Changing subset=pid on an existing procfs instance is not safe. If a
full procfs mount has entries hidden by overmounts, switching it to
subset=pid would hide the top-level procfs entries from lookup and
readdir while leaving the existing overmounts reachable.
Reject attempts to change the subset=pid state during reconfigure before
applying any other procfs mount options, so a failed reconfigure cannot
partially update the instance.
Signed-off-by: Alexey Gladkov <legion@kernel.org>
Link: https://patch.msgid.link/13295f40f642af5d6e7038d681d43132ad80f7b2.1777278334.git.legion@kernel.org
Reviewed-by: Aleksa Sarai <aleksa@amutable.com>
Signed-off-by: Christian Brauner <brauner@kernel.org>
Mateusz Guzik <mjguzik@gmail.com> says:
The file is a mess with a hand-rolled linked list in a desperate need of
a clean up.
The code to emit /proc/filesystems is used frequently because libselinux
reads the file, which in turn is linked into numerous frequently used
programs (even ones you would not suspect, like sed!). In order to
combat that pre-gen the string instead of pointer-chasing and printfing
one by-one.
open+read+close cycle single-threaded (ops/s):
before: 442732
after: 1063462 (+140%)
Additionally scalability is also improved thanks to bypassing ref
maintenance on open/close.
open+read+close cycle with 20 processes (ops/s):
before: 606177
after: 3300576 (+444%)
The main bottleneck afterwards is the spurious lockref trip on open.
* patches from https://patch.msgid.link/20260425220844.1763933-1-mjguzik@gmail.com:
fs: cache the string generated by reading /proc/filesystems
fs: RCU-ify filesystems list
proc: allow to mark /proc files permanent outside of fs/proc/
Link: https://patch.msgid.link/20260425220844.1763933-1-mjguzik@gmail.com
Signed-off-by: Christian Brauner <brauner@kernel.org>
Cache the mounters credentials and allow access to the net directories
contingent of the permissions of the mounter of proc.
Do not show /proc/self/net when proc is mounted with subset=pid option
and the mounter does not have CAP_NET_ADMIN. To avoid inadvertently
allowing access to /proc/<pid>/net, updating mounter credentials is not
supported.
Signed-off-by: Alexey Gladkov <legion@kernel.org>
Link: https://patch.msgid.link/d2466fe9085367f1e24693c437ecb8cff2789660.1777278334.git.legion@kernel.org
Reviewed-by: Aleksa Sarai <aleksa@amutable.com>
Signed-off-by: Christian Brauner <brauner@kernel.org>
It is being read surprisingly often (e.g., by mkdir, ls and even sed!).
This is lock-protected pointer chasing over a linked list to pay for
sprintf for every fs (32 on my boxen).
Instead cache the result.
While here make the file as permanent to avoid spurious ref trips in
procfs.
Signed-off-by: Mateusz Guzik <mjguzik@gmail.com>
Link: https://patch.msgid.link/20260425220844.1763933-4-mjguzik@gmail.com
Signed-off-by: Christian Brauner <brauner@kernel.org>
Now that FS_USERNS_MOUNT_RESTRICTED is a file_system_type flag,
sysfs_get_tree() is a trivial wrapper around kernfs_get_tree() with no
additional logic. Point sysfs_fs_context_ops.get_tree directly at
kernfs_get_tree() and remove the wrapper.
Link: https://patch.msgid.link/e8ac71fc96ad864c8b58fc0a8e5305550c01db25.1777278334.git.legion@kernel.org
Reviewed-by: Aleksa Sarai <aleksa@amutable.com>
Signed-off-by: Christian Brauner <brauner@kernel.org>
The drivers list was protected by an rwlock; every mount, every open
of /proc/filesystems and the legacy sysfs(2) syscall walked a
hand-rolled singly-linked list under it. /proc/filesystems is
especially hot because libselinux causes programs as mundane as
mkdir, ls and sed to open and read it on every invocation.
Convert the list to an RCU-protected hlist and switch the writer side
to a plain spinlock. Writers keep their existing non-sleeping
section while readers walk under rcu_read_lock() with no lock traffic:
- register_filesystem()/unregister_filesystem() take
file_systems_lock, publish via hlist_{add_tail,del_init}_rcu()
and invalidate the cached /proc/filesystems string.
unregister_filesystem() keeps its synchronize_rcu() after
dropping the lock so in-flight readers are drained before the
module (and its embedded file_system_type) can go away.
- __get_fs_type(), list_bdev_fs_names() and the
fs_index()/fs_name()/fs_maxindex() helpers walk the list under
rcu_read_lock(). fs_name() continues to drop the read-side
lock after try_module_get() and accesses ->name outside the RCU
section; the module reference pins the embedded file_system_type
across the boundary.
struct file_system_type::next becomes struct hlist_node list; no
in-tree caller references the old ->next field outside
fs/filesystems.c.
Link: https://patch.msgid.link/20260425220844.1763933-3-mjguzik@gmail.com
Signed-off-by: Christian Brauner <brauner@kernel.org>
Whether a filesystem's mounts need to undergo a visibility check in user
namespaces is a static property of the filesystem type, not a runtime
property of each superblock instance. Both proc and sysfs always set
SB_I_USERNS_VISIBLE on their superblocks unconditionally (sysfs does so
on first creation, and subsequent mounts reuse the same superblock).
Move this flag from sb->s_iflags (SB_I_USERNS_VISIBLE) to
file_system_type->fs_flags (FS_USERNS_MOUNT_RESTRICTED) so the intent
is expressed at the filesystem type level where it belongs.
All check sites are updated to test sb->s_type->fs_flags instead of
sb->s_iflags. The SB_I_NOEXEC and SB_I_NODEV flags remain on the
superblock as they are runtime properties set during fill_super.
Link: https://patch.msgid.link/72887c5b6204dc3adf5a53104f0be6bd8bc4f6cd.1777278334.git.legion@kernel.org
Reviewed-by: Aleksa Sarai <aleksa@amutable.com>
Signed-off-by: Christian Brauner <brauner@kernel.org>
Add proc_make_permanent() function to mark PDE as permanent to speed up
open/read/close (one alloc/free and lock/unlock less).
Enable it for built-in code and for compiled-in modules.
This function becomes nop magically in modular code.
Note, note, note!
If built-in code creates and deletes PDEs dynamically (not in init
hook), then proc_make_permanent() must not be used.
It is intended for simple code:
static int __init xxx_module_init(void)
{
g_pde = proc_create_single();
proc_make_permanent(g_pde);
return 0;
}
static void __exit xxx_module_exit(void)
{
remove_proc_entry(g_pde);
}
If module is built-in then exit hook never executed and PDE is
permanent so it is OK to mark it as such.
If module is module then rmmod will yank PDE, but proc_make_permanent()
is nop and core /proc code will do everything right.
Signed-off-by: Alexey Dobriyan <adobriyan@gmail.com>
Link: https://patch.msgid.link/20260425220844.1763933-2-mjguzik@gmail.com
Signed-off-by: Christian Brauner <brauner@kernel.org>
Instead of wading through all the mounts in the mount namespace rbtree
to find fully visible procfs and sysfs mounts, be honest about them
being special cruft and record them in a separate per-mount namespace
list.
Link: https://patch.msgid.link/684859a8e0ac929cb89c1fbe16ce15b30c70eb1f.1777278334.git.legion@kernel.org
Reviewed-by: Aleksa Sarai <aleksa@amutable.com>
Signed-off-by: Christian Brauner <brauner@kernel.org>
1. iunique does not take the hash lock as of:
3f19b2ab97 ("vfs, afs, ext4: Make the inode hash table RCU searchable")
2. s_inode_list_lock is no longer taken under the hash lock as of:
c918f15420 ("fs: call inode_sb_list_add() outside of inode hash lock")
Signed-off-by: Mateusz Guzik <mjguzik@gmail.com>
Link: https://patch.msgid.link/20260423170431.1483370-1-mjguzik@gmail.com
Reviewed-by: Jan Kara <jack@suse.cz>
Signed-off-by: Christian Brauner <brauner@kernel.org>
Mateusz Guzik <mjguzik@gmail.com> says:
The stock kernel support partial lockless in handling in that iput() can
decrement any value > 1. Any ref acquire however requires the spinlock.
With this patchset ref acquires when the value was already at least 1
also become lockless. That is, only transitions 0->1 and 1->0 take the
lock.
I verified when nfs calls into the hash taking the lock is typically
avoided. Similarly, btrfs likes to igrab() and avoids the lock.
However, I have to fully admit I did not perform any benchmarks. While
cleaning stuff up I noticed lockless operation is almost readily
available so I went for it.
Clean-up wise, the icount_read_once() stuff lines up with
inode_state_read_once(). The prefix is different but I opted to not
change it due to igrab(), ihold() et al.
* patches from https://patch.msgid.link/20260421182538.1215894-1-mjguzik@gmail.com:
fs: allow lockless ->i_count bumps as long as it does not transition 0->1
fs: relocate and tidy up ihold()
fs: add icount_read_once() and stop open-coding ->i_count loads
Link: https://patch.msgid.link/20260421182538.1215894-1-mjguzik@gmail.com
Signed-off-by: Christian Brauner <brauner@kernel.org>
With this change only 0->1 and 1->0 transitions need the lock.
I verified all places which look at the refcount either only care about
it staying 0 (and have the lock enforce it) or don't hold the inode lock
to begin with (making the above change irrelevant to their correcness or
lack thereof).
I also confirmed nfs and btrfs like to call into these a lot and now
avoid the lock in the common case, shaving off some atomics.
Signed-off-by: Mateusz Guzik <mjguzik@gmail.com>
Link: https://patch.msgid.link/20260421182538.1215894-4-mjguzik@gmail.com
Reviewed-by: Jan Kara <jack@suse.cz>
Signed-off-by: Christian Brauner <brauner@kernel.org>
The placement was illogical, move it next to igrab().
Take this opportunity to add docs and an assert on the refcount. While
its modification remains gated with a WARN_ON, the new assert will also
dump the inode state which might aid debugging.
No functional changes.
Signed-off-by: Mateusz Guzik <mjguzik@gmail.com>
Link: https://patch.msgid.link/20260421182538.1215894-3-mjguzik@gmail.com
Reviewed-by: Jan Kara <jack@suse.cz>
Signed-off-by: Christian Brauner <brauner@kernel.org>
Similarly to inode_state_read_once(), it makes the caller spell out
they acknowledge instability of the returned value.
Signed-off-by: Mateusz Guzik <mjguzik@gmail.com>
Link: https://patch.msgid.link/20260421182538.1215894-2-mjguzik@gmail.com
Reviewed-by: Jan Kara <jack@suse.cz>
Signed-off-by: Christian Brauner <brauner@kernel.org>
Retry doesn't work here, since bo will be freed on error, leading to
UAF. However, now that we do the alloc & init before the attach, we can
now combine this as one unit and have the init do the alloc for us. This
should make the retry safe.
Reported by Sashiko.
v2: Fix up the error unwind (CI)
Closes: https://sashiko.dev/#/patchset/20260506184332.86743-2-matthew.auld%40intel.com
Fixes: eb289a5f6c ("drm/xe: Convert xe_dma_buf.c for exhaustive eviction")
Signed-off-by: Matthew Auld <matthew.auld@intel.com>
Cc: Thomas Hellström <thomas.hellstrom@linux.intel.com>
Cc: Matthew Brost <matthew.brost@intel.com>
Cc: <stable@vger.kernel.org> # v6.18+
Reviewed-by: Thomas Hellström <thomas.hellstrom@linux.intel.com>
Link: https://patch.msgid.link/20260508102635.149172-4-matthew.auld@intel.com
(cherry picked from commit 4796694182)
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
There look to be some nasty races here when triggering the
invalidate_mappings hook:
1) We do xe_bo_alloc() followed by the attach, before the actual full bo
init step in xe_dma_buf_init_obj(). However the bo is visible on the
attachments list after the attach. This is bad since exporter driver,
say amdgpu, can at any time call back into our invalidate_mappings hook,
with an empty/bogus bo, leading to potential bugs/crashes.
2) Similar to 1) but here we get a UAF, when the invalidate_mappings
hook is triggered. For example, we get as far as xe_bo_init_locked()
but this fails in some way. But here the bo will be freed on error, but
we still have it attached from dma-buf pov, so if the
invalidate_mappings is now triggered then the bo we access is gone and
we trigger UAF and more bugs/crashes.
To fix this, move the attach step until after we actually have a fully
set up buffer object. Note that the bo is not published to userspace
until later, so not sure what the comment "Don't publish the bo
until we have a valid attachment", is referring to.
We have at least two different customers reporting hitting a NULL ptr
deref in evict_flags when importing something from amdgpu, followed by
triggering the evict flow. Hit rate is also pretty low, which would
hint at some kind of race, so something like 1) or 2) might explain
this.
v2:
- Shuffle the order of the ops slightly (no functional change)
- Improve the comment to better explain the ordering (Matt B)
Assisted-by: Gemini:gemini-3 #debug
Link: https://gitlab.freedesktop.org/drm/xe/kernel/-/work_items/7903
Link: https://gitlab.freedesktop.org/drm/xe/kernel/-/work_items/4055
Fixes: dd08ebf6c3 ("drm/xe: Introduce a new DRM driver for Intel GPUs")
Signed-off-by: Matthew Auld <matthew.auld@intel.com>
Cc: Thomas Hellström <thomas.hellstrom@linux.intel.com>
Cc: Matthew Brost <matthew.brost@intel.com>
Cc: <stable@vger.kernel.org> # v6.8+
Reviewed-by: Matthew Brost <matthew.brost@intel.com>
Acked-by: Thomas Hellström <thomas.hellstrom@linux.intel.com>
Link: https://patch.msgid.link/20260508102635.149172-3-matthew.auld@intel.com
(cherry picked from commit af1f2ad0c5)
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
The blitter engines' MEM_COPY and MEM_SET instructions were added as
part of the same hardware change that introduced service copy engines
(i.e., BCS1-BCS8) which is why the driver checks for service copy engine
presence when deciding whether to use these instructions or the older
XY_* instructions. However when making this decision the driver should
consider which engines are part of the hardware architecture, not which
engines are present/usable on the current device. For graphics IP
versions that architecturally include service copy engines (i.e.,
everything Xe2 and later, plus PVC's Xe_HPC) we should use MEM_SET and
MEM_COPY even in if all of the service copy engines wind up getting
fused off. I.e., we need to decide based on whether the platform's
graphics descriptor contains these engines, rather than whether the
usable engine mask contains them. This logic got broken when
gt->info.__engine_mask was removed, although in practice that mistake
has been harmless so far because there haven't been any hardware
SKUs that fuse off all of the service copy engines yet.
Replace the incorrect has_service_copy_support() function with a GT
feature flag that tracks more accurately whether the new blitter
instructions are usable. In addition to fixing incorrect logic if all
service copies are fused off, the flag also makes it more obvious what
the calling code is trying to do; previously it wasn't terribly obvious
why "has service copy engines" was being used as the condition for using
different instructions on all copy engine types.
The new feature flag is named 'has_xe2_blt_instructions' because we
expect this flag to be set for all Xe2 and later platforms (i.e.,
everything officially supported by the Xe driver). Technically there's
also one Xe1-era platform (PVC) that supports these engines/instructions
and will set this flag, but this still seems to be the most clear and
understandable name for the flag.
Fixes: 61549a2ee5 ("drm/xe: Drop __engine_mask")
Cc: Balasubramani Vivekanandan <balasubramani.vivekanandan@intel.com>
Reviewed-by: Balasubramani Vivekanandan <balasubramani.vivekanandan@intel.com>
Link: https://patch.msgid.link/20260507-xe2_copy-v1-1-26506381b821@intel.com
Signed-off-by: Matt Roper <matthew.d.roper@intel.com>
(cherry picked from commit 09b3998429)
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
xe_bo_recompute_purgeable_state() walks all VMAs of a BO to determine
whether the BO can be made purgeable. This makes VMA create/destroy and
madvise updates O(n) in the number of mappings.
Replace the walk with BO-local counters protected by the BO dma-resv
lock:
- vma_count tracks the number of VMAs mapping the BO.
- willneed_count tracks active WILLNEED holders, including WILLNEED
VMAs and active dma-buf exports for non-imported BOs.
A DONTNEED BO is promoted back to WILLNEED on a 0->1 transition of
willneed_count. A BO is demoted to DONTNEED on a 1->0 transition only
when it still has VMAs, preserving the previous behaviour where a BO
with no mappings keeps its current madvise state.
PURGED remains terminal, preserving the existing "once purged, always
purged" rule.
Fixes: 4f44961eab ("drm/xe/vm: Prevent binding of purged buffer objects")
v2:
- Use early return for imported BOs in all four helpers to avoid
nesting (Matt B).
- Group purgeability state into a purgeable sub-struct on struct
xe_bo (Matt B).
- Reword xe_bo_willneed_put_locked() kernel-doc to explain that a 1->0
transition means all remaining active VMAs are DONTNEED (Matt B).
v3:
- Move DONTNEED/PURGED reject from vma_lock_and_validate() into
xe_vma_create(), gated on attr->purgeable_state == WILLNEED.
Fixes vm_bind bypass and partial-unbind rejection on DONTNEED
BOs (Matt B).
- Drop .check_purged from MAP and REMAP; keep it for PREFETCH and
add a comment why (Matt B).
- Skip BO validation in vma_lock_and_validate() for non-WILLNEED
VMA remnants so cleanup/remap paths do not repopulate
DONTNEED/PURGED BOs.
Suggested-by: Thomas Hellström <thomas.hellstrom@linux.intel.com>
Cc: Matthew Brost <matthew.brost@intel.com>
Cc: Thomas Hellström <thomas.hellstrom@linux.intel.com>
Cc: Himal Prasad Ghimiray <himal.prasad.ghimiray@intel.com>
Signed-off-by: Arvind Yadav <arvind.yadav@intel.com>
Reviewed-by: Matthew Brost <matthew.brost@intel.com>
Link: https://patch.msgid.link/20260506132027.2556046-1-arvind.yadav@intel.com
Signed-off-by: Himal Prasad Ghimiray <himal.prasad.ghimiray@intel.com>
(cherry picked from commit 23fb2ea56c)
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
cpuset_can_attach() currently adds the bandwidth of all migrating
SCHED_DEADLINE tasks to sum_migrate_dl_bw. If the source and destination
cpuset effective CPU masks do not overlap, the whole sum is then
reserved in the destination root domain.
set_cpus_allowed_dl(), however, subtracts bandwidth from the source
root domain only when the affinity change really moves the task between
root domains. A DL task can move between cpusets that are still in the
same root domain, so including that task in sum_migrate_dl_bw can reserve
destination bandwidth without a matching source-side subtraction.
Share the root-domain move test with set_cpus_allowed_dl(). Keep
nr_migrate_dl_tasks counting all migrating deadline tasks for cpuset DL
task accounting, but add to sum_migrate_dl_bw only for tasks that need a
root-domain bandwidth move. Keep using the destination cpuset effective
CPU mask and leave the broader can_attach()/attach() transaction model
unchanged.
Fixes: 2ef269ef1a ("cgroup/cpuset: Free DL BW in case can_attach() fails")
Cc: stable@vger.kernel.org # v6.10+
Signed-off-by: Guopeng Zhang <zhangguopeng@kylinos.cn>
Reviewed-by: Waiman Long <longman@redhat.com>
Acked-by: Juri Lelli <juri.lelli@redhat.com>
Tested-by: Juri Lelli <juri.lelli@redhat.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
If a firmware node is allocated on the stack (for instance: temporary
software node whose life-time we control) or on the heap - but using a
non-zeroing allocation function - and initialized using fwnode_init(),
its secondary pointer will contain uninitialized memory which likely
will be neither NULL nor IS_ERR() and so may end up being dereferenced
(for example: in dev_to_swnode()). Set fwnode->secondary to NULL on
initialization. While at it: initialize the remaining fields of struct
fwnode_handle too just to be sure.
Cc: stable@vger.kernel.org
Fixes: 01bb86b380 ("driver core: Add fwnode_init()")
Reviewed-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Reviewed-by: Rafael J. Wysocki (Intel) <rafael@kernel.org>
Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Signed-off-by: Bartosz Golaszewski <bartosz.golaszewski@oss.qualcomm.com>
Link: https://patch.msgid.link/20260511074927.9473-1-bartosz.golaszewski@oss.qualcomm.com
[ Fix typo in commit message. - Danilo ]
Signed-off-by: Danilo Krummrich <dakr@kernel.org>
The old format to define pinctrl settings for imx in DT has two hierarchy
levels. The first level are function device nodes. The second level are
pingroups which contain a property fsl,pins. The original ntention was to
define all pin functions in a single dtsi file and just reference the
correct ones in the board files.
The commit ("5fcdf6a7ed pinctrl: imx: Allow parsing DT without function
nodes") already make moden i.MX chip support flatten layout.
Make legacy chipes (more than 15 years) support this flatten layout also.
Fixes: e948cbdc41 ("ARM: dts: imx: remove redundant intermediate node in pinmux hierarchy")
Tested-by: Sébastien Szymanski <sebastien.szymanski@armadeus.com>
Signed-off-by: Frank Li <Frank.Li@nxp.com>
Signed-off-by: Linus Walleij <linusw@kernel.org>
When decoding osd_state and osd_weight from an incoming osdmap in
osdmap_decode(), both are decoded for each osd, i.e., map->max_osd
times. The ceph_decode_need() check only accounts for
sizeof(*map->osd_weight) once. This can potentially result in an
out-of-bounds memory access if the incoming message is corrupted such
that the max_osd value exceeds the actual content of the osdmap message.
This patch fixes the issue by changing the corresponding part in the
ceph_decode_need() check to account for
map->max_osd*sizeof(*map->osd_weight).
Cc: stable@vger.kernel.org
Fixes: dcbc919a5d ("libceph: switch osdmap decoding to use ceph_decode_entity_addr")
Signed-off-by: Raphael Zimmer <raphael.zimmer@tu-ilmenau.de>
Reviewed-by: Ilya Dryomov <idryomov@gmail.com>
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
This venerable comment got detached from its context when the code moved
in commit 394158559d ("x86: move all the pgd_list handling to one
place"). Put it back next to its context. It was originally on
pgd_list_add() but it actually describes pgd_list so put it there.
While moving it, update it to strip away stale and superfluous info.
pageattr.c doesn't exist any more. pgd_list is now required
for all x86 architectures. Also be slightly more precise about what PGDs
are in this list.
[ dhansen: tweak and trim the updated comment a bit ]
Signed-off-by: Brendan Jackman <jackmanb@google.com>
Signed-off-by: Dave Hansen <dave.hansen@linux.intel.com>
Link: https://patch.msgid.link/20260323-pgd_list-comment-v2-1-77ccf2dc77e8@google.com
The receiver shutdown timer handler, batadv_tp_receiver_shutdown(), is
responsible for releasing the tp_vars reference it holds. However, the
existing logic for coordinating this release with batadv_tp_stop_all() was
flawed.
timer_shutdown_sync() guarantees the timer will not fire again after it
returns, but it returns non-zero only when the timer was pending at the
time of the call. If the timer had already expired (and
batadv_tp_stop_all() would unsucessfully try to rearm itself),
batadv_tp_stop_all() skips its batadv_tp_vars_put(), and
batadv_tp_receiver_shutdown() fails to put its own reference as well.
Fix this by introducing a new atomic variable receiving that is set to 1
when the receiver is initialized and cleared atomically with atomic_xchg()
by whichever side claims it first. Only the side that observes the
transition from 1 to 0 is responsible for releasing the tp_vars timer
reference, eliminating the uncertainty.
Cc: stable@kernel.org
Fixes: 3d3cf6a731 ("batman-adv: stop tp_meter sessions during mesh teardown")
Signed-off-by: Sven Eckelmann <sven@narfation.org>
batadv_tp_sender_shutdown() unconditionally decrements the "sending"
atomic counter. If multiple paths (e.g. timeout, user cancel, and
normal finish) call this function, the counter can underflow to -1.
Since the sender logic treats any non-zero value as "still sending",
a negative value causes the sender kthread to loop indefinitely.
This leads to a use-after-free when the interface is removed while
the zombie thread is still active.
Fix this by using atomic_xchg() to ensure the counter only transitions
from 1 to 0 once.
Fixes: 33a3bb4a33 ("batman-adv: throughput meter implementation")
Cc: stable@kernel.org
Reported-by: Yuan Tan <yuantan098@gmail.com>
Reported-by: Yifan Wu <yifanwucs@gmail.com>
Reported-by: Juefei Pu <tomapufckgml@gmail.com>
Reported-by: Xin Liu <bird@lzu.edu.cn>
Signed-off-by: Luxiao Xu <rakukuip@gmail.com>
Signed-off-by: Ren Wei <n05ec@lzu.edu.cn>
[sven: added missing change in batadv_tp_send]
Signed-off-by: Sven Eckelmann <sven@narfation.org>
io_uring_try_cancel_requests() dropped ctx->uring_lock before calling
io_kill_timeouts(), which walks each timeout's link chain via
io_match_task() to test REQ_F_INFLIGHT. With chain mutation now
serialized by ctx->uring_lock, that walk needs the lock too.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
io_link_timeout_fn() is the hrtimer callback that fires when a linked
timeout expires. It currently calls io_remove_next_linked(prev) under
ctx->timeout_lock to splice the timeout request out of the link chain.
This is the only chain-mutation site that runs without ctx->uring_lock,
because hrtimer callbacks cannot take a mutex. Defer the splicing until
the task_work callback.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
io_wq_free_work() calls io_req_find_next() from io-wq worker context,
which reads and clears req->link without holding any lock. This can
potentially race with other paths that mutate the same chain under
ctx->uring_lock.
Take ctx->uring_lock around the io_req_find_next() call. Only requests
with IO_REQ_LINK_FLAGS reach this path, which is not the hot path.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
When a controller connects, nvme_start_ctrl() emits the
"NVME_EVENT=connected" uevent and sets the NVME_CTRL_STARTED_ONCE flag.
Currently, the uevent is emitted before the flag is set. This creates a
race condition for userspace tools (like udev rules) that might rely on
the "connected" event to configure other attributes.
Swap the order of operations in nvme_start_ctrl() so that the
NVME_CTRL_STARTED_ONCE flag is set before the uevent is sent. This
guarantees that the admin_timeout can already be changed when userspace
is notified.
Reviewed-by: Sagi Grimberg <sagi@grimberg.me>
Reviewed-by: Hannes Reinecke <hare@kernel.org>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Daniel Wagner <dwagner@suse.de>
Signed-off-by: Maurizio Lombardi <mlombard@redhat.com>
Signed-off-by: Keith Busch <kbusch@kernel.org>
Since every platform driver can be forced to match a device that doesn't
match its list of device IDs because of device_match_driver_override(),
platform drivers that rely on the existence of a device's ACPI companion
object should verify its presence.
Accordingly, add requisite ACPI_COMPANION() or ACPI_HANDLE() checks
against NULL to 13 platform drivers handling core ACPI devices.
Also change the value returned by the ACPI thermal zone driver when
the device's ACPI companion is not present to -ENODEV for consistency
with the other drivers.
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Reviewed-by: Hans de Goede <johannes.goede@oss.qualcomm.com>
Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Link: https://patch.msgid.link/4516068.ejJDZkT8p0@rafael.j.wysocki
Cc: 7.0+ <stable@vger.kernel.org> # 7.0+
When an already-exiting task oopses, make_task_dead() currently calls
do_task_dead() with preemption enabled. That is forbidden:
do_task_dead() calls __schedule(), which has a comment saying "WARNING:
must be called with preemption disabled!".
If an oopsing task is preempted in do_task_dead(), between becoming
TASK_DEAD and entering the scheduler explicitly, bad things happen:
finish_task_switch() assumes that once the scheduler has switched away
from a TASK_DEAD task, the task can never run again and its stack is no
longer needed; but that assumption apparently doesn't hold if the dead
task was preempted (the SM_PREEMPT case).
This means that the scheduler ends up repeatedly dropping references on
the dead task's stack, which can lead to use-after-free or double-free
of the entire task stack; in other words, two tasks can end up running
on the same stack, resulting in various kinds of memory corruption.
(This does not just affect "recursively oopsing" tasks; it is enough to
oops once during task exit, for example in a file_operations::release
handler)
Fixes: 7f80a2fd7d ("exit: Stop poorly open coding do_task_dead in make_task_dead")
Cc: stable@kernel.org
Signed-off-by: Jann Horn <jannh@google.com>
Acked-by: Peter Zijlstra <peterz@infradead.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Merge the mmc fixes for v7.1-rc[n] into the next branch, to allow them to
get tested together with the mmc changes that are targeted for the next
release.
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
Add voltage switching infrastructure for UHS-I modes by integrating both
regulator framework (for supply voltage control) and pinctrl state
switching (for pin drive strength optimization).
- Add regulator supply parsing and voltage switching callback
- Add optional pinctrl state switching between "default" (3.3V) and
"state_uhs" (1.8V) configurations
- Enable coordinated voltage and pin configuration changes for UHS modes
This provides complete voltage switching support while maintaining
backward compatibility when pinctrl states are not defined.
Tested-by: Anand Moon <linux.amoon@gmail.com>
Tested-by: Trevor Gamblin <tgamblin@baylibre.com>
Acked-by: Adrian Hunter <adrian.hunter@intel.com>
Reviewed-by: Troy Mitchell <troy.mitchell@linux.dev>
Tested-by: Vincent Legoll <legoll@online.fr>
Signed-off-by: Iker Pedrosa <ikerpedrosam@gmail.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
Ensure SD card pins receive clock signals by enabling pad clock
generation and overriding automatic clock gating. Required for all SD
operation modes.
The SDHC_GEN_PAD_CLK_ON setting in LEGACY_CTRL_REG is safe for both SD
and eMMC operation as both protocols use the same physical MMC interface
pins and require proper clock signal generation at the hardware level
for signal integrity and timing.
Additional SD-specific clock overrides (SDHC_OVRRD_CLK_OEN and
SDHC_FORCE_CLK_ON) are conditionally applied only for SD-only
controllers to handle removable card scenarios.
Tested-by: Anand Moon <linux.amoon@gmail.com>
Acked-by: Adrian Hunter <adrian.hunter@intel.com>
Tested-by: Trevor Gamblin <tgamblin@baylibre.com>
Reviewed-by: Troy Mitchell <troy.mitchell@linux.dev>
Tested-by: Vincent Legoll <legoll@online.fr>
Signed-off-by: Iker Pedrosa <ikerpedrosam@gmail.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
Document pinctrl properties to support voltage-dependent pin
configuration switching for UHS-I SD card modes.
Add optional pinctrl-names property with two states:
- "default": For 3.3V operation with standard drive strength
- "state_uhs": For 1.8V operation with optimized drive strength
These pinctrl states allow the SDHCI driver to coordinate voltage
switching with pin configuration changes, ensuring proper signal
integrity during UHS-I mode transitions.
Acked-by: Conor Dooley <conor.dooley@microchip.com>
Signed-off-by: Iker Pedrosa <ikerpedrosam@gmail.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
Yazhou Tang says:
====================
bpf: Fix call offset truncation and OOB read in bpf_patch_call_args()
From: Yazhou Tang <tangyazhou518@outlook.com>
This patchset addresses a silent truncation bug in the BPF verifier that
occurs when a bpf-to-bpf call involves a massive relative jump offset.
Additionally, it fixes a pre-existing out-of-bounds (OOB) read issue
in the interpreter fallback path.
Because the BPF instruction set utilizes a 32-bit imm field for bpf-to-bpf
calls, implicitly downcasting it to the 16-bit insn->off in bpf_patch_call_args()
causes incorrect call targets or subprog ID resolution for large BPF programs.
While fixing this by swapping the imm and off fields, it was discovered that
the original code also had a load-time OOB read vulnerability when the stack
depth exceeds MAX_BPF_STACK during JIT fallback.
Patch 1/3 fixes the pre-existing OOB read in bpf_patch_call_args(). It
changes the function to return an int and explicitly rejects the JIT
fallback if the stack depth exceeds MAX_BPF_STACK, preventing a potential
stack buffer overflow.
Patch 2/3 fixes the s16 truncation bug.
1. Keep the original imm field unchanged and use the off field to store
the interpreter function index.
2. Adjust the JMP_CALL_ARGS case in ___bpf_prog_run() accordingly.
3. Restore the legacy xlated dump layout in bpf_insn_prepare_dump().
Patch 3/3 introduces a selftest for this fix.
---
Change log:
v10:
1. Make the error log in patch 1/3 more clear. (Kuohai)
2. Drop bpftool and disasm_helpers.c changes, and instead restore the
legacy xlated dump layout in bpf_insn_prepare_dump(). This avoids
requiring bpftool compatibility handling. (Quentin and Alexei)
v9: https://lore.kernel.org/bpf/20260429171904.107244-1-tangyazhou@zju.edu.cn/
1. Modify the selftest in patch 3/3: use __clobber_all in inline asm.
(Sashiko AI reviewer)
v8: https://lore.kernel.org/bpf/20260429105608.92741-1-tangyazhou@zju.edu.cn/
1. Update cfg_partition_funcs() in bpftool to use insn->imm for call target
calculation. (Sashiko AI reviewer)
2. Modify the selftest in patch 3/3: add a large padding before the call
instruction, preventing the kernel panic on kernel without the fix.
(Sashiko AI reviewer)
3. Modify the selftest in patch 3/3: make it more clear.
v7: https://lore.kernel.org/bpf/20260421144504.823756-1-tangyazhou@zju.edu.cn/
1. Rebase the patchset to the bpf-next tree to resolve the apply conflict.
(Alexei)
2. Add Patch 1/3 to properly fix a pre-existing OOB read in bpf_patch_call_args().
(Sashiko AI reviewer)
v6: https://lore.kernel.org/bpf/20260412170334.716778-1-tangyazhou@zju.edu.cn/
1. Use a different but clearer approach to resolve this issue: keeping
the original imm field unchanged and using the off field to store the
interpreter function index. (Kuohai)
2. Update the related dumper code and remove a previous workaround in the
selftests disasm helpers, which is no longer needed after this fix.
v5: https://lore.kernel.org/bpf/20260326090133.221957-1-tangyazhou@zju.edu.cn/
1. Some minor changes in commit messages. (AI Reviewer)
v4: https://lore.kernel.org/bpf/20260326063329.10031-1-tangyazhou@zju.edu.cn/
1. Remove some redundant commit messages of patch 2/3. (Emil)
2. Change the number of instructions in padding_subprog() from 200,000
to 32,765, which is the minimum number of instructions required to
trigger the verifier failure. (Emil)
v3: https://lore.kernel.org/bpf/20260323122254.98540-1-tangyazhou@zju.edu.cn/
1. Resend to fix a typo in v2 and add "Fixes" tag. The rest of the changes
are identical to v2.
v2 (incorrect): https://lore.kernel.org/bpf/20260323081748.106603-1-tangyazhou@zju.edu.cn/
1. Move the s16 boundary check from fixup_call_args() to bpf_patch_call_args(),
and change the return type of bpf_patch_call_args() to int. (Emil)
2. Add Patch 3/3 to fix the incorrect subprog ID in dumped bpf_pseudo_call
instructions, which is caused by the same truncation issue. (Puranjay)
3. Refine the new selftest for clarity and add detailed comments explaining
the test design. (Emil)
v1: https://lore.kernel.org/bpf/20260316190220.113417-1-tangyazhou@zju.edu.cn/
====================
Link: https://patch.msgid.link/20260506094714.419842-1-tangyazhou@zju.edu.cn
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Add a selftest to verify the verifier and JIT behavior when handling
bpf-to-bpf calls with relative jump offsets exceeding the s16 boundary.
The test utilizes an inline assembly block with ".rept 32765" to generate
a massive dummy subprogram. By placing this padding between the main
program and the target subprogram, it forces the verifier to process a
bpf-to-bpf call where the imm field exceeds the s16 range.
- When JIT is enabled, it asserts that the program is successfully loaded
and executes correctly to return the expected value. Since the fix
does not change the JIT behavior, the test passes whether the fix is
applied or not.
- When JIT is disabled, it also asserts that the program is successfully
loaded and executes correctly to return the expected value 3.
- Before the fix, the verifier rewrites the call instruction with a
truncated offset (here 32768 -> -32768) and lets it pass. When the
program is executed, the call instruction will go to a wrong target
(the landing pad) instead of the intended subprogram, then return -1
and fail.
- After the fix, the verifier correctly handles the large offset and
allows it to pass. The program then executes correctly to return the
expected value 3.
Co-developed-by: Tianci Cao <ziye@zju.edu.cn>
Signed-off-by: Tianci Cao <ziye@zju.edu.cn>
Co-developed-by: Shenghao Yuan <shenghaoyuan0928@163.com>
Signed-off-by: Shenghao Yuan <shenghaoyuan0928@163.com>
Signed-off-by: Yazhou Tang <tangyazhou518@outlook.com>
Acked-by: Xu Kuohai <xukuohai@huawei.com>
Link: https://lore.kernel.org/r/20260506094714.419842-4-tangyazhou@zju.edu.cn
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Currently, the BPF instruction set allows bpf-to-bpf calls (or internal
calls, pseudo calls) to use a 32-bit imm field to represent the relative
jump offset.
However, when JIT is disabled or falls back to the interpreter, the
verifier invokes bpf_patch_call_args() to rewrite the call instruction.
In this function, the 32-bit imm is downcast to s16 and stored in the off
field.
void bpf_patch_call_args(struct bpf_insn *insn, u32 stack_depth)
{
stack_depth = max_t(u32, stack_depth, 1);
insn->off = (s16) insn->imm;
insn->imm = interpreters_args[(round_up(stack_depth, 32) / 32) - 1] -
__bpf_call_base_args;
insn->code = BPF_JMP | BPF_CALL_ARGS;
}
If the original imm exceeds the s16 range (i.e., a jump offset greater
than 32767 instructions), this downcast silently truncates the offset,
resulting in an incorrect call target.
Fix this by:
1. In bpf_patch_call_args(), keeping the imm field unchanged and using the
off field to store the index of the interpreter function.
2. In ___bpf_prog_run() for the JMP_CALL_ARGS case, retrieving the
interpreter function pointer from the interpreters_args array using the
off field as the index, and passing the original imm to calculate the
last argument of the interpreter function.
After these changes, the truncation issue is resolved, and __bpf_call_base_args
is also no longer needed and can be removed, which makes the code cleaner.
Performance: In ___bpf_prog_run() for the JMP_CALL_ARGS case, changing the
retrieval of the interpreter function pointer from pointer addition to
direct array indexing improves performance. The possible reason is that the
latter has better instruction-level parallelism. See the v5 discussion [1]
for more details.
[1] https://lore.kernel.org/bpf/f120c3c4-6999-414a-b514-518bb64b4758@zju.edu.cn/
To avoid requiring bpftool changes, keep the new imm/off encoding internal
and restore the legacy xlated dump layout in bpf_insn_prepare_dump().
For bpf-to-bpf call offsets that do not fit in s16, export off as 0 instead
of a truncated and misleading value.
Fixes: 1ea47e01ad ("bpf: add support for bpf_call to interpreter")
Fixes: 7105e828c0 ("bpf: allow for correlation of maps and helpers in dump")
Suggested-by: Xu Kuohai <xukuohai@huaweicloud.com>
Suggested-by: Puranjay Mohan <puranjay@kernel.org>
Co-developed-by: Tianci Cao <ziye@zju.edu.cn>
Signed-off-by: Tianci Cao <ziye@zju.edu.cn>
Co-developed-by: Shenghao Yuan <shenghaoyuan0928@163.com>
Signed-off-by: Shenghao Yuan <shenghaoyuan0928@163.com>
Signed-off-by: Yazhou Tang <tangyazhou518@outlook.com>
Link: https://lore.kernel.org/r/20260506094714.419842-3-tangyazhou@zju.edu.cn
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
The interpreters_args array only accommodates stack depths up to
MAX_BPF_STACK (512 bytes). However, do_misc_fixups() may allow a larger
stack depth if JIT is requested.
If JIT compilation later fails and falls back to the interpreter, the
verifier invokes bpf_patch_call_args() with this oversized stack depth.
This causes a load-time out-of-bounds (OOB) read when calculating the
interpreter function pointer index.
Fix this by changing bpf_patch_call_args() to return an int and explicitly
rejecting the JIT fallback (returning -EINVAL) if the stack depth exceeds
MAX_BPF_STACK.
Fixes: 1ea47e01ad ("bpf: add support for bpf_call to interpreter")
Co-developed-by: Tianci Cao <ziye@zju.edu.cn>
Signed-off-by: Tianci Cao <ziye@zju.edu.cn>
Co-developed-by: Shenghao Yuan <shenghaoyuan0928@163.com>
Signed-off-by: Shenghao Yuan <shenghaoyuan0928@163.com>
Signed-off-by: Yazhou Tang <tangyazhou518@outlook.com>
Acked-by: Xu Kuohai <xukuohai@huawei.com>
Link: https://lore.kernel.org/r/20260506094714.419842-2-tangyazhou@zju.edu.cn
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Instead of assigning the pci_device_id members using a list (which is
hard to read as you need to look at the order of the members in that
struct in parallel) use the PCI_VDEVICE() convenience macro to compact
the initialisation while improving readability.
Also drop trailing zeros that the compiler will care about then.
The change doesn't introduce binary changes to the compiled driver,
verified on both ARCH=x86 and ARCH=arm64.
Signed-off-by: Uwe Kleine-König (The Capable Hub) <u.kleine-koenig@baylibre.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
When uart_flush_buffer() runs before the DMA completion IRQ is delivered,
the following race can occur (all steps serialized by uart_port_lock):
1. DMA starts: tx_remaining = N, kfifo contains N bytes
2. DMA completes in hardware; IRQ is pending but not yet delivered
3. uart_flush_buffer() acquires the port lock and calls kfifo_reset(),
making kfifo_len() = 0 while tx_remaining remains N
4. uart_flush_buffer() releases the port lock
5. DMA IRQ fires; handle_tx_dma() acquires the port lock and calls
uart_xmit_advance(uport, tx_remaining) on an empty kfifo
uart_xmit_advance() increments kfifo->out by tx_remaining. Since
kfifo_reset() already set both in and out to 0, out wraps past in,
causing kfifo_len() to return UART_XMIT_SIZE - tx_remaining. The next
start_tx_dma() call then submits a DMA transfer of stale buffer data.
Fix this by snapshotting kfifo_len() at the start of handle_tx_dma()
and skipping uart_xmit_advance() when fifo_len < tx_remaining, which
indicates the kfifo was reset by a preceding flush.
Fixes: 2aaa43c707 ("tty: serial: qcom-geni-serial: add support for serial engine DMA")
Cc: stable <stable@kernel.org>
Signed-off-by: Viken Dadhaniya <viken.dadhaniya@oss.qualcomm.com>
Reviewed-by: Bartosz Golaszewski <bartosz.golaszewski@oss.qualcomm.com>
Link: https://patch.msgid.link/20260506-serial-dma-stale-tx-buf-v1-1-e3ccb360d719@oss.qualcomm.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
lpuart_start_rx_dma() allocates sport->rx_ring.buf with kzalloc() and
then maps a scatterlist via dma_map_sg(). On three subsequent error
paths the function returns directly without releasing those resources:
- when dma_map_sg() returns 0 (-EINVAL):
ring->buf is leaked.
- when dmaengine_slave_config() fails:
ring->buf and the DMA mapping are leaked.
- when dmaengine_prep_dma_cyclic() returns NULL:
ring->buf and the DMA mapping are leaked.
The sole cleanup path, lpuart_dma_rx_free(), is only reached when
lpuart_dma_rx_use is set, and the caller lpuart_rx_dma_startup() clears
that flag on failure of lpuart_start_rx_dma(). So these resources are
permanently leaked on every failure in this function. Repeated port
open/close or termios changes under error conditions will slowly consume
memory and leave stale streaming DMA mappings behind.
Fix it by introducing two error labels that unmap the scatterlist and
free the ring buffer as appropriate. While here, replace the misleading
-EFAULT (bad userspace pointer) returned when dmaengine_prep_dma_cyclic()
fails with the more accurate -ENOMEM, matching how other dmaengine users
in the tree treat this failure.
No functional change on the success path.
Fixes: 5887ad43ee ("tty: serial: fsl_lpuart: Use cyclic DMA for Rx")
Cc: stable <stable@kernel.org>
Signed-off-by: Shitalkumar Gandhi <shitalkumar.gandhi@cambiumnetworks.com>
Reviewed-by: Frank Li <Frank.Li@nxp.com>
Link: https://patch.msgid.link/20260420135903.2062024-1-shitalkumar.gandhi@cambiumnetworks.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
mmc_davinci_irq() returns early only when both host->cmd and
host->data are NULL:
if (host->cmd == NULL && host->data == NULL) {
...
return IRQ_NONE;
}
So we may legitimately reach the rest of the handler with
host->data == NULL (and therefore data == NULL). The DATDNE branch
already guards against this with an explicit "if (data != NULL)"
check, but the subsequent TOUTRD ("read data timeout") and
CRCWR/CRCRD ("data CRC error") branches dereference data
unconditionally:
if (qstatus & MMCST0_TOUTRD) {
data->error = -ETIMEDOUT; <-- NULL deref
...
davinci_abort_data(host, data);
}
if (qstatus & (MMCST0_CRCWR | MMCST0_CRCRD)) {
data->error = -EILSEQ; <-- NULL deref
...
}
If either bit is set in qstatus while host->data is NULL, the kernel
will crash inside the IRQ handler. smatch flags this:
drivers/mmc/host/davinci_mmc.c:933 mmc_davinci_irq() error: we
previously assumed 'data' could be null (see line 914)
Gate both branches on a non-NULL data, matching the existing pattern
used by the DATDNE branch.
No functional change for callers where data is non-NULL, which is
the only case in which these branches did meaningful work before
this change.
Signed-off-by: Stepan Ionichev <sozdayvek@gmail.com>
Reviewed-by: Bartosz Golaszewski <bartosz.golaszewski@oss.qualcomm.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
This reverts commit 7f991e3f9b ("nvme: add quirk
NVME_QUIRK_IGNORE_DEV_SUBNQN for 144d:a808")
The incorrect implementations of SUBNQN is a known issue in a massive number of
NVMe units. However, the warning "nvme nvmex: missing or invalid SUBNQN field."
is usually appropriate and will not affect performance or behavior etc. That is
because the support for SUBNQN is mandatory if the controller supports NVMe
revision 1.2.1 or greater, and it reported itself without a SUBNQN field which
breaks compliance with the specification. It should be not quirked by the Linux
Kernel.
Signed-off-by: Alan Cui <me@alancui.cc>
Signed-off-by: Keith Busch <kbusch@kernel.org>
Shivam Kumar found via vulnerability testing:
When data digest is enabled on an NVMe/TCP connection and a digest
mismatch occurs on a non-final H2C_DATA PDU during an R2T-based
data transfer, the digest error handler in nvmet_tcp_try_recv_ddgst()
calls nvmet_req_uninit() — which performs percpu_ref_put() on the
submission queue — but does NOT mark the command as completed. It
does not set cqe->status, does not modify rbytes_done, and does not
clear any flag. When the subsequent fatal error triggers queue
teardown, nvmet_tcp_uninit_data_in_cmds() iterates all commands,
checks nvmet_tcp_need_data_in() for each one, and finds that the
already-uninited command still appears to need data (because
rbytes_done < transfer_len and cqe->status == 0). It therefore calls
nvmet_req_uninit() a second time on the same command — a double
percpu_ref_put against a single percpu_ref_get.
Reported-by: Shivam Kumar <kumar.shivam43666@gmail.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Sagi Grimberg <sagi@grimberg.me>
Signed-off-by: Keith Busch <kbusch@kernel.org>
nvme_free_host_mem() frees dev->hmb_sgt via dma_free_noncontiguous()
but never clears the pointer afterward. This leads to a use-after-free
if nvme_free_host_mem() is called twice in the same error path.
This can happen during nvme_probe() when nvme_setup_host_mem() succeeds
in allocating the HMB (setting dev->hmb_sgt) but nvme_set_host_mem()
fails with an I/O error:
nvme_setup_host_mem()
nvme_alloc_host_mem_single() -> sets dev->hmb_sgt
nvme_set_host_mem() -> fails with -EIO
nvme_free_host_mem() -> frees hmb_sgt, but does NOT NULL it
return error
nvme_probe() error path:
nvme_free_host_mem() -> dev->hmb_sgt is stale, use-after-free
The second call dereferences the freed sgt, causing a NULL pointer
dereference in iommu_dma_free_noncontiguous() when it accesses
sgt->sgl->dma_address (the backing memory has been freed and zeroed).
This is reproducible on Thunderbolt-attached NVMe devices (e.g., OWC
Envoy Express behind a Dell WD22TB4 dock) where the device intermittently
returns I/O errors during HMB setup due to PCIe link instability.
BUG: kernel NULL pointer dereference, address: 0000000000000010
RIP: 0010:iommu_dma_free_noncontiguous+0x22/0x80
Call Trace:
<TASK>
dma_free_noncontiguous+0x3b/0x130
nvme_free_host_mem+0x30/0xf0 [nvme]
nvme_probe.cold+0xcc/0x275 [nvme]
local_pci_probe+0x43/0xa0
pci_device_probe+0xeea/0x290
really_probe+0xf9/0x3b0
__driver_probe_device+0x8b/0x170
driver_probe_device+0x24/0xd0
__driver_attach_async_helper+0x6b/0x110
async_run_entry_fn+0x37/0x170
process_one_work+0x1ac/0x3d0
worker_thread+0x1b8/0x360
kthread+0xf7/0x130
ret_from_fork+0x2d8/0x3a0
ret_from_fork_asm+0x1a/0x30
</TASK>
Fix this by setting dev->hmb_sgt to NULL after freeing it, so the
second call takes the multi-descriptor path which safely handles the
already-cleaned-up state.
Fixes: 63a5c7a4b4 ("nvme-pci: use dma_alloc_noncontigous if possible")
Signed-off-by: Chia-Lin Kao (AceLan) <acelan.kao@canonical.com>
Signed-off-by: Keith Busch <kbusch@kernel.org>
From a security standpoint we should not allow to print out the DH-HMAC-CHAP
secrets, but at the same time having them is useful for debugging
authentication failures.
So add a Kconfig option NVME_TARGET_AUTH_DEBUG to only enable debugging
if explictly requested at build time.
Reviewed-by: Sagi Grimberg <sagi@grimberg.me>
Signed-off-by: Hannes Reinecke <hare@kernel.org>
Signed-off-by: Keith Busch <kbusch@kernel.org>
The local bio is always NULL, so we'd leak the bio if the integrity
mapping failed. Just get it directly from the request.
Fixes: d0d1d52231 ("blk-map: provide the bdev to bio if one exists")
Reviewed-by: Sagi Grimberg <sagi@grimberg.me>
Reviewed-by: John Garry <john.g.garry@oracle.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Keith Busch <kbusch@kernel.org>
The warning is a bit alarming, and it only prints for the very first
non-sgl capable device that receives a passthrough command. Just log an
informational message on initial discovery for every device.
Reviewed-by: Jens Axboe <axboe@kernel.dk>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Keith Busch <kbusch@kernel.org>
