All 3 properties are required by sram.yaml. Fixes the dtbs_check warning:
sram@900000: '#address-cells' is a required property
sram@900000: '#size-cells' is a required property
sram@900000: 'ranges' is a required property
All 3 properties are required by sram.yaml. Fixes the dtbs_check warning:
sram@900000: '#address-cells' is a required property
sram@900000: '#size-cells' is a required property
sram@900000: 'ranges' is a required property
Use the general touchscreen method to config the max pressure for
touch tsc2046(data sheet suggest 8 bit pressure), otherwise, for
ABS_PRESSURE, when config the same max and min value, weston will
meet the following issue,
[17:19:39.183] event1 - ADS7846 Touchscreen: is tagged by udev as: Touchscreen
[17:19:39.183] event1 - ADS7846 Touchscreen: kernel bug: device has min == max on ABS_PRESSURE
[17:19:39.183] event1 - ADS7846 Touchscreen: was rejected
[17:19:39.183] event1 - not using input device '/dev/input/event1'
This will then cause the APP weston-touch-calibrator can't list touch devices.
root@imx6ul7d:~# weston-touch-calibrator
could not load cursor 'dnd-move'
could not load cursor 'dnd-copy'
could not load cursor 'dnd-none'
No devices listed.
And accroding to binding Doc, "ti,x-max", "ti,y-max", "ti,pressure-max"
belong to the deprecated properties, so remove them. Also for "ti,x-min",
"ti,y-min", "ti,x-plate-ohms", the value set in dts equal to the default
value in driver, so are redundant, also remove here.
In most cases this is related to fsl,err006687-workaround-present, which
requires a GPIO interrupt next a GIC interrupt.
This fixes the dtbs_check warning:
imx6dl-mba6a.dtb: ethernet@2188000: More than one condition true in oneOf schema:
{'$filename': 'Documentation/devicetree/bindings/net/fsl,fec.yaml',
[...]
This was fixed in initialize_cpsch before, but not in initialize_nocpsch.
Factor sdma bitmap initialization into a helper function to apply the
correct implementation in both cases without duplicating it.
v2: Added a range check
Reported-by: Ellis Michael <ellis@ellismichael.com> Signed-off-by: Felix Kuehling <Felix.Kuehling@amd.com> Reviewed-by: Graham Sider <Graham.Sider@amd.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
[why]
vid stream control is double bufferred, if we don't wait for video
stream enable set to 0, we may get temporary image corruption
showing on the stream when setting PIXEL_TO_SYMBOL_FIFO_ENABLE to 0.
Reviewed-by: Ariel Bernstein <Eric.Bernstein@amd.com> Acked-by: Jasdeep Dhillon <jdhillon@amd.com> Signed-off-by: Wenjing Liu <wenjing.liu@amd.com> Tested-by: Daniel Wheeler <daniel.wheeler@amd.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
The sequence for Source DP PHY CTS automation is [2][1]:
1- Emulate successful Link Training(LT)
2- Short HPD and change link rates and number of lanes by LT.
(This is same flow for Link Layer CTS)
3- Short HPD and change PHY test pattern and swing/pre-emphasis
levels (This step should not trigger LT)
The problem is with DP PHY compliance setup as follow:
At step 3, before writing TRAINING_LANEx_SET/LINK_QUAL_PATTERN_SET
to declare the pattern/swing requested by scope, we write link
config in LINK_BW_SET/LANE_COUNT_SET on a port that has LTTPR.
As LTTPR snoops aux transaction, LINK_BW_SET/LANE_COUNT_SET writes
indicate a LT will start [Check DP 2.0 E11 -Sec 3.6.8.2 & 3.6.8.6.3],
and LTTPR will reset the link and stop sending DP signals to
DPTX/Scope causing the measurements to fail. Note that step 3 will
not trigger LT and DP link will never recovered by the
Aux Emulator/Scope.
The reset of link can be tested with a monitor connected to LTTPR
port simply by writing to LINK_BW_SET or LANE_COUNT_SET as follow
This single aux write causes the screen to blank, sending short HPD to
DPTX, setting LINK_STATUS_UPDATE = 1 in DPCD 0x204, and triggering LT.
As stated in [1]:
"Before any TX electrical testing can be performed, the link between a
DPTX and DPRX (in this case, a piece of test equipment), including all
LTTPRs within the path, shall be trained as defined in this Standard."
In addition, changing Phy pattern/Swing/Pre-emphasis (Step 3) uses the
same link rate and lane count applied on step 2, so no need to redo LT.
The fix is to not rewrite link config in step 3, and just writes
TRAINING_LANEx_SET and LINK_QUAL_PATTERN_SET
drm bridges added by meson_encoder_hdmi_init and meson_encoder_cvbs_init
were not manually removed at module unload time, which caused dangling
references to freed memory to remain linked in the global bridge_list.
When loading the driver modules back in, the same functions would again
call drm_bridge_add, and when traversing the global bridge_list, would
end up peeking into freed memory.
Once again KASAN revealed the problem:
[ +0.000095] =============================================================
[ +0.000008] BUG: KASAN: use-after-free in __list_add_valid+0x9c/0x120
[ +0.000018] Read of size 8 at addr ffff00003da291f0 by task modprobe/2483
[ +0.000013] The buggy address belongs to the object at ffff00003da29000
which belongs to the cache kmalloc-1k of size 1024
[ +0.000008] The buggy address is located 496 bytes inside of
1024-byte region [ffff00003da29000, ffff00003da29400)
Fix by keeping track of which encoders were initialised in the meson_drm
structure and manually removing their bridges at aggregate driver's unbind
time.
Because component_master_del wasn't being called when unloading the
meson_drm module, the aggregate device would linger forever in the global
aggregate_devices list. That means when unloading and reloading the
meson_dw_hdmi module, component_add would call into
try_to_bring_up_aggregate_device and find the unbound meson_drm aggregate
device.
This would in turn dereference some of the aggregate_device's struct
entries which point to memory automatically freed by the devres API when
unbinding the aggregate device from meson_drv_unbind, and trigger an
use-after-free bug:
[ +0.000013] The buggy address belongs to the object at ffff000006731600
which belongs to the cache kmalloc-256 of size 256
[ +0.000009] The buggy address is located 136 bytes inside of
256-byte region [ffff000006731600, ffff000006731700)
[ +0.000011] Memory state around the buggy address:
[ +0.000007] ffff000006731580: fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc
[ +0.000007] ffff000006731600: fa fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb
[ +0.000007] >ffff000006731680: fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb
[ +0.000007] ^
[ +0.000006] ffff000006731700: fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc
[ +0.000007] ffff000006731780: fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc
[ +0.000006] ==================================================================
Fix by adding 'remove' driver callback for meson-drm, and explicitly deleting the
aggregate device.
[ +0.000014] The buggy address belongs to the object at ffff000020c39000
which belongs to the cache kmalloc-4k of size 4096
[ +0.000008] The buggy address is located 1504 bytes inside of
4096-byte region [ffff000020c39000, ffff000020c3a000)
The reason this is happening is unloading meson-dw-hdmi will cause the
component API to take down the aggregate device, which in turn will cause
all devres-managed memory to be freed, including the struct dw_hdmi
allocated in dw_hdmi_probe. This struct embeds a struct drm_bridge that is
added at the end of the function, and which is later on picked up in
meson_encoder_hdmi_init.
However, when attaching the bridge to the encoder created in
meson_encoder_hdmi_init, it's linked to the encoder's bridge chain, from
where it never leaves, even after devres_release_group is called when the
driver's components are unbound and the embedding structure freed.
Then, when calling drm_dev_put in the aggregate driver's unbind function,
drm_bridge_detach is called for every single bridge linked to the encoder,
including the one whose memory had already been deallocated.
Fix by calling component_unbind_all after drm_dev_put.
[Why]
Hostvm should be enabled/disabled accordding to
the status of riommu_active, but hostvm always
be disabled on DCN31 which causes underflow
[How]
Set correct hostvm flag on DCN31
Reviewed-by: Charlene Liu <Charlene.Liu@amd.com> Acked-by: Wayne Lin <wayne.lin@amd.com> Signed-off-by: Sherry Wang <Yao.Wang1@amd.com> Tested-by: Daniel Wheeler <daniel.wheeler@amd.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
Signed-off-by: Philip Yang <Philip.Yang@amd.com> Suggested-by: Felix Kuehling <felix.kuehling@amd.com> Reviewed-by: Christian König <christian.koenig@amd.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
Some Intel-based platforms rely on a topology file that hard-codes the
use of MCLK0. This is incorrect in 10% of the cases. Rather than
generating yet another set of topology files, this patch adds a kernel
module parameter to override the topology value.
In hindsight, we should never have allowed mclks to be specified in
topology, this is a hardware-level information that should not have
been visible in the topology.
Future patches will try to set this value automagically, e.g. by
parsing the NHLT content.
Signed-off-by: Pierre-Louis Bossart <pierre-louis.bossart@linux.intel.com> Reviewed-by: Kai Vehmanen <kai.vehmanen@linux.intel.com> Reviewed-by: Bard Liao <yung-chuan.liao@linux.intel.com> Link: https://lore.kernel.org/r/20220919115350.43104-3-pierre-louis.bossart@linux.intel.com Signed-off-by: Mark Brown <broonie@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
On a MSI S270 with Fedora 37 x86_64 / systemd-251.4 the module does not
properly autoload.
This is likely caused by issues with how systemd-udevd handles the single
quote char (') which is part of the sys_vendor / chassis_vendor strings
on this laptop. As a workaround remove the single quote char + everything
behind it from the sys_vendor + chassis_vendor matches. This fixes
the module not autoloading.
Error 0x06 (invalid command parameter) is reported by hp-wmi module
when reading the current thermal profile and then proceed to set it
back. The failing condition occurs in Linux NixOS after user
configures the thermal profile to ‘quiet mode’ in Windows. Quiet Fan
Mode is supported in Windows but was not supported in hp-wmi module.
This fix adds support for PLATFORM_PROFILE_QUIET in hp-wmi module for
HP notebooks other than HP Omen series. Quiet thermal profile is not
supported in HP Omen series notebooks.
Signed-off-by: Jorge Lopez <jorge.lopez2@hp.com> Link: https://lore.kernel.org/r/20220912192603.4001-1-jorge.lopez2@hp.com Reviewed-by: Hans de Goede <hdegoede@redhat.com> Signed-off-by: Hans de Goede <hdegoede@redhat.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
cros_ec_handle_event in the cros_ec driver can notify the PM of wake
events. When a device is suspended, cros_ec_handle_event will not check
MKBP events. Instead, received MKBP events are checked during resume by
cros_ec_report_events_during_suspend. But
cros_ec_report_events_during_suspend cannot notify the PM if received
events are wake events, causing wake events to not be reported if
received while the device is suspended.
Update cros_ec_report_events_during_suspend to notify the PM of wake
events during resume by calling pm_wakeup_event.
This commit fixes vertical timings of the VEC (composite output) modes
to accurately represent the 525-line ("NTSC") and 625-line ("PAL") ITU-R
standards.
Previous timings were actually defined as 502 and 601 lines, resulting
in non-standard 62.69 Hz and 52 Hz signals being generated,
respectively.
The USB-audio driver matches per interface, and as default, it
registers the card instance at the very first instance. This can be a
problem for the devices that have multiple interfaces to be probed, as
the udev rule isn't applied properly for the later appearing
interfaces. Although we introduced the delayed_register option and
the quirks for covering those shortcomings, it's nothing but a
workaround for specific devices.
This patch is an another attempt to fix the problem in a more generic
way. Now the driver checks the whole USB device descriptor at the
very first time when an interface is attached to a sound card. It
looks at each matching interface in the descriptor and remembers the
last matching one. The snd_card_register() is invoked only when this
last interface is probed.
After this change, the quirks for the delayed registration become
superfluous, hence they are removed along with the patch. OTOH, the
delayed_register option is still kept, as it might be useful for some
corner cases (e.g. a special driver overtakes the interface probe from
the standard driver, and the last interface probe may miss).
[Why]
There is no CG(Clock Gating)/PG(Power Gating) requirement on SRIOV VF.
For multi VF, VF should not enable any CG/PG features.
For one VF, PF will program CG/PG related registers.
[How]
Do not set any cg/pg flag bit at early init under sriov.
Acked-by: Christian König <christian.koenig@amd.com> Signed-off-by: Yifan Zha <Yifan.Zha@amd.com> Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
There are two events that signal a real change of the link state: HPD going
high means the sink is newly connected or wants the source to re-read the
EDID, RX sense going low is a indication that the link has been disconnected.
Ignore the other two events that also trigger interrupts, but don't need
immediate attention: HPD going low does not necessarily mean the link has
been lost and should not trigger a immediate read of the status. RX sense
going high also does not require a detect cycle, as HPD going high is the
right point in time to read the EDID.
When userspace tries to map the dmabuf and if for some reason
(e.g. OOM) the creation of the sg table fails, ubuf->sg needs to be
set to NULL. Otherwise, when the userspace subsequently closes the
dmabuf fd, we'd try to erroneously free the invalid sg table from
release_udmabuf resulting in the following crash reported by syzbot:
The definition of MIN_I64 in bw_fixed.c can cause gcc to whinge about
integer overflow, because it is treated as a positive value, which is
then negated. The temporary positive value is not necessarily
representable.
This causes the following warning:
../drivers/gpu/drm/amd/amdgpu/../display/dc/dml/calcs/bw_fixed.c:30:19:
warning: integer overflow in expression ‘-9223372036854775808’ of type
‘long long int’ results in ‘-9223372036854775808’ [-Woverflow]
30 | (int64_t)(-(1LL << 63))
| ^
Writing out (-MAX_I64 - 1) works instead.
Signed-off-by: David Gow <davidgow@google.com> Signed-off-by: Tales Aparecida <tales.aparecida@gmail.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
A NULL check for bridge->encoder shows that it may be NULL, but it
already been dereferenced on all paths leading to the check.
812 if (!bridge->encoder) {
Dereference the pointer bridge->encoder.
810 drm_connector_attach_encoder(<9611->connector, bridge->encoder);
Komeda driver relies on the generic DRM atomic helper functions to handle
commits. It only implements an atomic_commit_tail hook for the
mode_config_helper_funcs and even that one is pretty close to the generic
implementation with the exception of additional dma_fence signalling.
What the generic helper framework doesn't do is waiting for the actual
hardware to signal that the commit parameters have been written into the
appropriate registers. As we signal CRTC events only on the irq handlers,
we need to flush the configuration and wait for the hardware to respond.
Add the Komeda specific implementation for atomic_commit_hw_done() that
flushes and waits for flip done before calling drm_atomic_helper_commit_hw_done().
The fix was prompted by a patch from Carsten Haitzler where he was trying to
solve the same issue but in a different way that I think can lead to wrong
event signaling to userspace.
The strlen() function returns a size_t which is an unsigned int on 32-bit
arches and an unsigned long on 64-bit arches. But in the drm_copy_field()
function, the strlen() return value is assigned to an 'int len' variable.
Later, the len variable is passed as copy_from_user() third argument that
is an unsigned long parameter as well.
In theory, this can lead to an integer overflow via type conversion. Since
the assignment happens to a signed int lvalue instead of a size_t lvalue.
In practice though, that's unlikely since the values copied are set by DRM
drivers and not controlled by userspace. But using a size_t for len is the
correct thing to do anyways.
nouveau_bo_alloc() allocates a memory chunk for "nvbo" with kzalloc().
When some error occurs, "nvbo" should be released. But when
WARN_ON(pi < 0)) equals true, the function return ERR_PTR without
releasing the "nvbo", which will lead to a memory leak.
We should release the "nvbo" with kfree() if WARN_ON(pi < 0)) equals true.
The code is ogranized in a way that all related parts
to the certain platform quirk go together. This is not
the case for AMD NAVI. Shuffle code to make it happen.
While at it, drop the frequency definition and use
hard coded value as it's done for other platforms and
add a comment to the PCI ID list.
Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com> Acked-by: Jarkko Nikula <jarkko.nikula@linux.intel.com> Signed-off-by: Wolfram Sang <wsa@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
Martynas reported bpf_get_func_ip returning +4 address when
CONFIG_X86_KERNEL_IBT option is enabled.
When CONFIG_X86_KERNEL_IBT is enabled we'll have endbr instruction
at the function entry, which screws return value of bpf_get_func_ip()
helper that should return the function address.
There's short term workaround for kprobe_multi bpf program made by
Alexei [1], but we need this fixup also for bpf_get_attach_cookie,
that returns cookie based on the entry_ip value.
Moving the fixup in the fprobe handler, so both bpf_get_func_ip
and bpf_get_attach_cookie get expected function address when
CONFIG_X86_KERNEL_IBT option is enabled.
Also renaming kprobe_multi_link_handler entry_ip argument to fentry_ip
so it's clearer this is an ftrace __fentry__ ip.
While waiting for memory in thread1, the socket is released with its wait
queue because thread2 has closed it. This caused by tcp_bpf_send_verdict
didn't increase the f_count of psock->sk_redir->sk_socket->file in thread1.
We should check if SOCK_DEAD flag is set on wakeup in sk_stream_wait_memory
before accessing the wait queue.
Suggested-by: Jakub Sitnicki <jakub@cloudflare.com> Signed-off-by: Liu Jian <liujian56@huawei.com> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: John Fastabend <john.fastabend@gmail.com> Cc: Eric Dumazet <edumazet@google.com> Link: https://lore.kernel.org/bpf/20220823133755.314697-2-liujian56@huawei.com Signed-off-by: Sasha Levin <sashal@kernel.org>
On 32-bit platforms, long is 32 bits, so (long)UINT_MAX is less than
(long)SHT4X_MIN_POLL_INTERVAL, which means the clamping operation is
bogus. Fix this by clamping at INT_MAX, so that the upperbound is the
same on all platforms.
Instead of using the default value 33 (pci), set US_CYC_CNT init based
on Programming guide:
If available, set chipset bus clock with fallback to cpu clock/3.
Any change to the hardware timestamps configuration triggers nic restart,
which breaks transmition and reception of network packets for a while.
But there is no need to fully restart the device because while configuring
hardware timestamps. The code for changing configuration runs after all
of the initialisation, when the NIC is actually up and running. This patch
changes the code that ioctl will only update configuration registers and
will not trigger carrier status change, but in case of timestamps for
all rx packetes it fallbacks to close()/open() sequnce because of
synchronization issues in the hardware. Tested on BCM57504.
Cc: Richard Cochran <richardcochran@gmail.com> Signed-off-by: Vadim Fedorenko <vfedorenko@novek.ru> Reviewed-by: Michael Chan <michael.chan@broadcom.com> Link: https://lore.kernel.org/r/20220922191038.29921-1-vfedorenko@novek.ru Signed-off-by: Jakub Kicinski <kuba@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
syzbot is reporting cancel_delayed_work() without INIT_DELAYED_WORK() at
l2cap_chan_del() [1], for CONF_NOT_COMPLETE flag (which meant to prevent
l2cap_chan_del() from calling cancel_delayed_work()) is cleared by timer
which fires before l2cap_chan_del() is called by closing file descriptor
created by socket(AF_BLUETOOTH, SOCK_STREAM, BTPROTO_L2CAP).
l2cap_bredr_sig_cmd(L2CAP_CONF_REQ) and l2cap_bredr_sig_cmd(L2CAP_CONF_RSP)
are calling l2cap_ertm_init(chan), and they call l2cap_chan_ready() (which
clears CONF_NOT_COMPLETE flag) only when l2cap_ertm_init(chan) succeeded.
l2cap_sock_init() does not call l2cap_ertm_init(chan), and it instead sets
CONF_NOT_COMPLETE flag by calling l2cap_chan_set_defaults(). However, when
connect() is requested, "command 0x0409 tx timeout" happens after 2 seconds
from connect() request, and CONF_NOT_COMPLETE flag is cleared after 4
seconds from connect() request, for l2cap_conn_start() from
l2cap_info_timeout() callback scheduled by
Fix this problem by initializing delayed works used by L2CAP_MODE_ERTM
mode as soon as l2cap_chan_create() allocates a channel, like I did in
commit be8597239379f0f5 ("Bluetooth: initialize skb_queue_head at
l2cap_chan_create()").
Link: https://syzkaller.appspot.com/bug?extid=83672956c7aa6af698b3 Reported-by: syzbot <syzbot+83672956c7aa6af698b3@syzkaller.appspotmail.com> Signed-off-by: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp> Signed-off-by: Luiz Augusto von Dentz <luiz.von.dentz@intel.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
In monitor mode we should be able to received all packets even if it's not
destined to us. But after scan, the configuration was wrongly set, so we
fix it.
We should reset mstat->airtime_ac along with clear up the entries in the
hardware WLAN table for the Rx and Rx accumulative airtime. Otherwsie, the
value msta->airtime_ac - [tx, rx]_last may be a negative and that is not
the actual airtime the device took in the last run.
Reported-by: YN Chen <YN.Chen@mediatek.com> Signed-off-by: Sean Wang <sean.wang@mediatek.com> Signed-off-by: Felix Fietkau <nbd@nbd.name> Signed-off-by: Sasha Levin <sashal@kernel.org>
mhi_alloc_controller() allocates a memory space for mhi_ctrl. When gets
some error, mhi_ctrl should be freed with mhi_free_controller(). But
when ath11k_mhi_read_addr_from_dt() fails, the function returns without
calling mhi_free_controller(), which will lead to a memory leak.
We can fix it by calling mhi_free_controller() when
ath11k_mhi_read_addr_from_dt() fails.
By using a ratio of delay to poll_enabled_time that is not integer
time_remaining underflows and does not exit the loop as expected.
As delay could be derived from DT and poll_enabled_time is defined
in the driver this can easily happen.
Use a signed iterator to make sure that the loop exits once
the remaining time is negative.
HarrrisonPeak, CyclonePeak, SnowFieldPeak and SandyPeak controllers
are marked to support HCI_QUIRK_LE_STATES.
Signed-off-by: Kiran K <kiran.k@intel.com> Signed-off-by: Chethan T N <chethan.tumkur.narayan@intel.com> Signed-off-by: Luiz Augusto von Dentz <luiz.von.dentz@intel.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
During stress tests with adding VF to namespace and changing vf's
trust there was a race between iavf_reset_task and iavf_close.
Sometimes when IAVF_FLAG_AQ_DISABLE_QUEUES from iavf_close was sent
to PF after reset and before IAVF_AQ_GET_CONFIG was sent then PF
returns error IAVF_NOT_SUPPORTED to disable queues request and
following requests. There is need to get_config before other
aq_required will be send but iavf_close clears all flags, if
get_config was not sent before iavf_close, then it will not be send
at all.
In case when IAVF_FLAG_AQ_GET_OFFLOAD_VLAN_V2_CAPS was sent before
IAVF_FLAG_AQ_DISABLE_QUEUES then there was rtnl_lock deadlock
between iavf_close and iavf_adminq_task until iavf_close timeouts
and disable queues was sent after iavf_close ends.
There was also a problem with sending delete/add filters.
Sometimes when filters was not yet added to PF and in
iavf_close all filters was set to remove there might be a try
to remove nonexistent filters on PF.
Add aq_required_tmp to save aq_required flags and send them after
disable_queues will be handled. Clear flags given to iavf_down
different than IAVF_FLAG_AQ_GET_CONFIG as this flag is necessary
to sent other aq_required. Remove some flags that we don't
want to send as we are in iavf_close and we want to disable
interface. Remove filters which was not yet sent and send del
filters flags only when there are filters to remove.
Signed-off-by: Michal Jaron <michalx.jaron@intel.com> Signed-off-by: Mateusz Palczewski <mateusz.palczewski@intel.com> Tested-by: Konrad Jankowski <konrad0.jankowski@intel.com> Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
When the system shuts down, SMMU driver will be stopped and
will not assist in IOVA translations. SMMU driver expects all
of its consumers to shutdown before shutting down itself.
WCN6750 being one of the consumer device should not perform any
DMA operations after the SMMU has shutdown which will otherwise
result in SMMU faults.
SMMU driver will call the shutdown() callback of all its
consumer devices and the consumers shall stop further DMA
activity after the invocation of their respective shutdown()
callbacks.
Register the shutdown() callback to the platform core for WCN6750.
Change will not impact other AHB ath11k devices.
Currently if ipcomp_alloc_scratches() fails to allocate memory
ipcomp_scratches holds obsolete address. So when we try to free the
percpu scratches using ipcomp_free_scratches() it tries to vfree non
existent vm area. Described below:
static void * __percpu *ipcomp_alloc_scratches(void)
{
...
scratches = alloc_percpu(void *);
if (!scratches)
return NULL;
ipcomp_scratches does not know about this allocation failure.
Therefore holding the old obsolete address.
...
}
The IP_UNICAST_IF socket option is used to set the outgoing interface
for outbound packets.
The IP_UNICAST_IF socket option was added as it was needed by the
Wine project, since no other existing option (SO_BINDTODEVICE socket
option, IP_PKTINFO socket option or the bind function) provided the
needed characteristics needed by the IP_UNICAST_IF socket option. [1]
The IP_UNICAST_IF socket option works well for unconnected sockets,
that is, the interface specified by the IP_UNICAST_IF socket option
is taken into consideration in the route lookup process when a packet
is being sent. However, for connected sockets, the outbound interface
is chosen when connecting the socket, and in the route lookup process
which is done when a packet is being sent, the interface specified by
the IP_UNICAST_IF socket option is being ignored.
This inconsistent behavior was reported and discussed in an issue
opened on systemd's GitHub project [2]. Also, a bug report was
submitted in the kernel's bugzilla [3].
To understand the problem in more detail, we can look at what happens
for UDP packets over IPv4 (The same analysis was done separately in
the referenced systemd issue).
When a UDP packet is sent the udp_sendmsg function gets called and
the following happens:
1. The oif member of the struct ipcm_cookie ipc (which stores the
output interface of the packet) is initialized by the ipcm_init_sk
function to inet->sk.sk_bound_dev_if (the device set by the
SO_BINDTODEVICE socket option).
2. If the IP_PKTINFO socket option was set, the oif member gets
overridden by the call to the ip_cmsg_send function.
3. If no output interface was selected yet, the interface specified
by the IP_UNICAST_IF socket option is used.
4. If the socket is connected and no destination address is
specified in the send function, the struct ipcm_cookie ipc is not
taken into consideration and the cached route, that was calculated in
the connect function is being used.
Thus, for a connected socket, the IP_UNICAST_IF sockopt isn't taken
into consideration.
This patch corrects the behavior of the IP_UNICAST_IF socket option
for connect()ed sockets by taking into consideration the
IP_UNICAST_IF sockopt when connecting the socket.
In order to avoid reconnecting the socket, this option is still
ignored when applied on an already connected socket until connect()
is called again by the Richard Gobert.
Change the __ip4_datagram_connect function, which is called during
socket connection, to take into consideration the interface set by
the IP_UNICAST_IF socket option, in a similar way to what is done in
the udp_sendmsg function.
The RX and TX byte/packet statistics in this driver could be overflowed
relatively quickly on a 32-bit platform. Switch these stats to use the
u64_stats infrastructure to avoid this.
The APIC supports two modes, legacy APIC (or xAPIC), and Extended APIC
(or x2APIC). X2APIC mode is mostly compatible with legacy APIC, but
it disables the memory-mapped APIC interface in favor of one that uses
MSRs. The APIC mode is controlled by the EXT bit in the APIC MSR.
The MMIO/xAPIC interface has some problems, most notably the APIC LEAK
[1]. This bug allows an attacker to use the APIC MMIO interface to
extract data from the SGX enclave.
Introduce support for a new feature that will allow the BIOS to lock
the APIC in x2APIC mode. If the APIC is locked in x2APIC mode and the
kernel tries to disable the APIC or revert to legacy APIC mode a GP
fault will occur.
Introduce support for a new MSR (IA32_XAPIC_DISABLE_STATUS) and handle
the new locked mode when the LEGACY_XAPIC_DISABLED bit is set by
preventing the kernel from trying to disable the x2APIC.
On platforms with the IA32_XAPIC_DISABLE_STATUS MSR, if SGX or TDX are
enabled the LEGACY_XAPIC_DISABLED will be set by the BIOS. If
legacy APIC is required, then it SGX and TDX need to be disabled in the
BIOS.
[1]: https://aepicleak.com/aepicleak.pdf
Signed-off-by: Daniel Sneddon <daniel.sneddon@linux.intel.com> Signed-off-by: Dave Hansen <dave.hansen@linux.intel.com> Acked-by: Dave Hansen <dave.hansen@linux.intel.com> Tested-by: Neelima Krishnan <neelima.krishnan@intel.com> Link: https://lkml.kernel.org/r/20220816231943.1152579-1-daniel.sneddon@linux.intel.com Signed-off-by: Sasha Levin <sashal@kernel.org>
As we are now enabling full end-to-end flow control to the Thunderbolt
networking driver, in order for it to work properly on second generation
Thunderbolt hardware (Falcon Ridge), we need to add back the workaround
that was removed with commit 53f13319d131 ("thunderbolt: Get rid of E2E
workaround"). However, this time we only apply it for Falcon Ridge
controllers as a form of an additional quirk. For non-Falcon Ridge this
does nothing.
While there fix a typo 'reqister' -> 'register' in the comment.
Signed-off-by: Mika Westerberg <mika.westerberg@linux.intel.com> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Sasha Levin <sashal@kernel.org>
syzbot is reporting uninit value at ath9k_htc_rx_msg() [1], for
ioctl(USB_RAW_IOCTL_EP_WRITE) can call ath9k_hif_usb_rx_stream() with
pkt_len = 0 but ath9k_hif_usb_rx_stream() uses
__dev_alloc_skb(pkt_len + 32, GFP_ATOMIC) based on an assumption that
pkt_len is valid. As a result, ath9k_hif_usb_rx_stream() allocates skb
with uninitialized memory and ath9k_htc_rx_msg() is reading from
uninitialized memory.
Since bytes accessed by ath9k_htc_rx_msg() is not known until
ath9k_htc_rx_msg() is called, it would be difficult to check minimal valid
pkt_len at "if (pkt_len > 2 * MAX_RX_BUF_SIZE) {" line in
ath9k_hif_usb_rx_stream().
We have two choices. One is to workaround by adding __GFP_ZERO so that
ath9k_htc_rx_msg() sees 0 if pkt_len is invalid. The other is to let
ath9k_htc_rx_msg() validate pkt_len before accessing. This patch chose
the latter.
Note that I'm not sure threshold condition is correct, for I can't find
details on possible packet length used by this protocol.
When memory poison consumption machine checks fire, MCE notifier
handlers like nfit_handle_mce() record the impacted physical address
range which is reported by the hardware in the MCi_MISC MSR. The error
information includes data about blast radius, i.e. how many cachelines
did the hardware determine are impacted. A recent change
7917f9cdb503 ("acpi/nfit: rely on mce->misc to determine poison granularity")
updated nfit_handle_mce() to stop hard coding the blast radius value of
1 cacheline, and instead rely on the blast radius reported in 'struct
mce' which can be up to 4K (64 cachelines).
It turns out that apei_mce_report_mem_error() had a similar problem in
that it hard coded a blast radius of 4K rather than reading the blast
radius from the error information. Fix apei_mce_report_mem_error() to
convey the proper poison granularity.
tcp_md5sig_pool_populated can be read while another thread
changes its value.
The race has no consequence because allocations
are protected with tcp_md5sig_mutex.
This patch adds READ_ONCE() and WRITE_ONCE() to document
the race and silence KCSAN.
Reported-by: Abhishek Shah <abhishek.shah@columbia.edu> Signed-off-by: Eric Dumazet <edumazet@google.com> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Sasha Levin <sashal@kernel.org>
Currently queue_userspace_packet will call kfree_skb for all frames,
whether or not an error occurred. This can result in a single dropped
frame being reported as multiple drops in dropwatch. This functions
caller may also call kfree_skb in case of an error. This patch will
consume the skbs instead and allow caller's to use kfree_skb.
Frames sent to userspace can be reported as dropped in
ovs_dp_process_packet, however, if they are dropped in the netlink code
then netlink_attachskb will report the same frame as dropped.
This patch checks for error codes which indicate that the frame has
already been freed.
On the CPSW and ICSS peripherals, there is a possibility that the MDIO
interface returns corrupt data on MDIO reads or writes incorrect data
on MDIO writes. There is also a possibility for the MDIO interface to
become unavailable until the next peripheral reset.
The workaround is to configure the MDIO in manual mode and disable the
MDIO state machine and emulate the MDIO protocol by reading and writing
appropriate fields in MDIO_MANUAL_IF_REG register of the MDIO controller
to manipulate the MDIO clock and data pins.
More details about the errata i2329 and the workaround is available in:
https://www.ti.com/lit/er/sprz487a/sprz487a.pdf
Add implementation to disable MDIO state machine, configure MDIO in manual
mode and achieve MDIO read and writes via MDIO Bitbanging
Signed-off-by: Ravi Gunasekaran <r-gunasekaran@ti.com> Reported-by: kernel test robot <lkp@intel.com> Reviewed-by: Andrew Lunn <andrew@lunn.ch> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Sasha Levin <sashal@kernel.org>
When the user changes the number of queues via ethtool, the driver
allocates new rings. This allocation did not initialize tx_tstamps. This
results in the tx_tstamps field being zero (due to kcalloc allocation), and
would result in a NULL pointer dereference when attempting a transmit
timestamp on the new ring.
Signed-off-by: Jacob Keller <jacob.e.keller@intel.com> Tested-by: Gurucharan <gurucharanx.g@intel.com> (A Contingent worker at Intel) Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
When bpftool is linked against libcap, the library runs a "constructor"
function to compute the number of capabilities of the running kernel
[0], at the beginning of the execution of the program. As part of this,
it performs multiple calls to prctl(). Some of these may fail, and set
errno to a non-zero value:
# strace -e prctl ./bpftool version
prctl(PR_CAPBSET_READ, CAP_MAC_OVERRIDE) = 1
prctl(PR_CAPBSET_READ, 0x30 /* CAP_??? */) = -1 EINVAL (Invalid argument)
prctl(PR_CAPBSET_READ, CAP_CHECKPOINT_RESTORE) = 1
prctl(PR_CAPBSET_READ, 0x2c /* CAP_??? */) = -1 EINVAL (Invalid argument)
prctl(PR_CAPBSET_READ, 0x2a /* CAP_??? */) = -1 EINVAL (Invalid argument)
prctl(PR_CAPBSET_READ, 0x29 /* CAP_??? */) = -1 EINVAL (Invalid argument)
** fprintf added at the top of main(): we have errno == 1
./bpftool v7.0.0
using libbpf v1.0
features: libbfd, libbpf_strict, skeletons
+++ exited with 0 +++
This has been addressed in libcap 2.63 [1], but until this version is
available everywhere, we can fix it on bpftool side.
Let's clean errno at the beginning of the main() function, to make sure
that these checks do not interfere with the batch mode, where we error
out if errno is set after a bpftool command.
Currently host can send two WMI commands at once. There is possibility to
cause SMMU issues or corruption, if host wants to initiate 2 DMA
transfers, it is possible when copy complete interrupt for first DMA
reaches host, CE has already updated SRRI (Source ring read index) for
both DMA transfers and is in the middle of 2nd DMA. Host uses SRRI
(Source ring read index) to interpret how many DMA’s have been completed
and tries to unmap/free both the DMA entries. Hence now it is limiting to
one.Because CE is still in the middle of 2nd DMA which can cause these
issues when handling two DMA transfers.
This change will not impact other targets, as it is only for WCN3990.
Use-after-free occurred when the laundromat tried to free expired
cpntf_state entry on the s2s_cp_stateids list after inter-server
copy completed. The sc_cp_list that the expired copy state was
inserted on was already freed.
When COPY completes, the Linux client normally sends LOCKU(lock_state x),
FREE_STATEID(lock_state x) and CLOSE(open_state y) to the source server.
The nfs4_put_stid call from nfsd4_free_stateid cleans up the copy state
from the s2s_cp_stateids list before freeing the lock state's stid.
However, sometimes the CLOSE was sent before the FREE_STATEID request.
When this happens, the nfsd4_close_open_stateid call from nfsd4_close
frees all lock states on its st_locks list without cleaning up the copy
state on the sc_cp_list list. When the time the FREE_STATEID arrives the
server returns BAD_STATEID since the lock state was freed. This causes
the use-after-free error to occur when the laundromat tries to free
the expired cpntf_state.
This patch adds a call to nfs4_free_cpntf_statelist in
nfsd4_close_open_stateid to clean up the copy state before calling
free_ol_stateid_reaplist to free the lock state's stid on the reaplist.
Signed-off-by: Dai Ngo <dai.ngo@oracle.com> Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
This was discussed with Chuck as part of this patch set. Returning
nfserr_resource was decided to not be the best error message here, and
he suggested changing to nfserr_serverfault instead.
Clang produces a false positive when building with CONFIG_FORTIFY_SOURCE=y
and CONFIG_UBSAN_BOUNDS=y when operating on an array with a dynamic
offset. Work around this by using a direct assignment of an empty
instance. Avoids this warning:
../include/linux/fortify-string.h:309:4: warning: call to __write_overflow_field declared with 'warn
ing' attribute: detected write beyond size of field (1st parameter); maybe use struct_group()? [-Wat
tribute-warning]
__write_overflow_field(p_size_field, size);
^
which was isolated to the memset() call in xen_load_idt().
Note that this looks very much like another bug that was worked around:
https://github.com/ClangBuiltLinux/linux/issues/1592
Cc: Juergen Gross <jgross@suse.com> Cc: Boris Ostrovsky <boris.ostrovsky@oracle.com> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: Ingo Molnar <mingo@redhat.com> Cc: Borislav Petkov <bp@alien8.de> Cc: Dave Hansen <dave.hansen@linux.intel.com> Cc: x86@kernel.org Cc: "H. Peter Anvin" <hpa@zytor.com> Cc: xen-devel@lists.xenproject.org Reviewed-by: Boris Ostrovsky <boris.ostrovsky@oracle.com> Link: https://lore.kernel.org/lkml/41527d69-e8ab-3f86-ff37-6b298c01d5bc@oracle.com Signed-off-by: Kees Cook <keescook@chromium.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
Dell Inspiron 14 2-in-1 has two ACPI nodes under GPP1 both with _ADR of
0, both without _HID. It's ambiguous which the kernel should take, but
it seems to take "DEV0". Unfortunately "DEV0" is missing the device
property `StorageD3Enable` which is present on "NVME".
To avoid this causing problems for suspend, add a quirk for this system
to behave like `StorageD3Enable` property was found.
Link: https://bugzilla.kernel.org/show_bug.cgi?id=216440 Reported-and-tested-by: Luya Tshimbalanga <luya@fedoraproject.org> Signed-off-by: Mario Limonciello <mario.limonciello@amd.com> Reviewed-by: Hans de Goede <hdegoede@redhat.com> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
The .data.rel.ro.local section has the same semantics as .data.rel.ro
here, so include it in the .rodata section of the decompressor.
Additionally since the .printk_index section isn't usable outside of
the core kernel, discard it in the decompressor. Avoids these warnings:
arm-linux-gnueabi-ld: warning: orphan section `.data.rel.ro.local' from `arch/arm/boot/compressed/fdt_rw.o' being placed in section `.data.rel.ro.local'
arm-linux-gnueabi-ld: warning: orphan section `.printk_index' from `arch/arm/boot/compressed/fdt_rw.o' being placed in section `.printk_index'
Reported-by: kernel test robot <lkp@intel.com> Link: https://lore.kernel.org/linux-mm/202209080545.qMIVj7YM-lkp@intel.com Cc: Russell King <linux@armlinux.org.uk> Cc: linux-arm-kernel@lists.infradead.org Signed-off-by: Kees Cook <keescook@chromium.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
When CPU 0 is offline and intel_powerclamp is used to inject
idle, it generates kernel BUG:
BUG: using smp_processor_id() in preemptible [00000000] code: bash/15687
caller is debug_smp_processor_id+0x17/0x20
CPU: 4 PID: 15687 Comm: bash Not tainted 5.19.0-rc7+ #57
Call Trace:
<TASK>
dump_stack_lvl+0x49/0x63
dump_stack+0x10/0x16
check_preemption_disabled+0xdd/0xe0
debug_smp_processor_id+0x17/0x20
powerclamp_set_cur_state+0x7f/0xf9 [intel_powerclamp]
...
...
Here CPU 0 is the control CPU by default and changed to the current CPU,
if CPU 0 offlined. This check has to be performed under cpus_read_lock(),
hence the above warning.
Use get_cpu() instead of smp_processor_id() to avoid this BUG.
When value < time_unit, the parameter of ilog2() will be zero and
the return value is -1. u64(-1) is too large for shift exponent
and then will trigger shift-out-of-bounds:
shift exponent 18446744073709551615 is too large for 32-bit type 'int'
Call Trace:
rapl_compute_time_window_core
rapl_write_data_raw
set_time_window
store_constraint_time_window_us
Signed-off-by: Chao Qin <chao.qin@intel.com> Acked-by: Zhang Rui <rui.zhang@intel.com> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
Clang is especially sensitive about argument type matching when using
__overloaded functions (like memcmp(), etc). Help it see that function
pointers are just "void *". Avoids this error:
arch/mips/bcm47xx/prom.c:89:8: error: no matching function for call to 'memcmp'
if (!memcmp(prom_init, prom_init + mem, 32))
^~~~~~
include/linux/string.h:156:12: note: candidate function not viable: no known conversion from 'void (void)' to 'const void *' for 1st argument extern int memcmp(const void *,const void *,__kernel_size_t);
Users may disable HWP in firmware, in which case intel_pstate wouldn't load
unless the CPU model is explicitly supported.
Add TIGERLAKE to the list of CPUs that can register intel_pstate while not
advertising the HWP capability. Without this change, an TIGERLAKE in no-HWP
mode could only use the acpi_cpufreq frequency scaling driver.
See also commits: d8de7a44e11f: cpufreq: intel_pstate: Add Skylake servers support fbdc21e9b038: cpufreq: intel_pstate: Add Icelake servers support in no-HWP mode 706c5328851d: cpufreq: intel_pstate: Add Cometlake support in no-HWP mode
Reported by: M. Cargi Ari <cagriari@pm.me> Signed-off-by: Doug Smythies <dsmythies@telus.net> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
On a Packard Bell Dot SC (Intel Atom N2600 model) there is a FPDT table
which contains invalid physical addresses, with high bits set which fall
outside the range of the CPU-s supported physical address range.
Calling acpi_os_map_memory() on such an invalid phys address leads to
the below WARN_ON in ioremap triggering resulting in an oops/stacktrace.
Add code to verify the physical address before calling acpi_os_map_memory()
to fix / avoid the oops.
Signed-off-by: Hans de Goede <hdegoede@redhat.com> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
With CONFIG_FORTIFY=y and CONFIG_UBSAN_LOCAL_BOUNDS=y enabled, we observe
a runtime panic while running Android's Compatibility Test Suite's (CTS)
android.hardware.input.cts.tests. This is stemming from a strlen()
call in hidinput_allocate().
__compiletime_strlen() is implemented in terms of __builtin_object_size(),
then does an array access to check for NUL-termination. A quirk of
__builtin_object_size() is that for strings whose values are runtime
dependent, __builtin_object_size(str, 1 or 0) returns the maximum size
of possible values when those sizes are determinable at compile time.
Example:
static const char *v = "FOO BAR";
static const char *y = "FOO BA";
unsigned long x (int z) {
// Returns 8, which is:
// max(__builtin_object_size(v, 1), __builtin_object_size(y, 1))
return __builtin_object_size(z ? v : y, 1);
}
So when FORTIFY_SOURCE is enabled, the current implementation of
__compiletime_strlen() will try to access beyond the end of y at runtime
using the size of v. Mixed with UBSAN_LOCAL_BOUNDS we get a fault.
hidinput_allocate() has a local C string whose value is control flow
dependent on a switch statement, so __builtin_object_size(str, 1)
evaluates to the maximum string length, making all other cases fault on
the last character check. hidinput_allocate() could be cleaned up to
avoid runtime calls to strlen() since the local variable can only have
literal values, so there's no benefit to trying to fortify the strlen
call site there.
Perform a __builtin_constant_p() check against index 0 earlier in the
macro to filter out the control-flow-dependant case. Add a KUnit test
for checking the expected behavioral characteristics of FORTIFY_SOURCE
internals.
Toshiba Satellite Z830 needs the quirk video_disable_backlight_sysfs_if
for proper backlight control after suspend/resume cycles.
Toshiba Portege Z830 is simply the same laptop rebranded for certain
markets (I looked through the manual to other language sections to confirm
this) and thus also needs this quirk.
Thanks to Hans de Goede for suggesting this fix.
Link: https://www.spinics.net/lists/platform-driver-x86/msg34394.html Suggested-by: Hans de Goede <hdegoede@redhat.com> Signed-off-by: Arvid Norlander <lkml@vorpal.se> Reviewed-by: Hans de Goede <hdegoede@redhat.com> Tested-by: Arvid Norlander <lkml@vorpal.se> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
Fix the wrong lowest perf value reading which is used for new
des_perf calculation by governor requested, the incorrect min_perf will
get incorrect des_perf to be set , that will cause the system frequency
changing unexpectedly.
Reviewed-by: Huang Rui <ray.huang@amd.com> Acked-by: Viresh Kumar <viresh.kumar@linaro.org> Signed-off-by: Perry Yuan <Perry.Yuan@amd.com> Signed-off-by: Su Jinzhou <jinzhou.su@amd.com> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
The fill_page_cache_func() function allocates couple of pages to store
kvfree_rcu_bulk_data structures. This is a lightweight (GFP_NORETRY)
allocation which can fail under memory pressure. The function will,
however keep retrying even when the previous attempt has failed.
This retrying is in theory correct, but in practice the allocation is
invoked from workqueue context, which means that if the memory reclaim
gets stuck, these retries can hog the worker for quite some time.
Although the workqueues subsystem automatically adjusts concurrency, such
adjustment is not guaranteed to happen until the worker context sleeps.
And the fill_page_cache_func() function's retry loop is not guaranteed
to sleep (see the should_reclaim_retry() function).
And we have seen this function cause workqueue lockups:
Originally, we thought that the root cause of this lockup was several
retries with direct reclaim, but this is not yet confirmed. Furthermore,
we have seen similar lockups without any heavy memory pressure. This
suggests that there are other factors contributing to these lockups.
However, it is not really clear that endless retries are desireable.
So let's make the fill_page_cache_func() function back off after
allocation failure.
Cc: Uladzislau Rezki (Sony) <urezki@gmail.com> Cc: "Paul E. McKenney" <paulmck@kernel.org> Cc: Frederic Weisbecker <frederic@kernel.org> Cc: Neeraj Upadhyay <quic_neeraju@quicinc.com> Cc: Josh Triplett <josh@joshtriplett.org> Cc: Steven Rostedt <rostedt@goodmis.org> Cc: Mathieu Desnoyers <mathieu.desnoyers@efficios.com> Cc: Lai Jiangshan <jiangshanlai@gmail.com> Cc: Joel Fernandes <joel@joelfernandes.org> Signed-off-by: Michal Hocko <mhocko@suse.com> Reviewed-by: Uladzislau Rezki (Sony) <urezki@gmail.com> Signed-off-by: Paul E. McKenney <paulmck@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
Kernels built with PREEMPT_RCU=y and RCU_STRICT_GRACE_PERIOD=y trigger
irq-work from rcu_read_unlock(), and the resulting irq-work handler
invokes rcu_preempt_deferred_qs_handle(). The point of this triggering
is to force grace periods to end quickly in order to give tools like KASAN
a better chance of detecting RCU usage bugs such as leaking RCU-protected
pointers out of an RCU read-side critical section.
However, this irq-work triggering is unconditional. This works, but
there is no point in doing this irq-work unless the current grace period
is waiting on the running CPU or task, which is not the common case.
After all, in the common case there are many rcu_read_unlock() calls
per CPU per grace period.
This commit therefore triggers the irq-work only when the current grace
period is waiting on the running CPU or task.
This change was tested as follows on a four-CPU system:
This procedure produces results in this per-CPU set of files:
/sys/kernel/debug/tracing/trace_stat/function*
Sample output from one of these files is as follows:
Function Hit Time Avg s^2
-------- --- ---- --- ---
rcu_preempt_deferred_qs_handle 838746 182650.3 us 0.217 us 0.004 us
The baseline sum of the "Hit" values (the number of calls to this
function) was 3,319,015. With this commit, that sum was 1,140,359,
for a 2.9x reduction. The worst-case variance across the CPUs was less
than 25%, so this large effect size is statistically significant.
This patch fixes a race between queue_work() in
_dlm_lowcomms_commit_msg() and srcu_read_unlock(). The queue_work() can
take the final reference of a dlm_msg and so msg->idx can contain
garbage which is signaled by the following warning:
I reproduced this warning with dlm_locktorture test which is currently
not upstream. However this patch fix the issue by make a additional
refcount between dlm_lowcomms_new_msg() and dlm_lowcomms_commit_msg().
In case of the race the kref_put() in dlm_lowcomms_commit_msg() will be
the final put.
Signed-off-by: Alexander Aring <aahringo@redhat.com> Signed-off-by: David Teigland <teigland@redhat.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
An instance of Client does not implicitly close /dev/tpm* handle, once it
gets destroyed. Close the file handle in the class destructor
Client.__del__().
Fixes: 6ea3dfe1e0732 ("selftests: add TPM 2.0 tests") Cc: Shuah Khan <shuah@kernel.org> Cc: linux-kselftest@vger.kernel.org Cc: Jarkko Sakkinen <jarkko@kernel.org> Signed-off-by: Stefan Berger <stefanb@linux.ibm.com> Reviewed-by: Jarkko Sakkinen <jarkko@kernel.org> Signed-off-by: Jarkko Sakkinen <jarkko@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
f2fs_inode_info.cp_task was introduced for FS_CP_DATA_IO accounting
since commit b0af6d491a6b ("f2fs: add app/fs io stat").
However, cp_task usage coverage has been increased due to below
commits:
commit 040d2bb318d1 ("f2fs: fix to avoid deadloop if data_flush is on")
commit 186857c5a14a ("f2fs: fix potential recursive call when enabling data_flush")
So that, if data_flush mountoption is on, when data flush was
triggered from background, the IO from data flush will be accounted
as checkpoint IO type incorrectly.
In order to fix this issue, this patch splits cp_task into two:
a) cp_task: used for IO accounting
b) wb_task: used to avoid deadlock
Fixes: 040d2bb318d1 ("f2fs: fix to avoid deadloop if data_flush is on") Fixes: 186857c5a14a ("f2fs: fix potential recursive call when enabling data_flush") Signed-off-by: Chao Yu <chao@kernel.org> Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
The following scenarios exist.
process A: process B:
->f2fs_drop_extent_tree ->f2fs_update_extent_cache_range
->f2fs_update_extent_tree_range
->write_lock
->set_inode_flag
->is_inode_flag_set
->__free_extent_tree // Shouldn't
// have been
// cleaned up
// here
->write_lock
In this case, the "FI_NO_EXTENT" flag is set between
f2fs_update_extent_tree_range and is_inode_flag_set
by other process. it leads to clearing the whole exten
tree which should not have happened. And we fix it by
move the setting it to the range of write_lock.
If an error is detected as a result of user-space process accessing a
corrupt memory location, the CPU may take an abort. Then the platform
firmware reports kernel via NMI like notifications, e.g. NOTIFY_SEA,
NOTIFY_SOFTWARE_DELEGATED, etc.
For NMI like notifications, commit 7f17b4a121d0 ("ACPI: APEI: Kick the
memory_failure() queue for synchronous errors") keep track of whether
memory_failure() work was queued, and make task_work pending to flush out
the queue so that the work is processed before return to user-space.
The code use init_mm to check whether the error occurs in user space:
if (current->mm != &init_mm)
The condition is always true, becase _nobody_ ever has "init_mm" as a real
VM any more.
In addition to abort, errors can also be signaled as asynchronous
exceptions, such as interrupt and SError. In such case, the interrupted
current process could be any kind of thread. When a kernel thread is
interrupted, the work ghes_kick_task_work deferred to task_work will never
be processed because entry_handler returns to call ret_to_kernel() instead
of ret_to_user(). Consequently, the estatus_node alloced from
ghes_estatus_pool in ghes_in_nmi_queue_one_entry() will not be freed.
After around 200 allocations in our platform, the ghes_estatus_pool will
run of memory and ghes_in_nmi_queue_one_entry() returns ENOMEM. As a
result, the event failed to be processed.
sdei: event 805 on CPU 113 failed with error: -2
Finally, a lot of unhandled events may cause platform firmware to exceed
some threshold and reboot.
The condition should generally just do
if (current->mm)
as described in active_mm.rst documentation.
Then if an asynchronous error is detected when a kernel thread is running,
(e.g. when detected by a background scrubber), do not add task_work to it
as the original patch intends to do.
Fixes: 7f17b4a121d0 ("ACPI: APEI: Kick the memory_failure() queue for synchronous errors") Signed-off-by: Shuai Xue <xueshuai@linux.alibaba.com> Reviewed-by: Tony Luck <tony.luck@intel.com> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
Counterintuitively, mod_timer(..., jiffies + 1) will cause the timer to
fire not in the next jiffy, but in two jiffies. The way to cause
the timer to fire in the next jiffy is with mod_timer(..., jiffies).
Doing so then lets us bump the upper bound back up again.
Fixes: 50ee7529ec45 ("random: try to actively add entropy rather than passively wait for it") Fixes: 829d680e82a9 ("random: cap jitter samples per bit to factor of HZ") Cc: Dominik Brodowski <linux@dominikbrodowski.net> Cc: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Cc: Sultan Alsawaf <sultan@kerneltoast.com> Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
projects / linux / log