On 3/2/22 07:03, Christoph Hellwig wrote:
On Tue, Mar 01, 2022 at 09:05:24PM -0800, Bart Van Assche wrote:
Hmm ... even with that patch applied, I still see the crash reported in my
previous email. After I observed that crash I did a clean kernel build to
make sure that the kernel binaries used in my test match the source code.
I still can't reproduce it at all. With this patchset on Jens'
for-5.18/block branch I do get a pre-existing crash in
nvmf_connect_admin_queue, and on Jens' for-next tree that has all the
latest fixes from Linus' tree I only see the CM lockdep warning you
reported.
FYI, this is a branch with the patches applied ontop of the for-next
branch:
http://git.infradead.org/users/hch/block.git/shortlog/refs/heads/freeze-for-next
Hi Christoph,
Thanks for having published a merge of block-for-next and the branch
with this patch series. That makes it easy for me to replicate your
kernel tree. I can reproduce the null-ptr-deref with the freeze-for-next
branch but not with Jens' block-for-next branch (commit e70f36e84f9b
("Merge branch 'for-5.18/block' into for-next")). This is what appears
in the kernel log on my test setup for the freeze-for-next branch
(commit acac349e5516 ("block: move rq_qos_exit() into disk_release()"):
BUG: KASAN: null-ptr-deref in __blk_account_io_start+0x28/0xa0
Maybe we are using different kernel configurations? I'm using
CONFIG_NVME_MULTIPATH=n. I guess that you are using CONFIG_NVME_MULTIPATH=y?
Bart.
