Re: [syzbot] WARNING in trc_read_check_handler

On Sat, Dec 04, 2021 at 10:50:47AM +0100, Dmitry Vyukov wrote:

On Wed, 1 Dec 2021 at 22:09, Paul E. McKenney [off-list ref] wrote:

quoted

On Wed, Dec 01, 2021 at 12:50:07PM -0800, syzbot wrote:

quoted

syzbot suspects this issue was fixed by commit:

commit 96017bf9039763a2e02dcc6adaa18592cd73a39d
Author: Paul E. McKenney [off-list ref]
Date:   Wed Jul 28 17:53:41 2021 +0000

    rcu-tasks: Simplify trc_read_check_handler() atomic operations

bisection log:  https://syzkaller.appspot.com/x/bisect.txt?x=1281d89db00000
start commit:   5319255b8df9 selftests/bpf: Skip verifier tests that fail ..
git tree:       bpf-next
kernel config:  https://syzkaller.appspot.com/x/.config?x=9290a409049988d4
dashboard link: https://syzkaller.appspot.com/bug?extid=fe9d8c955bd1d0f02dc1
syz repro:      https://syzkaller.appspot.com/x/repro.syz?x=14990477300000
C reproducer:   https://syzkaller.appspot.com/x/repro.c?x=105ebd84b00000

If the result looks correct, please mark the issue as fixed by replying with:

#syz fix: rcu-tasks: Simplify trc_read_check_handler() atomic operations

quoted

For information about bisection process see: https://goo.gl/tpsmEJ#bisection

Give or take.  There were quite a few related bugs, so some or all of
the following commits might also have helped:

cbe0d8d91415c rcu-tasks: Wait for trc_read_check_handler() IPIs
18f08e758f34e rcu-tasks: Add trc_inspect_reader() checks for exiting critical section
46aa886c483f5 rcu-tasks: Fix IPI failure handling in trc_wait_for_one_reader

Thanks for checking. If we don't have one exact fix, let's go with
what syzbot suggested. At this point it does not matter much since all
of them are in most trees I assume. We just need to close the bug with
something.

#syz fix: rcu-tasks: Simplify trc_read_check_handler() atomic operations

Fair enough!

quoted

Quibbles aside, it is nice to get an automated email about having fixed
a bug as opposed to having added one.  ;-)

Yes, but one is not possible without the other :-)

But of course it is possible!  For example, syzkaller might find a bug
that was already fixed, and then before notifying me about the bug, you
see the fix.  For example, by failing to reproduce a mainline bug on -rcu.

Not that I particularly want to be auto-spammed about bugs that I have
already fixed, mind you!  ;-)

							Thanx, Paul