Clang recently added support for -Wattribute-alias [1], which results in
the same warnings that necessitated commit bee20031772a ("disable
-Wattribute-alias warning for SYSCALL_DEFINEx()") for GCC.
kernel/time/itimer.c:325:1: error: alias and aliasee have different types 'long (unsigned int)' and 'long (typeof (__builtin_choose_expr((__builtin_types_compatible_p(typeof ((unsigned int)0), typeof (0LL)) || __builtin_types_compatible_p(typeof ((unsigned int)0), typeof (0ULL))), 0LL, 0L)))' (aka 'long (long)') [-Werror,-Wattribute-alias]
325 | SYSCALL_DEFINE1(alarm, unsigned int, seconds)
| ^
include/linux/syscalls.h:225:36: note: expanded from macro 'SYSCALL_DEFINE1'
225 | #define SYSCALL_DEFINE1(name, ...) SYSCALL_DEFINEx(1, _##name, __VA_ARGS__)
| ^
include/linux/syscalls.h:236:2: note: expanded from macro 'SYSCALL_DEFINEx'
236 | __SYSCALL_DEFINEx(x, sname, __VA_ARGS__)
| ^
include/linux/syscalls.h:251:18: note: expanded from macro '__SYSCALL_DEFINEx'
251 | __attribute__((alias(__stringify(__se_sys##name)))); \
| ^
kernel/time/itimer.c:325:1: note: aliasee is declared here
include/linux/syscalls.h:225:36: note: expanded from macro 'SYSCALL_DEFINE1'
225 | #define SYSCALL_DEFINE1(name, ...) SYSCALL_DEFINEx(1, _##name, __VA_ARGS__)
| ^
include/linux/syscalls.h:236:2: note: expanded from macro 'SYSCALL_DEFINEx'
236 | __SYSCALL_DEFINEx(x, sname, __VA_ARGS__)
| ^
include/linux/syscalls.h:255:18: note: expanded from macro '__SYSCALL_DEFINEx'
255 | asmlinkage long __se_sys##name(__MAP(x,__SC_LONG,__VA_ARGS__)) \
| ^
<scratch space>:16:1: note: expanded from here
16 | __se_sys_alarm
| ^
Disable the warnings in the same way for clang-23 and newer. Disable the
warning about unknown warning options to avoid breaking the build for
versions of clang-23 that do not have -Wattribute-alias, such as ones
deployed by vendors like Android or CI systems or when bisecting LLVM
between llvmorg-23-init and release/23.x.
Arnd Bergmann [Thu, 28 May 2026 13:31:47 +0000 (15:31 +0200)]
Merge tag 'qcomtee-fix-for-v7.1' of git://git.kernel.org/pub/scm/linux/kernel/git/jenswi/linux-tee into arm/fixes
QCOMTEE fix for v7.1
Adding a missing va_end in early return qcomtee_object_user_init()
* tag 'qcomtee-fix-for-v7.1' of git://git.kernel.org/pub/scm/linux/kernel/git/jenswi/linux-tee:
tee: qcomtee: add missing va_end in early return qcomtee_object_user_init()
Arnd Bergmann [Thu, 28 May 2026 13:28:19 +0000 (15:28 +0200)]
Merge tag 'optee-fix-for-v7.1' of git://git.kernel.org/pub/scm/linux/kernel/git/jenswi/linux-tee into arm/fixes
OP-TEE fix for v7.1
Prevent possible use after free in supplicant communication.
* tag 'optee-fix-for-v7.1' of git://git.kernel.org/pub/scm/linux/kernel/git/jenswi/linux-tee:
tee: optee: prevent use-after-free when the client exits before the supplicant
Marco Scardovi [Tue, 26 May 2026 17:02:46 +0000 (19:02 +0200)]
gpio: rockchip: teardown bugs and resource leaks
Address several teardown issues and resource leaks in the driver's remove
path and error handling:
1. Debounce clock reference leak: The debounce clock (bank->db_clk) is
obtained using of_clk_get() which increments the clock's reference
count, but clk_put() is never called. Register a devm action to
cleanly release it on unbind. Note that of_clk_get(..., 1) remains
necessary over devm_clk_get() because the DT binding does not define
clock-names, precluding name-based lookup.
2. Unregistered chained IRQ handler: The chained IRQ handler is not
disconnected in remove(). If a stray interrupt fires after the driver
is removed, the kernel attempts to execute a stale handler, leading
to a panic. Fix this by clearing the handler in remove().
3. IRQ domain leak: The linear IRQ domain and its generic chips are
allocated manually during probe but never removed. Remove the IRQ
domain during driver teardown to free the associated generic chips
and mappings.
Fixes: 936ee2675eee ("gpio/rockchip: add driver for rockchip gpio") Assisted-by: Antigravity:gemini-3.5-flash Signed-off-by: Marco Scardovi <scardracs@disroot.org> Link: https://patch.msgid.link/20260526171050.12785-3-scardracs@disroot.org
[Bartosz: don't emit an error message on devres allocation failure] Signed-off-by: Bartosz Golaszewski <bartosz.golaszewski@oss.qualcomm.com>
Marco Scardovi [Tue, 26 May 2026 17:02:45 +0000 (19:02 +0200)]
gpio: rockchip: convert bank->clk to devm_clk_get_enabled()
The bank->clk was previously obtained via of_clk_get() and manually
prepared/enabled. However, it was missing a corresponding clk_put() in
both the error paths and the remove function, leading to a reference leak.
Convert the allocation to devm_clk_get_enabled(), which also properly
propagates failures from clk_prepare_enable() that were previously ignored.
The GPIO bank device uses the same OF node as the previous of_clk_get()
call, so devm_clk_get_enabled(dev, NULL) correctly resolves the same
clock provider entry.
Fix the reference leak and simplify the code by removing the manual
clk_disable_unprepare() calls in the probe error paths and in the
remove function.
Fixes: 936ee2675eee ("gpio/rockchip: add driver for rockchip gpio") Assisted-by: Antigravity:gemini-3.5-flash Signed-off-by: Marco Scardovi <scardracs@disroot.org> Link: https://patch.msgid.link/20260526171050.12785-2-scardracs@disroot.org Signed-off-by: Bartosz Golaszewski <bartosz.golaszewski@oss.qualcomm.com>
Dan Carpenter [Mon, 25 May 2026 07:15:16 +0000 (10:15 +0300)]
gpio: virtuser: Fix uninitialized data bug in gpio_virtuser_direction_do_write()
If *ppos is non-zero (user-space write split over multiple calls to
write()) then simple_write_to_buffer() won't initialize the start of the
buffer. Really, non-zero values for *ppos aren't going to work at all.
Check for that and return -EINVAL at the start of the function.
Fixes: 91581c4b3f29 ("gpio: virtuser: new virtual testing driver for the GPIO API") Signed-off-by: Dan Carpenter <error27@gmail.com> Link: https://patch.msgid.link/ahP3BJWWy-m_qI0X@stanley.mountain Signed-off-by: Bartosz Golaszewski <bartosz.golaszewski@oss.qualcomm.com>
gpio: shared: fix lockdep false positive by removing unneeded lock
By the time gpio_device_teardown_shared() is called, the parent device
is gone from the global list of GPIO devices and all outstanding SRCU
read-side critical sections have completed. That means that no
concurrent gpio_find_and_request() can call
gpio_shared_add_proxy_lookup() for this device at this time. There's
also no risk of the parent device being re-bound to the driver before
the unbinding completes (including the child devices).
Lockdep produces a false-positive report about a possible circular
dependency as it doesn't know the ordering guarantee. Not taking the
ref->lock in gpio_device_teardown_shared() silences it and is safe to do.
gpio: shared: fix deadlock on shared proxy's parent removal
Commit 710abda58055 ("gpio: shared: call gpio_chip::of_xlate() if set")
used the mutex embedded in struct gpio_shared_entry to protect the
offset field which now can be modified after assignment. The critical
section however is too wide and introduced a potential deadlock on the
removal of the shared GPIO proxy's parent.
Make the critical section shorter - only protect the offset when it's
being read.
While at it: mention the fact that the entry lock is now also used to
protect against concurrent access to the offset field in the structure's
documentation.
gpio: adnp: fix flow control regression caused by scoped_guard()
scoped_guard() is implemented as a for loop. Using it to protect code
using the continue statement changes the flow as we now only break out
of the hidden loop inside scoped_guard(), not the original for loop. Use
a regular code block instead.
Fixes: c7fe19ed3973 ("gpio: adnp: use lock guards for the I2C lock") Reported-by: David Lechner <dlechner@baylibre.com> Closes: https://lore.kernel.org/all/cde2abb2-4cc8-4fc9-b34a-0c5d2b95779f@baylibre.com/ Reviewed-by: Linus Walleij <linusw@kernel.org> Link: https://patch.msgid.link/20260522073527.9812-1-bartosz.golaszewski@oss.qualcomm.com Signed-off-by: Bartosz Golaszewski <bartosz.golaszewski@oss.qualcomm.com>
gpio: shared: undo the vote of the proxy on GPIO free
When the user of a shared GPIO managed by gpio-shared-proxy calls
gpiod_put() to release it, we never undo the potential "vote" for
driving the shared line "high". In the free() callback, check if this
proxy voted for "high" and - if so - decrease the number of votes and
potentially revert the value to low if this is the last user.
Arnd Bergmann [Thu, 28 May 2026 13:20:29 +0000 (15:20 +0200)]
Merge tag 'tee-fixes-for-v7.1' of git://git.kernel.org/pub/scm/linux/kernel/git/jenswi/linux-tee into arm/fixes
TEE fixes for v7.1
Fixing:
- params_from_user() cleanup in error path in tee_ioctl_supp_recv()
- possible tee_shm leak in error path in register_shm_helper()
- padding in struct tee_ioctl_object_invoke_arg
* tag 'tee-fixes-for-v7.1' of git://git.kernel.org/pub/scm/linux/kernel/git/jenswi/linux-tee:
tee: fix params_from_user() error path in tee_ioctl_supp_recv
tee: shm: fix shm leak in register_shm_helper()
tee: fix tee_ioctl_object_invoke_arg padding
Li RongQing [Thu, 28 May 2026 12:40:43 +0000 (08:40 -0400)]
KVM: x86: Use fls() instead of ffs() for rmaps histogram bucketing
The kvm_mmu_rmaps_stat_show() function implements a logarithmic histogram
to collect statistics on the distribution of pte_list_count sizes. However,
it currently uses ffs() (Find First Set), which looks for the least
significant set bit.
Using ffs() leads to severe statistical distortion for any count that is
not a power of two. For example, if a rmap has a pte_list_count of 6
(binary 0110), ffs(6) returns 2, forcing this entry to be erroneously
counted towards the bucket representing a count of 2. In contrast, it
conceptually belongs to the bucket spanning the 4 to 7 range.
Replace ffs() with fls() (Find Last Set) to correctly identify the most
significant set bit (effectively computing floor(log2(count)) + 1). This Ensure
that intermediate counts are correctly categorized into their appropriate
power-of-two order-of-magnitude buckets rather than being severely
under-reported.
Claudiu Beznea [Fri, 22 May 2026 10:57:17 +0000 (13:57 +0300)]
pinctrl: renesas: rzv2m: Use -ENOTSUPP instead of -EOPNOTSUPP
The pinctrl and GPIO core code make exceptions for the -ENOTSUPP error
code. One such example is gpio_set_config_with_argument_optional(), which
returns success when gpio_set_config_with_argument() returns -ENOTSUPP, but
reports failure for all other error codes.
Returning -EOPNOTSUPP from the pinctrl driver on the unsupported pinctrl
operation may lead to boot failures when pinctrl drivers implements
struct gpio_chip::set_config, the system uses GPIO hogs, and the
struct gpio_chip::set_config implementation returns -EOPNOTSUPP for the
unsupported operations.
Currently, the driver does not implement struct gpio_chip::set_config().
To avoid future failures, return -ENOTSUPP from
rzv2m_pinctrl_pinconf_set().
rzv2m_pinctrl_pinconf_group_get() is used when dumping pinctrl
configuration. pinconf_generic_dump_one(), which calls it, makes
exceptions for the -EINVAL and -ENOTSUPP error codes. The documentation
for struct pinconf_ops::pin_config_group_get states that it "should
return -ENOTSUPP and -EINVAL using the same rules as pin_config_get()".
The documentation for struct pinconf_ops::pin_config_get states:
"get the config of a certain pin, if the requested config is not available
on this controller this should return -ENOTSUPP and if it is available but
disabled it should return -EINVAL".
Return -ENOTSUPP for the unsupported pinctrl operation.
Li RongQing [Thu, 28 May 2026 03:15:45 +0000 (23:15 -0400)]
KVM: x86: Fix wrong return value type in guest_cpuid_has()
The function guest_cpuid_has() is declared with a return type of 'bool'.
However, when kvm_find_cpuid_entry_index() fails to locate a valid CPUID
entry, the code incorrectly returns 'NULL' instead of 'false'.
Bluetooth: hci_sync: Set HCI_CMD_DRAIN_WORKQUEUE during device close
Since hci_dev_close_sync() can now be called during the reset path, we
should also set HCI_CMD_DRAIN_WORKQUEUE. This avoids queuing timeouts
while the hdev workqueue is being drained.
Fixes: 877afadad2dc ("Bluetooth: When HCI work queue is drained, only queue chained work") Signed-off-by: Heitor Alves de Siqueira <halves@igalia.com> Signed-off-by: Luiz Augusto von Dentz <luiz.von.dentz@intel.com>
Bluetooth: hci_core: Rework hci_dev_do_reset() to use hci_sync functions
The current HCI reset function in hci_core.c duplicates most of the work
done by hci_dev_close_sync(), and doesn't handle LE, advertising or
discovery.
Instead of porting these to hci_dev_do_reset(), directly call the
close/open functions from hci_sync to reset the hdev. MGMT now notifies
when a user performs a reset.
Suggested-by: Luiz Augusto von Dentz <luiz.von.dentz@intel.com> Signed-off-by: Heitor Alves de Siqueira <halves@igalia.com> Signed-off-by: Luiz Augusto von Dentz <luiz.von.dentz@intel.com>
Muhammad Bilal [Wed, 27 May 2026 04:59:18 +0000 (04:59 +0000)]
Bluetooth: ISO: serialize iso_sock_clear_timer with socket lock
iso_sock_close() calls iso_sock_clear_timer() before acquiring
lock_sock(sk).
iso_sock_clear_timer() reads iso_pi(sk)->conn twice without the
socket lock held:
if (!iso_pi(sk)->conn)
return;
cancel_delayed_work(&iso_pi(sk)->conn->timeout_work);
Concurrently, iso_conn_del() executes under lock_sock(sk) and calls
iso_chan_del(), which sets iso_pi(sk)->conn to NULL and may result in
the final reference to the connection being dropped:
iso_pi(sk)->conn is not stable across the unlock window, causing a
NULL pointer dereference or use-after-free.
Serialize iso_sock_clear_timer() with the socket lock by moving it
inside lock_sock()/release_sock(), matching the pattern used in
iso_conn_del() and all other call sites.
Fixes: ccf74f2390d60a2f9a75ef496d2564abb478f46a ("Bluetooth: Add BTPROTO_ISO socket type") Cc: stable@vger.kernel.org Signed-off-by: Muhammad Bilal <meatuni001@gmail.com> Signed-off-by: Luiz Augusto von Dentz <luiz.von.dentz@intel.com>
Muhammad Bilal [Wed, 27 May 2026 04:59:17 +0000 (04:59 +0000)]
Bluetooth: ISO: fix UAF in iso_recv_frame
iso_recv_frame reads conn->sk under iso_conn_lock but releases the lock
before using sk, with no reference held. A concurrent iso_sock_kill()
can free sk in that window, causing use-after-free on sk->sk_state and
sock_queue_rcv_skb().
Fix by replacing the bare pointer read with iso_sock_hold(conn), which
calls sock_hold() while the spinlock is held, atomically elevating the
refcount before the lock drops. Add a drop_put label so sock_put() is
called on all exit paths where the hold succeeded.
Fixes: ccf74f2390d60a2f9a75ef496d2564abb478f46a ("Bluetooth: Add BTPROTO_ISO socket type") Cc: stable@vger.kernel.org Signed-off-by: Muhammad Bilal <meatuni001@gmail.com> Signed-off-by: Luiz Augusto von Dentz <luiz.von.dentz@intel.com>
Bluetooth: L2CAP: Fix possible crash on l2cap_ecred_conn_rsp
If dcid is received for an already-assigned destination CID the spec
requires that both channels to be discarded, but calling l2cap_chan_del
may invalidate the tmp cursor created by list_for_each_entry_safe and
in fact it is the wrong procedure as the chan->dcid may be assigned
previously it really needs to be disconnected.
Calling l2cap_chan_clone directly may still lead to l2cap_chan_del so
instead schedule l2cap_chan_timeout with delay 0 to close the channel
asynchronously.
Fixes: 15f02b910562 ("Bluetooth: L2CAP: Add initial code for Enhanced Credit Based Mode") Signed-off-by: Luiz Augusto von Dentz <luiz.von.dentz@intel.com>
Zhenghang Xiao [Tue, 26 May 2026 10:51:52 +0000 (18:51 +0800)]
Bluetooth: l2cap: clear chan->ident on ECRED reconfiguration success
l2cap_ecred_reconf_rsp() returns early on success without clearing
chan->ident. Every other L2CAP response handler (l2cap_ecred_conn_rsp,
l2cap_le_connect_rsp, l2cap_config_rsp) clears chan->ident after a
successful transaction to prevent the channel from matching subsequent
responses with the recycled ident value.
A remote attacker that completed a reconfiguration as the peer can
replay a failure response with the stale ident, causing the kernel to
match and destroy the already-established channel via
l2cap_chan_del(chan, ECONNRESET).
Clear chan->ident for all matching channels on success, and harden the
failure path by using l2cap_chan_hold_unless_zero() consistent with
other L2CAP handlers (l2cap_le_command_rej, __l2cap_get_chan_by_ident).
Fixes: 15f02b910562 ("Bluetooth: L2CAP: Add initial code for Enhanced Credit Based Mode") Signed-off-by: Zhenghang Xiao <kipreyyy@gmail.com> Signed-off-by: Luiz Augusto von Dentz <luiz.von.dentz@intel.com>
Santhosh Kumar K [Wed, 27 May 2026 17:37:36 +0000 (23:07 +0530)]
spi: spi-mem: avoid mutating op template in spi_mem_supports_op()
spi_mem_supports_op() accepts a const struct spi_mem_op pointer but
casts away const internally to call spi_mem_adjust_op_freq(). This
mutates the caller's op template, which causes stale max_freq values
when callers reuse persistent templates - subsequent calls won't
re-apply the device frequency cap since spi_mem_adjust_op_freq()
skips non-zero values.
Fix by operating on a stack-local copy instead.
Fixes: a4f8e70d75dd ("spi: spi-mem: add spi_mem_adjust_op_freq() in spi_mem_supports_op()") Cc: Tianyu Xu <xtydtc@gmail.com> Cc: stable@vger.kernel.org Signed-off-by: Santhosh Kumar K <s-k6@ti.com> Reviewed-by: Miquel Raynal <miquel.raynal@bootlin.com> Link: https://patch.msgid.link/20260527173736.2243004-1-s-k6@ti.com Signed-off-by: Mark Brown <broonie@kernel.org>
Mingyu Wang [Sat, 23 May 2026 13:52:10 +0000 (21:52 +0800)]
fs/fcntl: fix SOFTIRQ-unsafe lock order in fasync signaling
A SOFTIRQ-safe to SOFTIRQ-unsafe lock order deadlock can occur in
send_sigio() and send_sigurg() when a process group receives a signal.
When FASYNC is configured for a process group (PIDTYPE_PGID), both
functions use read_lock(&tasklist_lock) to traverse the task list.
However, they are frequently called from softirq context:
- send_sigio() via input_inject_event -> kill_fasync
- send_sigurg() via tcp_check_urg -> sk_send_sigurg (NET_RX_SOFTIRQ)
The deadlock is caused by the rwlock writer fairness mechanism:
1. CPU 0 (process context) holds read_lock(&tasklist_lock) in do_wait().
2. CPU 1 (process context) attempts write_lock(&tasklist_lock) in
fork() or exit() and spins, which blocks all new readers.
3. CPU 0 is interrupted by a softirq (e.g., TCP URG packet reception).
4. The softirq calls send_sigurg() and attempts to acquire
read_lock(&tasklist_lock), deadlocking because CPU 1 is waiting.
Since PID hashing and do_each_pid_task() traversals are already
RCU-protected, the read_lock on tasklist_lock is no longer strictly
required for safe traversal. Fix this by replacing tasklist_lock with
rcu_read_lock(), aligning the process group signaling path with the
single-PID path. This also mitigates a potential remote denial of
service vector via TCP URG packets.
Merge patch series "fs/pipe: reduce pipe->mutex contention by pre-allocating outside the lock"
Breno Leitao <leitao@debian.org> says:
While profiling Meta's caching code[1], I found pipe->mutex contention
on the hot path. anon_pipe_write() currently calls alloc_page() once
per page while holding pipe->mutex. The allocation can sleep doing
direct reclaim and runs memcg charging, which extends the critical
section and stalls any concurrent reader on the same mutex.
This series pre-allocates pages outside pipe->mutex in
anon_pipe_write(): for writes that span more than one full page, up
to PIPE_PREALLOC_MAX (8) pages are allocated via a per-page
alloc_page() loop before the mutex is taken. anon_pipe_get_page()
then drains the prealloc array first, falls back to the per-pipe
tmp_page[] cache, and only enters the allocator under the mutex for
the leftover pages (writes larger than PIPE_PREALLOC_MAX, single-page
writes that skip prealloc, or shortfalls when the prealloc loop
fails). Leftover prealloc pages are recycled into tmp_page[] before
unlock and any remainder is put_page()'d after unlock, keeping the
allocator out of the critical section on both sides.
alloc_pages_bulk_mempolicy() looked tempting but the bulk allocator
refuses __GFP_ACCOUNT under memcg -- it returns at most one page
when memcg_kmem_online() && (gfp & __GFP_ACCOUNT), see commit 8dcb3060d81d ("memcg: page_alloc: skip bulk allocator for
__GFP_ACCOUNT"). A per-page loop keeps memcg accounting and the
task NUMA mempolicy honoured uniformly without open-coding the
charge.
I also vibe-coded a microbenchmark to validate the change. It sweeps
writers x readers over {1,2,5} x {1,5,10} with 64KB writes against a
1 MB pipe and prints throughput + latency percentiles per config.
Measured on arm64 and also on x86 using virtme-ng (16 vCPUs, 64KB
writes, 1 MB pipe). The numbers below were collected on v1
(alloc_pages_bulk()); v2's per-page loop preserves the dominant
"allocation outside the mutex" win and is expected to land in the same
range.
Throughput improves +6% to +28% and average write latency drops 5%
to 22% across every configuration.
== Under memory pressure (--memory-pressure, 6s per config) ==
stress-ng --vm 2 --vm-bytes 50% --vm-keep is forked alongside the
sweep so the alloc_page() calls inside anon_pipe_write() routinely
hit direct reclaim -- exactly the regime the patch targets.
Throughput improves +21% to +48% and average write latency drops
17% to 33% -- a noticeably bigger win than the no-pressure run.
That tracks: when alloc_page() has to dip into reclaim, the cost
of holding pipe->mutex across it is highest, and pulling the
allocation out of the critical section pays the most.
* patches from https://patch.msgid.link/20260524-fix_pipe-v3-0-bb4a75d23a90@debian.org:
selftests/pipe: add pipe_bench microbenchmark
fs/pipe: pre-allocate pages outside pipe->mutex in anon_pipe_write
Breno Leitao [Sun, 24 May 2026 14:44:59 +0000 (07:44 -0700)]
selftests/pipe: add pipe_bench microbenchmark
Add a small selftest that stresses pipe->mutex contention by spawning N
writer threads that hammer a single pipe with multi-page writes, plus M
reader threads that drain. Each writer records its own write() latency
samples into a log2-bucketed histogram; main aggregates and prints
total writes, throughput, average and percentile (p50/p99) latencies,
and the maximum observed latency.
Pass --memory-pressure to fork stress-ng (--vm 4 --vm-bytes 80%
--vm-method all) for the duration of the run, so alloc_page() in
anon_pipe_write() routinely hits direct reclaim. The flag fails
fast if stress-ng is not on $PATH.
Program print something like the following, for different writes,
readers, msgsizes and memory pressure:
Breno Leitao [Sun, 24 May 2026 14:44:58 +0000 (07:44 -0700)]
fs/pipe: pre-allocate pages outside pipe->mutex in anon_pipe_write
anon_pipe_write() takes pipe->mutex (aka "mutex protecting the whole
thing") and then, from the per-iteration anon_pipe_get_page() helper,
used to call alloc_page(GFP_HIGHUSER | __GFP_ACCOUNT) once per page
while still holding it.
That allocation can sleep doing direct reclaim and/or runs memcg
charging, which extends the critical section and stalls a concurrent
reader on the very same mutex.
Just pre-alloc the required pages before the lock in an array and just pop
them inside the lock.
This can improve the pipe throughput up to 48% and reduce the
latency in 33%, easily seen when there is memory pressure and direct
reclaim.
Mateusz Guzik [Fri, 22 May 2026 14:21:52 +0000 (16:21 +0200)]
fs: retire stale comment in fget_task_next()
The routine originally showed up in e9a53aeb5e0a838f ("file: Implement
task_lookup_next_fd_rcu"), afterwards it got renamed and started
entering RCU on its own in 8fd3395ec9051a52 ("get rid of
...lookup...fdget_rcu() family").
fs/qnx6: fix pointer arithmetic in directory iteration
The conversion to qnx6_get_folio() in commit b2aa61556fcf
("qnx6: Convert qnx6_get_page() to qnx6_get_folio()")
introduced a regression in directory iteration. The pointer 'de'
and the 'limit' address were calculated using byte offsets from
a char pointer without scaling by the size of a QNX6 directory
entry.
This causes the driver to read from incorrect memory offsets,
leading to "invalid direntry size" errors and premature
termination of directory scans.
Fix this by casting 'kaddr' to 'struct qnx6_dir_entry *' before
applying the offset and last_entry(...) increments. This allows the
compiler to correctly scale the pointer arithmetic by the 32-byte
stride of the directory entry structure.
Fixes: b2aa61556fcf ("qnx6: Convert qnx6_get_page() to qnx6_get_folio()") Cc: stable@vger.kernel.org Signed-off-by: Arpith Kalaginanavoor <arpithk@nvidia.com> Link: https://patch.msgid.link/20260526123858.1683035-1-arpithk@nvidia.com Signed-off-by: Christian Brauner (Amutable) <brauner@kernel.org>
NeilBrown [Mon, 25 May 2026 06:23:45 +0000 (16:23 +1000)]
VFS: fix possible failure to unlock in nfsd4_create_file()
atomic_create() in fs/namei.c drops the reference to the dentry
when it returns an error.
This behaviour was imported into dentry_create() so that it
will drop the reference if an error is returned from atomic_create(),
though not if vfs_create() returns an error (in the case where
->atomic_create is not supported).
The caller - nfsd4_create_file() - is made aware of this by checking
path->dentry, which will either be a counted reference to a dentry, or
an error pointer.
However the change to use start_creating()/end_creating() (which landed
shortly before the dentry_create() change landed, though was likely
developed around the same time) means that nfsd4_create_file() *needs* a
valid dentry so that it can unlock the parent.
The net result is that if NFSD exports a filesystem which uses
->atomic_create, and if a call to ->atomic_create returns an error, then
nfsd4_create_file() will pass an error pointer to end_creating()
and the parent will not be unlocked.
Fix this by changing dentry_create() to make sure path->dentry is always
a valid dentry, never an error-pointer. The actual error is already
returned a different way.
Note that if ->atomic_create() returns a different dentry (which may not
be possible in practice) we are guaranteed (because it is only ever
provided by d_spliace_alias()) that it will have the same d_parent and
so it will have the same effect when passed to end_creating().
Fixes: 64a989dbd144 ("VFS/knfsd: Teach dentry_create() to use atomic_open()") Signed-off-by: NeilBrown <neil@brown.name> Link: https://patch.msgid.link/177969022571.3379282.16448744624428323496@noble.neil.brown.name Reviewed-by: Jeff Layton <jlayton@kernel.org> Reviewed-by: Benjamin Coddington <bcodding@hammerspace.com> Reviewed-by: Jori Koolstra <jkoolstra@xs4all.nl> Signed-off-by: Christian Brauner (Amutable) <brauner@kernel.org>
Qingshuang Fu [Wed, 27 May 2026 10:00:24 +0000 (18:00 +0800)]
fs: fix spelling mistakes in comment
Fix three spelling errors in the comment for an internal file structure
allocation function:
- happend → happened
- over → exceed (grammatical fix)
- int → in
Changes since v1:
- Fix comma after e.g.
- Fix incorrect use of "imbalance"
Paolo Abeni [Thu, 28 May 2026 12:05:31 +0000 (14:05 +0200)]
Merge branch 'dpll-zl3073x-various-fixes'
Ivan Vecera says:
====================
dpll: zl3073x: various fixes
Three fixes for the zl3073x DPLL driver.
Patch 1 exports __dpll_device_change_ntf() for use by drivers that
need to send device change notifications from within callbacks
already running under dpll_lock.
Patch 2 replaces the change_work workqueue mechanism with direct
calls to __dpll_device_change_ntf(), eliminating a race condition
where the work handler could dereference a freed dpll_dev pointer
during device teardown.
Patch 3 moves the freq_monitor flag from per-DPLL to per-device
scope to match the hardware behavior where frequency measurement
registers are shared across all DPLL channels.
====================
Ivan Vecera [Tue, 26 May 2026 07:45:25 +0000 (09:45 +0200)]
dpll: zl3073x: make frequency monitor a per-device attribute
The frequency monitoring feature uses shared hardware registers
that measure input reference frequencies independently of
individual DPLL channels. However, the freq_monitor flag was
incorrectly placed in the per-DPLL structure, causing each
channel to track its own enable/disable state independently.
Since the DPLL core calls measured_freq_get() only for the first
pin registration, the measured_freq_check() in the periodic worker
was gated by the per-DPLL freq_monitor flag of whichever channel
happens to be checked. If the first DPLL channel had frequency
monitoring disabled while another had it enabled, measurements
were never reported.
Move freq_monitor from struct zl3073x_dpll to struct zl3073x_dev
so all DPLL channels share a single flag, matching the hardware
behavior. Update freq_monitor_set() to notify other DPLL devices
about the change (like phase_offset_avg_factor_set() already does)
and remove the mode-dependent guard in zl3073x_dpll_changes_check()
since all input pin monitoring (pin state, phase offset, FFO, and
measured frequency) works correctly in all DPLL modes.
Ivan Vecera [Tue, 26 May 2026 07:45:24 +0000 (09:45 +0200)]
dpll: zl3073x: use __dpll_device_change_ntf() and remove change_work
The change_work was introduced to send device change notifications
from DPLL device callbacks without deadlocking on dpll_lock, since
the callbacks are already invoked under that lock. Now that
__dpll_device_change_ntf() is exported for callers that already
hold dpll_lock, use it directly and remove the change_work
infrastructure entirely.
This eliminates a race condition where change_work could be
re-scheduled after cancel_work_sync() during device teardown,
potentially causing the handler to dereference a freed or NULL
dpll_dev pointer.
Ivan Vecera [Tue, 26 May 2026 07:45:23 +0000 (09:45 +0200)]
dpll: export __dpll_device_change_ntf() for use under dpll_lock
Export __dpll_device_change_ntf() so that drivers can send device
change notifications from within device callbacks, which are already
called under dpll_lock. Using dpll_device_change_ntf() in that
context would deadlock.
Add lockdep_assert_held() to catch misuse without the lock held.
Merge patch series "fs: replace __get_free_pages() call with kmalloc()"
Mike Rapoport (Microsoft) <rppt@kernel.org> says:
This is a (small) part of larger work of replacing page allocator calls
with kmalloc.
* patches from https://patch.msgid.link/20260523-b4-fs-v1-0-275e36a83f0e@kernel.org:
bfs: replace get_zeroed_page() with kzalloc()
binfmt_misc: replace __get_free_page() with kmalloc()
configfs: replace __get_free_pages() with kzalloc()
fs/namespace: use __getname() to allocate mntpath buffer
fs/select: replace __get_free_page() with kmalloc()
fuse: replace __get_free_page() with kmalloc()
isofs: replace __get_free_page() with kmalloc()
jbd2: replace __get_free_pages() with kmalloc()
jfs: replace __get_free_page() with kmalloc()
libfs: simple_transaction_get(): replace get_zeroed_page() with kzalloc()
NFSD: replace __get_free_page() with kmalloc() in nfsd_buffered_readdir()
NFS: remove unused page and page2 in nfs4_replace_transport()
NFS: replace __get_free_page() with kmalloc() in nfs_show_devname()
nilfs2: replace get_zeroed_page() with kzalloc()
ocfs2/dlm: replace __get_free_page() with kmalloc()
proc: replace __get_free_page() with kmalloc()
quota: allocate dquot_hash with kmalloc()
fs/namespace: use __getname() to allocate mntpath buffer
mnt_warn_timestamp_expiry() allocates memory for a path with
__get_free_page() although there is a dedicated helper for allocation of
file paths: __getname().
Replace __get_free_page() for allocation of a path buffer with __getname().
Christian Brauner <brauner@kernel.org> says:
Pass PATH_MAX (not PAGE_SIZE) to d_path() to match the size that
__getname() actually allocates, and drop the now-unnecessary NULL check
around __putname() since __putname() handles NULL. Both per Jan Kara's
review feedback, acked by the author.
Joerg Roedel [Thu, 28 May 2026 07:53:17 +0000 (09:53 +0200)]
MAINTAINERS: Add Vasant Hegde to reviewers of AMD IOMMU
Vasant has a long history of providing valuable feedback and testing
results for the AMD IOMMU code. Still, too often he gets not Cc'ed on
code changes, so make his reviewer status official.
====================
net/handshake: anchor request lifetime to a pinned file reference
handshake_nl_accept_doit() has accumulated four follow-on fixes
since 3b3009ea8abb ("net/handshake: Create a NETLINK service for
handling handshake requests"): 7ea9c1ec66bc, 7798b59409c3, fe67b063f687, and dabac51b8102. Each was a local refcount or
NULL-check correction; none moved where the file reference is
owned, and the same code keeps producing the same class of bug.
Reworking the ownership is what breaks the pattern.
For the duration of a request, sock->file has no single owner.
Submit publishes the request without taking a file reference;
accept_doit acquires one inside the handler, after the request
has already left the pending list. The consumer can drop its
own reference at any time, including the moment between
handshake_req_next() popping the request and accept_doit
reaching get_file(). The submit-side sock_hold() pins only
struct sock; struct socket and sock->file remain under the
consumer's control via the file descriptor.
This series places the file reference under unambiguous
ownership. handshake_req_submit() pins it on the request and
completion or cancel drops it (patches 4-5); the submit-side
sock_hold() then becomes redundant, and dropping it also closes
a publish-before-pin race the late sock_hold itself opened
(patch 6). The handshake_complete() API and its consumers move
to a uniform negative-errno sign convention (patch 3), with the
matching sign correction in nvme-tcp (patch 2). Patch 1
hardens hn_lock for BH context, the netns-exit drain fix
builds on the new file-pin infrastructure (patch 8), and new
KUnit file-count assertions verify the refcount contract
(patch 7).
Three things in this restructuring want a careful look. In
handshake_complete(), the fput() of the request's file
reference has to come after hp_done() -- fput() can transitively
run handshake_sk_destruct() and free the request, so the patch
stashes hr_file in a local first. handshake_sk_destruct()
itself is kept on purpose: it owns rhashtable removal and
kfree, and remains the backstop if a consumer path bypasses
handshake_complete() entirely. Third, handshake_req_next() now
returns its request with an extra get_file() held under
hn_lock; accept_doit must consume that reference (FD_PREPARE on
success, explicit fput on the fdf.err path), and any future
caller has to honor the same contract.
Chuck Lever [Mon, 25 May 2026 16:51:22 +0000 (12:51 -0400)]
net/handshake: Drain pending requests at net namespace exit
The arguments to list_splice_init() in handshake_net_exit() are
reversed. The call moves the local empty "requests" list onto
hn->hn_requests, leaving the local list empty, so the subsequent
drain loop runs zero iterations. Pending handshake requests that
had not yet been accepted are not torn down when the net namespace
is destroyed; each one keeps a reference on a socket file and on
the handshake_req allocation.
Pass the source and destination in the documented order
(list_splice_init(list, head) moves list onto head) so the pending
list is transferred to the local scratch list and drained through
handshake_complete().
Fixing the splice direction exposes a list-corruption race. After
the splice each req->hr_list still has non-empty link pointers,
threading the stack-local scratch list rather than hn_requests.
A concurrent handshake_req_cancel() -- for example, from sunrpc's
TLS timeout on a kernel socket whose netns reference was not
taken -- finds the request through the rhashtable, calls
remove_pending(), and sees !list_empty(&req->hr_list).
__remove_pending_locked() then list_del_init()s an entry off the
scratch list while the drain iterates, corrupting it. The same
call arriving after the drain loop has run list_del() on an
entry hits LIST_POISON instead.
Have remove_pending() check HANDSHAKE_F_NET_DRAINING under
hn_lock and report not-found when drain is in progress. The
drain has already taken ownership; handshake_complete()'s existing
test_and_set on HANDSHAKE_F_REQ_COMPLETED still arbitrates
between drain and cancel for who calls the consumer's hp_done. Use
list_del_init() rather than list_del() in the drain so req->hr_list
does not carry LIST_POISON after drain releases the entry.
The DRAINING guard in remove_pending() makes cancel return false,
but cancel still falls through to test_and_set_bit on
HANDSHAKE_F_REQ_COMPLETED and drops the request's hr_file reference.
Without another pin, if that is the last reference, sk_destruct frees
the request while it is still linked on the drain loop's local list.
Pin each request's hr_file under hn_lock before releasing the list,
and drop that drain pin after the loop finishes with the request.
Fixes: 3b3009ea8abb ("net/handshake: Create a NETLINK service for handling handshake requests") Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Reviewed-by: Hannes Reinecke <hare@kernel.org> Link: https://patch.msgid.link/20260525-handshake-file-pin-v3-8-66c616906ead@oracle.com Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Chuck Lever [Mon, 25 May 2026 16:51:21 +0000 (12:51 -0400)]
net/handshake: Verify file-reference balance in submit paths
The new file-reference contract on struct handshake_req is silently
breakable: a missing get_file() at submit or a missing fput() on an
error path leaves the file leaked but does not crash the test, so
the existing absence-of-crash checks pass either way.
Snapshot file_count(filp) before each handshake_req_submit() in
the submit-success, EAGAIN, EBUSY, and cancel tests, and assert
the expected balance after submit and again after cancel. The
already-completed cancel test also asserts the post-complete
balance, which pins down that handshake_complete() drops the
reference and that the subsequent cancel does not double-fput.
The destroy test gets the same treatment before __fput_sync(),
which double-checks that cancel's fput() ran and the only
remaining reference is the one sock_alloc_file() established.
Chuck Lever [Mon, 25 May 2026 16:51:20 +0000 (12:51 -0400)]
net/handshake: Close the submit-side sock_hold race
handshake_req_submit() publishes the request via
handshake_req_hash_add() and __add_pending_locked(), drops
hn_lock, and calls handshake_genl_notify() (which can sleep)
before taking sock_hold() on req->hr_sk. A fast tlshd ACCEPT
followed by DONE can drive handshake_complete()'s sock_put()
into the window between the spin_unlock and the late
sock_hold(); on a system where the consumer's fd held the
only sk reference, the late sock_hold() then operates on an
sk whose refcount has reached zero.
The preceding two patches install an explicit file reference
on struct handshake_req. That file pins sock->file, which
pins the embedded struct socket, which defers inet_release()'s
sock_put(). As long as hr_file is held, sk cannot reach refcount
zero from the consumer side, and the submit-side sock_hold()
with its matching sock_put() calls in handshake_complete() and
handshake_req_cancel() is now redundant.
Drop all three. The file reference already keeps each request's
socket alive, and the lifetime story is contained in a single
get_file()/fput() pair.
Fixes: 3b3009ea8abb ("net/handshake: Create a NETLINK service for handling handshake requests") Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Reviewed-by: Hannes Reinecke <hare@kernel.org> Link: https://patch.msgid.link/20260525-handshake-file-pin-v3-6-66c616906ead@oracle.com Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Chuck Lever [Mon, 25 May 2026 16:51:19 +0000 (12:51 -0400)]
net/handshake: hand off the pinned file reference to accept_doit
handshake_req_next() removes the request from the per-net
pending list and drops hn_lock before handshake_nl_accept_doit()
reads req->hr_sk->sk_socket and dereferences sock->file (once in
FD_PREPARE() and again in get_file()). In that window a
consumer running tls_handshake_cancel() followed by sockfd_put()
(svc_sock_free) or __fput_sync() (xs_reset_transport) releases
sock->file. sock_release() then runs sock_orphan(), zeroing
sk_socket, and frees the struct socket. The accept-side code
either reads NULL through sk_socket or chases freed memory.
The submit-side sock_hold() does not prevent this. sk_refcnt
protects struct sock, but struct socket and sock->file are
independently refcounted via the file descriptor the consumer
owns. Pinning sk leaves sock and sock->file unprotected.
Retarget the accept-side dereferences at req->hr_file, which was
pinned at submit time, instead of req->hr_sk->sk_socket->file.
Pinning on its own is not sufficient: a consumer that cancels
between handshake_req_next() returning and accept_doit reaching
FD_PREPARE() takes the !remove_pending() branch in
handshake_req_cancel() and drops hr_file before the accept side
takes its own reference. Hand off an additional file reference
inside handshake_req_next(), under hn_lock, so the accept side
operates on a reference that no concurrent handshake_req_cancel()
can revoke. FD_PREPARE() consumes that handed-off reference,
either by transferring it to the new fd in fd_publish() or by
dropping it in the cleanup destructor on error; the explicit
get_file() that previously balanced FD_PREPARE() is therefore
redundant and goes away.
Update handshake_req_cancel_test2 and _test3 to simulate the
FD_PREPARE() consumption with an fput() so the kunit file-count
assertions stay balanced.
Reported-by: Chris Mason <clm@meta.com> Fixes: 3b3009ea8abb ("net/handshake: Create a NETLINK service for handling handshake requests") Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Reviewed-by: Hannes Reinecke <hare@kernel.org> Link: https://patch.msgid.link/20260525-handshake-file-pin-v3-5-66c616906ead@oracle.com Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Chuck Lever [Mon, 25 May 2026 16:51:18 +0000 (12:51 -0400)]
net/handshake: Take a long-lived file reference at submit
handshake_nl_accept_doit() needs the file pointer backing
req->hr_sk->sk_socket to survive the window between
handshake_req_next() and the subsequent FD_PREPARE() and get_file().
The submit-side sock_hold() does not provide that. sk_refcnt keeps
struct sock alive, but struct socket is owned by sock->file: when
the consumer fputs the last file reference, sock_release() tears
the socket down regardless of any sock_hold.
Add an hr_file pointer to struct handshake_req and acquire an
explicit reference on sock->file during handshake_req_submit().
handshake_complete() and handshake_req_cancel() release the
reference on the completion-bit-winning path.
The submit error path must also release the file reference, but
after rhashtable insertion a concurrent handshake_req_cancel() can
discover the request and race the error path. Gate the error-path
cleanup -- sk_destruct restoration, fput, and request destruction
-- with test_and_set_bit(HANDSHAKE_F_REQ_COMPLETED), the same
serialization handshake_complete() and handshake_req_cancel()
already use. When cancel has already claimed ownership, the submit
error path returns without touching the request; socket teardown
handles final destruction.
The accept-side dereferences are not yet retargeted; that change
comes in the next patch.
Chuck Lever [Mon, 25 May 2026 16:51:17 +0000 (12:51 -0400)]
net/handshake: Pass negative errno through handshake_complete()
handshake_complete() declares status as unsigned int and
tls_handshake_done() negates that value (-status) before handing
it to the TLS consumer. Consumers match on negative errno
constants -- xs_tls_handshake_done() has
switch (status) {
case 0:
case -EACCES:
case -ETIMEDOUT:
lower_transport->xprt_err = status;
break;
default:
lower_transport->xprt_err = -EACCES;
}
so the API as designed expects callers to pass positive errno
values that the tlshd shim then negates.
Three internal callers in handshake_nl_accept_doit(), the
net-exit drain, and a kunit test follow kernel convention and
pass negative errnos -- -EIO, -ETIMEDOUT, -ETIMEDOUT. The
implicit conversion to unsigned int turns -ETIMEDOUT into
0xFFFFFF92; the subsequent -status in tls_handshake_done()
wraps back to 110, the consumer's switch falls through, and
the xprt reports -EACCES on what should be -ETIMEDOUT or -EIO.
Fix the API rather than the call sites. The natural kernel
convention is negative errno in, negative errno out. Change
handshake_complete() and hp_done to take int status, drop the
negation in tls_handshake_done(), and negate once in
handshake_nl_done_doit() where status arrives from the wire
as an unsigned netlink attribute. The three internal callers
were already correct under that convention and need no change.
At the same wire boundary, declare MAX_ERRNO as the netlink
policy upper bound for HANDSHAKE_A_DONE_STATUS. Attribute
validation rejects out-of-range values before
handshake_nl_done_doit() runs, and negating a bounded u32 there
stays within int range -- closing the UBSAN-visible signed-
integer overflow that an unconstrained u32 would invoke.
Fixes: 3b3009ea8abb ("net/handshake: Create a NETLINK service for handling handshake requests") Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Reviewed-by: Hannes Reinecke <hare@kernel.org> Link: https://patch.msgid.link/20260525-handshake-file-pin-v3-3-66c616906ead@oracle.com Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Chuck Lever [Mon, 25 May 2026 16:51:16 +0000 (12:51 -0400)]
nvme-tcp: store negative errno in queue->tls_err
nvme_tcp_tls_done() assigns queue->tls_err in three branches. The
ENOKEY lookup failure and the EOPNOTSUPP initializer both store
negative errnos. The third branch, reached when the handshake
layer reports a non-zero status, stores -status.
The handshake layer delivers status to the consumer callback as a
negative errno; the other in-tree consumers --
xs_tls_handshake_done() and the nvmet target callback -- treat
their status argument that way. The extra negation in
nvme_tcp_tls_done() flips the sign, leaving tls_err as a positive
value (for instance, +EIO), which nvme_tcp_start_tls() then
returns to its caller.
Drop the extra negation so queue->tls_err uniformly carries a
negative errno on failure.
Fixes: be8e82caa685 ("nvme-tcp: enable TLS handshake upcall") Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Reviewed-by: Hannes Reinecke <hare@kernel.org> Reviewed-by: Alistair Francis <alistair.francis@wdc.com> Link: https://patch.msgid.link/20260525-handshake-file-pin-v3-2-66c616906ead@oracle.com Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Chuck Lever [Mon, 25 May 2026 16:51:15 +0000 (12:51 -0400)]
net/handshake: Use spin_lock_bh for hn_lock
nvmet_tcp_state_change(), a socket callback that runs in BH context,
can reach handshake_req_cancel() via nvmet_tcp_schedule_release_queue()
and tls_handshake_cancel(). handshake_req_cancel() acquires
hn->hn_lock with plain spin_lock(). If a process-context thread on
the same CPU holds hn->hn_lock when a softirq invokes the cancel path,
the lock attempt deadlocks. This is the only caller that invokes
tls_handshake_cancel() from BH context; every other consumer calls it
from process context.
Deferring the cancel to process context in the NVMe target is not
straightforward: nvmet_tcp_schedule_release_queue() must call
tls_handshake_cancel() atomically with its state transition to
DISCONNECTING. If the cancel were deferred, the handshake completion
callback could fire in the window before the cancel runs, observe the
unexpected state, and return without dropping its kref on the queue.
Reworking that interlock is considerably more invasive than hardening
the handshake lock. Convert all hn->hn_lock acquisitions from
spin_lock/spin_unlock to spin_lock_bh/spin_unlock_bh so the lock is
never taken with softirqs enabled.
Minh Nguyen [Tue, 26 May 2026 04:12:39 +0000 (11:12 +0700)]
net: skbuff: fix missing zerocopy reference in pskb_carve helpers
pskb_carve_inside_header() and pskb_carve_inside_nonlinear() both copy
the old skb_shared_info header into a new buffer via memcpy(), which
includes the destructor_arg pointer (uarg) for MSG_ZEROCOPY skbs.
Neither function calls net_zcopy_get() for the new shinfo, creating an
unaccounted holder: every skb_shared_info with destructor_arg set will
call skb_zcopy_clear() once when freed, but the corresponding
net_zcopy_get() was never called for the new copy. Repeated calls
drive uarg->refcnt to zero prematurely, freeing ubuf_info_msgzc while
TX skbs still hold live destructor_arg pointers.
KASAN reports use-after-free on a freed ubuf_info_msgzc:
BUG: KASAN: slab-use-after-free in skb_release_data+0x77b/0x810
Read of size 8 at addr ffff88801574d3e8 by task poc/220
Allocated by task 219:
msg_zerocopy_realloc+0x157/0x7b0
tcp_sendmsg_locked+0x2892/0x3ba0
Freed by task 219:
ip_recv_error+0x74a/0xb10
tcp_recvmsg+0x475/0x530
The skb consuming the late access still referenced the same uarg via
shinfo->destructor_arg copied by pskb_carve_inside_nonlinear() without
a refcount bump. This has been verified to be reliably exploitable: a
working proof-of-concept achieves full root privilege escalation from
an unprivileged local user on a default kernel configuration.
The fix follows the pattern of pskb_expand_head() which has the same
memcpy/cloned structure. For pskb_carve_inside_header(), net_zcopy_get()
is placed after skb_orphan_frags() succeeds, so the orphan error path
needs no cleanup. For pskb_carve_inside_nonlinear(), net_zcopy_get() is
placed after all failure points and just before skb_release_data(), so
no error path needs cleanup at all -- matching pskb_expand_head() more
closely and avoiding the need for a balancing net_zcopy_put().
Fixes: 6fa01ccd8830 ("skbuff: Add pskb_extract() helper function") Cc: stable@vger.kernel.org Assisted-by: Claude:claude-sonnet-4-6 Signed-off-by: Minh Nguyen <minhnguyen.080505@gmail.com> Reviewed-by: Willem de Bruijn <willemb@google.com> Link: https://patch.msgid.link/20260526041240.329462-1-minhnguyen.080505@gmail.com Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Some drivers want to use topology name, but currently each drivers are
setting it by own method.
This patch adds new snd_soc_card_set_topology_name() and do it by
same method.
Almost all driver doesn't set topology name, let's remove fixed name
array, and use devm_kasprintf() instead.
Ankit Nautiyal [Wed, 27 May 2026 04:10:50 +0000 (09:40 +0530)]
drm/i915/dp: Account for AS_SDP guardband only when enabled
Currently the intel_dp_sdp_min_guardband() accounts for AS_SDP for all
platforms that support adaptive sync SDP even for configurations where
it cannot be enabled. Instead account for adaptive sync SDP guardband
only when it is enabled.
Ankit Nautiyal [Wed, 27 May 2026 04:10:49 +0000 (09:40 +0530)]
drm/i915/dp: Enable AS SDP whenever VRR is possible or PR !async
Currently AS SDP is only configured when VRR is enabled.
With optimized guardband, we also need to account for wakeup time and other
relevant details that depend on the AS SDP position whenever AS SDP is
enabled. If a feature enabling AS SDP gets turned on later (after modeset),
the guardband might not be sufficient and may need to increase, triggering
a full modeset.
Additionally, for Panel Replay with Aux-less ALPM where the sink does
not support asynchronous video timing in PR active, the source must
keep transmitting Adaptive-Sync SDPs while PR is active.
So, always send AS SDP whenever there is a possibility to use it for VRR
OR for Panel Replay for synchronization.
v2: Check if AS SDP can be used for synchronization for VRR or PR. (Ville)
v3: Use intel_psr_needs_alpm_aux_less() instead of
intel_alpm_is_alpm_aux_less() to avoid including the LOBF case. (Ville)
Modify the commit message and subject.
Ankit Nautiyal [Wed, 27 May 2026 04:10:47 +0000 (09:40 +0530)]
drm/i915/dp: Compute and include coasting vtotal for AS SDP
DP v2.1 allows the source to temporarily suspend Adaptive-Sync SDP
transmission while Panel Replay is active when the sink supports
asynchronous video timing.
In such cases, the sink relies on the last transmitted AS SDP timing
information to maintain the refresh rate. To support this behavior,
compute and populate the coasting vtotal field in the AS SDP payload.
Include coasting vtotal in AS SDP packing, unpacking, and comparison,
and set it during late AS SDP configuration for PR with Aux-less ALPM
when asynchronous video timing is supported.
Note:
The coasting vtotal value is fully under driver control i.e. the HW does
not overwrite these payload bytes. HW only samples the PR_ALPM_CTL[AS SDP
Transmission in Active Disable] bit during PR active state and reflects it
in the AS SDP payload at the appropriate time.
Ankit Nautiyal [Wed, 27 May 2026 04:10:45 +0000 (09:40 +0530)]
drm/i915/dp: Set relevant Downspread Ctrl DPCD bits for PR + Auxless ALPM
If a Panel Replay capable sink, supports Async Video timing in
PR active state, then source does not necessarily need to send AS SDPs
during PR active.
However, if asynchronous video timing is not supported, then for PR with
Aux-less ALPM, the source must transmit Adaptive-Sync SDPs for video
timing synchronization while PR is active.
If the source needs to send AS SDP during PR active, this requires setting
DPCD 0x0107[6] (FIXED_VTOTAL_AS_SDP_EN_IN_PR_ACTIVE). This applies whether
VRR is enabled (AVT/FAVT) or fixed-timing mode is used.
This bit defines AS SDP timing behavior during PR Active, even if AS SDPs
are briefly suspended.
Program the relevant Downspread Ctrl DPCD bits accordingly.
v2: Instead of Panel Replay check simply use AS SDP enable check. (Ville)
v3: Since the bit is defined in context of Panel Replay and AS SDP, add
a check for both. (Ville)
v4: Extract pr_with_as_sdp logic into helper function. (Ville)
Ankit Nautiyal [Wed, 27 May 2026 04:10:43 +0000 (09:40 +0530)]
drm/i915/display: Add helper for AS SDP transmission time selection
AS SDP may be transmitted at T1 or T2 depending on Panel Replay and
Adaptive Sync SDP configuration as per DP 2.1. Current we are using
T1 only, but future PR/AS SDP modes/features may require T2 or dynamic
selection.
Introduce a helper to return the appropriate AS SDP transmission time so
that a single value is consistently used for programming PR_ALPM.
For now this returns T1.
v2: Avoid adding new member to crtc_state; use a helper. (Ville)
v3: Clarify why AS SDP transmission time is fixed to T1. (Ville)
v4: Return u8 from intel_dp_as_sdp_transmission_time(). (Ville)
Ankit Nautiyal [Wed, 27 May 2026 04:10:42 +0000 (09:40 +0530)]
drm/i915/psr: Write the PR config DPCDs in burst mode
Replace the consecutive single-byte writes to PANEL_REPLAY_CONFIG and
CONFIG2 with one drm_dp_dpcd_write() burst starting at PANEL_REPLAY_CONFIG,
reducing AUX transactions.
v2: Drop extra conditions, and optimize variables. (Ville)
v3: Drop the error check after write. (Ville)
Ankit Nautiyal [Wed, 27 May 2026 04:10:40 +0000 (09:40 +0530)]
drm/i915/dp: Add member to intel_dp to store AS SDP v2 support
eDP v1.5a advertises support for Adaptive Sync SDP and with that the
support for AS SDP v2 is mandatory.
DP v2.1 SCR advertises support for FAVT payload fields parsing in DPCD
0x2214 Bit 2. This indicates the support for Adaptive-Sync SDP version 2
(AS SDP v2), which allows the source to set the version in HB2[4:0] and the
payload length in HB3[5:0] of the AS SDP header.
DP v2.1 SCR also introduces ASYNC_VIDEO_TIMING_NOT_SUPPORTED_IN_PR in the
Panel Replay Capability DPCD 0x00b1 (Bit 3). When this bit is set, the sink
does not support asynchronous video timing while in a Panel Replay Active
state and the source is required to keep transmitting Adaptive-Sync
SDPs. The spec mandates that such sinks shall support AS SDP v2.
Infer AS SDP v2 support from these capabilities and store it in
struct intel_dp for use by subsequent feature enablement changes.
v2:
- Include parsing ASYNC_VIDEO_TIMING_NOT_SUPPORTED_IN_PR bit to
determine AS SDP v2 support. (Ville)
v3:
- Use helper to determine asynch video timing support.
v4:
- Add AS SDP v2 support for eDP as per v1.5a.
- Add a check for Panel Replay support before checking for Async video
timing support in PR
- Add a TODO for Display ID and PCON considerations. (Ville)
Claudio Imbrenda [Wed, 27 May 2026 14:43:55 +0000 (16:43 +0200)]
KVM: s390: Implement KVM_PRE_FAULT_MEMORY
Implement and enable the KVM_PRE_FAULT_MEMORY ioctl for s390.
Faulted-in pages will be marked as accessed, unlike x86, otherwise they
will trigger a minor fault when accessed. Avoiding such faults is one of
the points of KVM_PRE_FAULT_MEMORY.
Claudio Imbrenda [Wed, 27 May 2026 14:43:54 +0000 (16:43 +0200)]
KVM: s390: Track page size in struct guest_fault
Until now, the members of struct guest_fault are always accessed while
holding the required locks, and thus the ptep and crstep pointers can
be dereferenced safely.
There will be some new cases where callers of kvm_s390_faultin_gfn()
need to know the size of the page used to solve the fault, at which
point no locks are held anymore, and dereferencing the crstep field
is not possible.
Introduce a new crste_region3 flag for struct guest_fault to indicate
whether the crstep used to solve the fault was a region 3 entry with FC=1
(large pud).
This allows to disambiguate all three possible scenarios:
* If ptep is not NULL, the fault was solved with a pte.
* If ptep is NULL and crste_region3 is 0, a segment entry with FC=1
(large pmd) was used.
* If ptep is NULL and crste_region3 is 1, a region 3 entry with FC=1
(large pud) was used.
This series fixes an RX packet corruption issue observed when SMMU is
disabled on the hibmcge driver. The fixes include disabling PCI Relaxed
Ordering and correcting the order of DMA barrier operations in the RX
data sync path.
====================
Tomas Glozar [Wed, 27 May 2026 14:49:28 +0000 (16:49 +0200)]
Documentation/rtla: Add -A/--aligned option
Cover the newly added -A/--aligned option that aligns timerlat threads
using the corresponding feature of the timerlat tracer.
A note is added to clarify what alignment means, similar to the note in
the tracer implementation in commit 4245bf4dc58f ("tracing/osnoise: Add
option to align tlat threads").
Tomas Glozar [Wed, 27 May 2026 14:49:26 +0000 (16:49 +0200)]
rtla/timerlat: Add -A/--aligned CLI option
Add a new option, -A/--aligned, that enables timerlat thread alignment
implemented on the kernel-side in commit 4245bf4dc58f ("tracing/osnoise:
Add option to align tlat threads"). The option takes an argument,
representing alignment between timerlat threads in microseconds.
The feature is modeled after the option of the same name in the
cyclictest tool.
Tomas Glozar [Thu, 28 May 2026 10:32:54 +0000 (12:32 +0200)]
rtla/tests: Add unit tests for CLI option callbacks
In addition to testing all tool_parse_args() functions, test also all
callbacks used for parsing custom option formats.
The callbacks represent a middle layer between the parsing functions
and utility functions dedicated to checking specific argument formats,
for example, scheduling class and duration. Callback tests are run
before parsing functions to make sure any issue in the former is
reported before it is encountered through the latter.
Tests verify both successful parsing and proper rejection of invalid
inputs (via exit tests). To enable testing static callbacks, a pragma
once guard is added to timerlat.h for safe inclusion by cli_p.h.
Add dependency of UNIT_TESTS_IN on LIBSUBCMD_INCLUDES, as the new test
file tests/unit/cli_opt_callback.c includes cli_p.h which includes
subcmd/parse-options.h.
Tomas Glozar [Thu, 28 May 2026 10:32:53 +0000 (12:32 +0200)]
rtla/tests: Add unit tests for _parse_args() functions
Add a test suite for the _parse_args() function of each tool that checks
the params structures (struct common_params, struct osnoise_params,
struct timerlat_params) returned by them for correctness.
One test case is added per option, as well as a few special cases for
tricky combinations of options. Test cases are ordered the same as the
option arrays and help message to allow easy checking of whether all
options are covered.
This should help clarify what the proper command line behavior of RTLA
is in case there are holes in the documentation and verify that the
intended behavior is implemented correctly.
A few necessary changes to the unit tests were done as part of this
commit:
- Unit tests now also link to libsubcmd and its dependencies.
- A new global variable in_unit_test is added to RTLA's CLI interface,
causing it to skip check for root if running in unit tests. This
allows the CLI unit tests to run as non-root, like existing unit
tests.
There is quite a lot of duplication, some of it is mitigated with macros,
but partially it is intentional so that future changes in behavior are
tracked across tools.
Tomas Glozar [Thu, 28 May 2026 10:32:52 +0000 (12:32 +0200)]
rtla: Parse cmdline using libsubcmd
Instead of using getopt_long() directly to parse the command line
arguments given to an RTLA tool, use libsubcmd's parse_options().
Utilizing libsubcmd for parsing command line arguments has several
benefits:
- A help message is automatically generated by libsubcmd from the
specification, removing the need of writing it by hand.
- Options are sorted into groups based on which part of tracing (CPU,
thread, auto-analysis, tuning, histogram) they relate to.
- Common parsing patterns for numerical and boolean values now share
code, with the target variable being stored in the option array.
To avoid duplication of the option parsing logic, RTLA-specific
macros defining struct option values are created:
- RTLA_OPT_* for options common to all tools
- OSNOISE_OPT_* and TIMERLAT_OPT_* for options specific to
osnoise/timerlat tools
- HIST_OPT_* macros for options specific to histogram-based tools.
Individual *_parse_args() functions then construct an array out of
these macros that is then passed to libsubcmd's parse_options().
All code specific to command line options parsing is moved out of the
individual tool files into a new file, cli.c, which also contains the
contents of the rtla.c file. A private header, cli_p.h, is added
alongside the public header cli.h, so that unit tests are able to test
statically declared option callbacks.
Minor changes:
- The return value of tool-level help option changes to 129, as this is
the value set by libsubcmd; this is reflected in affected test cases.
The implementation of help for command-level and tracer-level help
is set to 129 as well for consistency, and the change is reflected in
exit value documentation.
- Related to the above, {rtla,osnoise,timerlat}_usage() are marked
__noreturn and exit() is removed from after they are called for
cleaner code.
- The error messages for invalid argument for options --dma-latency and
-E/--entries were corrected, fixing off-by-one in the limits.
Note that unsetting options (using --no-<opt> syntax) is currently not
implemented for options that use custom callbacks. For --irq and
--thread, it will never be implemented, as they conflict with already
existing --no-irq and --no-thread with a different meaning.
Tomas Glozar [Thu, 28 May 2026 10:32:51 +0000 (12:32 +0200)]
tools subcmd: allow parsing distinct --opt and --no-opt
libsubcmd automatically generates for every option --opt an equivalent
negated option, --no-opt, to unset the option. Vice versa, for every
option declared as --no-opt, a shorthand --opt is declared for
convenience.
Add a flag, PARSE_OPT_NOAUTONEG, to disable this behavior. This new flag
behaves similarly to the already existing PARSE_OPT_NONEG, only it does
not reject the --no-opt variant, but leaves it undefined. That is useful
when there is a conflicting distinct --no-opt option in the syntax of
the tool.
PARSE_OPT_NOAUTONEG is enabled per-option, allowing to unset other
options that do not have this conflict.
Tomas Glozar [Thu, 28 May 2026 10:32:50 +0000 (12:32 +0200)]
tools subcmd: support optarg as separate argument
In addition to "-ovalue" and "--opt=value" syntax, allow also "-o value"
and "--opt value" for options with optional argument when the newly
added PARSE_OPT_OPTARG_ALLOW_NEXT flag is set.
This behavior is turned off by default since it does not make sense for
tools using non-option command line arguments. Consider the ambiguity
of "cmd -d x", where "-d x" can mean either "-d with argument of x" or
"-d without argument, followed by non-option argument x". This is not an
issue in the case that the tool takes no non-option arguments.
To implement this, a new local variable, force_defval, is created in
get_value(), along with a comment explaining the logic.
Tomas Glozar [Thu, 28 May 2026 10:32:49 +0000 (12:32 +0200)]
rtla: Add libsubcmd dependency
In preparation for migrating RTLA to libsubcmd, build libsubcmd from the
appropriate directory next to the RTLA build proper, and link the
resulting object to RTLA.
libsubcmd uses str_error_r() and strlcpy() at several places. To support
these, also link the respective libraries from tools/lib.
For completeness, also add tools/include to include path. This will
allow other userspace functions and macros shipped with the kernel to be
used in RTLA; perf and bpftool, two other users of libsubcmd, already do
that.
To prevent a name conflict, rename RTLA's run_command() function to
run_tool_command(), and replace RTLA's own container_of implementation
with the one in tools/include/linux/container_of.h.
Tomas Glozar [Tue, 26 May 2026 10:25:23 +0000 (12:25 +0200)]
rtla/tests: Add runtime tests for restoring continue flag
In case an action preceding the continue action fails, not only
the continue flag should not be set, it should be unset if it was set
from a previous run of actions_perform().
Add a runtime test to both osnoise and timerlat tools that checks that
this works properly by creating a temporary file.
Tomas Glozar [Tue, 26 May 2026 10:25:22 +0000 (12:25 +0200)]
rtla/tests: Run runtime tests in temporary directory
Create a temporary directory before each test case to serve as working
directory during the duration of the test.
This prevents littering of the original working directory as well as
allows tests to use it to avoid path conflicts.
In order not to break already existing tests, also add a new "testdir"
variable containing the directory where the test file is located. This
is then used to locate artifacts used during testing like BPF programs
and scripts for checking the tracer threads.
Tomas Glozar [Tue, 26 May 2026 10:25:21 +0000 (12:25 +0200)]
rtla/tests: Add unit test for restoring continue flag
In case an action preceding the continue action fails, not only
the continue flag should not be set, it should be unset if it was set
from a previous run of actions_perform().
Add a unit test to check if this is implemented correctly.
Tomas Glozar [Tue, 26 May 2026 10:25:20 +0000 (12:25 +0200)]
rtla/actions: Restore continue flag in actions_perform()
Currently, actions_perform() only ever sets the continue flag (when
performing the continue action), but never resets it. That leads to
RTLA continuing tracing even if the continue action was not performed in
the current iteration.
should print Spike! at most once, because after hitting the threshold
for the first time, /tmp/a exists, the shell action will fail, and the
continue action is not performed. However, unless /tmp/a exists before
the measurement, it will print Spike! until stopped, as the continue
flag stays set.
Set the continue flag to false in the beginning of actions_perform() to
make RTLA continue only if the action was actually performed.
Jijie Shao [Mon, 25 May 2026 14:45:25 +0000 (22:45 +0800)]
net: hibmcge: move dma_rmb() after dma_sync_single_for_cpu() in RX path
The dma_rmb() barrier was placed before dma_sync_single_for_cpu(), which
is incorrect. DMA sync must complete first to make the buffer accessible
to the CPU, then the rmb barrier ensures subsequent descriptor reads
observe the latest data written by the hardware.
Reorder the operations so dma_sync_single_for_cpu() is called before
dma_rmb() to guarantee the driver reads consistent data from the DMA
buffer.
Jijie Shao [Mon, 25 May 2026 14:45:24 +0000 (22:45 +0800)]
net: hibmcge: disable Relaxed Ordering to fix RX packet corruption
When SMMU is disabled, the hibmcge driver may receive corrupted packets.
The hardware writes packet data and descriptors to the same page, but
with Relaxed Ordering enabled, PCI write transactions may not be
strictly ordered. This can cause the driver to observe a valid
descriptor before the corresponding packet data is fully written.
Fix this by clearing PCI_EXP_DEVCTL_RELAX_EN in the PCI bridge control
register to ensure strict write ordering between packet data and
descriptors.
====================
net/sched: Fix packet loops in mirred and netem
This patchset adds a 2-bit per-skb tc_depth counter that travels with
the packet. The existing per-CPU mirred nest tracking loses state
when a packet is deferred through the backlog or moves between CPUs
via XPS/RPS. A per-skb field covers both cases.
Patch 1 adds the tc_depth field in a padding hole in sk_buff.
Patches 2-3 revert the check_netem_in_tree() fix and its tests,
which broke legitimate multi-netem configurations.
Patch 4 uses tc_depth to stop netem duplicate recursion.
Patch 5 uses tc_depth to catch mirred ingress redirect loops.
Patch 6 fixes the infinite loop in the mirred egress blockcast case.
Patch 7 fixes drop stats in early return error scenarios in tcf_mirred_act
for redirect (caught by Sashiko [1]).
Patches 8-9 add mirred and netem test cases.
Victor Nogueira [Mon, 25 May 2026 12:25:56 +0000 (08:25 -0400)]
selftests/tc-testing: Add netem test case exercising loops
Add a netem nested duplicate test case to validate that it won't
cause an infinite loop
Acked-by: Jamal Hadi Salim <jhs@mojatatu.com> Acked-by: Stephen Hemminger <stephen@networkplumber.org> Signed-off-by: Victor Nogueira <victor@mojatatu.com> Link: https://patch.msgid.link/20260525122556.973584-10-jhs@mojatatu.com Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Acked-by: Jamal Hadi Salim <jhs@mojatatu.com> Acked-by: Stephen Hemminger <stephen@networkplumber.org> Signed-off-by: Victor Nogueira <victor@mojatatu.com> Link: https://patch.msgid.link/20260525122556.973584-9-jhs@mojatatu.com Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Victor Nogueira [Mon, 25 May 2026 12:25:54 +0000 (08:25 -0400)]
net/sched: act_mirred: Fix return code in early mirred redirect error paths
Since retval is set as TC_ACT_STOLEN in the mirred redirect case, returning
retval in cases where redirect failed will make the callers not register
the skb as being dropped.
Fix this by returning TC_ACT_SHOT instead in such scenarios.
Fixes: 16085e48cb48 ("net/sched: act_mirred: Create function tcf_mirred_to_dev and improve readability") Reported-by: Sashiko <sashiko-bot@kernel.org> Closes: https://sashiko.dev/#/patchset/20260413082027.2244884-1-hxzene%40gmail.com Signed-off-by: Victor Nogueira <victor@mojatatu.com> Link: https://patch.msgid.link/20260525122556.973584-8-jhs@mojatatu.com Signed-off-by: Paolo Abeni <pabeni@redhat.com>
net/sched: act_mirred: Fix blockcast recursion bypass leading to stack overflow
tcf_mirred_act() checks sched_mirred_nest against MIRRED_NEST_LIMIT (4)
to prevent deep recursion. However, when the action uses blockcast
(tcfm_blockid != 0), the function returns at the tcf_blockcast() call
BEFORE reaching the counter increment. As a result, the recursion
counter never advances and the limit check is entirely bypassed.
When two devices share a TC egress block with a mirred blockcast rule,
a packet egressing on device A is mirrored to device B via blockcast;
device B's egress TC re-enters tcf_mirred_act() via blockcast and
mirrors back to A, creating an unbounded recursion loop:
This recursion continues until the kernel stack overflows.
The bug is reachable from an unprivileged user via
unshare(CLONE_NEWUSER | CLONE_NEWNET): user namespaces grant
CAP_NET_ADMIN in the new network namespace, which is sufficient to
create dummy devices, attach clsact qdiscs with shared blocks, and
install mirred blockcast filters.
BUG: TASK stack guard page was hit at ffffc90000b7fff8
Oops: stack guard page: 0000 [#1] SMP KASAN NOPTI
CPU: 2 UID: 1000 PID: 169 Comm: poc Not tainted 7.0.0-rc7-next-20260410
RIP: 0010:xas_find+0x17/0x480
Call Trace:
xa_find+0x17b/0x1d0
tcf_mirred_act+0x640/0x1060
tcf_action_exec+0x400/0x530
basic_classify+0x128/0x1d0
tcf_classify+0xd83/0x1150
tc_run+0x328/0x620
__dev_queue_xmit+0x797/0x3100
tcf_mirred_to_dev+0x7b1/0xf70
tcf_mirred_act+0x68a/0x1060
[repeating ~30+ times until stack overflow]
Kernel panic - not syncing: Fatal exception in interrupt
Fix this by incrementing sched_mirred_nest before calling
tcf_blockcast() and decrementing it on return, mirroring the
non-blockcast path. This ensures subsequent recursive entries see the
updated counter and are correctly limited by MIRRED_NEST_LIMIT.
When mirred redirects to ingress (from either ingress or egress) the loop
state from sched_mirred_dev array dev is lost because of 1) the packet
deferral into the backlog and 2) the fact the sched_mirred_dev array is
cleared. In such cases, if there was a loop we won't discover it.
Here's a simple test to reproduce:
ip a add dev port0 10.10.10.11/24
tc qdisc add dev port0 clsact
tc filter add dev port0 egress protocol ip \
prio 10 matchall action mirred ingress redirect dev port1
tc qdisc add dev port1 clsact
tc filter add dev port1 ingress protocol ip \
prio 10 matchall action mirred egress redirect dev port0
ping -c 1 -W0.01 10.10.10.10
Fixes: fe946a751d9b ("net/sched: act_mirred: add loop detection") Tested-by: Victor Nogueira <victor@mojatatu.com> Reviewed-by: Stephen Hemminger <stephen@networkplumber.org> Signed-off-by: Jamal Hadi Salim <jhs@mojatatu.com> Link: https://patch.msgid.link/20260525122556.973584-6-jhs@mojatatu.com Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Jamal Hadi Salim [Mon, 25 May 2026 12:25:51 +0000 (08:25 -0400)]
net/sched: fix packet loop on netem when duplicate is on
When netem duplicates a packet it re-enqueues the copy at the root qdisc.
If another netem sits in the tree the copy can be duplicated
again, recursing until the stack or memory is exhausted.
The original duplication guard temporarily zeroed q->duplicate around
the re-enqueue, but that does not cover all cases because it is
per-qdisc state shared across all concurrent enqueue paths
and is not safe without additional locking.
Use the skb tc_depth field introduced in an earlier patch:
- increment it on the duplicate before re-enqueue
- skip duplication for any skb whose tc_depth is already non-zero.
This marks the packet itself rather than mutating qdisc state,
therefore it is safe regardless of tree topology or concurrency.
Fixes: 0afb51e72855 ("[PKT_SCHED]: netem: reinsert for duplication") Reported-by: William Liu <will@willsroot.io> Reported-by: Savino Dicanosa <savy@syst3mfailure.io> Closes: https://lore.kernel.org/netdev/8DuRWwfqjoRDLDmBMlIfbrsZg9Gx50DHJc1ilxsEBNe2D6NMoigR_eIRIG0LOjMc3r10nUUZtArXx4oZBIdUfZQrwjcQhdinnMis_0G7VEk=@willsroot.io/ Co-developed-by: Victor Nogueira <victor@mojatatu.com> Signed-off-by: Victor Nogueira <victor@mojatatu.com> Reviewed-by: William Liu <will@willsroot.io> Reviewed-by: Stephen Hemminger <stephen@networkplumber.org> Signed-off-by: Jamal Hadi Salim <jhs@mojatatu.com> Link: https://patch.msgid.link/20260525122556.973584-5-jhs@mojatatu.com Signed-off-by: Paolo Abeni <pabeni@redhat.com>
projects / thirdparty / kernel / linux.git / log
