| CVE |
Vendors |
Products |
Updated |
CVSS v3.1 |
| In the Linux kernel, the following vulnerability has been resolved:
octeontx2-pf: avoid double free of pool->stack on AQ init failure
otx2_pool_aq_init() frees pool->stack when mailbox sync or retry
allocation fails, but leaves the pointer unchanged. Later,
otx2_sq_aura_pool_init() unwinds the partial setup through
otx2_aura_pool_free(), which frees pool->stack again. The CN20K-specific
cn20k_pool_aq_init() implementation has the same bug in
its corresponding error path.
Set pool->stack to NULL immediately after the local free so the shared
cleanup path does not free the same stack again while cleaning up
partially initialized pool state.
The bug was first flagged by an experimental analysis tool we are
developing for kernel memory-management bugs while analyzing
v6.13-rc1. The tool is still under development and is not yet publicly
available. Manual inspection confirms that the bug is still present in
v7.1-rc3.
Runtime validation was not performed because reproducing this path
requires OcteonTX2/CN20K hardware. |
| In the Linux kernel, the following vulnerability has been resolved:
fuse: clear intr_entry in fuse_resend and fuse_remove_pending_req
When fuse_resend() moves a request from fpq->processing back to
fiq->pending, it sets FR_PENDING and clears FR_SENT but does not
remove the requests intr_entry from fiq->interrupts. If the
request had FR_INTERRUPTED set from a prior signal, intr_entry
remains dangling on fiq->interrupts. When the requesting task
then receives a fatal signal, fuse_remove_pending_req() sees
FR_PENDING=1, removes the request from fiq->pending and frees it
via the refcount path, also without cleaning intr_entry. The
stale intr_entry causes use-after-free when fuse_read_interrupt()
iterates fiq->interrupts:
- list_del_init(&req->intr_entry) -> UAF write on freed slab
- req->in.h.unique -> UAF read, data leaked to userspace
Remove intr_entry from fiq->interrupts in fuse_resend() for
interrupted requests before they are placed back on fiq->pending.
Add a WARN_ON if the intr_entry is not empty on request destruction. |
| In the Linux kernel, the following vulnerability has been resolved:
crypto: qat - remove unused character device and IOCTLs
The QAT driver exposes a character device (qat_adf_ctl) with IOCTLs
for device configuration, start, stop, status query and enumeration.
These IOCTLs are not part of any public uAPI header and have no known
in-tree or out-of-tree users. Device lifecycle is already managed via
sysfs.
The ioctl interface also increases the attack surface and is the
subject of a number of bug reports.
Remove the character device, the IOCTL definitions, and the related
data structures (adf_dev_status_info, adf_user_cfg_key_val,
adf_user_cfg_section, adf_user_cfg_ctl_data). Drop the now-unused
adf_cfg_user.h header and strip adf_ctl_drv.c down to the minimal
module_init/module_exit hooks for workqueue, AER, and crypto/compression
algorithm registration.
Clean up leftover dead code that was only reachable from the removed
IOCTL paths: adf_cfg_del_all(), adf_devmgr_verify_id(),
adf_devmgr_get_num_dev(), adf_devmgr_get_dev_by_id(),
adf_get_vf_real_id() and the unused ADF_CFG macros.
Additionally, drop the entry associated to QAT IOCTLs in
ioctl-number.rst. |
| In the Linux kernel, the following vulnerability has been resolved:
iio: event: Fix event FIFO reset race
`iio_event_getfd()` creates the event file descriptor with
`anon_inode_getfd()`, which allocates a new fd, creates the anonymous
file and installs it in the process fd table before returning to the
caller.
The IIO code resets the event FIFO after `anon_inode_getfd()` has returned,
but before `IIO_GET_EVENT_FD_IOCTL` has copied the fd number to userspace.
But since fd tables are shared between threads, another thread can guess
the newly allocated fd number and issue a `read()` on it as soon as the fd
has been installed.
This means the `kfifo_to_user()` in `iio_event_chrdev_read()` can run in
parallel with the `kfifo_reset_out()` in `iio_event_getfd()`.
The kfifo documentation says that `kfifo_reset_out()` is only safe when it
is called from the reader thread and there is only one concurrent reader.
Otherwise it is dangerous and must be handled in the same way as
`kfifo_reset()`.
If that happens, `kfifo_to_user()` can advance the FIFO `out` index based
on state from before the reset, after the reset has already moved the `out`
index to the current `in` index. That can leave the FIFO with an `out`
index past the `in` index. A later `read()` can then see an underflowed
FIFO length and copy more data than the event FIFO buffer contains. This
can result in an out-of-bounds read and leak adjacent kernel memory to
userspace.
Move the FIFO reset before `anon_inode_getfd()`. At that point the event fd is
marked busy, but the new fd has not been installed yet, so userspace cannot
access it while the FIFO is reset. |
| In the Linux kernel, the following vulnerability has been resolved:
6lowpan: fix NHC entry use-after-free on error path
lowpan_nhc_do_uncompression() looks up an NHC descriptor while holding
lowpan_nhc_lock. If the descriptor has no uncompress callback, the error
path drops the lock before printing nhc->name.
lowpan_nhc_del() removes descriptors under the same lock and then relies
on synchronize_net() before the owning module can be unloaded. That only
waits for net RX RCU readers. lowpan_header_decompress() is also exported
and can be reached from callers that are not necessarily covered by the net
core RX critical section, for example the Bluetooth 6LoWPAN L2CAP receive
path.
This leaves a race where one task drops lowpan_nhc_lock in the error path,
another task unregisters and frees the matching descriptor after
synchronize_net() returns, and the first task then dereferences nhc->name
for the warning.
With the post-unlock window widened, KASAN reports:
BUG: KASAN: slab-use-after-free in lowpan_nhc_do_uncompression+0x1f4/0x220
Read of size 8
lowpan_nhc_do_uncompression
lowpan_header_decompress
Fix this by printing the warning before dropping lowpan_nhc_lock, so the
descriptor name is read while unregister is still excluded. The malformed
packet is still rejected with -ENOTSUPP. |
| In the Linux kernel, the following vulnerability has been resolved:
staging: vme_user: bound slave read/write to the kern_buf size
The SLAVE-path helpers buffer_to_user() and buffer_from_user() copy
'count' bytes into/out of the fixed-size kern_buf (size_buf ==
PCI_BUF_SIZE == 0x20000, 128 KiB) using *ppos as the offset, without
bounding *ppos + count against size_buf.
vme_user_write()/vme_user_read() only clamp count to the VME window size
(image_size = vme_get_size(resource)), which VME_SET_SLAVE sets from the
user-supplied slave.size -- validated against the VME address space (up
to VME_A32_MAX = 4 GiB), not against PCI_BUF_SIZE. When the window
exceeds 128 KiB, a write()/read() copies past the kern_buf allocation.
Clamp count against size_buf in both helpers, with an early return when
*ppos is already at/after the buffer end. *ppos is >= 0 here (the caller
rejects negative offsets), so size_buf - *ppos cannot wrap. This mirrors
the existing clamp in the MASTER-path helpers resource_to_user() /
resource_from_user(), and matches the read()/write() convention of a
short transfer at end-of-buffer.
Found by static analysis (CodeQL taint tracking + CBMC bounded model
checking) and confirmed dynamically under KASAN with the vme_fake bridge:
BUG: KASAN: slab-out-of-bounds in _copy_from_user+0x2d/0x80
Write of size 262144 at addr ffff888004100000 by task trigger/68
_copy_from_user+0x2d/0x80
vme_user_write+0x13e/0x240 [vme_user]
vfs_write+0x1b8/0x7a0
ksys_write+0xb8/0x150 |
| In the Linux kernel, the following vulnerability has been resolved:
staging: media: ipu7: fix double-free and use-after-free in error paths
In both ipu7_isys_init() and ipu7_psys_init(), pdata is allocated and
then passed to ipu7_bus_initialize_device(), which stores it in
adev->pdata. The ipu7_bus_release() function frees adev->pdata when the
device's reference count drops to zero.
Two error paths incorrectly call kfree(pdata) after the device teardown
has already freed it:
1. When ipu7_mmu_init() fails: put_device() is called, which drops the
reference count to zero and triggers ipu7_bus_release() ->
kfree(pdata). The subsequent kfree(pdata) is a double-free.
2. When ipu7_bus_add_device() fails: it calls auxiliary_device_uninit()
internally, which calls put_device() -> ipu7_bus_release() ->
kfree(pdata). The subsequent kfree(pdata) is again a double-free.
Note that the kfree(pdata) when ipu7_bus_initialize_device() itself
fails is correct, because in that case auxiliary_device_init() failed
and the release function was never set up, so pdata must be freed
manually.
Additionally, the error code was not saved before calling put_device(),
causing ERR_CAST() to dereference the already-freed adev pointer when
constructing the return value. Fix this by saving the error from
dev_err_probe() before put_device() and returning ERR_PTR() instead.
Remove the redundant kfree(pdata) calls and fix the use-after-free in
the return values of the two affected error paths. |
| In the Linux kernel, the following vulnerability has been resolved:
staging: rtl8723bs: fix OOB read in OnAssocRsp() IE loop
The IE parsing loop in OnAssocRsp() advances by (pIE->length + 2) each
iteration but only guards on i < pkt_len. When a malicious AP sends an
AssocResponse whose last IE has only one byte remaining in the frame
(the element_id byte lands at pkt_len-1), the loop reads pIE->length
from pframe[pkt_len], which is one byte past the allocated receive buffer.
Additionally, even when the header bytes are in bounds, pIE->length
itself can extend the data window beyond pkt_len, silently passing a
truncated IE to the handler functions.
Add two guards at the top of the loop body:
1. Break if fewer than sizeof(*pIE) bytes remain (can't read header).
2. Break if the IE's declared data extends past pkt_len. |
| In the Linux kernel, the following vulnerability has been resolved:
staging: rtl8723bs: fix OOB reads in rtw_get_sec_ie(), rtw_get_wapi_ie(), and rtw_get_wps_attr()
Three IE/attribute parsing functions have missing bounds checks.
rtw_get_sec_ie() and rtw_get_wapi_ie() iterate over a raw IE buffer
without verifying that the header bytes (tag + length) are within the
remaining buffer before reading them. Additionally, rtw_get_sec_ie()
compares the 4-byte WPA OUI at cnt+2 without checking that at least
6 bytes remain, and rtw_get_wapi_ie() compares a 4-byte WAPI OUI at
cnt+6 without checking that at least 10 bytes remain.
rtw_get_wps_attr() reads wps_ie[0] and wps_ie+2 unconditionally at
entry, before verifying that wps_ielen is large enough to contain
the 6-byte WPS IE header (element_id + length + 4-byte OUI). Inside
the attribute loop, get_unaligned_be16() is called on attr_ptr and
attr_ptr+2 without checking that 4 bytes remain in the buffer.
Add a cnt+2 bounds check before each loop body in rtw_get_sec_ie()
and rtw_get_wapi_ie(), guard each multi-byte comparison with a minimum
IE length requirement, add a wps_ielen < 6 early return in
rtw_get_wps_attr(), and add a 4-byte bounds check in its inner loop. |
| In the Linux kernel, the following vulnerability has been resolved:
ksmbd: fix use-after-free of a deferred file_lock on SMB2_CLOSE then SMB2_CANCEL
Commit f580d27e8928 ("ksmbd: fix use-after-free of a deferred file_lock on
double SMB2_CANCEL") made smb2_cancel() skip a work whose state is
KSMBD_WORK_CANCELLED, so its cancel_fn cannot be fired a second time. But
KSMBD_WORK has three states (ACTIVE, CANCELLED, CLOSED), and the same
freeing producer path is reached for CLOSED too:
SMB2_CLOSE on the locking handle -> set_close_state_blocked_works() sets
the deferred work's state to KSMBD_WORK_CLOSED and wakes the smb2_lock()
worker. The worker takes the non-ACTIVE early-exit, locks_free_lock()s
the file_lock and, because the state is not KSMBD_WORK_CANCELLED, takes
the STATUS_RANGE_NOT_LOCKED branch with "goto out2" -- which, like the
cancelled branch, skips release_async_work(). The work stays on
conn->async_requests with a live cancel_fn = smb2_remove_blocked_lock
pointing at the freed file_lock.
A subsequent SMB2_CANCEL for the same AsyncId then passes the
KSMBD_WORK_CANCELLED-only guard (its state is KSMBD_WORK_CLOSED), so
smb2_cancel() fires cancel_fn again over the freed file_lock -- the same
use-after-free fixed, via SMB2_CLOSE instead of a first SMB2_CANCEL:
BUG: KASAN: slab-use-after-free in __locks_delete_block
__locks_delete_block
locks_delete_block
ksmbd_vfs_posix_lock_unblock
smb2_remove_blocked_lock
smb2_cancel <- 2nd SMB2_CANCEL fires cancel_fn
handle_ksmbd_work
Allocated by ...: locks_alloc_lock <- smb2_lock
Freed by ...: locks_free_lock <- smb2_lock (non-ACTIVE early-exit)
... cache file_lock_cache of size 192
Reproduced on mainline 7.1-rc7 (which already contains f580d27e8928) with
KASAN by an authenticated SMB client; the double-SMB2_CANCEL control is
silent on that kernel, so the splat is attributable to the CLOSE trigger.
Only an ACTIVE deferred work may have its cancel_fn fired: both terminal
states (CANCELLED and CLOSED) reach the smb2_lock() early-exit that frees
the file_lock and skips release_async_work(). Guard on KSMBD_WORK_ACTIVE
so any non-active work is skipped. |
| In the Linux kernel, the following vulnerability has been resolved:
audit: Fix data races of skb_queue_len() readers on audit_queue
Multiple readers access audit_queue.qlen via skb_queue_len() without
holding the queue lock or using READ_ONCE(), while kauditd writes to
this field via the skb_dequeue() → __skb_unlink() path with WRITE_ONCE()
protected by a spinlock. This constitutes data races.
All affected skb_queue_len(&audit_queue) call sites:
- kauditd_thread() wait_event_freezable() condition
- audit_receive_msg() AUDIT_GET handler (s.backlog assignment)
- audit_receive() backlog check
- audit_log_start() backlog check and pr_warn()
KCSAN reports the following conflicting access pattern (one example):
==================================================================
BUG: KCSAN: data-race in audit_log_start / skb_dequeue
write (marked) to 0xffffffff8512ee20 of 4 bytes by task 661 on cpu 57:
skb_dequeue+0x70/0xf0
kauditd_send_queue+0x71/0x220
kauditd_thread+0x1cb/0x430
kthread+0x1c2/0x210
ret_from_fork+0x162/0x1a0
ret_from_fork_asm+0x1a/0x30
read to 0xffffffff8512ee20 of 4 bytes by task 36586 on cpu 1:
audit_log_start+0x2a0/0x6b0
audit_core_dumps+0x64/0xa0
do_coredump+0x14b/0x1260
get_signal+0xeb2/0xf70
arch_do_signal_or_restart+0x41/0x170
exit_to_user_mode_loop+0xa2/0x1c0
do_syscall_64+0x1a3/0x1c0
entry_SYSCALL_64_after_hwframe+0x76/0xe0
value changed: 0x00000001 -> 0x00000000
==================================================================
Resolve the race by switching to lockless helper skb_queue_len_lockless(),
which internally uses READ_ONCE() and properly pairs with the WRITE_ONCE()
write accesses already present on the writer side.
[PM: line length tweak] |
| In the Linux kernel, the following vulnerability has been resolved:
mm: shrinker: fix shrinker_info teardown race with expansion
expand_shrinker_info() iterates all visible memcgs under shrinker_mutex,
including memcgs that have not finished ->css_online() yet.
Once pn->shrinker_info has been published, teardown must stay serialized
with expand_shrinker_info() until that memcg is either fully online or no
longer visible to iteration. Today alloc_shrinker_info() breaks that rule
by dropping shrinker_mutex before freeing a partially initialized
shrinker_info array, which may cause the following race:
CPU0 CPU1
==== ====
css_create
--> list_add_tail_rcu(&css->sibling, &parent_css->children);
online_css
--> mem_cgroup_css_online
--> alloc_shrinker_info
--> alloc node0 info
rcu_assign_pointer(C->node0->shrinker_info, old0)
alloc node1 info -> FAIL -> goto err
mutex_unlock(shrinker_mutex)
shrinker_alloc()
--> shrinker_memcg_alloc
--> mutex_lock(shrinker_mutex)
expand_shrinker_info
--> mem_cgroup_iter see the memcg
expand_one_shrinker_info
--> old0 = C->node0->shrinker_info
memcpy(new->unit, old0->unit, ...);
free_shrinker_info
--> kvfree(old0);
/* double free !! */
kvfree_rcu(old0, rcu);
The same problem exists later in mem_cgroup_css_online(). If
alloc_shrinker_info() succeeds but a subsequent objcg allocation fails,
the free_objcg -> free_shrinker_info() unwind path tears down the already
published pn->shrinker_info arrays without shrinker_mutex. The
expand_one_shrinker_info() can race with that teardown in the same way,
leading to use-after-free or double-free of the old shrinker_info.
Fix this by serializing shrinker_info teardown with shrinker_mutex, and by
keeping alloc_shrinker_info() error cleanup inside the locked section. |
| In the Linux kernel, the following vulnerability has been resolved:
coresight: ultrasoc-smb: Fix OOB write in smb_sync_perf_buffer()
When the SMB sink is used as a perf AUX sink, smb_update_buffer() calls
smb_sync_perf_buffer() to copy hardware trace data into the perf AUX ring
buffer pages. It derives pg_idx = head >> PAGE_SHIFT from @head, which is
handle->head, and indexes dst_pages[pg_idx]. The pg_idx %= nr_pages
normalization is only applied after the first loop iteration.
This leaves the initial page index underived from the buffer size, which
can result in an out-of-bounds write past dst_pages[] when head exceeds
the AUX buffer size.
Normalize head modulo the AUX buffer size before deriving the page index
and offset, mirroring tmc_etr_sync_perf_buffer(). |
| In the Linux kernel, the following vulnerability has been resolved:
ksmbd: require source read access for duplicate extents
FSCTL_DUPLICATE_EXTENTS_TO_FILE passes the source file directly to
vfs_clone_file_range() or vfs_copy_file_range() without checking the SMB
access mask granted to the source handle. A handle opened with attribute
access can consequently be used to copy file contents into an
attacker-readable destination.
Require FILE_READ_DATA on the source handle before either VFS operation,
matching other ksmbd data-copy paths. |
| In the Linux kernel, the following vulnerability has been resolved:
smb: client: mask server-provided mode to 07777 in modefromsid
When modefromsid is active, parse_dacl() applies the server-provided
sub_auth[2] value from the NFS mode SID to cf_mode without masking to
07777. Apply the correct masking, same as in the read path. |
| In the Linux kernel, the following vulnerability has been resolved:
HID: hid-goodix-spi: validate report size to prevent stack buffer overflow
goodix_hid_set_raw_report() builds a protocol frame in a 128-byte stack
buffer (tmp_buf), writing an 11-12 byte header followed by the
caller-supplied report data. The HID core caps report size at
HID_MAX_BUFFER_SIZE (16384) by default, while the driver does not set
hid_ll_driver.max_buffer_size and performs no bounds checking before
copying the payload:
memcpy(tmp_buf + tx_len, buf, len);
A hidraw SET_REPORT ioctl with a report larger than ~116 bytes
overflows the stack buffer.
Add a size check after constructing the header, rejecting reports that
would exceed the buffer capacity.
Discovered by Atuin - Automated Vulnerability Discovery Engine. |
| In the Linux kernel, the following vulnerability has been resolved:
partitions: aix: bound the pp_count scan to the ppe array
aix_partition() reads the physical volume descriptor into a fixed-size
struct pvd and then scans its physical-partition-extent array:
int numpps = be16_to_cpu(pvd->pp_count);
...
for (i = 0; i < numpps; i += 1) {
struct ppe *p = pvd->ppe + i;
...
lp_ix = be16_to_cpu(p->lp_ix);
pvd points at a single kmalloc()'d struct pvd whose ppe[] member holds a
fixed ARRAY_SIZE(pvd->ppe) (1016) entries, but the loop runs up to the
on-disk pp_count. pp_count is an unvalidated __be16 read straight from
the descriptor, so a crafted AIX image with pp_count larger than 1016
drives the loop to read pvd->ppe[i] past the end of the allocation (up
to 65535 entries, ~2 MB out of bounds).
The partition scan runs without mounting anything, when a block device
with a crafted AIX/IBM partition table appears (an attacker-supplied
image attached with losetup -P, or a device auto-scanned by udev), via
msdos_partition() -> aix_partition().
Clamp the scan to the number of entries the ppe[] array can hold. |
| In the Linux kernel, the following vulnerability has been resolved:
exfat: bound uniname advance in exfat_find_dir_entry()
In exfat_find_dir_entry(), each TYPE_EXTEND (file name) entry advances the
output pointer by a fixed amount while the loop guard only tracks the
accumulated name length:
if (++order == 2)
uniname = p_uniname->name;
else
uniname += EXFAT_FILE_NAME_LEN;
len = exfat_extract_uni_name(ep, entry_uniname);
name_len += len;
unichar = *(uniname+len);
*(uniname+len) = 0x0;
uniname grows by EXFAT_FILE_NAME_LEN (15) per name entry, but name_len
grows only by the actual extracted length, which is shorter when a name
fragment contains an early NUL. The only guard is
`name_len >= MAX_NAME_LENGTH`, so a crafted directory with many short
name fragments lets uniname run far past the
p_uniname->name[MAX_NAME_LENGTH + 3] buffer while name_len stays small,
causing an out-of-bounds read and write at *(uniname+len).
The sibling extractor exfat_get_uniname_from_ext_entry() already stops
on a short fragment (the lockstep `len != EXFAT_FILE_NAME_LEN` guard
added in commit d42334578eba ("exfat: check if filename entries exceeds
max filename length")); exfat_find_dir_entry() never got the
equivalent. Track the per-entry write offset as a count and reject a
fragment once the offset, or the offset plus the extracted length, would
exceed MAX_NAME_LENGTH, before forming the output pointer. |
| In the Linux kernel, the following vulnerability has been resolved:
iommufd: Use sizeof(*hdr) instead of sizeof(hdr) in veventq read
The bound-check in iommufd_veventq_fops_read() for the normal vEVENT
path uses sizeof(hdr) where the surrounding code uses sizeof(*hdr):
if (!vevent_for_lost_events_header(cur) &&
sizeof(hdr) + cur->data_len > count - done) {
hdr is declared as struct iommufd_vevent_header *, so sizeof(hdr)
evaluates to the size of the pointer. Surrounding code uses
sizeof(*hdr) consistently:
if (done >= count || sizeof(*hdr) > count - done) {
...
if (copy_to_user(buf + done, hdr, sizeof(*hdr))) {
...
done += sizeof(*hdr);
struct iommufd_vevent_header is currently 8 bytes (two __u32 fields,
flags and sequence), so on 64-bit (sizeof(void *) == 8) the two
expressions happen to be equal and the check works as intended.
On 32-bit (sizeof(void *) == 4) the check under-counts the header by
4 bytes: a vEVENT whose data_len causes 8 + cur->data_len to exceed
count - done while 4 + cur->data_len does not will pass the check,
then the loop will copy_to_user 8 bytes of header followed by data_len
bytes of payload, writing past the user-supplied buffer.
It is also a latent bug for any future expansion of struct
iommufd_vevent_header beyond sizeof(void *) on 64-bit; the check
should not depend on the type happening to match the host pointer
width.
Use sizeof(*hdr) to match the rest of the function and the actual
amount that will be copied. |
| In the Linux kernel, the following vulnerability has been resolved:
svcrdma: wake sq waiters when the transport closes
Threads parked in svc_rdma_sq_wait() on sc_sq_ticket_wait or
sc_send_wait can hang indefinitely in TASK_UNINTERRUPTIBLE state
across transport teardown, pinning svc_xprt references and
blocking svc_rdma_free().
The close path sets XPT_CLOSE before invoking xpo_detach and both
wait_event predicates include an XPT_CLOSE term, but the
predicates are re-evaluated only on wakeup. sc_sq_ticket_wait has
no completion-driven wake path; it is advanced solely by the
chained ticket handoff inside svc_rdma_sq_wait() itself. Without
an explicit wake at close, parked threads never observe
XPT_CLOSE, hold their svc_xprt_get reference forever, and
svc_rdma_free() blocks on xpt_ref dropping to zero.
Two close entry points reach this transport. Local teardown runs
svc_rdma_detach() from svc_handle_xprt() -> svc_delete_xprt() ->
xpo_detach() on a worker thread. A remote disconnect arrives at
svc_rdma_cma_handler(), which calls svc_xprt_deferred_close():
that sets XPT_CLOSE and enqueues the transport but does not
access either RDMA waitqueue, so a worker already parked in
svc_rdma_sq_wait() never re-evaluates its predicate. With every
worker parked on this transport, no thread is available to run
the local teardown either, and the wake site there is
unreachable.
Introduce svc_rdma_xprt_deferred_close(), a thin svcrdma wrapper
that calls svc_xprt_deferred_close() and then wakes both
sc_sq_ticket_wait and sc_send_wait. Convert the svcrdma producers
that called svc_xprt_deferred_close() directly:
svc_rdma_cma_handler(), qp_event_handler(),
svc_rdma_post_send_err(), svc_rdma_wc_send(), the sendto drop
path, the rw completion error paths, and the recvfrom flush and
read-list error paths.
Wake both waitqueues from svc_rdma_detach() as well. The
synchronous svc_xprt_close() path (backchannel ENOTCONN, device
removal via svc_rdma_xprt_done) reaches detach without flowing
through svc_xprt_deferred_close() and therefore does not invoke
the new helper.
[ cel: add svc_rdma_xprt_deferred_close() to complete the fix ] |
