Re: [PATCH 5.13 000/800] 5.13.2-rc1 review

[Greg Kroah-Hartman]

On Wed, Jul 14, 2021 at 02:07:10PM +0000, Holger Kiehl wrote:
> On Wed, 14 Jul 2021, Greg Kroah-Hartman wrote:
> 
> > On Wed, Jul 14, 2021 at 01:26:26PM +0000, Holger Kiehl wrote:
> > > On Wed, 14 Jul 2021, Holger Kiehl wrote:
> > > 
> > > > On Wed, 14 Jul 2021, Greg Kroah-Hartman wrote:
> > > > 
> > > > > On Wed, Jul 14, 2021 at 05:39:43AM +0000, Holger Kiehl wrote:
> > > > > > Hello,
> > > > > > 
> > > > > > On Mon, 12 Jul 2021, Greg Kroah-Hartman wrote:
> > > > > > 
> > > > > > > This is the start of the stable review cycle for the 5.13.2 release.
> > > > > > > There are 800 patches in this series, all will be posted as a response
> > > > > > > to this one.  If anyone has any issues with these being applied, please
> > > > > > > let me know.
> > > > > > > 
> > > > > > > Responses should be made by Wed, 14 Jul 2021 06:02:46 +0000.
> > > > > > > Anything received after that time might be too late.
> > > > > > > 
> > > > > > With this my system no longer boots:
> > > > > > 
> > > > > >    [  OK  ] Reached target Swap.
> > > > > >    [   75.213852] NMI watchdog: Watchdog detected hard LOCKUP on cpu 0
> > > > > >    [   75.213926] NMI watchdog: Watchdog detected hard LOCKUP on cpu 2
> > > > > >    [   75.213962] NMI watchdog: Watchdog detected hard LOCKUP on cpu 4
> > > > > >    [FAILED] Failed to start Wait for udev To Complete Device Initialization.
> > > > > >    See 'systemctl status systemd-udev-settle.service' for details.
> > > > > >             Starting Activation of DM RAID sets...
> > > > > >    [      ] (1 of 2) A start job is running for Activation of DM RAID sets (..min ..s / no limit)
> > > > > >    [      ] (2 of 2) A start job is running for Monitoring of LVM2 mirrors, snapshots etc. using dmeventd or progress polling (..min ..s / no limit)
> > > > > > 
> > > > > > System is a Fedora 34 with all updates applied. Two other similar
> > > > > > systems with AMD CPUs (Ryzen 4750G + 3400G) this does not happen
> > > > > > and boots fine. The system where it does not boot has an Intel
> > > > > > Xeon E3-1285L v4 CPU. All of them use a dm_crypt root filesystem.
> > > > > > 
> > > > > > Any idea which patch I should drop to see if it boots again. I already
> > > > > > dropped
> > > > > > 
> > > > > >    [PATCH 5.13 743/800] ASoC: Intel: sof_sdw: add quirk support for Brya and BT-offload
> > > > > > 
> > > > > > and I just see that this one should also be dropped:
> > > > > > 
> > > > > >    [PATCH 5.13 768/800] hugetlb: address ref count racing in prep_compound_gigantic_page
> > > > > > 
> > > > > > Will still need to test this.
> > > > > 
> > > > > Can you run 'git bisect' to see what commit causes the problem?
> > > > > 
> > > > Yes, will try to do that. I think it will take some time ...
> > > > 
> > > With the help of Pavel Machek and Jiri Slaby I was able 'git bisect'
> > > this to:
> > > 
> > >    yoda:/usr/src/kernels/linux-5.13.y# git bisect good
> > >    a483f513670541227e6a31ac7141826b8c785842 is the first bad commit
> > >    commit a483f513670541227e6a31ac7141826b8c785842
> > >    Author: Jan Kara <jack@xxxxxxx>
> > >    Date:   Wed Jun 23 11:36:33 2021 +0200
> > > 
> > >        bfq: Remove merged request already in bfq_requests_merged()
> > > 
> > >        [ Upstream commit a921c655f2033dd1ce1379128efe881dda23ea37 ]
> > > 
> > >        Currently, bfq does very little in bfq_requests_merged() and handles all
> > >        the request cleanup in bfq_finish_requeue_request() called from
> > >        blk_mq_free_request(). That is currently safe only because
> > >        blk_mq_free_request() is called shortly after bfq_requests_merged()
> > >        while bfqd->lock is still held. However to fix a lock inversion between
> > >        bfqd->lock and ioc->lock, we need to call blk_mq_free_request() after
> > >        dropping bfqd->lock. That would mean that already merged request could
> > >        be seen by other processes inside bfq queues and possibly dispatched to
> > >        the device which is wrong. So move cleanup of the request from
> > >        bfq_finish_requeue_request() to bfq_requests_merged().
> > > 
> > >        Acked-by: Paolo Valente <paolo.valente@xxxxxxxxxx>
> > >        Signed-off-by: Jan Kara <jack@xxxxxxx>
> > >        Link: https://lore.kernel.org/r/20210623093634.27879-2-jack@xxxxxxx
> > >        Signed-off-by: Jens Axboe <axboe@xxxxxxxxx>
> > >        Signed-off-by: Sasha Levin <sashal@xxxxxxxxxx>
> > > 
> > >     block/bfq-iosched.c | 41 +++++++++++++----------------------------
> > >     1 file changed, 13 insertions(+), 28 deletions(-)
> > > 
> > > Holger
> > 
> > Wonderful!
> > 
> > So if you drop that, all works well?  I'll go drop that from the queues
> > now.
> > 
> Yes. Just double checked it took a plain 5.13.1, patched it with
> patch-5.13.2-rc1.xz and then reverted
> 
>    PATCH-5.13-259-800-bfq-Remove-merged-request-already-in-bfq_requests_merged
> 
> and it booted fine with no problems. Tested several times.
> Just wonder why it only happens on the Intel Broadwell CPU.
> Maybe it is the 128MB eDRAM L4 Cache ...

Wondeful!

Could you test 5.14-rc1 to verify if this problem is there or not?  If
it is, the developers need to know this so that they can work to fix the
regression.

thanks,

greg k-h