** Description changed:
[SRU Justification]
[Impact]
The current epoll implementation in the 5.15 kernel utilizes a read-write
semaphore (rwlock_t) to protect the ready event list. While this allows
multiple producers to concurrently add items, it introduces a scheduling
priority inversion vulnerability.
If a high-priority consumer (such as a real-time thread calling epoll_wait) is
blocked waiting for the exclusive write lock, it can be indefinitely stalled
by
a low-priority producer holding the read lock. This results in
un-deterministic
system stalls and latency spikes.
The fix involves replacing rwlock_t with a standard spinlock_t one-to-one, and
removing the now-redundant lockless helper functions (list_add_tail_lockless
and chain_epi_lockless). This ensures that under real-time configurations,
priority inheritance works correctly across the epoll subsystem, eliminating
priority inversion.
[Fix]
Backport upstream commit:
0c43094f8cc9 ("eventpoll: Replace rwlock with spinlock")
[Test Plan]
Due to the nature of scheduling priority inversion, reproducing this bug
reliably on demand is highly impractical. Because this race condition relies
on erratic, non-deterministic scheduling micro-windows, a standard
deterministic reproduction script cannot be provided.
Therefore, validation relies on verifying that the replacement locking
mechanism functions correctly, introduces no regressions, and scales safely
under synthetic load.
- There is a test kernel available in the following ppa:
+ There is a test kernel available in the following PPA:
https://launchpad.net/~munirsid/+archive/ubuntu/lp2154194
[Where Problems Could Occur]
- There is a trade-off in raw throughput for highly specific, synthetic
workloads.
- As seen in the upstream commit description [0], in artificial benchmarks where
- hundreds of threads continuously spam epoll events, throughput can drop by
~38%
- due to serialization around the new spinlock.
+ There could be a performance degradation with highly specific, synthetic
+ workloads on the GA kernel. As seen in the upstream commit description [0],
+ in artificial benchmarks where hundreds of threads continuously spam epoll
+ events, throughput can drop due to serialization around the new spinlock.
However, testing with realistic workloads (via perf bench epoll wait) actually
demonstrates a performance improvement on x86 architectures.
- The regression potential for real-world production environments is considered
- low, as typical workloads do not exhibit continuous, uninterrupted
- event-spamming behavior. Moreover, the fix is strictly isolated to
- fs/eventpoll.c and alters no external kernel APIs.
+ The regression potential for real-world production environments is low, as
+ typical workloads do not exhibit continuous, uninterrupted event-spamming
+ behavior. Moreover, the fix is strictly isolated to fs/eventpoll.c and alters
+ no external kernel APIs.
[Other Info]
This bug was addressed upstream and has already been integrated into Noble and
subsequent releases.
[0] -
https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git/commit/?id=0c43094f8cc9d3d99d835c0ac9c4fe1ccc62babd
--
You received this bug notification because you are a member of Ubuntu
Bugs, which is subscribed to Ubuntu.
https://bugs.launchpad.net/bugs/2154194
Title:
[Jammy] Priority inversion problem in epoll for rt kernel
To manage notifications about this bug go to:
https://bugs.launchpad.net/ubuntu/+source/linux/+bug/2154194/+subscriptions
--
ubuntu-bugs mailing list
[email protected]
https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs
