As the description before netdev_run_todo, we cannot call free_netdev
before rtnl_unlock, fix it by reorder the code.
Signed-off-by: yangerkun <yangerkun@huawei.com> Reviewed-by: Oliver Hartkopp <socketcan@hartkopp.net> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Sasha Levin <sashal@kernel.org>
When queueing a signal, we increment both the users count of pending
signals (for RLIMIT_SIGPENDING tracking) and we increment the refcount
of the user struct itself (because we keep a reference to the user in
the signal structure in order to correctly account for it when freeing).
That turns out to be fairly expensive, because both of them are atomic
updates, and particularly under extreme signal handling pressure on big
machines, you can get a lot of cache contention on the user struct.
That can then cause horrid cacheline ping-pong when you do these
multiple accesses.
So change the reference counting to only pin the user for the _first_
pending signal, and to unpin it when the last pending signal is
dequeued. That means that when a user sees a lot of concurrent signal
queuing - which is the only situation when this matters - the only
atomic access needed is generally the 'sigpending' count update.
This was noticed because of a particularly odd timing artifact on a
dual-socket 96C/192T Cascade Lake platform: when you get into bad
contention, on that machine for some reason seems to be much worse when
the contention happens in the upper 32-byte half of the cacheline.
As a result, the kernel test robot will-it-scale 'signal1' benchmark had
an odd performance regression simply due to random alignment of the
'struct user_struct' (and pointed to a completely unrelated and
apparently nonsensical commit for the regression).
Avoiding the double increments (and decrements on the dequeueing side,
of course) makes for much less contention and hugely improved
performance on that will-it-scale microbenchmark.
Quoting Feng Tang:
"It makes a big difference, that the performance score is tripled! bump
from original 17000 to 54000. Also the gap between 5.0-rc6 and
5.0-rc6+Jiri's patch is reduced to around 2%"
[ The "2% gap" is the odd cacheline placement difference on that
platform: under the extreme contention case, the effect of which half
of the cacheline was hot was 5%, so with the reduced contention the
odd timing artifact is reduced too ]
It does help in the non-contended case too, but is not nearly as
noticeable.
Reported-and-tested-by: Feng Tang <feng.tang@intel.com> Cc: Eric W. Biederman <ebiederm@xmission.com> Cc: Huang, Ying <ying.huang@intel.com> Cc: Philip Li <philip.li@intel.com> Cc: Andi Kleen <andi.kleen@intel.com> Cc: Jiri Olsa <jolsa@redhat.com> Cc: Peter Zijlstra <peterz@infradead.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
The dtbs_check should be a phony target, but currently it is not
specified so.
'make dtbs_check' works even if a file named 'dtbs_check' exists
because it depends on another phony target, scripts_dtc, but we
should not rely on it.
For some reason, device may be in one situation which can't handle
FS request, so STS_RESOURCE is always returned and the FS request
will be added to hctx->dispatch. However passthrough request may
be required at that time for fixing the problem. If passthrough
request is added to scheduler queue, there isn't any chance for
blk-mq to dispatch it given we prioritize requests in hctx->dispatch.
Then the FS IO request may never be completed, and IO hang is caused.
So passthrough request has to be added to hctx->dispatch directly
for fixing the IO hang.
Fix this issue by inserting passthrough request into hctx->dispatch
directly together withing adding FS request to the tail of
hctx->dispatch in blk_mq_dispatch_rq_list(). Actually we add FS request
to tail of hctx->dispatch at default, see blk_mq_request_bypass_insert().
Then it becomes consistent with original legacy IO request
path, in which passthrough request is always added to q->queue_head.
Cc: Dongli Zhang <dongli.zhang@oracle.com> Cc: Christoph Hellwig <hch@infradead.org> Cc: Ewan D. Milne <emilne@redhat.com> Signed-off-by: Ming Lei <ming.lei@redhat.com> Signed-off-by: Jens Axboe <axboe@kernel.dk> Signed-off-by: Sasha Levin <sashal@kernel.org>
The SDMA engine used by TEMAC halts operation when it has finished
processing of the last buffer descriptor in the buffer ring.
Unfortunately, no interrupt event is generated when this happens,
so we need to setup another mechanism to make sure DMA operation is
restarted when enough buffers have been added to the ring.
Fixes: 7391458cef35 ("net: add Xilinx ll_temac device driver") Signed-off-by: Esben Haabendal <esben@geanix.com> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Sasha Levin <sashal@kernel.org>
In order to handle such failures more graceful, this change splits the
receive loop into one for consuming the received buffers, and one for
allocating new buffers.
When GFP_ATOMIC allocations fail, the receive will continue with the
buffers that is still there, and with the expectation that the allocations
will succeed in a later call to receive.
Fixes: 7391458cef35 ("net: add Xilinx ll_temac device driver") Signed-off-by: Esben Haabendal <esben@geanix.com> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Sasha Levin <sashal@kernel.org>
It is possible that the interrupt handler fires and frees up space in
the TX ring in between checking for sufficient TX ring space and
stopping the TX queue in temac_start_xmit. If this happens, the
queue wake from the interrupt handler will occur before the queue is
stopped, causing a lost wakeup and the adapter's transmit hanging.
To avoid this, after stopping the queue, check again whether there is
sufficient space in the TX ring. If so, wake up the queue again.
This is a port of the similar fix in axienet driver,
commit 83f057204bf5 ("net: axienet: Fix race condition causing TX hang").
Fixes: 77ff44a61a76 ("net: ll_temac: fix checksum offload logic") Signed-off-by: Esben Haabendal <esben@geanix.com> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Sasha Levin <sashal@kernel.org>
local->sta_mtx is held in __ieee80211_check_fast_rx_iface().
No need to use list_for_each_entry_rcu() as it also requires
a cond argument to avoid false lockdep warnings when not used in
RCU read-side section (with CONFIG_PROVE_RCU_LIST).
Therefore use list_for_each_entry();
The KS8851 requires that packet RX and TX are mutually exclusive.
Currently, the driver hopes to achieve this by disabling interrupt
from the card by writing the card registers and by disabling the
interrupt on the interrupt controller. This however is racy on SMP.
Replace this approach by expanding the spinlock used around the
ks_start_xmit() TX path to ks_irq() RX path to assure true mutual
exclusion and remove the interrupt enabling/disabling, which is
now not needed anymore. Furthermore, disable interrupts also in
ks_net_stop(), which was missing before.
Note that a massive improvement here would be to re-use the KS8851
driver approach, which is to move the TX path into a worker thread,
interrupt handling to threaded interrupt, and synchronize everything
with mutexes, but that would be a much bigger rework, for a separate
patch.
Signed-off-by: Marek Vasut <marex@denx.de> Cc: David S. Miller <davem@davemloft.net> Cc: Lukas Wunner <lukas@wunner.de> Cc: Petr Stetiar <ynezz@true.cz> Cc: YueHaibing <yuehaibing@huawei.com> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Sasha Levin <sashal@kernel.org>
usbnet creates network interfaces with min_mtu = 0 and
max_mtu = ETH_MAX_MTU.
These values are not modified by qmi_wwan when the network interface
is created initially, allowing, for example, to set mtu greater than 1500.
When a raw_ip switch is done (raw_ip set to 'Y', then set to 'N') the mtu
values for the network interface are set through ether_setup, with
min_mtu = ETH_MIN_MTU and max_mtu = ETH_DATA_LEN, not allowing anymore to
set mtu greater than 1500 (error: mtu greater than device maximum).
The patch restores the original min/max mtu values set by usbnet after a
raw_ip switch.
Signed-off-by: Daniele Palmas <dnlplm@gmail.com> Acked-by: Bjørn Mork <bjorn@mork.no> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Sasha Levin <sashal@kernel.org>
We may end up with a NULL reg_rule after the loop in
handle_channel_custom() if the bandwidth didn't fit,
check if this is the case and bail out if so.
Fix a varargs-related bug in print_synth_event() which resulted in
strange output and oopses on 32-bit x86 systems. The problem is that
trace_seq_printf() expects the varargs to match the format string, but
print_synth_event() was always passing u64 values regardless. This
results in unspecified behavior when unpacking with va_arg() in
trace_seq_printf().
Add a function that takes the size into account when calling
trace_seq_printf().
It's possible that there is scheduled work left while the device is
already being removed, which can cause a kernel crash. Adding a flag
will avoid this.
Signed-off-by: Hanno Zulla <kontakt@hanno.de> Signed-off-by: Benjamin Tissoires <benjamin.tissoires@redhat.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
It's required to call hid_hw_stop() once hid_hw_start() was called
previously, so error cases need to handle this. Also, hid_hw_close() is
not necessary during removal.
Signed-off-by: Hanno Zulla <kontakt@hanno.de> Signed-off-by: Benjamin Tissoires <benjamin.tissoires@redhat.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
The struct *bigben was allocated via devm_kzalloc() and then used as a
parameter in input_ff_create_memless(). This caused a double kfree
during removal of the device, since both the managed resource API and
ml_ff_destroy() in drivers/input/ff-memless.c would call kfree() on it.
Signed-off-by: Hanno Zulla <kontakt@hanno.de> Signed-off-by: Benjamin Tissoires <benjamin.tissoires@redhat.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
If the BIOS default timeout for the watchdog is too small userspace may
not have enough time to configure new timeout after opening the device
before the system is already reset. For this reason program default
timeout of 30 seconds in the driver probe and allow userspace to change
this from command line or through module parameter (wdat_wdt.timeout).
Reported-by: Jean Delvare <jdelvare@suse.de> Signed-off-by: Mika Westerberg <mika.westerberg@linux.intel.com> Reviewed-by: Jean Delvare <jdelvare@suse.de> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
Magic Keyboards with more recent firmware (0x0100) report Fn key differently.
Without this patch, Fn key may not behave as expected and may not be
configurable via hid_apple fnmode module parameter.
In case the WDAT interface is broken, give the user an option to
ignore it to let a native driver bind to the watchdog device instead.
Signed-off-by: Jean Delvare <jdelvare@suse.de> Acked-by: Mika Westerberg <mika.westerberg@linux.intel.com> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
Use a more meaningful variable name for the invalidation request
that is distinct from the tmp variable that gets overwritten when
acquiring the invalidation semaphore.
Fixes: 3223f1aa5cfc ("drm/amdgpu: invalidate mmhub semaphore workaround in gmc9/gmc10") Signed-off-by: Felix Kuehling <Felix.Kuehling@amd.com> Reviewed-by: Christian König <christian.koenig@amd.com> Reviewed-by: Yong Zhao <Yong.Zhao@amd.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
Before releasing the global mutex, we only unlink the hashtable
from the hash list, its proc file is still not unregistered at
this point. So syzbot could trigger a race condition where a
parallel htable_create() could register the same file immediately
after the mutex is released.
Move htable_remove_proc_entry() back to mutex protection to
fix this. And, fold htable_destroy() into htable_put() to make
the code slightly easier to understand.
Reported-and-tested-by: syzbot+d195fd3b9a364ddd6731@syzkaller.appspotmail.com Fixes: 18cd20bd4d05 ("netfilter: xt_hashlimit: reduce hashlimit_mutex scope for htable_put()") Signed-off-by: Cong Wang <xiyou.wangcong@gmail.com> Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
In smc_ib_remove_dev() check if the provided ib device was actually
initialized for SMC before.
Reported-by: syzbot+84484ccebdd4e5451d91@syzkaller.appspotmail.com Fixes: 0bb7e7b2f4d9 ("smc: introduce SMC as an IB-client") Signed-off-by: Karsten Graul <kgraul@linux.ibm.com> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Fixes: 38506f4bbc9d ("bpf, sockmap: convert to generic sk_msg interface") Signed-off-by: Eric Dumazet <edumazet@google.com> Reported-by: syzbot+1938db17e275e85dc328@syzkaller.appspotmail.com Cc: Daniel Borkmann <daniel@iogearbox.net> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Commit 3ab2fb65945d ("iommu/amd: Re-factor guest virtual APIC
(de-)activation code") accidentally left out the ir_data pointer when
calling modity_irte_ga(), which causes the function amd_iommu_update_ga()
to return prematurely due to struct amd_ir_data.ref is NULL and
the "is_run" bit of IRTE does not get updated properly.
This results in bad I/O performance since IOMMU AVIC always generate GA Log
entry and notify IOMMU driver and KVM when it receives interrupt from the
PCI pass-through device instead of directly inject interrupt to the vCPU.
Fixes by passing ir_data when calling modify_irte_ga() as done previously.
VMD subdevices are created with a PCI domain ID of 0x10000 or
higher.
These subdevices are also handled like all other PCI devices by
dmar_pci_bus_notifier().
However, when dmar_alloc_pci_notify_info() take records of such devices,
it will truncate the domain ID to a u16 value (in info->seg).
The device at (e.g.) 10000:00:02.0 is then treated by the DMAR code as if
it is 0000:00:02.0.
In the unlucky event that a real device also exists at 0000:00:02.0 and
also has a device-specific entry in the DMAR table,
dmar_insert_dev_scope() will crash on:
BUG_ON(i >= devices_cnt);
That's basically a sanity check that only one PCI device matches a
single DMAR entry; in this case we seem to have two matching devices.
Fix this by ignoring devices that have a domain number higher than
what can be looked up in the DMAR table.
This problem was carefully diagnosed by Jian-Hong Pan.
Signed-off-by: Lu Baolu <baolu.lu@linux.intel.com> Signed-off-by: Daniel Drake <drake@endlessm.com> Fixes: 2b8daa95e1b78 ("iommu/vt-d: Update DRHD/RMRR/ATSR device scope caches when PCI hotplug happens") Signed-off-by: Joerg Roedel <jroedel@suse.de> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
When base address in RHSA structure doesn't match base address in
each DRHD structure, the base address in last DRHD is printed out.
This doesn't make sense when there are multiple DRHD units, fix it
by printing the buggy RHSA's base address.
Signed-off-by: Lu Baolu <baolu.lu@linux.intel.com> Signed-off-by: Zhenzhong Duan <zhenzhong.duan@gmail.com> Fixes: 6eb296e2f6f89 ("intel-iommu: Set a more specific taint flag for invalid BIOS DMAR tables") Signed-off-by: Joerg Roedel <jroedel@suse.de> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Since commit 050e3a21f8eb ("perf bench futex: Use cpumaps") the default
number of threads the benchmark uses got changed from number of online
CPUs to zero:
$ perf bench futex wake
# Running 'futex/wake' benchmark:
Run summary [PID 15930]: blocking on 0 threads (at [private] futex 0x558b8ee4bfac), waking up 1 at a time.
[Run 1]: Wokeup 0 of 0 threads in 0.0000 ms
[...]
[Run 10]: Wokeup 0 of 0 threads in 0.0000 ms
Wokeup 0 of 0 threads in 0.0004 ms (+-40.82%)
Restore the old behavior by grabbing the number of online CPUs via
cpu->nr:
$ perf bench futex wake
# Running 'futex/wake' benchmark:
Run summary [PID 18356]: blocking on 8 threads (at [private] futex 0xb3e62c), waking up 1 at a time.
[Run 1]: Wokeup 8 of 8 threads in 0.0260 ms
[...]
[Run 10]: Wokeup 8 of 8 threads in 0.0270 ms
Wokeup 8 of 8 threads in 0.0419 ms (+-24.35%)
Fixes: 050e3a21f8eb ("perf bench futex: Use cpumaps") Signed-off-by: Tommi Rantala <tommi.t.rantala@nokia.com> Tested-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com> Cc: Darren Hart <dvhart@infradead.org> Cc: Davidlohr Bueso <dave@stgolabs.net> Cc: Jiri Olsa <jolsa@redhat.com> Cc: Mark Rutland <mark.rutland@arm.com> Cc: Namhyung Kim <namhyung@kernel.org> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Link: http://lore.kernel.org/lkml/20200305083714.9381-3-tommi.t.rantala@nokia.com Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
If a GPIO we are trying to use is not available and we are deferring
the probe, don't output an error message.
This seems to have been the intent of commit be318bab97dc
("i2c: gpio: Add support for named gpios in DT") but the error was
still output due to not checking the updated 'retdesc'.
Fixes: be318bab97dc ("i2c: gpio: Add support for named gpios in DT") Signed-off-by: Hamish Martin <hamish.martin@alliedtelesis.co.nz> Acked-by: Linus Walleij <linus.walleij@linaro.org> Signed-off-by: Wolfram Sang <wsa@the-dreams.de> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
This does three inter-related things to clarify the usage of the
platform device dma_mask field. In the process, fix the bug introduced
by 741b752e9c7b ("driver core: initialize a default DMA mask for
platform device") that caused Artem Tashkinov's laptop to not boot with
newer Fedora kernels.
This does:
- First off, rename the field to "platform_dma_mask" to make it
greppable.
We have way too many different random fields called "dma_mask" in
various data structures, where some of them are actual masks, and
some of them are just pointers to the mask. And the structures all
have pointers to each other, or embed each other inside themselves,
and "pdev" sometimes means "platform device" and sometimes it means
"PCI device".
So to make it clear in the code when you actually use this new field,
give it a unique name (it really should be something even more unique
like "platform_device_dma_mask", since it's per platform device, not
per platform, but that gets old really fast, and this is unique
enough in context).
To further clarify when the field gets used, initialize it when we
actually start using it with the default value.
- Then, use this field instead of the random one-off allocation in
platform_device_register_full() that is now unnecessary since we now
already have a perfectly fine allocation for it in the platform
device structure.
- The above then allows us to fix the actual bug, where the error path
of platform_device_register_full() would unconditionally free the
platform device DMA allocation with 'kfree()'.
That kfree() was dont regardless of whether the allocation had been
done earlier with the (now removed) kmalloc, or whether
setup_pdev_dma_masks() had already been used and the dma_mask pointer
pointed to the mask that was part of the platform device.
It seems most people never triggered the error path, or only triggered
it from a call chain that set an explicit pdevinfo->dma_mask value (and
thus caused the unnecessary allocation that was "cleaned up" in the
error path) before calling platform_device_register_full().
Robin Murphy points out that in Artem's case the wdat_wdt driver failed
in platform_device_add(), and that was the one that had called
platform_device_register_full() with pdevinfo.dma_mask = 0, and would
have caused that kfree() of pdev.dma_mask corrupting the heap.
A later unrelated kmalloc() then oopsed due to the heap corruption.
Fixes: 741b752e9c7b ("driver core: initialize a default DMA mask for platform device") Reported-bisected-and-tested-by: Artem S. Tashkinov <aros@gmx.com> Reviewed-by: Robin Murphy <robin.murphy@arm.com> Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Signed-off-by: Christoph Hellwig <hch@lst.de> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
From commit a3376c4051a8 ("drm/i915/gvt: Use vgpu_lock to protect per
vgpu access") the vgpu idr destroy is moved later than vgpu resource
destroy, then it would fail to stop timer for schedule policy clean
which to check vgpu idr for any left vGPU. So this trys to destroy
vgpu idr earlier.
Cc: Colin Xu <colin.xu@intel.com> Fixes: a3376c4051a8 ("drm/i915/gvt: Use vgpu_lock to protect per vgpu access") Acked-by: Colin Xu <colin.xu@intel.com> Signed-off-by: Zhenyu Wang <zhenyuw@linux.intel.com> Link: http://patchwork.freedesktop.org/patch/msgid/20200229055445.31481-1-zhenyuw@linux.intel.com Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
kref_init starts with the reference count at 1, which will be balanced
by the pinctrl_put in pinctrl_unregister. The additional kref_get in
pinctrl_claim_hogs will increase this count to 2 and cause the hogs to
not get freed when pinctrl_unregister is called.
Commit 081157f765adb ("drm/i915/gvt: Enable gfx virtualiztion for CFL")
added the support on CFL. The vgpu emulation hotplug support on CFL was
supposed to be included in that patch. Without the vgpu emulation
hotplug support, the dma-buf based display gives us a blur face.
So fix this issue by adding the vgpu emulation hotplug support on CFL.
The functions vring_new_virtqueue() and __vring_new_virtqueue() are used
with split rings, and any allocations within these functions are managed
outside of the .we_own_ring flag. The commit 600570009bbc ("virtio_ring:
allocate desc state for split ring separately") allocates the desc state
within the __vring_new_virtqueue() but frees it only when the .we_own_ring
flag is set. This leads to a memory leak when freeing such allocated
virtqueues with the vring_del_virtqueue() function.
Fix this by moving the desc_state free code outside the flag and only
for split rings. Issue was discovered during testing with remoteproc
and virtio_rpmsg.
Fixes: 600570009bbc ("virtio_ring: allocate desc state for split ring separately") Signed-off-by: Suman Anna <s-anna@ti.com> Link: https://lore.kernel.org/r/20200224212643.30672-1-s-anna@ti.com Signed-off-by: Michael S. Tsirkin <mst@redhat.com> Acked-by: Jason Wang <jasowang@redhat.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
In the gxl driver, the sdio cmd and clk pins are inverted. It has not caused
any issue so far because devices using these pins always take both pins
so the resulting configuration is OK.
A transmission scheduling for an interface which is currently dropped by
batadv_iv_ogm_iface_disable could still be in progress. The B.A.T.M.A.N. V
is simply cancelling the workqueue item in an synchronous way but this is
not possible with B.A.T.M.A.N. IV because the OGM submissions are
intertwined.
Instead it has to stop submitting the OGM when it detect that the buffer
pointer is set to NULL.
intel_iommu_iova_to_phys() has a bug when it translates an IOVA for a huge
page onto its corresponding physical address. This commit fixes the bug by
accomodating the level of page entry for the IOVA and adds IOVA's lower
address to the physical address.
Cc: <stable@vger.kernel.org> Acked-by: Lu Baolu <baolu.lu@linux.intel.com> Reviewed-by: Moritz Fischer <mdf@kernel.org> Signed-off-by: Yonghyun Hwang <yonghyun@google.com> Fixes: af23e596b2694 ("VT-d: Changes to support KVM") Signed-off-by: Joerg Roedel <jroedel@suse.de> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
dmar_drhd_units is traversed using list_for_each_entry_rcu()
outside of an RCU read side critical section but under the
protection of dmar_global_lock. Hence add corresponding lockdep
expression to silence the following false-positive warnings:
Quoting from the comment describing the WARN functions in
include/asm-generic/bug.h:
* WARN(), WARN_ON(), WARN_ON_ONCE, and so on can be used to report
* significant kernel issues that need prompt attention if they should ever
* appear at runtime.
*
* Do not use these macros when checking for invalid external inputs
The (buggy) firmware tables which the dmar code was calling WARN_TAINT
for really are invalid external inputs. They are not under the kernel's
control and the issues in them cannot be fixed by a kernel update.
So logging a backtrace, which invites bug reports to be filed about this,
is not helpful.
Some distros, e.g. Fedora, have tools watching for the kernel backtraces
logged by the WARN macros and offer the user an option to file a bug for
this when these are encountered. The WARN_TAINT in warn_invalid_dmar()
+ another iommu WARN_TAINT, addressed in another patch, have lead to over
a 100 bugs being filed this way.
This commit replaces the WARN_TAINT("...") calls, with
pr_warn(FW_BUG "...") + add_taint(TAINT_FIRMWARE_WORKAROUND, ...) calls
avoiding the backtrace and thus also avoiding bug-reports being filed
about this against the kernel.
Fixes: 6eb296e2f6f8 ("intel-iommu: Set a more specific taint flag for invalid BIOS DMAR tables") Fixes: 5e558059eaad ("iommu/vt-d: Parse ANDD records") Signed-off-by: Hans de Goede <hdegoede@redhat.com> Signed-off-by: Joerg Roedel <jroedel@suse.de> Acked-by: Lu Baolu <baolu.lu@linux.intel.com> Cc: stable@vger.kernel.org Link: https://lore.kernel.org/r/20200309140138.3753-2-hdegoede@redhat.com BugLink: https://bugzilla.redhat.com/show_bug.cgi?id=1564895 Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The way cookie_init_hw_msi_region() allocates the iommu_dma_msi_page
structures doesn't match the way iommu_put_dma_cookie() frees them.
The former performs a single allocation of all the required structures,
while the latter tries to free them one at a time. It doesn't quite
work for the main use case (the GICv3 ITS where the range is 64kB)
when the base granule size is 4kB.
This leads to a nice slab corruption on teardown, which is easily
observable by simply creating a VF on a SRIOV-capable device, and
tearing it down immediately (no need to even make use of it).
Fortunately, this only affects systems where the ITS isn't translated
by the SMMU, which are both rare and non-standard.
Fix it by allocating iommu_dma_msi_page structures one at a time.
Fixes: eaa95d7632f8a ("iommu/dma: Handle IOMMU API reserved regions") Signed-off-by: Marc Zyngier <maz@kernel.org> Reviewed-by: Eric Auger <eric.auger@redhat.com> Cc: Robin Murphy <robin.murphy@arm.com> Cc: Joerg Roedel <jroedel@suse.de> Cc: Will Deacon <will@kernel.org> Cc: stable@vger.kernel.org Reviewed-by: Robin Murphy <robin.murphy@arm.com> Signed-off-by: Joerg Roedel <jroedel@suse.de> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
There are two implemented bits in the PPIN_CTL MSR:
Bit 0: LockOut (R/WO)
Set 1 to prevent further writes to MSR_PPIN_CTL.
Bit 1: Enable_PPIN (R/W)
If 1, enables MSR_PPIN to be accessible using RDMSR.
If 0, an attempt to read MSR_PPIN will cause #GP.
So there are four defined values:
0: PPIN is disabled, PPIN_CTL may be updated
1: PPIN is disabled. PPIN_CTL is locked against updates
2: PPIN is enabled. PPIN_CTL may be updated
3: PPIN is enabled. PPIN_CTL is locked against updates
Code would only enable the X86_FEATURE_INTEL_PPIN feature for case "2".
When it should have done so for both case "2" and case "3".
Fix the final test to just check for the enable bit. Also fix some of
the other comments in this function.
Fixes: ed2a921c6a76 ("x86/mce: Include the PPIN in MCE records when available") Signed-off-by: Tony Luck <tony.luck@intel.com> Signed-off-by: Borislav Petkov <bp@suse.de> Cc: <stable@vger.kernel.org> Link: https://lkml.kernel.org/r/20200226011737.9958-1-tony.luck@intel.com Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Enable the sampling check in kernel/events/core.c::perf_event_open(),
which returns the more appropriate -EOPNOTSUPP.
BEFORE:
$ sudo perf record -a -e instructions,l3_request_g1.caching_l3_cache_accesses true
Error:
The sys_perf_event_open() syscall returned with 22 (Invalid argument) for event (l3_request_g1.caching_l3_cache_accesses).
/bin/dmesg | grep -i perf may provide additional information.
With nothing relevant in dmesg.
AFTER:
$ sudo perf record -a -e instructions,l3_request_g1.caching_l3_cache_accesses true
Error:
l3_request_g1.caching_l3_cache_accesses: PMU Hardware doesn't support sampling/overflow-interrupts. Try 'perf stat'
Fixes: 418df02dbe80 ("perf/x86/amd: Add support for AMD NB and L2I "uncore" counters") Signed-off-by: Kim Phillips <kim.phillips@amd.com> Signed-off-by: Borislav Petkov <bp@suse.de> Acked-by: Peter Zijlstra <peterz@infradead.org> Cc: stable@vger.kernel.org Link: https://lkml.kernel.org/r/20200311191323.13124-1-kim.phillips@amd.com Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
If the hardware receives an oversized packet with too many rx fragments,
skb_shinfo(skb)->frags can overflow and corrupt memory of adjacent pages.
This becomes especially visible if it corrupts the freelist pointer of
a slab page.
Cc: stable@vger.kernel.org Signed-off-by: Felix Fietkau <nbd@nbd.name> Signed-off-by: Kalle Valo <kvalo@codeaurora.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Function i2c_dw_pci_remove() -> pci_free_irq_vectors() ->
pci_disable_msi() -> free_msi_irqs() will throw a BUG_ON() for MSI
enabled device since the driver has not released the requested IRQ before
calling the pci_free_irq_vectors().
Here driver requests an IRQ using devm_request_irq() but automatic
release happens only after remove callback. Fix this by explicitly
freeing the IRQ before calling pci_free_irq_vectors().
Fixes: 96cafae3f525 ("i2c: designware-pci: Switch over to MSI interrupts") Cc: stable@vger.kernel.org # v5.4+ Signed-off-by: Jarkko Nikula <jarkko.nikula@linux.intel.com> Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com> Signed-off-by: Wolfram Sang <wsa@the-dreams.de> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
There is a race and a buffer overflow corrupting a kernel memory while
reading an EFI variable with a size more than 1024 bytes via the older
sysfs method. This happens because accessing struct efi_variable in
efivar_{attr,size,data}_read() and friends is not protected from
a concurrent access leading to a kernel memory corruption and, at best,
to a crash. The race scenario is the following:
CPU0: CPU1:
efivar_attr_read()
var->DataSize = 1024;
efivar_entry_get(... &var->DataSize)
down_interruptible(&efivars_lock)
efivar_attr_read() // same EFI var
var->DataSize = 1024;
efivar_entry_get(... &var->DataSize)
down_interruptible(&efivars_lock)
virt_efi_get_variable()
// returns EFI_BUFFER_TOO_SMALL but
// var->DataSize is set to a real
// var size more than 1024 bytes
up(&efivars_lock)
virt_efi_get_variable()
// called with var->DataSize set
// to a real var size, returns
// successfully and overwrites
// a 1024-bytes kernel buffer
up(&efivars_lock)
This can be reproduced by concurrent reading of an EFI variable which size
is more than 1024 bytes:
ts# for cpu in $(seq 0 $(nproc --ignore=1)); do ( taskset -c $cpu \
cat /sys/firmware/efi/vars/KEKDefault*/size & ) ; done
Fix this by using a local variable for a var's data buffer size so it
does not get overwritten.
The dmidecode program fails to properly decode the SMBIOS data supplied
by OVMF/UEFI when running in an SEV guest. The SMBIOS area, under SEV, is
encrypted and resides in reserved memory that is marked as EFI runtime
services data.
As a result, when memremap() is attempted for the SMBIOS data, it
can't be mapped as regular RAM (through try_ram_remap()) and, since
the address isn't part of the iomem resources list, it isn't mapped
encrypted through the fallback ioremap().
Add a new __ioremap_check_other() to deal with memory types like
EFI_RUNTIME_SERVICES_DATA which are not covered by the resource ranges.
This allows any runtime services data which has been created encrypted,
to be mapped encrypted too.
[ bp: Move functionality to a separate function. ]
Commit 9af00b7d114f made sure OF devices get an OF style modalias with
I2C events. It assumed all in-tree users were converted, yet it missed
some Macintosh drivers.
Add an OF module device table for all windfarm drivers to make them
automatically load again.
Fixes: 9af00b7d114f ("i2c: core: report OF style module alias for devices registered via OF") Link: https://bugzilla.kernel.org/show_bug.cgi?id=199471 Reported-by: Erhard Furtner <erhard_f@mailbox.org> Tested-by: Erhard Furtner <erhard_f@mailbox.org> Acked-by: Michael Ellerman <mpe@ellerman.id.au> (powerpc) Signed-off-by: Wolfram Sang <wsa@the-dreams.de> Cc: stable@kernel.org # v4.17+ Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
After FS_IOC_REMOVE_ENCRYPTION_KEY removes a key, it syncs the
filesystem and tries to get and put all inodes that were unlocked by the
key so that unused inodes get evicted via fscrypt_drop_inode().
Normally, the inodes are all clean due to the sync.
However, after the filesystem is sync'ed, userspace can modify and close
one of the files. (Userspace is *supposed* to close the files before
removing the key. But it doesn't always happen, and the kernel can't
assume it.) This causes the inode to be dirtied and have i_count == 0.
Then, fscrypt_drop_inode() failed to consider this case and indicated
that the inode can be dropped, causing the write to be lost.
On f2fs, other problems such as a filesystem freeze could occur due to
the inode being freed while still on f2fs's dirty inode list.
Fix this bug by making fscrypt_drop_inode() only drop clean inodes.
I've written an xfstest which detects this bug on ext4, f2fs, and ubifs.
vtimes may wrap and time_before/after64() should be used to determine
whether a given vtime is before or after another. iocg_is_idle() was
incorrectly using plain "<" comparison do determine whether done_vtime
is before vtime. Here, the only thing we're interested in is whether
done_vtime matches vtime which indicates that there's nothing in
flight. Let's test for inequality instead.
Although the IRQ assignment in ipmi_si driver is optional,
platform_get_irq() spews error messages unnecessarily:
ipmi_si dmi-ipmi-si.0: IRQ index 0 not found
Fix this by switching to platform_get_irq_optional().
Cc: stable@vger.kernel.org # 5.4.x Cc: John Donnelly <john.p.donnelly@oracle.com> Fixes: 682a742ed79d ("driver core: platform: Add an error message to platform_get_irq*()") Reported-and-tested-by: Patrick Vo <patrick.vo@hpe.com> Signed-off-by: Takashi Iwai <tiwai@suse.de>
Message-Id: <20200205093146.1352-1-tiwai@suse.de> Signed-off-by: Corey Minyard <cminyard@mvista.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Devices are formatted in multiple of tracks.
For an Extent Space Efficient (ESE) volume we get errors when accessing
unformatted tracks. In this case the driver either formats the track on
the flight for write requests or returns zero data for read requests.
In case a request spans multiple tracks, the indication of an unformatted
track presented for the first track is incorrectly applied to all tracks
covered by the request. As a result, tracks containing data will be handled
as empty, resulting in zero data being returned on read, or overwriting
existing data with zero on write.
Fix by determining the track that gets the NRF error.
For write requests only format the track that is surely not formatted.
For Read requests all tracks before have returned valid data and should not
be touched.
All tracks after the unformatted track might be formatted or not. Those are
returned to the blocklayer to build a new request.
When using alias devices there is a chance that multiple write requests
trigger a format of the same track which might lead to data loss. Ensure
that a track is formatted only once by maintaining a list of currently
processed tracks.
Fixes: 4fb6bf848fbc ("s390/dasd: Add dynamic formatting support for ESE volumes") Cc: stable@vger.kernel.org # 5.3+ Signed-off-by: Stefan Haberland <sth@linux.ibm.com> Reviewed-by: Jan Hoeppner <hoeppner@linux.ibm.com> Reviewed-by: Peter Oberparleiter <oberpar@linux.ibm.com> Signed-off-by: Jens Axboe <axboe@kernel.dk> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Normal, synchronous requests will have their args allocated on the stack.
After the FR_FINISHED bit is set by receiving the reply from the userspace
fuse server, the originating task may return and reuse the stack frame,
resulting in an Oops if the args structure is dereferenced.
Fix by setting a flag in the request itself upon initializing, indicating
whether it has an asynchronous ->end() callback.
When an EVMCS enabled L1 guest on KVM will tries doing enlightened VMEnter
with EVMCS GPA = 0 the host crashes because the
evmcs_gpa != vmx->nested.hv_evmcs_vmptr
condition in nested_vmx_handle_enlightened_vmptrld() will evaluate to
false (as nested.hv_evmcs_vmptr is zeroed after init). The crash will
happen on vmx->nested.hv_evmcs pointer dereference.
Another problematic EVMCS ptr value is '-1' but it only causes host crash
after nested_release_evmcs() invocation. The problem is exactly the same as
with '0', we mistakenly think that the EVMCS pointer hasn't changed and
thus nested.hv_evmcs_vmptr is valid.
Resolve the issue by adding an additional !vmx->nested.hv_evmcs
check to nested_vmx_handle_enlightened_vmptrld(), this way we will
always be trying kvm_vcpu_map() when nested.hv_evmcs is NULL
and this is supposed to catch all invalid EVMCS GPAs.
Also, initialize hv_evmcs_vmptr to '0' in nested_release_evmcs()
to be consistent with initialization where we don't currently
set hv_evmcs_vmptr to '-1'.
"f3 a5" is a "rep movsw" instruction, which should not be intercepted
at all. Commit a455ab392774 ("KVM: emulate: clean up initializations in
init_decode_cache") reduced the number of fields cleared by
init_decode_cache() claiming that they are being cleared elsewhere,
'intercept', however, is left uncleared if the instruction does not have
any of the "slow path" flags (NotImpl, Stack, Op3264, Sse, Mmx, CheckPerm,
NearBranch, No16 and of course Intercept itself).
Fixes: a455ab392774 ("KVM: emulate: clean up initializations in init_decode_cache") Fixes: 705955e37f4b ("KVM: nVMX: Don't emulate instructions in guest mode") Cc: stable@vger.kernel.org Suggested-by: Paolo Bonzini <pbonzini@redhat.com> Signed-off-by: Vitaly Kuznetsov <vkuznets@redhat.com> Reviewed-by: Sean Christopherson <sean.j.christopherson@intel.com> Signed-off-by: Paolo Bonzini <pbonzini@redhat.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
with the way fs/namei.c:do_last() had been done, ->atomic_open()
instances needed to recognize the case when existing file got
found with O_EXCL|O_CREAT, either by falling back to finish_no_open()
or failing themselves. gfs2 one didn't.
several iterations of ->atomic_open() calling conventions ago, we
used to need fput() if ->atomic_open() failed at some point after
successful finish_open(). Now (since 2016) it's not needed -
struct file carries enough state to make fput() work regardless
of the point in struct file lifecycle and discarding it on
failure exits in open() got unified. Unfortunately, I'd missed
the fact that we had an instance of ->atomic_open() (cifs one)
that used to need that fput(), as well as the stale comment in
finish_open() demanding such late failure handling. Trivially
fixed...
Fixes: b033dbb52b3c "do_last(): take fput() on error after opening to out:" Cc: stable@kernel.org # v4.7+ Signed-off-by: Al Viro <viro@zeniv.linux.org.uk> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Before rebooting the box, a "ssh sync" is called to the test machine to see
if it is alive or not. But if the test machine is in a partial state, that
ssh may never actually finish, and the ktest test hangs.
Add a 10 second timeout to the sync test, which will fail after 10 seconds
and then cause the test to reboot the test machine.
Cc: stable@vger.kernel.org Fixes: aa8896990f75d ("ktest.pl: Powercycle the box on reboot if no connection can be made") Signed-off-by: Steven Rostedt (VMware) <rostedt@goodmis.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Since the semaphore fence may be signaled from inside an interrupt
handler from inside a request holding its request->lock, we cannot then
enter into the engine->active.lock for processing the semaphore priority
bump as we may traverse our call tree and end up on another held
request.
Variable grph_obj_type is being assigned twice, one of these is
redundant so remove it.
Addresses-Coverity: ("Evaluation order violation") Signed-off-by: Colin Ian King <colin.king@canonical.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com> Cc: <nobuhiro1.iwamatsu@toshiba.co.jp> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
wq_select_unbound_cpu() is designed for unbound workqueues only, but
it's wrongly called when using a bound workqueue too.
Fixing this ensures work queued to a bound workqueue with
cpu=WORK_CPU_UNBOUND always runs on the local CPU.
Before, that would happen only if wq_unbound_cpumask happened to include
it (likely almost always the case), or was empty, or we got lucky with
forced round-robin placement. So restricting
/sys/devices/virtual/workqueue/cpumask to a small subset of a machine's
CPUs would cause some bound work items to run unexpectedly there.
Fixes: 434754ccedba ("workqueue: schedule WORK_CPU_UNBOUND work on wq_unbound_cpumask CPUs") Cc: stable@vger.kernel.org # v4.5+ Signed-off-by: Hillf Danton <hdanton@sina.com>
[dj: massage changelog] Signed-off-by: Daniel Jordan <daniel.m.jordan@oracle.com> Cc: Tejun Heo <tj@kernel.org> Cc: Lai Jiangshan <jiangshanlai@gmail.com> Cc: linux-kernel@vger.kernel.org Signed-off-by: Tejun Heo <tj@kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
If .next function does not change position index,
following .show function will repeat output related
to current position index.
Without patch:
# dd if=/proc/net/ip_tables_matches # original file output
conntrack
conntrack
conntrack
recent
recent
icmp
udplite
udp
tcp
0+1 records in
0+1 records out
65 bytes copied, 5.4074e-05 s, 1.2 MB/s
# dd if=/proc/net/ip_tables_matches bs=62 skip=1
dd: /proc/net/ip_tables_matches: cannot skip to specified offset
cp <<< end of last line
tcp <<< and then unexpected whole last line once again
0+1 records in
0+1 records out
7 bytes copied, 0.000102447 s, 68.3 kB/s
If .next function does not change position index,
following .show function will repeat output related
to current position index.
Without the patch:
# dd if=/proc/net/xt_recent/SSH # original file outpt
src=127.0.0.4 ttl: 0 last_seen: 6275444819 oldest_pkt: 1 6275444819
src=127.0.0.2 ttl: 0 last_seen: 6275438906 oldest_pkt: 1 6275438906
src=127.0.0.3 ttl: 0 last_seen: 6275441953 oldest_pkt: 1 6275441953
0+1 records in
0+1 records out
204 bytes copied, 6.1332e-05 s, 3.3 MB/s
Read after lseek into middle of last line (offset 140 in example below)
generates expected end of last line and then unexpected whole last line
once again
# dd if=/proc/net/xt_recent/SSH bs=140 skip=1
dd: /proc/net/xt_recent/SSH: cannot skip to specified offset
127.0.0.3 ttl: 0 last_seen: 6275441953 oldest_pkt: 1 6275441953
src=127.0.0.3 ttl: 0 last_seen: 6275441953 oldest_pkt: 1 6275441953
0+1 records in
0+1 records out
132 bytes copied, 6.2487e-05 s, 2.1 MB/s
Quoting from the comment describing the WARN functions in
include/asm-generic/bug.h:
* WARN(), WARN_ON(), WARN_ON_ONCE, and so on can be used to report
* significant kernel issues that need prompt attention if they should ever
* appear at runtime.
*
* Do not use these macros when checking for invalid external inputs
The (buggy) firmware tables which the dmar code was calling WARN_TAINT
for really are invalid external inputs. They are not under the kernel's
control and the issues in them cannot be fixed by a kernel update.
So logging a backtrace, which invites bug reports to be filed about this,
is not helpful.
Since nobody else is going to restart our hw_queue for us, the
blk_mq_start_stopped_hw_queues() is in virtblk_done() is not sufficient
necessarily sufficient to ensure that the queue will get started again.
In case of global resource outage (-ENOMEM because mapping failure,
because of swiotlb full) our virtqueue may be empty and we can get
stuck with a stopped hw_queue.
Let us not stop the queue on arbitrary errors, but only on -EONSPC which
indicates a full virtqueue, where the hw_queue is guaranteed to get
started by virtblk_done() before when it makes sense to carry on
submitting requests. Let us also remove a stale comment.
Signed-off-by: Halil Pasic <pasic@linux.ibm.com> Cc: Jens Axboe <axboe@kernel.dk> Fixes: a37d38ad0930 ("virtio_ring: fix return code on DMA mapping fails") Link: https://lore.kernel.org/r/20200213123728.61216-2-pasic@linux.ibm.com Signed-off-by: Michael S. Tsirkin <mst@redhat.com> Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The logic for checking required NVM sections was recently fixed in
commit d04d93f3d9fa ("iwlwifi: mvm: fix NVM check for 3168
devices"). However, with that fixed the else is now taken for 3168
devices and within the else clause there is a mandatory check for the
PHY_SKU section. This causes the parsing to fail for 3168 devices.
The PHY_SKU section is really only mandatory for the IWL_NVM_EXT
layout (the phy_sku parameter of iwl_parse_nvm_data is only used when
the NVM type is IWL_NVM_EXT). So this changes the PHY_SKU section
check so that it's only mandatory for IWL_NVM_EXT.
Fixes: d04d93f3d9fa ("iwlwifi: mvm: fix NVM check for 3168 devices") Signed-off-by: Dan Moulding <dmoulding@me.com> Signed-off-by: Kalle Valo <kvalo@codeaurora.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
nft will loop forever if the kernel doesn't support an expression:
1. nft_expr_type_get() appends the family specific name to the module list.
2. -EAGAIN is returned to nfnetlink, nfnetlink calls abort path.
3. abort path sets ->done to true and calls request_module for the
expression.
4. nfnetlink replays the batch, we end up in nft_expr_type_get() again.
5. nft_expr_type_get attempts to append family-specific name. This
one already exists on the list, so we continue
6. nft_expr_type_get adds the generic expression name to the module
list. -EAGAIN is returned, nfnetlink calls abort path.
7. abort path encounters the family-specific expression which
has 'done' set, so it gets removed.
8. abort path requests the generic expression name, sets done to true.
9. batch is replayed.
If the expression could not be loaded, then we will end up back at 1),
because the family-specific name got removed and the cycle starts again.
Note that userspace can SIGKILL the nft process to stop the cycle, but
the desired behaviour is to return an error after the generic expr name
fails to load the expression.
Fixes: 32fe74758c077 ("netfilter: nf_tables: autoload modules from the abort path") Signed-off-by: Florian Westphal <fw@strlen.de> Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
PF_EXITING is set earlier than actual removal from css_set when a task
is exitting. This can confuse cgroup.procs readers who see no PF_EXITING
tasks, however, rmdir is checking against css_set membership so it can
transitionally fail with EBUSY.
Fix this by listing tasks that weren't unlinked from css_set active
lists.
It may happen that other users of the task iterator (without
CSS_TASK_ITER_PROCS) spot a PF_EXITING task before cgroup_exit(). This
is equal to the state before commit 0940dd91a4dc ("cgroup: Include dying
leaders with live threads in PROCS iterations") but it may be reviewed
later.
Reported-by: Suren Baghdasaryan <surenb@google.com> Fixes: 0940dd91a4dc ("cgroup: Include dying leaders with live threads in PROCS iterations") Signed-off-by: Michal Koutný <mkoutny@suse.com> Signed-off-by: Tejun Heo <tj@kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
If seq_file .next fuction does not change position index,
read after some lseek can generate unexpected output:
1) dd bs=1 skip output of each 2nd elements
$ dd if=/sys/fs/cgroup/cgroup.procs bs=8 count=1
2
3
4
5
1+0 records in
1+0 records out
8 bytes copied, 0,000267297 s, 29,9 kB/s
[test@localhost ~]$ dd if=/sys/fs/cgroup/cgroup.procs bs=1 count=8
2
4 <<< NB! 3 was skipped
6 <<< ... and 5 too
8 <<< ... and 7
8+0 records in
8+0 records out
8 bytes copied, 5,2123e-05 s, 153 kB/s
This happen because __cgroup_procs_start() makes an extra
extra cgroup_procs_next() call
2) read after lseek beyond end of file generates whole last line.
3) read after lseek into middle of last line generates
expected rest of last line and unexpected whole line once again.
Additionally patch removes an extra position index changes in
__cgroup_procs_start()
It is currently possible for a PHY device to be suspended as part of a
network device driver's suspend call while it is still being attached to
that net_device, either via phy_suspend() or implicitly via phy_stop().
Later on, when the MDIO bus controller get suspended, we would attempt
to suspend again the PHY because it is still attached to a network
device.
This is both a waste of time and creates an opportunity for improper
clock/power management bugs to creep in.
Fixes: 76dcd2ee4473 ("net: phy: avoid suspending twice a PHY") Signed-off-by: Florian Fainelli <f.fainelli@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
By default, DSA drivers should configure CPU and DSA ports to their
maximum speed. In many configurations this is sufficient to make the
link work.
In some cases it is necessary to configure the link to run slower,
e.g. because of limitations of the SoC it is connected to. Or back to
back PHYs are used and the PHY needs to be driven in order to
establish link. In this case, phylink is used.
Only instantiate phylink if it is required. If there is no PHY, or no
fixed link properties, phylink can upset a link which works in the
default configuration.
Fixes: 5f71d95619ea ("net: dsa: Use PHYLINK for the CPU/DSA ports") Signed-off-by: Andrew Lunn <andrew@lunn.ch> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
This patch update {ipv4, ipv6}_addr_metric_test with
1. Set metric of address with peer route and see if the route added
correctly.
2. Modify metric and peer address for peer route and see if the route
changed correctly.
Signed-off-by: Hangbin Liu <liuhangbin@gmail.com> Reviewed-by: David Ahern <dsahern@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
When we modify the peer route and changed it to a new one, we should
remove the old route first. Before the fix:
+ ip addr add dev dummy1 2001:db8::1 peer 2001:db8::2
+ ip -6 route show dev dummy1
2001:db8::1 proto kernel metric 256 pref medium
2001:db8::2 proto kernel metric 256 pref medium
+ ip addr change dev dummy1 2001:db8::1 peer 2001:db8::3
+ ip -6 route show dev dummy1
2001:db8::1 proto kernel metric 256 pref medium
2001:db8::2 proto kernel metric 256 pref medium
After the fix:
+ ip addr change dev dummy1 2001:db8::1 peer 2001:db8::3
+ ip -6 route show dev dummy1
2001:db8::1 proto kernel metric 256 pref medium
2001:db8::3 proto kernel metric 256 pref medium
This patch depend on the previous patch "net/ipv6: need update peer route
when modify metric" to update new peer route after delete old one.
Signed-off-by: Hangbin Liu <liuhangbin@gmail.com> Reviewed-by: David Ahern <dsahern@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
When we modify the route metric, the peer address's route need also
be updated. Before the fix:
+ ip addr add dev dummy1 2001:db8::1 peer 2001:db8::2 metric 60
+ ip -6 route show dev dummy1
2001:db8::1 proto kernel metric 60 pref medium
2001:db8::2 proto kernel metric 60 pref medium
+ ip addr change dev dummy1 2001:db8::1 peer 2001:db8::2 metric 61
+ ip -6 route show dev dummy1
2001:db8::1 proto kernel metric 61 pref medium
2001:db8::2 proto kernel metric 60 pref medium
After the fix:
+ ip addr change dev dummy1 2001:db8::1 peer 2001:db8::2 metric 61
+ ip -6 route show dev dummy1
2001:db8::1 proto kernel metric 61 pref medium
2001:db8::2 proto kernel metric 61 pref medium
Fixes: cc389405590d ("net/ipv6: Add support for specifying metric of connected routes") Signed-off-by: Hangbin Liu <liuhangbin@gmail.com> Reviewed-by: David Ahern <dsahern@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
So far we have the unfortunate situation that mdio_bus_phy_may_suspend()
is called in suspend AND resume path, assuming that function result is
the same. After the original change this is no longer the case,
resulting in broken resume as reported by Geert.
To fix this call mdio_bus_phy_may_suspend() in the suspend path only,
and let the phy_device store the info whether it was suspended by
MDIO bus PM.
projects / kernel.git / log