CVE-2026-98072

In the Linux kernel, the following vulnerability has been resolved:

net/rds: use wq has sleeper() in release in xmit()

release in xmit() clears RDS IN XMIT with clear bit unlock() and then checks waitqueue active() to decide whether anyone needs waking. clear bit unlock() 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 rds conn shutdown() and rds tcp reset callbacks(), both in uninterruptible wait event() 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: release in xmit() did clear bit() followed by smp mb after atomic() until commit 1422f28826d2 ("rds: introduce acquire/release ordering in acquire/release in xmit()") folded both into clear bit unlock(), which strengthened the lock hand-off but silently dropped the full barrier the wake-up check depends on. The refill counterpart, release refill() in net/rds/ib recv.c, still carries its smp mb after atomic() for exactly this reason.

Use wq has sleeper(), which is waitqueue active() preceded by the required full barrier.