What | Removed | Added |
---|---|---|
CC | tiwai@suse.com |
Gah, that's bad. It's a kernel BUG() that was introduced by 4.4.21 stable, the backport of commit 9006a01829a50cfd6bbd4980910ed46e895e93d7 hrtimer: Catch illegal clockids But the commit was later reverted by the commit 82e88ff1ea948d83125a8aaa7c9809f03ccc500f hrtimer: Revert CLOCK_MONOTONIC_RAW support Revert commits: a6e707ddbdf1: KVM: arm/arm64: timer: Switch to CLOCK_MONOTONIC_RAW 9006a01829a5: hrtimer: Catch illegal clockids 9c808765e88e: hrtimer: Add support for CLOCK_MONOTONIC_RAW Marc found out, that there are fundamental issues with that patch series because __hrtimer_get_next_event() and hrtimer_forward() need support for CLOCK_MONOTONIC_RAW. Nothing which is easily fixed, so revert the whole lot. But this fix doesn't seem taken by the upstream. Interestingly, the check was revived by the later commit 336a9cde10d641e70bac67d90ae91b3190c3edca hrtimer: Catch invalid clockids again but now with a WARN() instead of BUG(). So we'll see the kernel Oops but not a panic, at least. The real culprit is, however, the call site in mac80211_hwsim.c. This was fixed by the commit 8fbcfeb8a9cc803464d6c166e7991913711c612c mac80211_hwsim: Replace bogus hrtimer clockid This has to be backported.