device section of usb_device_match() had an
unconditional "return 1" statement, which allowed user-space to bind
USB devices to the usbip_host device driver, if desired. However,
the aforementioned commit changed the default/fallback return
value to zero. This breaks device drivers such as usbip_host, so
this commit restores the legacy behaviour, but only if a device
driver does not have an id_table and a match() function.
In addition, if usb_device_match is called for a device driver
and device pair where the device does not match the id_table of the
device driver in question, then the device driver will be disqualified
for the device. This allows avoiding the default case of "return 1",
which prevents undesirable probe() calls to a driver even though
its id_table did not match the device.
Finally, this commit changes the specialised-driver-to-generic-driver
transition code so that when a device driver returns -ENODEV, a more
generic device driver is only considered if the current device driver
does not have an id_table and a match() function. This ensures that
"generic" drivers such as usbip_host will not be considered specialised
device drivers and will not cause the device to be locked in to the
generic device driver, when a more specialised device driver could be
tried.
All of these changes restore usbip functionality without regressions,
ensure that the specialised/generic device driver selection logic works
as expected with the usb and apple-mfi-fastcharge drivers, and do not
negatively affect the use of devices provided by dummy_hcd.
This commit resolves a minor bug in the selection/discovery of more
specific USB device drivers for devices that are currently bound to
generic USB device drivers.
The bug is related to the way a candidate USB device driver is
compared against the generic USB device driver. The code in
is_dev_usb_generic_driver() assumes that the device driver in question
is a USB device driver by calling to_usb_device_driver(dev->driver)
to downcast; however I have observed that this assumption is not always
true, through code instrumentation.
This commit avoids the incorrect downcast altogether by comparing
the USB device's driver (i.e., dev->driver) to the generic USB
device driver directly. This method was suggested by Alan Stern.
This bug was found while investigating Andrey Konovalov's report
indicating usbip device driver misbehaviour with the recently merged
generic USB device driver selection feature. The report is linked
below.
This commit resolves a bug in the selection/discovery of more
specific USB device drivers for devices that are currently bound to
generic USB device drivers.
The bug is in the logic that determines whether a device currently
bound to a generic USB device driver should be re-probed by a
more specific USB device driver or not. The code in
__usb_bus_reprobe_drivers() used to have the following lines:
As the reader will notice, the code checks whether the USB device in
consideration matches the identifier table (id_table) of a specific
USB device_driver (new_udriver), followed by a similar check, but this
time with the USB device driver's match function. However, the match
function's return value is not checked correctly. When match() returns
zero, it means that the specific USB device driver is *not* applicable
to the USB device in question, but the code then goes on to reprobe the
device with the new USB device driver under consideration. All this to
say, the logic is inverted.
This bug was found by code inspection and instrumentation while
investigating the root cause of the issue reported by Andrey Konovalov,
where usbip took over syzkaller's virtual USB devices in an undesired
manner. The report is linked below.
This commit reverts commit 7a2f2974f265 ("usbip: Implement a match
function to fix usbip").
In summary, commit d5643d2249b2 ("USB: Fix device driver race")
inadvertently broke usbip functionality, which I resolved in an incorrect
manner by introducing a match function to usbip, usbip_match(), that
unconditionally returns true.
However, the usbip_match function, as is, causes usbip to take over
virtual devices used by syzkaller for USB fuzzing, which is a regression
reported by Andrey Konovalov.
Furthermore, in conjunction with the fix of another bug, handled by another
patch titled "usbcore/driver: Fix specific driver selection" in this patch
set, the usbip_match function causes unexpected USB subsystem behaviour
when the usbip_host driver is loaded. The unexpected behaviour can be
qualified as follows:
- If commit 41160802ab8e ("USB: Simplify USB ID table match") is included
in the kernel, then all USB devices are bound to the usbip_host
driver, which appears to the user as if all USB devices were
disconnected.
- If the same commit (41160802ab8e) is not in the kernel (as is the case
with v5.8.10) then all USB devices are re-probed and re-bound to their
original device drivers, which appears to the user as a disconnection
and re-connection of USB devices.
Please note that this commit will make usbip non-operational again,
until yet another patch in this patch set is merged, titled
"usbcore/driver: Accommodate usbip".
commit 2b74b0a04d3e ("USB: gadget: f_ncm: add bounds checks to ncm_unwrap_ntb()")
adds important bounds checking however it unfortunately also introduces a
bug with respect to section 3.3.1 of the NCM specification.
wDatagramIndex[1] : "Byte index, in little endian, of the second datagram
described by this NDP16. If zero, then this marks the end of the sequence
of datagrams in this NDP16."
wDatagramLength[1]: "Byte length, in little endian, of the second datagram
described by this NDP16. If zero, then this marks the end of the sequence
of datagrams in this NDP16."
wDatagramIndex[1] and wDatagramLength[1] respectively then may be zero but
that does not mean we should throw away the data referenced by
wDatagramIndex[0] and wDatagramLength[0] as is currently the case.
Breaking the loop on (index2 == 0 || dg_len2 == 0) should come at the end
as was previously the case and checks for index2 and dg_len2 should be
removed since zero is valid.
I'm not sure how much testing the above patch received but for me right now
after enumeration ping doesn't work. Reverting the commit restores ping,
scp, etc.
The extra validation associated with wDatagramIndex[0] and
wDatagramLength[0] appears to be valid so, this change removes the incorrect
restriction on wDatagramIndex[1] and wDatagramLength[1] restoring data
processing between host and device.
Commit bedf9fc01ff1 ("mmc: sdhci: Workaround broken command queuing on
Intel GLK"), disabled command-queuing on Intel GLK based LENOVO models
because of it being broken due to what is believed to be a bug in
the BIOS.
It seems that the BIOS of some IRBIS models, including the IRBIS NB111
model has the same issue, so disable command queuing there too.
Fixes: bedf9fc01ff1 ("mmc: sdhci: Workaround broken command queuing on Intel GLK") BugLink: https://bugzilla.kernel.org/show_bug.cgi?id=209397 Reported-and-tested-by: RussianNeuroMancer <russianneuromancer@ya.ru> Signed-off-by: Hans de Goede <hdegoede@redhat.com> Acked-by: Adrian Hunter <adrian.hunter@intel.com> Link: https://lore.kernel.org/r/20200927104821.5676-1-hdegoede@redhat.com Cc: stable@vger.kernel.org Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
We use a device's allocation state tree to track ranges in a device used
for allocated chunks, and we set ranges in this tree when allocating a new
chunk. However after a device replace operation, we were not setting the
allocated ranges in the new device's allocation state tree, so that tree
is empty after a device replace.
This means that a fitrim operation after a device replace will trim the
device ranges that have allocated chunks and extents, as we trim every
range for which there is not a range marked in the device's allocation
state tree. It is also important during chunk allocation, since the
device's allocation state is used to determine if a range is already
allocated when allocating a new chunk.
This is trivial to reproduce and the following script triggers the bug:
$ cat reproducer.sh
#!/bin/bash
DEV1="/dev/sdg"
DEV2="/dev/sdh"
DEV3="/dev/sdi"
wipefs -a $DEV1 $DEV2 $DEV3 &> /dev/null
# Create a raid1 test fs on 2 devices.
mkfs.btrfs -f -m raid1 -d raid1 $DEV1 $DEV2 > /dev/null
mount $DEV1 /mnt/btrfs
echo "Starting to replace $DEV1 with $DEV3"
btrfs replace start -B $DEV1 $DEV3 /mnt/btrfs
echo
echo "Running fstrim"
fstrim /mnt/btrfs
echo
echo "Unmounting filesystem"
umount /mnt/btrfs
echo "Mounting filesystem in degraded mode using $DEV3 only"
wipefs -a $DEV1 $DEV2 &> /dev/null
mount -o degraded $DEV3 /mnt/btrfs
if [ $? -ne 0 ]; then
dmesg | tail
echo
echo "Failed to mount in degraded mode"
exit 1
fi
echo
echo "File foo data (expected all bytes = 0xab):"
od -A d -t x1 /mnt/btrfs/foo
umount /mnt/btrfs
When running the reproducer:
$ ./replace-test.sh
wrote 10485760/10485760 bytes at offset 0
10 MiB, 2560 ops; 0.0901 sec (110.877 MiB/sec and 28384.5216 ops/sec)
Starting to replace /dev/sdg with /dev/sdi
Running fstrim
Unmounting filesystem
Mounting filesystem in degraded mode using /dev/sdi only
mount: /mnt/btrfs: wrong fs type, bad option, bad superblock on /dev/sdi, missing codepage or helper program, or other error.
[19581.748641] BTRFS info (device sdg): dev_replace from /dev/sdg (devid 1) to /dev/sdi started
[19581.803842] BTRFS info (device sdg): dev_replace from /dev/sdg (devid 1) to /dev/sdi finished
[19582.208293] BTRFS info (device sdi): allowing degraded mounts
[19582.208298] BTRFS info (device sdi): disk space caching is enabled
[19582.208301] BTRFS info (device sdi): has skinny extents
[19582.212853] BTRFS warning (device sdi): devid 2 uuid 1f731f47-e1bb-4f00-bfbb-9e5a0cb4ba9f is missing
[19582.213904] btree_readpage_end_io_hook: 25839 callbacks suppressed
[19582.213907] BTRFS error (device sdi): bad tree block start, want 30490624 have 0
[19582.214780] BTRFS warning (device sdi): failed to read root (objectid=7): -5
[19582.231576] BTRFS error (device sdi): open_ctree failed
Failed to mount in degraded mode
So fix by setting all allocated ranges in the replace target device when
the replace operation is finishing, when we are holding the chunk mutex
and we can not race with new chunk allocations.
A test case for fstests follows soon.
Fixes: 1c11b63eff2a67 ("btrfs: replace pending/pinned chunks lists with io tree") CC: stable@vger.kernel.org # 5.2+ Reviewed-by: Nikolay Borisov <nborisov@suse.com> Reviewed-by: Johannes Thumshirn <johannes.thumshirn@wdc.com> Signed-off-by: Filipe Manana <fdmanana@suse.com> Signed-off-by: David Sterba <dsterba@suse.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
which is due to a failure in removing the double poll wait entry if we
hit a wakeup match. This can cause multiple invocations of the wakeup,
which isn't safe.
Commit 6cfcd5563b4f ("clocksource/drivers/timer-ti-dm: Fix suspend and
resume for am3 and am4") exposed a new issue for type2 dual mode timers
on at least omap5 where the clockevent will stop when the SoC starts
entering idle states during the boot.
Turns out we are wrongly first enabling the system timer and then
resetting it, while we must also re-enable it after reset. The current
sequence leaves the timer module in a partially initialized state. This
issue went unnoticed earlier with ti-sysc driver reconfiguring the timer
module until we fixed the issue of ti-sysc reconfiguring system timers.
Let's fix the issue by calling dmtimer_systimer_enable() from reset for
both type1 and type2 timers, and switch the order of reset and enable in
dmtimer_systimer_setup(). Let's also move dmtimer_systimer_enable() and
dmtimer_systimer_disable() to do this without adding forward declarations.
Fixes: 6cfcd5563b4f ("clocksource/drivers/timer-ti-dm: Fix suspend and resume for am3 and am4") Reported-by: H. Nikolaus Schaller <hns@goldelico.com> Signed-off-by: Tony Lindgren <tony@atomide.com> Signed-off-by: Daniel Lezcano <daniel.lezcano@linaro.org> Link: https://lore.kernel.org/r/20200817092428.6176-1-tony@atomide.com Signed-off-by: Sasha Levin <sashal@kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
dm_queue_split() is removed because __split_and_process_bio() _must_
handle splitting bios to ensure proper bio submission and completion
ordering as a bio is split.
Otherwise, multiple recursive calls to ->submit_bio will cause multiple
split bios to be allocated from the same ->bio_split mempool at the same
time. This would result in deadlock in low memory conditions because no
progress could be made (only one bio is available in ->bio_split
mempool).
This fix has been verified to still fix the loss of performance, due
to excess splitting, that commit 120c9257f5f1 provided.
Fixes: 120c9257f5f1 ("Revert "dm: always call blk_queue_split() in dm_process_bio()"") Cc: stable@vger.kernel.org # 5.0+, requires custom backport due to 5.9 changes Reported-by: Ming Lei <ming.lei@redhat.com> Signed-off-by: Mike Snitzer <snitzer@redhat.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
KVM currently assumes that an instruction abort can never be a write.
This is in general true, except when the abort is triggered by
a S1PTW on instruction fetch that tries to update the S1 page tables
(to set AF, for example).
This can happen if the page tables have been paged out and brought
back in without seeing a direct write to them (they are thus marked
read only), and the fault handling code will make the PT executable(!)
instead of writable. The guest gets stuck forever.
In these conditions, the permission fault must be considered as
a write so that the Stage-1 update can take place. This is essentially
the I-side equivalent of the problem fixed by 60e21a0ef54c ("arm64: KVM:
Take S1 walks into account when determining S2 write faults").
Update kvm_is_write_fault() to return true on IABT+S1PTW, and introduce
kvm_vcpu_trap_is_exec_fault() that only return true when no faulting
on a S1 fault. Additionally, kvm_vcpu_dabt_iss1tw() is renamed to
kvm_vcpu_abt_iss1tw(), as the above makes it plain that it isn't
specific to data abort.
If we cancel these requests, we'll leak the memory associated with the
filename. Add them to the table of ops that need cleaning, if
REQ_F_NEED_CLEANUP is set.
In register_mem_sect_under_node() the system_state's value is checked to
detect whether the call is made during boot time or during an hot-plug
operation. Unfortunately, that check against SYSTEM_BOOTING is wrong
because regular memory is registered at SYSTEM_SCHEDULING state. In
addition, memory hot-plug operation can be triggered at this system
state by the ACPI [1]. So checking against the system state is not
enough.
The consequence is that on system with interleaved node's ranges like this:
This can be seen on PowerPC LPAR after multiple memory hot-plug and
hot-unplug operations are done. At the next reboot the node's memory
ranges can be interleaved and since the call to link_mem_sections() is
made in topology_init() while the system is in the SYSTEM_SCHEDULING
state, the node's id is not checked, and the sections registered to
multiple nodes:
$ ls -l /sys/devices/system/memory/memory21/node*
total 0
lrwxrwxrwx 1 root root 0 Aug 24 05:27 node1 -> ../../node/node1
lrwxrwxrwx 1 root root 0 Aug 24 05:27 node2 -> ../../node/node2
In that case, the system is able to boot but if later one of theses
memory blocks is hot-unplugged and then hot-plugged, the sysfs
inconsistency is detected and this is triggering a BUG_ON():
This patch addresses the root cause by not relying on the system_state
value to detect whether the call is due to a hot-plug operation. An
extra parameter is added to link_mem_sections() detailing whether the
operation is due to a hot-plug operation.
[1] According to Oscar Salvador, using this qemu command line, ACPI
memory hotplug operations are raised at SYSTEM_SCHEDULING state:
In that case, we can see memory blocks assigned to multiple nodes in
sysfs:
$ ls -l /sys/devices/system/memory/memory21
total 0
lrwxrwxrwx 1 root root 0 Aug 24 05:27 node1 -> ../../node/node1
lrwxrwxrwx 1 root root 0 Aug 24 05:27 node2 -> ../../node/node2
-rw-r--r-- 1 root root 65536 Aug 24 05:27 online
-r--r--r-- 1 root root 65536 Aug 24 05:27 phys_device
-r--r--r-- 1 root root 65536 Aug 24 05:27 phys_index
drwxr-xr-x 2 root root 0 Aug 24 05:27 power
-r--r--r-- 1 root root 65536 Aug 24 05:27 removable
-rw-r--r-- 1 root root 65536 Aug 24 05:27 state
lrwxrwxrwx 1 root root 0 Aug 24 05:25 subsystem -> ../../../../bus/memory
-rw-r--r-- 1 root root 65536 Aug 24 05:25 uevent
-r--r--r-- 1 root root 65536 Aug 24 05:27 valid_zones
The same applies in the node's directory with a memory21 link in both
the node1 and node2's directory.
This is wrong but doesn't prevent the system to run. However when
later, one of these memory blocks is hot-unplugged and then hot-plugged,
the system is detecting an inconsistency in the sysfs layout and a
BUG_ON() is raised:
The root cause of this issue is that when node's memory is registered,
the range used can overlap another node's range, thus the memory block
is registered to multiple nodes in sysfs.
There are two issues here:
(a) The sysfs memory and node's layouts are broken due to these
multiple links
(b) The link errors in link_mem_sections() should not lead to a system
panic.
To address (a) register_mem_sect_under_node should not rely on the
system state to detect whether the link operation is triggered by a hot
plug operation or not. This is addressed by the patches 1 and 2 of this
series.
Issue (b) will be addressed separately.
This patch (of 2):
The memmap_context enum is used to detect whether a memory operation is
due to a hot-add operation or happening at boot time.
Make it general to the hotplug operation and rename it as
meminit_context.
There is no functional change introduced by this patch
Suggested-by: David Hildenbrand <david@redhat.com> Signed-off-by: Laurent Dufour <ldufour@linux.ibm.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Reviewed-by: David Hildenbrand <david@redhat.com> Reviewed-by: Oscar Salvador <osalvador@suse.de> Acked-by: Michal Hocko <mhocko@suse.com> Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Cc: "Rafael J . Wysocki" <rafael@kernel.org> Cc: Nathan Lynch <nathanl@linux.ibm.com> Cc: Scott Cheloha <cheloha@linux.ibm.com> Cc: Tony Luck <tony.luck@intel.com> Cc: Fenghua Yu <fenghua.yu@intel.com> Cc: <stable@vger.kernel.org> Link: https://lkml.kernel.org/r/20200915094143.79181-1-ldufour@linux.ibm.com Link: https://lkml.kernel.org/r/20200915132624.9723-1-ldufour@linux.ibm.com Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Currently to make sure that every page table entry is read just once
gup_fast walks perform READ_ONCE and pass pXd value down to the next
gup_pXd_range function by value e.g.:
static int gup_pud_range(p4d_t p4d, unsigned long addr, unsigned long end,
unsigned int flags, struct page **pages, int *nr)
...
pudp = pud_offset(&p4d, addr);
This function passes a reference on that local value copy to pXd_offset,
and might get the very same pointer in return. This happens when the
level is folded (on most arches), and that pointer should not be
iterated.
On s390 due to the fact that each task might have different 5,4 or
3-level address translation and hence different levels folded the logic
is more complex and non-iteratable pointer to a local copy leads to
severe problems.
Here is an example of what happens with gup_fast on s390, for a task
with 3-level paging, crossing a 2 GB pud boundary:
// addr = 0x1007ffff000, end = 0x10080001000
static int gup_pud_range(p4d_t p4d, unsigned long addr, unsigned long end,
unsigned int flags, struct page **pages, int *nr)
{
unsigned long next;
pud_t *pudp;
// pud_offset returns &p4d itself (a pointer to a value on stack)
pudp = pud_offset(&p4d, addr);
do {
// on second iteratation reading "random" stack value
pud_t pud = READ_ONCE(*pudp);
// next = 0x10080000000, due to PUD_SIZE/MASK != PGDIR_SIZE/MASK on s390
next = pud_addr_end(addr, end);
...
} while (pudp++, addr = next, addr != end); // pudp++ iterating over stack
return 1;
}
This happens since s390 moved to common gup code with commit d1874a0c2805 ("s390/mm: make the pxd_offset functions more robust") and
commit 1a42010cdc26 ("s390/mm: convert to the generic
get_user_pages_fast code").
s390 tried to mimic static level folding by changing pXd_offset
primitives to always calculate top level page table offset in pgd_offset
and just return the value passed when pXd_offset has to act as folded.
What is crucial for gup_fast and what has been overlooked is that
PxD_SIZE/MASK and thus pXd_addr_end should also change correspondingly.
And the latter is not possible with dynamic folding.
To fix the issue in addition to pXd values pass original pXdp pointers
down to gup_pXd_range functions. And introduce pXd_offset_lockless
helpers, which take an additional pXd entry value parameter. This has
already been discussed in
SWP_FS is used to make swap_{read,write}page() go through the
filesystem, and it's only used for swap files over NFS. So, !SWP_FS
means non NFS for now, it could be either file backed or device backed.
Something similar goes with legacy SWP_FILE.
So in order to achieve the goal of the original patch, SWP_BLKDEV should
be used instead.
FS corruption can be observed with SSD device + XFS + fragmented
swapfile due to CONFIG_THP_SWAP=y.
I reproduced the issue with the following details:
NULL pointer dereference is observed while exporting the dmabuf but
failed to allocate the 'struct file' which results into the dropping of
the allocated dentry corresponding to this file in the dmabuf fs, which
is ending up in dma_buf_release() and accessing the uninitialzed
dentry->d_fsdata.
It was missed when I was forking Loongson2ef from Loongson64 but
should be applied to Loongson2ef as march=loongson2f
will also enable Loongson MMI in GCC-9+.
Signed-off-by: Jiaxun Yang <jiaxun.yang@flygoat.com> Fixes: 71e2f4dd5a65 ("MIPS: Fork loongson2ef from loongson64") Reported-by: Thomas Bogendoerfer <tsbogend@alpha.franken.de> Cc: stable@vger.kernel.org # v5.8+ Signed-off-by: Thomas Bogendoerfer <tsbogend@alpha.franken.de> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
For some inexplicable reason I decided to call flush_scheduled_work()
instead of cancel_delayed_work_sync(). The problem with that is that
flush_scheduled_work() waits for *all* queued scheduled work to be
completed instead of just the work itself.
This can cause a deadlock if a CEC driver also schedules work that
takes the same lock. See the comments for flush_scheduled_work() in
linux/workqueue.h.
This is exactly what has been observed a few times.
This patch simply replaces flush_scheduled_work() by
cancel_delayed_work_sync().
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl> Cc: <stable@vger.kernel.org> # for v5.8 and up Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The overflow happens, because in btree_readpage_end_io_hook() we assume
that we have found a 4 byte checksum instead of the real possible 32
bytes we have for the checksums.
With the fix applied:
[ 35.726623] BTRFS: device fsid 815caf9a-dc43-4d2a-ac54-764b8333d765 devid 1 transid 5 /dev/loop0 scanned by syz-repro (215)
[ 35.738994] BTRFS info (device loop0): disk space caching is enabled
[ 35.738998] BTRFS info (device loop0): has skinny extents
[ 35.743337] BTRFS warning (device loop0): loop0 checksum verify failed on 1052672 wanted 0xf9c035fc8d239a54 found 0x67a25c14b7eabcf9 level 0
[ 35.743420] BTRFS error (device loop0): failed to read chunk root
[ 35.745899] BTRFS error (device loop0): open_ctree failed
Reported-by: syzbot+e864a35d361e1d4e29a5@syzkaller.appspotmail.com Fixes: d5178578bcd4 ("btrfs: directly call into crypto framework for checksumming") CC: stable@vger.kernel.org # 5.4+ Signed-off-by: Johannes Thumshirn <johannes.thumshirn@wdc.com> Signed-off-by: David Sterba <dsterba@suse.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
projects / linux-stable / log