This allows to replace a tcp option with nop padding to selectively disable
a particular tcp option.
Optstrip mode is chosen when userspace passes the exthdr expression with
neither a source nor a destination register attribute.
This is identical to xtables TCPOPTSTRIP extension.
The only difference is that TCPOPTSTRIP allows to pass in a bitmap
of options to remove rather than a single number.
Unlike TCPOPTSTRIP this expression can be used multiple times
in the same rule to get the same effect.
We could add a new nested attribute later on in case there is a
use case for single-expression-multi-remove.
The touchpad of this device is both connected via PS/2 and i2c. This causes
strange behavior when both driver fight for control. The easy fix is to
prevent the PS/2 driver from accessing the mouse port as the full feature
set of the touchpad is only supported in the i2c interface anyway.
The strange behavior in this case is, that when an external screen is
connected and the notebook is closed, the pointer on the external screen is
moving to the lower right corner. When the notebook is opened again, this
movement stops, but the touchpad clicks are unresponsive afterwards until
reboot.
Now i8042-x86ia64io.h is shared by X86 and IA64, but it can be shared
by more platforms (such as LoongArch) with ACPI firmware on which PNP
typed keyboard and mouse is configured in DSDT. So rename it to i8042-
acpipnpio.h.
This warning corresponds to this code in __queue_work:
/*
* For a draining wq, only works from the same workqueue are
* allowed. The __WQ_DESTROYING helps to spot the issue that
* queues a new work item to a wq after destroy_workqueue(wq).
*/
if (unlikely(wq->flags & (__WQ_DESTROYING | __WQ_DRAINING) &&
WARN_ON_ONCE(!is_chained_work(wq))))
return;
For this to trip, we must have a thread draining the inodedgc workqueue
and a second thread trying to queue inodegc work to that workqueue.
This can happen if freezing or a ro remount race with reclaim poking our
faux inodegc shrinker and another thread dropping an unlinked O_RDONLY
file:
Thread 0 Thread 1 Thread 2
xfs_inodegc_stop
xfs_inodegc_shrinker_scan
xfs_is_inodegc_enabled
<yes, will continue>
xfs_clear_inodegc_enabled
xfs_inodegc_queue_all
<list empty, do not queue inodegc worker>
xfs_inodegc_queue
<add to list>
xfs_is_inodegc_enabled
<no, returns>
In other words, everything between the access to inodegc_enabled state
and the decision to poke the inodegc workqueue requires some kind of
coordination to avoid the WQ_DRAINING state. We could perhaps introduce
a lock here, but we could also try to eliminate WQ_DRAINING from the
picture.
We could replace the drain_workqueue call with a loop that flushes the
workqueue and queues workers as long as there is at least one inode
present in the per-cpu inodegc llists. We've disabled inodegc at this
point, so we know that the number of queued inodes will eventually hit
zero as long as xfs_inodegc_start cannot reactivate the workers.
There are four callers of xfs_inodegc_start. Three of them come from the
VFS with s_umount held: filesystem thawing, failed filesystem freezing,
and the rw remount transition. The fourth caller is mounting rw (no
remount or freezing possible).
There are three callers ofs xfs_inodegc_stop. One is unmounting (no
remount or thaw possible). Two of them come from the VFS with s_umount
held: fs freezing and ro remount transition.
Hence, it is correct to replace the drain_workqueue call with a loop
that drains the inodegc llists.
Fixes: 6191cf3ad59f ("xfs: flush inodegc workqueue tasks before cancel") Signed-off-by: Darrick J. Wong <djwong@kernel.org> Reviewed-by: Dave Chinner <dchinner@redhat.com> Signed-off-by: Dave Chinner <david@fromorbit.com> Signed-off-by: Leah Rumancik <leah.rumancik@gmail.com> Acked-by: Darrick J. Wong <djwong@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
The fscounters scrub code doesn't work properly because it cannot
quiesce updates to the percpu counters in the filesystem, hence it
returns false corruption reports. This has been fixed properly in
one of the online repair patchsets that are under review by replacing
the xchk_disable_reaping calls with an exclusive filesystem freeze.
Disabling background gc isn't sufficient to fix the problem.
In other words, scrub doesn't need to call xfs_inodegc_stop, which is
just as well since it wasn't correct to allow scrub to call
xfs_inodegc_start when something else could be calling xfs_inodegc_stop
(e.g. trying to freeze the filesystem).
Neuter the scrubber for now, and remove the xchk_*_reaping functions.
Signed-off-by: Darrick J. Wong <djwong@kernel.org> Reviewed-by: Dave Chinner <dchinner@redhat.com> Signed-off-by: Dave Chinner <david@fromorbit.com> Signed-off-by: Leah Rumancik <leah.rumancik@gmail.com> Acked-by: Darrick J. Wong <djwong@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
Now that we've allegedly worked out the problem of the per-cpu inodegc
workers being scheduled on the wrong cpu, let's put in a debugging knob
to let us know if a worker ever gets mis-scheduled again.
Signed-off-by: Darrick J. Wong <djwong@kernel.org> Reviewed-by: Dave Chinner <dchinner@redhat.com> Signed-off-by: Dave Chinner <david@fromorbit.com> Signed-off-by: Leah Rumancik <leah.rumancik@gmail.com> Acked-by: Darrick J. Wong <djwong@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
I've been noticing odd racing behavior in the inodegc code that could
only be explained by one cpu adding an inode to its inactivation llist
at the same time that another cpu is processing that cpu's llist.
Preemption is disabled between get/put_cpu_ptr, so the only explanation
is scheduler mayhem. I inserted the following debug code into
xfs_inodegc_worker (see the next patch):
ASSERT(gc->cpu == smp_processor_id());
This assertion tripped during overnight tests on the arm64 machines, but
curiously not on x86_64. I think we haven't observed any resource leaks
here because the lockfree list code can handle simultaneous llist_add
and llist_del_all functions operating on the same list. However, the
whole point of having percpu inodegc lists is to take advantage of warm
memory caches by inactivating inodes on the last processor to touch the
inode.
The incorrect scheduling seems to occur after an inodegc worker is
subjected to mod_delayed_work(). This wraps mod_delayed_work_on with
WORK_CPU_UNBOUND specified as the cpu number. Unbound allows for
scheduling on any cpu, not necessarily the same one that scheduled the
work.
Because preemption is disabled for as long as we have the gc pointer, I
think it's safe to use current_cpu() (aka smp_processor_id) to queue the
delayed work item on the correct cpu.
Fixes: 7cf2b0f9611b ("xfs: bound maximum wait time for inodegc work") Signed-off-by: Darrick J. Wong <djwong@kernel.org> Reviewed-by: Dave Chinner <dchinner@redhat.com> Signed-off-by: Dave Chinner <david@fromorbit.com> Signed-off-by: Leah Rumancik <leah.rumancik@gmail.com> Acked-by: Darrick J. Wong <djwong@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
The current blocking mechanism for pushing the inodegc queue out to
disk can result in systems becoming unusable when there is a long
running inodegc operation. This is because the statfs()
implementation currently issues a blocking flush of the inodegc
queue and a significant number of common system utilities will call
statfs() to discover something about the underlying filesystem.
This can result in userspace operations getting stuck on inodegc
progress, and when trying to remove a heavily reflinked file on slow
storage with a full journal, this can result in delays measuring in
hours.
Avoid this problem by adding "push" function that expedites the
flushing of the inodegc queue, but doesn't wait for it to complete.
Convert xfs_fs_statfs() and xfs_qm_scall_getquota() to use this
mechanism so they don't block but still ensure that queued
operations are expedited.
Fixes: ab23a7768739 ("xfs: per-cpu deferred inode inactivation queues") Reported-by: Chris Dunlop <chris@onthe.net.au> Signed-off-by: Dave Chinner <dchinner@redhat.com>
[djwong: fix _getquota_next to use _inodegc_push too] Reviewed-by: Darrick J. Wong <djwong@kernel.org> Signed-off-by: Darrick J. Wong <djwong@kernel.org> Signed-off-by: Leah Rumancik <leah.rumancik@gmail.com> Acked-by: Darrick J. Wong <djwong@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
Currently inodegc work can sit queued on the per-cpu queue until
the workqueue is either flushed of the queue reaches a depth that
triggers work queuing (and later throttling). This means that we
could queue work that waits for a long time for some other event to
trigger flushing.
Hence instead of just queueing work at a specific depth, use a
delayed work that queues the work at a bound time. We can still
schedule the work immediately at a given depth, but we no long need
to worry about leaving a number of items on the list that won't get
processed until external events prevail.
Signed-off-by: Dave Chinner <dchinner@redhat.com> Reviewed-by: Darrick J. Wong <djwong@kernel.org> Signed-off-by: Darrick J. Wong <djwong@kernel.org> Signed-off-by: Leah Rumancik <leah.rumancik@gmail.com> Acked-by: Darrick J. Wong <djwong@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
In i2c_mux_gpio_probe_fw(), we should add fwnode_handle_put()
when break out of the iteration device_for_each_child_node()
as it will automatically increase and decrease the refcounter.
Fixes: 98b2b712bc85 ("i2c: i2c-mux-gpio: Enable this driver in ACPI land") Signed-off-by: Liang He <windhl@126.com> Signed-off-by: Wolfram Sang <wsa@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
If an error occurs after a successful irq_domain_add_linear() call, it
should be undone by a corresponding irq_domain_remove(), as already done
in the remove function.
Fix up backport of 136191703038 ("interconnect: Teach lockdep about icc_bw_lock order")
Add a missing include to fix the following build error:
drivers/interconnect/core.c: In function 'icc_init':
drivers/interconnect/core.c:1148:9: error: implicit declaration of function 'fs_reclaim_acquire' [-Werror=implicit-function-declaration]
1148 | fs_reclaim_acquire(GFP_KERNEL);
| ^~~~~~~~~~~~~~~~~~
drivers/interconnect/core.c:1150:9: error: implicit declaration of function 'fs_reclaim_release' [-Werror=implicit-function-declaration]
1150 | fs_reclaim_release(GFP_KERNEL);
| ^~~~~~~~~~~~~~~~~~
When users attempt to obtain the coalesce setting using the
ethtool command, current code always returns 0 for tx-usecs.
This is because I225/6 always uses a queue pair setting, hence
tx_coalesce_usecs does not return a value during the
igc_ethtool_get_coalesce() callback process. The pair queue
condition checking in igc_ethtool_get_coalesce() is removed by
this patch so that the user gets information of the value of tx-usecs.
Even if i225/6 is using queue pair setting, there is no harm in
notifying the user of the tx-usecs. The implementation of the current
code may have previously been a copy of the legacy code i210.
Since I225 has the queue pair setting enabled, tx-usecs will always adhere
to the user-set rx-usecs value. An error message will appear when the user
attempts to set the tx-usecs value for the input parameters because,
by default, they should only set the rx-usecs value.
This patch also adds the helper function to get the
previous rx coalesce value similar to tx coalesce.
How to test:
User can get the coalesce value using ethtool command.
bnxt_poll_nitroa0() invokes bnxt_rx_pkt() which can run a XDP program
which in turn can return XDP_REDIRECT. bnxt_rx_pkt() is also used by
__bnxt_poll_work() which flushes (xdp_do_flush()) the packets after each
round. bnxt_poll_nitroa0() lacks this feature.
xdp_do_flush() should be invoked before leaving the NAPI callback.
Invoke xdp_do_flush() after a redirect in bnxt_poll_nitroa0() NAPI.
Cc: Michael Chan <michael.chan@broadcom.com> Fixes: f18c2b77b2e4e ("bnxt_en: optimized XDP_REDIRECT support") Reviewed-by: Andy Gospodarek <gospo@broadcom.com> Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Reviewed-by: Michael Chan <michael.chan@broadcom.com> Acked-by: Jesper Dangaard Brouer <hawk@kernel.org> Signed-off-by: Paolo Abeni <pabeni@redhat.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
xdp_do_flush() should be invoked before leaving the NAPI poll function
after a XDP-redirect. This is not the case if the driver leaves via
the error path (after having a redirect in one of its previous
iterations).
Invoke xdp_do_flush() also in the error path.
Cc: Arthur Kiyanovski <akiyano@amazon.com> Cc: David Arinzon <darinzon@amazon.com> Cc: Noam Dagan <ndagan@amazon.com> Cc: Saeed Bishara <saeedb@amazon.com> Cc: Shay Agroskin <shayagr@amazon.com> Fixes: a318c70ad152b ("net: ena: introduce XDP redirect implementation") Acked-by: Arthur Kiyanovski <akiyano@amazon.com> Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Acked-by: Jesper Dangaard Brouer <hawk@kernel.org> Signed-off-by: Paolo Abeni <pabeni@redhat.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
It was brought up by Tetsuo that the following sequence:
write_seqlock_irqsave()
printk_deferred_enter()
could lead to a deadlock if the lockdep annotation within
write_seqlock_irqsave() triggers.
The problem is that the sequence counter is incremented before the lockdep
annotation is performed. The lockdep splat would then attempt to invoke
printk() but the reader side, of the same seqcount, could have a
tty_port::lock acquired waiting for the sequence number to become even again.
The other lockdep annotations come before the actual locking because "we
want to see the locking error before it happens". There is no reason why
seqcount should be different here.
Do the lockdep annotation first then perform the locking operation (the
sequence increment).
Fixes: 1ca7d67cf5d5a ("seqcount: Add lockdep functionality to seqcount/seqlock structures") Reported-by: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp> Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Signed-off-by: Ingo Molnar <mingo@kernel.org> Link: https://lore.kernel.org/r/20230920104627._DTHgPyA@linutronix.de Closes: https://lore.kernel.org/20230621130641.-5iueY1I@linutronix.de Signed-off-by: Sasha Levin <sashal@kernel.org>
Kyle Zeng reported that there is a race between IPSET_CMD_ADD and IPSET_CMD_SWAP
in netfilter/ip_set, which can lead to the invocation of `__ip_set_put` on a
wrong `set`, triggering the `BUG_ON(set->ref == 0);` check in it.
The race is caused by using the wrong reference counter, i.e. the ref counter instead
of ref_netlink.
In rds_rdma_cm_event_handler_cmn() check, if conn pointer exists
before dereferencing it as rdma_set_service_type() argument
Found by Linux Verification Center (linuxtesting.org) with SVACE.
Fixes: fd261ce6a30e ("rds: rdma: update rdma transport for tos") Signed-off-by: Artem Chernyshev <artem.chernyshev@red-soft.ru> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Sasha Levin <sashal@kernel.org>
[1]
$ teamd -t team0 -d -c '{"runner": {"name": "loadbalance"}}'
$ ip link add name t-dummy type dummy
$ ip link add link t-dummy name t-dummy.100 type vlan id 100
$ ip link add name t-nlmon type nlmon
$ ip link set t-nlmon master team0
$ ip link set t-nlmon nomaster
$ ip link set t-dummy up
$ ip link set team0 up
$ ip link set t-dummy.100 down
$ ip link set t-dummy.100 master team0
When enslave a vlan device to team device and team device type is changed
from non-ether to ether, header_ops of team device is changed to
vlan_header_ops. That is incorrect and will trigger null-ptr-deref
for vlan->real_dev in vlan_dev_hard_header() because team device is not
a vlan device.
Cache eth_header_ops in team_setup(), then assign cached header_ops to
header_ops of team net device when its type is changed from non-ether
to ether to fix the bug.
Fixes: 1d76efe1577b ("team: add support for non-ethernet devices") Suggested-by: Hangbin Liu <liuhangbin@gmail.com> Reviewed-by: Hangbin Liu <liuhangbin@gmail.com> Signed-off-by: Ziyang Xuan <william.xuanziyang@huawei.com> Reviewed-by: Jiri Pirko <jiri@nvidia.com> Reviewed-by: Eric Dumazet <edumazet@google.com> Link: https://lore.kernel.org/r/20230918123011.1884401-1-william.xuanziyang@huawei.com Signed-off-by: Paolo Abeni <pabeni@redhat.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
Currently the reset process in hns3 and firmware watchdog init process is
asynchronous. we think firmware watchdog initialization is completed
before hns3 clear the firmware interrupt source. However, firmware
initialization may not complete early.
so we add delay before hns3 clear firmware interrupt source and 5 ms delay
is enough to avoid second firmware reset interrupt.
Fixes: c1a81619d73a ("net: hns3: Add mailbox interrupt handling to PF driver") Signed-off-by: Jie Wang <wangjie125@huawei.com> Signed-off-by: Jijie Shao <shaojijie@huawei.com> Signed-off-by: Paolo Abeni <pabeni@redhat.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
Currently, the driver will enable unicast promisc for the function
once configure mac address fail. It's unreasonable when the failure
is caused by using same mac address with other functions. So only
enable unicast promisc when mac table full.
The device_version V3 hardware can't offload the checksum for IP in GRE
packets, but can do it for NvGRE. So default to disable the checksum and
GSO offload for GRE, but keep the ability to enable it when only using
NvGRE.
Fixes: 76ad4f0ee747 ("net: hns3: Add support of HNS3 Ethernet Driver for hip08 SoC") Signed-off-by: Jie Wang <wangjie125@huawei.com> Signed-off-by: Jijie Shao <shaojijie@huawei.com> Signed-off-by: Paolo Abeni <pabeni@redhat.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
Reading the 'spec_rstack_overflow' sysfs file can trigger an unnecessary
MSR write, and possibly even a (handled) exception if the microcode
hasn't been updated.
Avoid all that by just checking X86_FEATURE_IBPB_BRTYPE instead, which
gets set by srso_select_mitigation() if the updated microcode exists.
It's possible for interrupts to get significantly delayed to the point
that callers of intel_scu_ipc_dev_command() and friends can call the
function once, hit a timeout, and call it again while the interrupt
still hasn't been processed. This driver will get seriously confused if
the interrupt is finally processed after the second IPC has been sent
with ipc_command(). It won't know which IPC has been completed. This
could be quite disastrous if calling code assumes something has happened
upon return from intel_scu_ipc_dev_simple_command() when it actually
hasn't.
Let's avoid this scenario by simply returning -EBUSY in this case.
Hopefully higher layers will know to back off or fail gracefully when
this happens. It's all highly unlikely anyway, but it's better to be
correct here as we have no way to know which IPC the status register is
telling us about if we send a second IPC while the previous IPC is still
processing.
Cc: Prashant Malani <pmalani@chromium.org> Cc: Kuppuswamy Sathyanarayanan <sathyanarayanan.kuppuswamy@linux.intel.com> Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com> Reviewed-by: Mika Westerberg <mika.westerberg@linux.intel.com> Fixes: ed12f295bfd5 ("ipc: Added support for IPC interrupt mode") Signed-off-by: Stephen Boyd <swboyd@chromium.org> Link: https://lore.kernel.org/r/20230913212723.3055315-5-swboyd@chromium.org Reviewed-by: Ilpo Järvinen <ilpo.jarvinen@linux.intel.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>
Andy discovered this bug during patch review. The 'scu' argument to this
function shouldn't be overridden by the function itself. It doesn't make
any sense. Looking at the commit history, we see that commit f57fa18583f5 ("platform/x86: intel_scu_ipc: Introduce new SCU IPC API")
removed the setting of the scu to ipcdev in other functions, but not
this one. That was an oversight. Remove this line so that we stop
overriding the scu instance that is used by this function.
Reported-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com> Closes: https://lore.kernel.org/r/ZPjdZ3xNmBEBvNiS@smile.fi.intel.com Cc: Prashant Malani <pmalani@chromium.org> Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com> Reviewed-by: Mika Westerberg <mika.westerberg@linux.intel.com> Fixes: f57fa18583f5 ("platform/x86: intel_scu_ipc: Introduce new SCU IPC API") Signed-off-by: Stephen Boyd <swboyd@chromium.org> Link: https://lore.kernel.org/r/20230913212723.3055315-4-swboyd@chromium.org Reviewed-by: Ilpo Järvinen <ilpo.jarvinen@linux.intel.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>
It's possible for the completion in ipc_wait_for_interrupt() to timeout,
simply because the interrupt was delayed in being processed. A timeout
in itself is not an error. This driver should check the status register
upon a timeout to ensure that scheduling or interrupt processing delays
don't affect the outcome of the IPC return value.
Fix this problem by reading the status bit in all cases, regardless of
the timeout. If the completion times out, we'll assume the problem was
that the IPC_STATUS_BUSY bit was still set, but if the status bit is
cleared in the meantime we know that we hit some scheduling delay and we
should just check the error bit.
Cc: Prashant Malani <pmalani@chromium.org> Reviewed-by: Kuppuswamy Sathyanarayanan <sathyanarayanan.kuppuswamy@linux.intel.com> Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com> Reviewed-by: Mika Westerberg <mika.westerberg@linux.intel.com> Fixes: ed12f295bfd5 ("ipc: Added support for IPC interrupt mode") Signed-off-by: Stephen Boyd <swboyd@chromium.org> Link: https://lore.kernel.org/r/20230913212723.3055315-3-swboyd@chromium.org 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>
It's possible for the polling loop in busy_loop() to get scheduled away
for a long time.
status = ipc_read_status(scu); // status = IPC_STATUS_BUSY
<long time scheduled away>
if (!(status & IPC_STATUS_BUSY))
If this happens, then the status bit could change while the task is
scheduled away and this function would never read the status again after
timing out. Instead, the function will return -ETIMEDOUT when it's
possible that scheduling didn't work out and the status bit was cleared.
Bit polling code should always check the bit being polled one more time
after the timeout in case this happens.
Fix this by reading the status once more after the while loop breaks.
The readl_poll_timeout() macro implements all of this, and it is
shorter, so use that macro here to consolidate code and fix this.
There were some concerns with using readl_poll_timeout() because it uses
timekeeping, and timekeeping isn't running early on or during the late
stages of system suspend or early stages of system resume, but an audit
of the code concluded that this code isn't called during those times so
it is safe to use the macro.
Cc: Prashant Malani <pmalani@chromium.org> Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com> Reviewed-by: Mika Westerberg <mika.westerberg@linux.intel.com> Reviewed-by: Kuppuswamy Sathyanarayanan <sathyanarayanan.kuppuswamy@linux.intel.com> Fixes: e7b7ab3847c9 ("platform/x86: intel_scu_ipc: Sleeping is fine when polling") Signed-off-by: Stephen Boyd <swboyd@chromium.org> Link: https://lore.kernel.org/r/20230913212723.3055315-2-swboyd@chromium.org Reviewed-by: Ilpo Järvinen <ilpo.jarvinen@linux.intel.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>
Valid domain value is in range 1 to HV_PERF_DOMAIN_MAX. Current code has
check for domain value greater than or equal to HV_PERF_DOMAIN_MAX. But
the check for domain value 0 is missing.
Fix this issue by adding check for domain value 0.
Before:
# ./perf stat -v -e hv_24x7/CPM_ADJUNCT_INST,domain=0,core=1/ sleep 1
Using CPUID 00800200
Control descriptor is not initialized
Error:
The sys_perf_event_open() syscall returned with 5 (Input/output error) for
event (hv_24x7/CPM_ADJUNCT_INST,domain=0,core=1/).
/bin/dmesg | grep -i perf may provide additional information.
Result from dmesg:
[ 37.819387] hv-24x7: hcall failed: [0 0x60040000 0x100 0] => ret
0xfffffffffffffffc (-4) detail=0x2000000 failing ix=0
After:
# ./perf stat -v -e hv_24x7/CPM_ADJUNCT_INST,domain=0,core=1/ sleep 1
Using CPUID 00800200
Control descriptor is not initialized
Warning:
hv_24x7/CPM_ADJUNCT_INST,domain=0,core=1/ event is not supported by the kernel.
failed to read counter hv_24x7/CPM_ADJUNCT_INST,domain=0,core=1/
Currently, we assume the skb is associated with a device before calling
__ip_options_compile, which is not always the case if it is re-routed by
ipvs.
When skb->dev is NULL, dev_net(skb->dev) will become null-dereference.
This patch adds a check for the edge case and switch to use the net_device
from the rtable when skb->dev is NULL.
Fixes: ed0de45a1008 ("ipv4: recompile ip options in ipv4_link_failure") Suggested-by: David Ahern <dsahern@kernel.org> Signed-off-by: Kyle Zeng <zengyhkyle@gmail.com> Cc: Stephen Suryaputra <ssuryaextr@gmail.com> Cc: Vadim Fedorenko <vfedorenko@novek.ru> Reviewed-by: David Ahern <dsahern@kernel.org> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Sasha Levin <sashal@kernel.org>
When an XDP redirect happens before the link is ready, that
transmission will not finish and will timeout, causing an adapter
reset. If the redirects do not stop, the adapter will not stop
resetting.
Wait for the driver to signal that there's a carrier before allowing
transmissions to proceed.
Previous code was relying that when __IGC_DOWN is cleared, the NIC is
ready to transmit as all the queues are ready, what happens is that
the carrier presence will only be signaled later, after the watchdog
workqueue has a chance to run. And during this interval (between
clearing __IGC_DOWN and the watchdog running) if any transmission
happens the timeout is emitted (detected by igc_tx_timeout()) which
causes the reset, with the potential for the infinite loop.
Fixes: 4ff320361092 ("igc: Add support for XDP_REDIRECT action") Reported-by: Ferenc Fejes <ferenc.fejes@ericsson.com> Closes: https://lore.kernel.org/netdev/0caf33cf6adb3a5bf137eeaa20e89b167c9986d5.camel@ericsson.com/ Signed-off-by: Vinicius Costa Gomes <vinicius.gomes@intel.com> Tested-by: Ferenc Fejes <ferenc.fejes@ericsson.com> Reviewed-by: Maciej Fijalkowski <maciej.fijalkowski@intel.com> Tested-by: Naama Meir <naamax.meir@linux.intel.com> Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Sasha Levin <sashal@kernel.org>
The ionic device supports a maximum buffer length of 16 bits (see
ionic_rxq_desc or ionic_rxq_sg_elem). When adding new buffers to
the receive rings, the function ionic_rx_fill() uses 16bit math when
calculating the number of pages to allocate for an RX descriptor,
given the interface's MTU setting. If the system PAGE_SIZE >= 64KB,
and the buf_info->page_offset is 0, the remain_len value will never
decrement from the original MTU value and the frag_len value will
always be 0, causing additional pages to be allocated as scatter-
gather elements unnecessarily.
A similar math issue exists in ionic_rx_frags(), but no failures
have been observed here since a 64KB page should not normally
require any scatter-gather elements at any legal Ethernet MTU size.
Fixes: 4b0a7539a372 ("ionic: implement Rx page reuse") Signed-off-by: David Christensen <drc@linux.vnet.ibm.com> Reviewed-by: Shannon Nelson <shannon.nelson@amd.com> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Sasha Levin <sashal@kernel.org>
If port VLAN is configured on a VF then any other VLANs on top of this VF
are broken.
During i40e_ndo_set_vf_port_vlan() call the i40e driver reset the VF and
iavf driver asks PF (using VIRTCHNL_OP_GET_VF_RESOURCES) for VF capabilities
but this reset occurs too early, prior setting of vf->info.pvid field
and because this field can be zero during i40e_vc_get_vf_resources_msg()
then VIRTCHNL_VF_OFFLOAD_VLAN capability is reported to iavf driver.
This is wrong because iavf driver should not report VLAN offloading
capability when port VLAN is configured as i40e does not support QinQ
offloading.
Fix the issue by moving VF reset after setting of vf->port_vlan_id
field.
Without this patch:
$ echo 1 > /sys/class/net/enp2s0f0/device/sriov_numvfs
$ ip link set enp2s0f0 vf 0 vlan 3
$ ip link set enp2s0f0v0 up
$ ip link add link enp2s0f0v0 name vlan4 type vlan id 4
$ ip link set vlan4 up
...
$ ethtool -k enp2s0f0v0 | grep vlan-offload
rx-vlan-offload: on
tx-vlan-offload: on
$ dmesg -l err | grep iavf
[1292500.742914] iavf 0000:02:02.0: Failed to add VLAN filter, error IAVF_ERR_INVALID_QP_ID
With this patch:
$ echo 1 > /sys/class/net/enp2s0f0/device/sriov_numvfs
$ ip link set enp2s0f0 vf 0 vlan 3
$ ip link set enp2s0f0v0 up
$ ip link add link enp2s0f0v0 name vlan4 type vlan id 4
$ ip link set vlan4 up
...
$ ethtool -k enp2s0f0v0 | grep vlan-offload
rx-vlan-offload: off [requested on]
tx-vlan-offload: off [requested on]
$ dmesg -l err | grep iavf
Fixes: f9b4b6278d51 ("i40e: Reset the VF upon conflicting VLAN configuration") Signed-off-by: Ivan Vecera <ivecera@redhat.com> Reviewed-by: Jesse Brandeburg <jesse.brandeburg@intel.com> Tested-by: Rafal Romanowski <rafal.romanowski@intel.com> Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
VFs by default are able to see all tagged traffic regardless of trust
and VLAN filters configured.
Add new private flag vf-vlan-pruning that allows changing of default
VF behavior for tagged traffic. When the flag is turned on
untrusted VF will only be able to receive untagged traffic
or traffic with VLAN tags it has created interfaces for
The flag is off by default and can only be changed if
there are no VFs spawned on the PF. This flag will only be effective
when no PVID is set on VF and VF is not trusted.
Add new function that computes the correct VLAN ID for VF VLAN filters
based on trust, PVID, vf-vlan-prune-disable flag and current VLAN ID.
Testing Hints:
Test 1: vf-vlan-pruning == off
==============================
1. Set the private flag
> ethtool --set-priv-flag eth0 vf-vlan-pruning off (default setting)
2. Use scapy to send any VLAN tagged traffic and make sure the VF
receives all VLAN tagged traffic that matches its destination MAC
filters (unicast, multicast, and broadcast).
Test 2: vf-vlan-pruning == on
==============================
1. Set the private flag
> ethtool --set-priv-flag eth0 vf-vlan-pruning on
2. Use scapy to send any VLAN tagged traffic and make sure the VF does
not receive any VLAN tagged traffic that matches its destination MAC
filters (unicast, multicast, and broadcast).
3. Add a VLAN filter on the VF netdev
> ip link add link eth0v0 name vlan10 type vlan id 10
4. Bring the VLAN netdev up
> ip link set vlan10 up
4. Use scapy to send traffic with VLAN 10, VLAN 11 (anything not VLAN
10), and untagged traffic. Make sure the VF only receives VLAN 10
and untagged traffic when the link partner is sending.
Test 3: vf-vlan-pruning == off && VF is in a port VLAN
==============================
1. Set the private flag
> ethtool --set-priv-flag eth0 vf-vlan-pruning off (default setting)
2. Create a VF
> echo 1 > sriov_numvfs
3. Put the VF in a port VLAN
> ip link set eth0 vf 0 vlan 10
4. Use scapy to send traffic with VLAN 10 and VLAN 11 (anything not VLAN
10) and make sure the VF only receives untagged traffic when the link
partner is sending VLAN 10 tagged traffic as the VLAN tag is expected
to be stripped by HW for port VLANs and not visible to the VF.
Test 4: Change vf-vlan-pruning while VFs are created
==============================
echo 0 > sriov_numvfs
ethtool --set-priv-flag eth0 vf-vlan-pruning off
echo 1 > sriov_numvfs
ethtool --set-priv-flag eth0 vf-vlan-pruning on (expect failure)
Signed-off-by: Sylwester Dziedziuch <sylwesterx.dziedziuch@intel.com> Signed-off-by: Przemyslaw Patynowski <przemyslawx.patynowski@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>
Stable-dep-of: d0d362ffa33d ("i40e: Fix VF VLAN offloading when port VLAN is configured") Signed-off-by: Sasha Levin <sashal@kernel.org>
Prevent schedule operations for adminq during device remove and when
__IAVF_IN_REMOVE_TASK flag is set. Currently, the iavf_down function
adds operations for adminq that shouldn't be processed when the device
is in the __IAVF_REMOVE state.
Reproduction:
echo 4 > /sys/bus/pci/devices/0000:17:00.0/sriov_numvfs
ip link set dev ens1f0 vf 0 trust on
ip link set dev ens1f0 vf 1 trust on
ip link set dev ens1f0 vf 2 trust on
ip link set dev ens1f0 vf 3 trust on
ip link set dev ens1f0 vf 0 mac 00:22:33:44:55:66
ip link set dev ens1f0 vf 1 mac 00:22:33:44:55:67
ip link set dev ens1f0 vf 2 mac 00:22:33:44:55:68
ip link set dev ens1f0 vf 3 mac 00:22:33:44:55:69
The devm_clk_get() can return -EPROBE_DEFER error,
modify the error code to be -EINVAL is not correct, which
cause the -EPROBE_DEFER error is not correctly handled.
This patch is to fix the return error code.
Fixes: b86ef5367761 ("ASoC: fsl: Add Audio Mixer machine driver") Signed-off-by: Shengjiu Wang <shengjiu.wang@nxp.com> Reviewed-by: Daniel Baluta <daniel.baluta@nxp.com> Link: https://lore.kernel.org/r/1694757731-18308-1-git-send-email-shengjiu.wang@nxp.com Signed-off-by: Mark Brown <broonie@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
When a PTP ethernet raw frame with a size of more than 256 bytes followed
by a 0xff pattern is sent to __skb_flow_dissect, nhoff value calculation
is wrong. For example: hdr->message_length takes the wrong value (0xffff)
and it does not replicate real header length. In this case, 'nhoff' value
was overridden and the PTP header was badly dissected. This leads to a
kernel crash.
tls.sendmsg_large and tls.sendmsg_multiple are trying to send through
the self->cfd socket (only configured with TLS_RX) and to receive through
the self->fd socket (only configured with TLS_TX), so they're not using
kTLS at all. Swap the sockets.
Fixes: 7f657d5bf507 ("selftests: tls: add selftests for TLS sockets") Signed-off-by: Sabrina Dubroca <sd@queasysnail.net> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Sasha Levin <sashal@kernel.org>
Sysbot discovered that the queue and stack maps can deadlock if they are
being used from a BPF program that can be called from NMI context (such as
one that is attached to a perf HW counter event). To fix this, add an
in_nmi() check and use raw_spin_trylock() in NMI context, erroring out if
grabbing the lock fails.
Anonymous sets need to be populated once at creation and then they are
bound to rule since 938154b93be8 ("netfilter: nf_tables: reject unbound
anonymous set before commit phase"), otherwise transaction reports
EINVAL.
Userspace does not need to delete elements of anonymous sets that are
not yet bound, reject this with EOPNOTSUPP.
From flush command path, skip anonymous sets, they are expected to be
bound already. Otherwise, EINVAL is hit at the end of this transaction
for unbound sets.
For spdif input to report the locked rate correctly, even when no capture
is running, the HW and reference clock must be started as soon as
the dai is probed.
nft_trans_gc_queue_sync() enqueues the GC transaction and it allocates a
new one. If this allocation fails, then stop this GC sync run and retry
later.
Fixes: 5f68718b34a5 ("netfilter: nf_tables: GC transaction API to avoid race with control plane") Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
pipapo needs to enqueue GC transactions for catchall elements through
nft_trans_gc_queue_sync(). Add nft_trans_gc_catchall_sync() and
nft_trans_gc_catchall_async() to handle GC transaction queueing
accordingly.
Fixes: 5f68718b34a5 ("netfilter: nf_tables: GC transaction API to avoid race with control plane") Fixes: f6c383b8c31a ("netfilter: nf_tables: adapt set backend to use GC transaction API") Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
rbtree GC does not modify the datastructure, instead it collects expired
elements and it enqueues a GC transaction. Use a read spinlock instead
to avoid data contention while GC worker is running.
Fixes: f6c383b8c31a ("netfilter: nf_tables: adapt set backend to use GC transaction API") Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
New elements in this transaction might expired before such transaction
ends. Skip sync GC for such elements otherwise commit path might walk
over an already released object. Once transaction is finished, async GC
will collect such expired element.
Fixes: f6c383b8c31a ("netfilter: nf_tables: adapt set backend to use GC transaction API") Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org> Signed-off-by: Florian Westphal <fw@strlen.de> Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
Don't queue more gc work, else we may queue the same elements multiple
times.
If an element is flagged as dead, this can mean that either the previous
gc request was invalidated/discarded by a transaction or that the previous
request is still pending in the system work queue.
The latter will happen if the gc interval is set to a very low value,
e.g. 1ms, and system work queue is backlogged.
The sets refcount is 1 if no previous gc requeusts are queued, so add
a helper for this and skip gc run if old requests are pending.
Add a helper for this and skip the gc run in this case.
Fixes: f6c383b8c31a ("netfilter: nf_tables: adapt set backend to use GC transaction API") Signed-off-by: Florian Westphal <fw@strlen.de> Signed-off-by: Sasha Levin <sashal@kernel.org>
Use nf_tables_gc_list_lock spinlock, not nf_tables_destroy_list_lock to
protect the gc_list.
Fixes: 5f68718b34a5 ("netfilter: nf_tables: GC transaction API to avoid race with control plane") Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
Abort path is missing a synchronization point with GC transactions. Add
GC sequence number hence any GC transaction losing race will be
discarded.
Fixes: 5f68718b34a5 ("netfilter: nf_tables: GC transaction API to avoid race with control plane") Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
Netlink event path is missing a synchronization point with GC
transactions. Add GC sequence number update to netns release path and
netlink event path, any GC transaction losing race will be discarded.
Fixes: 5f68718b34a5 ("netfilter: nf_tables: GC transaction API to avoid race with control plane") Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org> Signed-off-by: Florian Westphal <fw@strlen.de> Signed-off-by: Sasha Levin <sashal@kernel.org>
Insertion must ignore duplicate but expired entries.
Moreover, there is a strange asymmetry in nft_pipapo_activate:
It refetches the current element, whereas the other ->activate callbacks
(bitmap, hash, rhash, rbtree) use elem->priv.
Same for .remove: other set implementations take elem->priv,
nft_pipapo_remove fetches elem->priv, then does a relookup,
remove this.
I suspect this was the reason for the change that prompted the
removal of the expired check in pipapo_get() in the first place,
but skipping exired elements there makes no sense to me, this helper
is used for normal get requests, insertions (duplicate check)
and deactivate callback.
In first two cases expired elements must be skipped.
For ->deactivate(), this gets called for DELSETELEM, so it
seems to me that expired elements should be skipped as well, i.e.
delete request should fail with -ENOENT error.
Fixes: 24138933b97b ("netfilter: nf_tables: don't skip expired elements during walk") Signed-off-by: Florian Westphal <fw@strlen.de> Signed-off-by: Sasha Levin <sashal@kernel.org>
Set on the NFT_SET_ELEM_DEAD_BIT flag on this element, instead of
performing element removal which might race with an ongoing transaction.
Enable gc when dynamic flag is set on since dynset deletion requires
garbage collection after this patch.
Fixes: d0a8d877da97 ("netfilter: nft_dynset: support for element deletion") Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
Use the GC transaction API to replace the old and buggy gc API and the
busy mark approach.
No set elements are removed from async garbage collection anymore,
instead the _DEAD bit is set on so the set element is not visible from
lookup path anymore. Async GC enqueues transaction work that might be
aborted and retried later.
rbtree and pipapo set backends does not set on the _DEAD bit from the
sync GC path since this runs in control plane path where mutex is held.
In this case, set elements are deactivated, removed and then released
via RCU callback, sync GC never fails.
Fixes: 3c4287f62044 ("nf_tables: Add set type for arbitrary concatenation of ranges") Fixes: 8d8540c4f5e0 ("netfilter: nft_set_rbtree: add timeout support") Fixes: 9d0982927e79 ("netfilter: nft_hash: add support for timeouts") Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
The set types rhashtable and rbtree use a GC worker to reclaim memory.
From system work queue, in periodic intervals, a scan of the table is
done.
The major caveat here is that the nft transaction mutex is not held.
This causes a race between control plane and GC when they attempt to
delete the same element.
We cannot grab the netlink mutex from the work queue, because the
control plane has to wait for the GC work queue in case the set is to be
removed, so we get following deadlock:
cpu 1 cpu2
GC work transaction comes in , lock nft mutex
`acquire nft mutex // BLOCKS
transaction asks to remove the set
set destruction calls cancel_work_sync()
cancel_work_sync will now block forever, because it is waiting for the
mutex the caller already owns.
This patch adds a new API that deals with garbage collection in two
steps:
1) Lockless GC of expired elements sets on the NFT_SET_ELEM_DEAD_BIT
so they are not visible via lookup. Annotate current GC sequence in
the GC transaction. Enqueue GC transaction work as soon as it is
full. If ruleset is updated, then GC transaction is aborted and
retried later.
2) GC work grabs the mutex. If GC sequence has changed then this GC
transaction lost race with control plane, abort it as it contains
stale references to objects and let GC try again later. If the
ruleset is intact, then this GC transaction deactivates and removes
the elements and it uses call_rcu() to destroy elements.
Note that no elements are removed from GC lockless path, the _DEAD bit
is set and pointers are collected. GC catchall does not remove the
elements anymore too. There is a new set->dead flag that is set on to
abort the GC transaction to deal with set->ops->destroy() path which
removes the remaining elements in the set from commit_release, where no
mutex is held.
To deal with GC when mutex is held, which allows safe deactivate and
removal, add sync GC API which releases the set element object via
call_rcu(). This is used by rbtree and pipapo backends which also
perform garbage collection from control plane path.
Since element removal from sets can happen from control plane and
element garbage collection/timeout, it is necessary to keep the set
structure alive until all elements have been deactivated and destroyed.
We cannot do a cancel_work_sync or flush_work in nft_set_destroy because
its called with the transaction mutex held, but the aforementioned async
work queue might be blocked on the very mutex that nft_set_destroy()
callchain is sitting on.
This gives us the choice of ABBA deadlock or UaF.
To avoid both, add set->refs refcount_t member. The GC API can then
increment the set refcount and release it once the elements have been
free'd.
Set backends are adapted to use the GC transaction API in a follow up
patch entitled:
("netfilter: nf_tables: use gc transaction API in set backends")
This is joint work with Florian Westphal.
Fixes: cfed7e1b1f8e ("netfilter: nf_tables: add set garbage collection helpers") Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
There is an asymmetry between commit/abort and preparation phase if the
following conditions are met:
1. set is a verdict map ("1.2.3.4 : jump foo")
2. timeouts are enabled
In this case, following sequence is problematic:
1. element E in set S refers to chain C
2. userspace requests removal of set S
3. kernel does a set walk to decrement chain->use count for all elements
from preparation phase
4. kernel does another set walk to remove elements from the commit phase
(or another walk to do a chain->use increment for all elements from
abort phase)
If E has already expired in 1), it will be ignored during list walk, so its use count
won't have been changed.
Then, when set is culled, ->destroy callback will zap the element via
nf_tables_set_elem_destroy(), but this function is only safe for
elements that have been deactivated earlier from the preparation phase:
lack of earlier deactivate removes the element but leaks the chain use
count, which results in a WARN splat when the chain gets removed later,
plus a leak of the nft_chain structure.
Update pipapo_get() not to skip expired elements, otherwise flush
command reports bogus ENOENT errors.
Fixes: 3c4287f62044 ("nf_tables: Add set type for arbitrary concatenation of ranges") Fixes: 8d8540c4f5e0 ("netfilter: nft_set_rbtree: add timeout support") Fixes: 9d0982927e79 ("netfilter: nft_hash: add support for timeouts") Signed-off-by: Florian Westphal <fw@strlen.de> Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
The event inject files add events for a specific trace array. For an
instance, if the file is opened and the instance is deleted, reading or
writing to the file will cause a use after free.
Up the ref count of the trace_array when a event inject file is opened.
Len Brown has reported that system suspend sometimes fail due to
inability to freeze a task working in ext4_trim_fs() for one minute.
Trimming a large filesystem on a disk that slowly processes discard
requests can indeed take a long time. Since discard is just an advisory
call, it is perfectly fine to interrupt it at any time and the return
number of discarded blocks until that moment. Do that when we detect the
task is being frozen.
Cc: stable@kernel.org Reported-by: Len Brown <lenb@kernel.org> Suggested-by: Dave Chinner <david@fromorbit.com>
References: https://bugzilla.kernel.org/show_bug.cgi?id=216322 Signed-off-by: Jan Kara <jack@suse.cz> Link: https://lore.kernel.org/r/20230913150504.9054-2-jack@suse.cz Signed-off-by: Theodore Ts'o <tytso@mit.edu> Signed-off-by: Sasha Levin <sashal@kernel.org>
Currently we set the group's trimmed bit in ext4_trim_all_free() based
on return value of ext4_try_to_trim_range(). However when we will want
to abort trimming because of suspend attempt, we want to return success
from ext4_try_to_trim_range() but not set the trimmed bit. Instead
implementing awkward propagation of this information, just move setting
of trimmed bit into ext4_try_to_trim_range() when the whole group is
trimmed.
There is no good reason for the s_last_trim_minblks to be atomic. There is
no data integrity needed and there is no real danger in setting and
reading it in a racy manner. Change it to be unsigned long, the same type
as s_clusters_per_group which is the maximum that's allowed.
Signed-off-by: Lukas Czerner <lczerner@redhat.com> Suggested-by: Andreas Dilger <adilger@dilger.ca> Reviewed-by: Andreas Dilger <adilger@dilger.ca> Link: https://lore.kernel.org/r/20211103145122.17338-1-lczerner@redhat.com Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Stable-dep-of: 45e4ab320c9b ("ext4: move setting of trimmed bit into ext4_try_to_trim_range()") Signed-off-by: Sasha Levin <sashal@kernel.org>
As commit 6920b3913235 ("ext4: add new helper interface
ext4_try_to_trim_range()") moves some code into the separate function
ext4_try_to_trim_range(), the use of the variable ret within that
function is more limited and can be adjusted as well.
Scope the use of the variable ret locally and drop dead assignments.
When a CRC error occurs, the HBA asserts an interrupt to indicate an
interface fatal error (PxIS.IFS). The ISR clears PxIE and PxIS, then
does error recovery. If the adapter receives another SDB FIS
with an error (PxIS.TFES) from the device before the start of the EH
recovery process, the interrupt signaling the new SDB cannot be
serviced as PxIE was cleared already. This in turn results in the HBA
inability to issue any command during the error recovery process after
setting PxCMD.ST to 1 because PxIS.TFES is still set.
According to AHCI 1.3.1 specifications section 6.2.2, fatal errors
notified by setting PxIS.HBFS, PxIS.HBDS, PxIS.IFS or PxIS.TFES will
cause the HBA to enter the ERR:Fatal state. In this state, the HBA
shall not issue any new commands.
To avoid this situation, introduce the function
ahci_port_clear_pending_irq() to clear pending interrupts before
executing a COMRESET. This follows the AHCI 1.3.1 - section 6.2.2.2
specification.
Signed-off-by: Szuying Chen <Chloe_Chen@asmedia.com.tw> Fixes: e0bfd149973d ("[PATCH] ahci: stop engine during hard reset") Cc: stable@vger.kernel.org Reviewed-by: Niklas Cassel <niklas.cassel@wdc.com> Signed-off-by: Damien Le Moal <dlemoal@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
When the trace event enable and filter files are opened, increment the
trace array ref counter, otherwise they can be accessed when the trace
array is being deleted. The ref counter keeps the trace array from being
deleted while those files are opened.
The tracing marker files are write-only streams with no meaningful
concept of file position. Using stream_open() to mark them as
stream-link indicates this and has the added advantage that a single
file descriptor can now be used from multiple threads without contention
thanks to clearing FMODE_ATOMIC_POS.
Note that this has the potential to break existing userspace by since
both lseek(2) and pwrite(2) will now return ESPIPE when previously lseek
would have updated the stored offset and pwrite would have appended to
the trace. A survey of libtracefs and several other projects found to
use trace_marker(_raw) [1][2][3] suggests that everyone limits
themselves to calling write(2) and close(2) on these file descriptors so
there is a good chance this will go unnoticed and the benefits of
reduced overhead and lock contention seem worth the risk.
Currently, when GETDEVICEINFO returns multiple locations where each
is a different IP but the server's identity is same as MDS, then
nfs4_set_ds_client() finds the existing nfs_client structure which
has the MDS's max_connect value (and if it's 1), then the 1st IP
on the DS's list will get dropped due to MDS trunking rules. Other
IPs would be added as they fall under the pnfs trunking rules.
For the list of IPs the 1st goes thru calling nfs4_set_ds_client()
which will eventually call nfs4_add_trunk() and call into
rpc_clnt_test_and_add_xprt() which has the check for MDS trunking.
The other IPs (after the 1st one), would call rpc_clnt_add_xprt()
which doesn't go thru that check.
nfs4_add_trunk() is called when MDS trunking is happening and it
needs to enforce the usage of max_connect mount option of the
1st mount. However, this shouldn't be applied to pnfs flow.
Instead, this patch proposed to treat MDS=DS as DS trunking and
make sure that MDS's max_connect limit does not apply to the
1st IP returned in the GETDEVICEINFO list. It does so by
marking the newly created client with a new flag NFS_CS_PNFS
which then used to pass max_connect value to use into the
rpc_clnt_test_and_add_xprt() instead of the existing rpc
client's max_connect value set by the MDS connection.
For example, mount was done without max_connect value set
so MDS's rpc client has cl_max_connect=1. Upon calling into
rpc_clnt_test_and_add_xprt() and using rpc client's value,
the caller passes in max_connect value which is previously
been set in the pnfs path (as a part of handling
GETDEVICEINFO list of IPs) in nfs4_set_ds_client().
However, when NFS_CS_PNFS flag is not set and we know we
are doing MDS trunking, comparing a new IP of the same
server, we then set the max_connect value to the
existing MDS's value and pass that into
rpc_clnt_test_and_add_xprt().
Fixes: dc48e0abee24 ("SUNRPC enforce creation of no more than max_connect xprts") Signed-off-by: Olga Kornievskaia <kolga@netapp.com> Signed-off-by: Anna Schumaker <Anna.Schumaker@Netapp.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
After receiving the location(s) of the DS server(s) in the
GETDEVINCEINFO, create the request for the clientid to such
server and indicate that the client is connecting to a DS.
Signed-off-by: Olga Kornievskaia <kolga@netapp.com> Signed-off-by: Anna Schumaker <Anna.Schumaker@Netapp.com>
Stable-dep-of: 806a3bc421a1 ("NFSv4.1: fix pnfs MDS=DS session trunking") Signed-off-by: Sasha Levin <sashal@kernel.org>
With IPv6, connect() can occasionally return EINVAL if a route is
unavailable. If this happens during I/O to a data server, we want to
report it using LAYOUTERROR as an inability to connect.
Fixes: dd52128afdde ("NFSv4.1/pnfs Ensure flexfiles reports all connection related errors") Signed-off-by: Trond Myklebust <trond.myklebust@hammerspace.com> Signed-off-by: Anna Schumaker <Anna.Schumaker@Netapp.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
Ensure that nfs_clear_request_commit() updates the correct counters when
it removes them from the commit list.
Fixes: ed5d588fe47f ("NFS: Try to join page groups before an O_DIRECT retransmission") Signed-off-by: Trond Myklebust <trond.myklebust@hammerspace.com> Signed-off-by: Anna Schumaker <Anna.Schumaker@Netapp.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
For DRM legacy gamma, AMD display manager applies implicit sRGB degamma
using a pre-defined sRGB transfer function. It works fine for DCN2
family where degamma ROM and custom curves go to the same color block.
But, on DCN3+, degamma is split into two blocks: degamma ROM for
pre-defined TFs and `gamma correction` for user/custom curves and
degamma ROM settings doesn't apply to cursor plane. To get DRM legacy
gamma working as expected, enable cursor degamma ROM for implict sRGB
degamma on HW with this configuration.
Cc: stable@vger.kernel.org Link: https://gitlab.freedesktop.org/drm/amd/-/issues/2803 Fixes: 96b020e2163f ("drm/amd/display: check attr flag before set cursor degamma on DCN3+") Signed-off-by: Melissa Wen <mwen@igalia.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The rsvp classifier has served us well for about a quarter of a century but has
has not been getting much maintenance attention due to lack of known users.
The offset is just 32bits here so this can potentially overflow if
somebody specifies a large value. Instead reduce the size to calculate
the last possible offset.
The error handling path incorrectly drops the reference to the user
fence BO resulting in potential reference count underflow.
Signed-off-by: Christian König <christian.koenig@amd.com> Reviewed-by: Alex Deucher <alexander.deucher@amd.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com> Cc: stable@vger.kernel.org Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Dropping bit 31:4 of page table base is wrong, it makes page table
base points to wrong address if phys addr is beyond 64GB; dropping
page_table_start/end bit 31:4 is unnecessary since dcn20_vmid_setup
will do that. Also, while we are at it, cleanup the assignments using
upper_32_bits()/lower_32_bits() and AMDGPU_GPU_PAGE_SHIFT.
Cc: stable@vger.kernel.org Link: https://gitlab.freedesktop.org/drm/amd/-/issues/2354 Fixes: 81d0bcf99009 ("drm/amdgpu: make display pinning more flexible (v2)") Acked-by: Harry Wentland <harry.wentland@amd.com> Reviewed-by: Alex Deucher <alexander.deucher@amd.com> Signed-off-by: Yifan Zhang <yifan1.zhang@amd.com> Co-developed-by: Hamza Mahfooz <hamza.mahfooz@amd.com> Signed-off-by: Hamza Mahfooz <hamza.mahfooz@amd.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
With the configuration PAGE_SIZE 64k and filesystem blocksize 64k,
a problem occurred when more than 13 million files were directly created
under a directory:
EXT4-fs error (device xx): ext4_dx_csum_set:492: inode #xxxx: comm xxxxx: dir seems corrupt? Run e2fsck -D.
EXT4-fs error (device xx): ext4_dx_csum_verify:463: inode #xxxx: comm xxxxx: dir seems corrupt? Run e2fsck -D.
EXT4-fs error (device xx): dx_probe:856: inode #xxxx: block 8188: comm xxxxx: Directory index failed checksum
When enough files are created, the fake_dirent->reclen will be 0xffff.
it doesn't equal to the blocksize 65536, i.e. 0x10000.
But it is not the same condition when blocksize equals to 4k.
when enough files are created, the fake_dirent->reclen will be 0x1000.
it equals to the blocksize 4k, i.e. 0x1000.
The problem seems to be related to the limitation of the 16-bit field
when the blocksize is set to 64k.
To address this, helpers like ext4_rec_len_{from,to}_disk has already
been introduced to complete the conversion between the encoded and the
plain form of rec_len.
So fix this one by using the helper, and all the other in this file too.
Cc: stable@kernel.org Fixes: dbe89444042a ("ext4: Calculate and verify checksums for htree nodes") Suggested-by: Andreas Dilger <adilger@dilger.ca> Suggested-by: Darrick J. Wong <djwong@kernel.org> Signed-off-by: Shida Zhang <zhangshida@kylinos.cn> Reviewed-by: Andreas Dilger <adilger@dilger.ca> Reviewed-by: Darrick J. Wong <djwong@kernel.org> Link: https://lore.kernel.org/r/20230803060938.1929759-1-zhangshida@kylinos.cn Signed-off-by: Theodore Ts'o <tytso@mit.edu> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The function pm8001_pci_resume() only calls pm8001_request_irq() without
calling pm8001_setup_irq(). This causes the IRQ allocation to fail, which
leads all drives being removed from the system.
Fix this issue by integrating the code for pm8001_setup_irq() directly
inside pm8001_request_irq() so that MSI-X setup is performed both during
normal initialization and resume operations.
Fixes: dbf9bfe61571 ("[SCSI] pm8001: add SAS/SATA HBA driver") Cc: stable@vger.kernel.org Signed-off-by: Damien Le Moal <dlemoal@kernel.org> Link: https://lore.kernel.org/r/20230911232745.325149-2-dlemoal@kernel.org Acked-by: Jack Wang <jinpu.wang@ionos.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The following processes run into a deadlock. CPU 41 was waiting for CPU 29
to handle a CSD request while holding spinlock "crashdump_lock", but CPU 29
was hung by that spinlock with IRQs disabled.
The lock is used to synchronize different sysfs operations, it doesn't
protect any resource that will be touched by an interrupt. Consequently
it's not required to disable IRQs. Replace the spinlock with a mutex to fix
the deadlock.
Signed-off-by: Junxiao Bi <junxiao.bi@oracle.com> Link: https://lore.kernel.org/r/20230828221018.19471-1-junxiao.bi@oracle.com Reviewed-by: Mike Christie <michael.christie@oracle.com> Cc: stable@vger.kernel.org Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
In AHCI 1.3.1, the register description for CAP.SSC:
"When cleared to ‘0’, software must not allow the HBA to initiate
transitions to the Slumber state via agressive link power management nor
the PxCMD.ICC field in each port, and the PxSCTL.IPM field in each port
must be programmed to disallow device initiated Slumber requests."
In AHCI 1.3.1, the register description for CAP.PSC:
"When cleared to ‘0’, software must not allow the HBA to initiate
transitions to the Partial state via agressive link power management nor
the PxCMD.ICC field in each port, and the PxSCTL.IPM field in each port
must be programmed to disallow device initiated Partial requests."
Ensure that we always set the corresponding bits in PxSCTL.IPM, such that
a device is not allowed to initiate transitions to power states which are
unsupported by the HBA.
DevSleep is always initiated by the HBA, however, for completeness, set the
corresponding bit in PxSCTL.IPM such that agressive link power management
cannot transition to DevSleep if DevSleep is not supported.
sata_link_scr_lpm() is used by libahci, ata_piix and libata-pmp.
However, only libahci has the ability to read the CAP/CAP2 register to see
if these features are supported. Therefore, in order to not introduce any
regressions on ata_piix or libata-pmp, create flags that indicate that the
respective feature is NOT supported. This way, the behavior for ata_piix
and libata-pmp should remain unchanged.
This change is based on a patch originally submitted by Runa Guo-oc.
Signed-off-by: Niklas Cassel <niklas.cassel@wdc.com> Fixes: 1152b2617a6e ("libata: implement sata_link_scr_lpm() and make ata_dev_set_feature() global") Cc: stable@vger.kernel.org Signed-off-by: Damien Le Moal <dlemoal@kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Reset the i2c controller when an i2c transfer timeout occurs.
The remaining interrupts and device should be reset to avoid
unpredictable controller behavior.
Fixes: 2e57b7cebb98 ("i2c: aspeed: Add multi-master use case support") Cc: <stable@vger.kernel.org> # v5.1+ Signed-off-by: Tommy Huang <tommy_huang@aspeedtech.com> Reviewed-by: Andi Shyti <andi.shyti@kernel.org> Signed-off-by: Wolfram Sang <wsa@kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The function tracefs_create_dir() was missing a lockdown check and was
called by the RV code. This gave an inconsistent behavior of this function
returning success while other tracefs functions failed. This caused the
inode being freed by the wrong kmem_cache.
Link: https://lkml.kernel.org/r/20230905182711.692687042@goodmis.org Link: https://lore.kernel.org/all/202309050916.58201dc6-oliver.sang@intel.com/ Cc: stable@vger.kernel.org Cc: Masami Hiramatsu <mhiramat@kernel.org> Cc: Mark Rutland <mark.rutland@arm.com> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Ajay Kaher <akaher@vmware.com> Cc: Ching-lin Yu <chinglinyu@google.com> Fixes: bf8e602186ec4 ("tracing: Do not create tracefs files if tracefs lockdown is in effect") Reported-by: kernel test robot <oliver.sang@intel.com> Signed-off-by: Steven Rostedt (Google) <rostedt@goodmis.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The option files update the options for a given trace array. For an
instance, if the file is opened and the instance is deleted, reading or
writing to the file will cause a use after free.
Up the ref count of the trace_array when an option file is opened.
The current_trace updates the trace array tracer. For an instance, if the
file is opened and the instance is deleted, reading or writing to the file
will cause a use after free.
Up the ref count of the trace array when current_trace is opened.
The tracing_max_latency file points to the trace_array max_latency field.
For an instance, if the file is opened and the instance is deleted,
reading or writing to the file will cause a use after free.
Up the ref count of the trace_array when tracing_max_latency is opened.
During the ino lookup ioctl we can end up calling btrfs_iget() to get an
inode reference while we are holding on a root's btree. If btrfs_iget()
needs to lookup the inode from the root's btree, because it's not
currently loaded in memory, then it will need to lock another or the
same path in the same root btree. This may result in a deadlock and
trigger the following lockdep splat:
WARNING: possible circular locking dependency detected 6.5.0-rc7-syzkaller-00004-gf7757129e3de #0 Not tainted
------------------------------------------------------
syz-executor277/5012 is trying to acquire lock: ffff88802df41710 (btrfs-tree-01){++++}-{3:3}, at: __btrfs_tree_read_lock+0x2f/0x220 fs/btrfs/locking.c:136
but task is already holding lock: ffff88802df418e8 (btrfs-tree-00){++++}-{3:3}, at: __btrfs_tree_read_lock+0x2f/0x220 fs/btrfs/locking.c:136
which lock already depends on the new lock.
the existing dependency chain (in reverse order) is:
When running delayed items we are holding a delayed node's mutex and then
we will attempt to modify a subvolume btree to insert/update/delete the
delayed items. However if have an error during the insertions for example,
btrfs_insert_delayed_items() may return with a path that has locked extent
buffers (a leaf at the very least), and then we attempt to release the
delayed node at __btrfs_run_delayed_items(), which requires taking the
delayed node's mutex, causing an ABBA type of deadlock. This was reported
by syzbot and the lockdep splat is the following:
WARNING: possible circular locking dependency detected 6.5.0-rc7-syzkaller-00024-g93f5de5f648d #0 Not tainted
------------------------------------------------------
syz-executor.2/13257 is trying to acquire lock: ffff88801835c0c0 (&delayed_node->mutex){+.+.}-{3:3}, at: __btrfs_release_delayed_node+0x9a/0xaa0 fs/btrfs/delayed-inode.c:256
but task is already holding lock: ffff88802a5ab8e8 (btrfs-tree-00){++++}-{3:3}, at: __btrfs_tree_lock+0x3c/0x2a0 fs/btrfs/locking.c:198
which lock already depends on the new lock.
the existing dependency chain (in reverse order) is:
ovl_{read,write}_iter() always call fdput(real) to put one or zero
refcounts of the real file, but for aio, whether it was submitted or not,
ovl_aio_put() also calls fdput(), which is not balanced. This is only a
problem in the less common case when FDPUT_FPUT flag is set.
To fix the problem use get_file() to take file refcount and use fput()
instead of fdput() in ovl_aio_put().