user/sven/linux.git/kernel, branch v5.17.1

rcu: Don't deboost before reporting expedited quiescent state

2022-03-28T08:03:21Z

commit 10c535787436d62ea28156a4b91365fd89b5a432 upstream. Currently rcu_preempt_deferred_qs_irqrestore() releases rnp->boost_mtx before reporting the expedited quiescent state. Under heavy real-time load, this can result in this function being preempted before the quiescent state is reported, which can in turn prevent the expedited grace period from completing. Tim Murray reports that the resulting expedited grace periods can take hundreds of milliseconds and even more than one second, when they should normally complete in less than a millisecond. This was fine given that there were no particular response-time constraints for synchronize_rcu_expedited(), as it was designed for throughput rather than latency. However, some users now need sub-100-millisecond response-time constratints. This patch therefore follows Neeraj's suggestion (seconded by Tim and by Uladzislau Rezki) of simply reversing the two operations. Reported-by: Tim Murray Reported-by: Joel Fernandes Reported-by: Neeraj Upadhyay Reviewed-by: Neeraj Upadhyay Reviewed-by: Uladzislau Rezki (Sony) Tested-by: Tim Murray Cc: Todd Kjos Cc: Sandeep Patil Cc: # 5.4.x Signed-off-by: Paul E. McKenney Signed-off-by: Greg Kroah-Hartman

configs/debug: restore DEBUG_INFO=y for overriding

2022-03-17T18:02:13Z

Previously, I failed to realize that Kees' patch [1] has not been merged into the mainline yet, and dropped DEBUG_INFO=y too eagerly from the mainline. As the results, "make debug.config" won't be able to flip DEBUG_INFO=n from the existing .config. This should close the gaps of a few weeks before Kees' patch is there, and work regardless of their merging status anyway. Link: https://lore.kernel.org/all/20220125075126.891825-1-keescook@chromium.org/ [1] Link: https://lkml.kernel.org/r/20220308153524.8618-1-quic_qiancai@quicinc.com Signed-off-by: Qian Cai Reported-by: Daniel Thompson Reviewed-by: Daniel Thompson Cc: Kees Cook Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds

Merge branch 'davidh' (fixes from David Howells)

2022-03-11T18:28:32Z

Merge misc fixes from David Howells: "A set of patches for watch_queue filter issues noted by Jann. I've added in a cleanup patch from Christophe Jaillet to convert to using formal bitmap specifiers for the note allocation bitmap. Also two filesystem fixes (afs and cachefiles)" * emailed patches from David Howells : cachefiles: Fix volume coherency attribute afs: Fix potential thrashing in afs writeback watch_queue: Make comment about setting ->defunct more accurate watch_queue: Fix lack of barrier/sync/lock between post and read watch_queue: Free the alloc bitmap when the watch_queue is torn down watch_queue: Fix the alloc bitmap size to reflect notes allocated watch_queue: Use the bitmap API when applicable watch_queue: Fix to always request a pow-of-2 pipe ring size watch_queue: Fix to release page in ->release() watch_queue, pipe: Free watchqueue state after clearing pipe ring watch_queue: Fix filter limit check

watch_queue: Make comment about setting ->defunct more accurate

2022-03-11T18:17:13Z

watch_queue_clear() has a comment stating that setting ->defunct to true preventing new additions as well as preventing notifications. Whilst the latter is true, the first bit is superfluous since at the time this function is called, the pipe cannot be accessed to add new event sources. Remove the "new additions" bit from the comment. Fixes: c73be61cede5 ("pipe: Add general notification queue support") Reported-by: Jann Horn Signed-off-by: David Howells Signed-off-by: Linus Torvalds

watch_queue: Fix lack of barrier/sync/lock between post and read

2022-03-11T18:17:13Z

There's nothing to synchronise post_one_notification() versus pipe_read(). Whilst posting is done under pipe->rd_wait.lock, the reader only takes pipe->mutex which cannot bar notification posting as that may need to be made from contexts that cannot sleep. Fix this by setting pipe->head with a barrier in post_one_notification() and reading pipe->head with a barrier in pipe_read(). If that's not sufficient, the rd_wait.lock will need to be taken, possibly in a ->confirm() op so that it only applies to notifications. The lock would, however, have to be dropped before copy_page_to_iter() is invoked. Fixes: c73be61cede5 ("pipe: Add general notification queue support") Reported-by: Jann Horn Signed-off-by: David Howells Signed-off-by: Linus Torvalds

watch_queue: Free the alloc bitmap when the watch_queue is torn down

2022-03-11T18:17:13Z

Free the watch_queue note allocation bitmap when the watch_queue is destroyed. Fixes: c73be61cede5 ("pipe: Add general notification queue support") Reported-by: Jann Horn Signed-off-by: David Howells Signed-off-by: Linus Torvalds

watch_queue: Fix the alloc bitmap size to reflect notes allocated

2022-03-11T18:17:12Z

Currently, watch_queue_set_size() sets the number of notes available in wqueue->nr_notes according to the number of notes allocated, but sets the size of the bitmap to the unrounded number of notes originally asked for. Fix this by setting the bitmap size to the number of notes we're actually going to make available (ie. the number allocated). Fixes: c73be61cede5 ("pipe: Add general notification queue support") Reported-by: Jann Horn Signed-off-by: David Howells Signed-off-by: Linus Torvalds

watch_queue: Use the bitmap API when applicable

2022-03-11T18:17:12Z

Use bitmap_alloc() to simplify code, improve the semantic and reduce some open-coded arithmetic in allocator arguments. Also change a memset(0xff) into an equivalent bitmap_fill() to keep consistency. Signed-off-by: Christophe JAILLET Signed-off-by: David Howells Signed-off-by: Linus Torvalds

watch_queue: Fix to always request a pow-of-2 pipe ring size

2022-03-11T18:17:12Z

The pipe ring size must always be a power of 2 as the head and tail pointers are masked off by AND'ing with the size of the ring - 1. watch_queue_set_size(), however, lets you specify any number of notes between 1 and 511. This number is passed through to pipe_resize_ring() without checking/forcing its alignment. Fix this by rounding the number of slots required up to the nearest power of two. The request is meant to guarantee that at least that many notifications can be generated before the queue is full, so rounding down isn't an option, but, alternatively, it may be better to give an error if we aren't allowed to allocate that much ring space. Fixes: c73be61cede5 ("pipe: Add general notification queue support") Reported-by: Jann Horn Signed-off-by: David Howells Signed-off-by: Linus Torvalds

watch_queue: Fix to release page in ->release()

2022-03-11T18:17:12Z

When a pipe ring descriptor points to a notification message, the refcount on the backing page is incremented by the generic get function, but the release function, which marks the bitmap, doesn't drop the page ref. Fix this by calling generic_pipe_buf_release() at the end of watch_queue_pipe_buf_release(). Fixes: c73be61cede5 ("pipe: Add general notification queue support") Reported-by: Jann Horn Signed-off-by: David Howells Signed-off-by: Linus Torvalds