Re: [PATCH 0/2] ublk: fix race between io_uring_cmd_complete_in_task and ublk_cancel_cmd

[Ming Lei <ming.lei@redhat.com>]

On Fri, Apr 25, 2025 at 12:10:31AM +0300, Jared Holzman wrote:
> On 23/04/2025 12:24, Ming Lei wrote:
> > Hello Jens,
> > 
> > The 2 patches try to fix race between between io_uring_cmd_complete_in_task
> > and ublk_cancel_cmd, please don't apply until Jared verifies them.
> > 
> > Jared, please test and see if the two can fix your crash issue on v6.15-rc3.
> > 
> > If you can't reproduce it on v6.15-rc3, please backport them to v6.14, and I
> > can help the backport if you need.
> > 
> > Thanks,
> > Ming
> > 
> > Ming Lei (2):
> >   ublk: call ublk_dispatch_req() for handling UBLK_U_IO_NEED_GET_DATA
> >   ublk: fix race between io_uring_cmd_complete_in_task and
> >     ublk_cancel_cmd
> > 
> >  drivers/block/ublk_drv.c | 51 ++++++++++++++++++++++++++--------------
> >  1 file changed, 34 insertions(+), 17 deletions(-)
> > 
> 
> Hi Ming,
> 
> It's a solid fix. I ran it through our automation and it passed 300 iterations without an issue. Previously we were getting crash after less than 20.
> 

Jared, thanks for the test and feedback!


> I also back-ported the patches to 6.14 and it works there too.
> 
> Will these fixes make it into 6.15? Or only 6.16?

I think it is fine for v6.15 if Jens agrees.

> 
> Also is there a 6.14 maintenance branch that could also be fixed or is it end-of-life already?

Both the two are simple enough, which may be backported to any stable tree,
IMO.


thanks,
Ming