CVE-2026-98072: net/rds: use wq_has_sleeper() in release_in_xmit()
In the Linux kernel, the following vulnerability has been resolved:
net/rds: use wqhassleeper() in releaseinxmit()
releaseinxmit() clears RDSINXMIT with clearbitunlock() and then checks waitqueueactive() to decide whether anyone needs waking. clearbitunlock() is only a release operation: it orders the critical section before the bit clear, but does not order the subsequent plain load of the wait queue head after it. The waiter side does the mirror image - it adds itself to the wait queue and then tests the bit. That is the classic store-buffering pattern: the releasing CPU can read the wait queue as empty while the waiting CPU still reads the bit as set, so the sleeper is never woken.
The waiters are rdsconnshutdown() and rdstcpresetcallbacks(), both in uninterruptible waitevent() with no timeout. A lost wake-up strands the shutdown worker on its single-threaded workqueue until some other sender releases the bit again - and on a connection that is being torn down precisely because it failed, there may never be another sender.
The barrier used to be there: releaseinxmit() did clearbit() followed by smpmbafteratomic() until commit 1422f28826d2 ("rds: introduce acquire/release ordering in acquire/releaseinxmit()") folded both into clearbitunlock(), which strengthened the lock hand-off but silently dropped the full barrier the wake-up check depends on. The refill counterpart, releaserefill() in net/rds/ibrecv.c, still carries its smpmbafteratomic() for exactly this reason.
Use wqhassleeper(), which is waitqueueactive() preceded by the required full barrier.
Affected Software
Remediation
Recommended actions to resolve this vulnerability, in priority order.
- Compensating control
In the Linux kernel RDS code, use wq_has_sleeper() instead of waitqueue_active() in release_in_xmit() when checking whether the wait queue needs waking.
Event History
Frequently Asked Questions
Which operations can block indefinitely?
The affected waiters are rds_conn_shutdown() and rds_tcp_reset_callbacks(). Both use uninterruptible wait_event() calls without a timeout, so a lost wake-up can leave them waiting indefinitely.
What would an affected system look like operationally?
A shutdown worker can become stranded on its single-threaded workqueue while tearing down an RDS connection. This is most likely to persist for a connection being torn down after failure, because there may be no later sender to release the bit and wake the waiter.
Can the blocked worker recover without intervention?
It can be woken if another sender later releases the relevant bit. The description notes that this may never happen for a failed connection that is being torn down.