Re: nvme-tls failing with (61b0ed3be81b6) nvme-tcp: parallelize I/O queue allocation and startup

[Chris Leech <cleech@redhat.com>]

On Thu, Nov 12, 2026 at 02:35:23PM -0800, Surabhi Gogte (she/her) wrote:
> On Thu, Nov 12, 2026 at 10:46 AM Chris Leech <cleech@redhat.com> wrote:
> >
> > Hello,
> >
> > I was testing some patchs and ran into trouble with nvme-tls that I
> > bisected back to this change (61b0ed3be81b6 nvme-tcp: parallelize I/O
> > queue allocation and startup).
> >
> > I haven't determined what exactly is happening yet, but blktests
> > nvme/062 and nvme/063 both fail with this applied and pass again if I
> > revert it.
> 
> Hi Chris,
> Can you check if the failures you saw are due to an older version of
> ktls-utils/tlshd being used?
> Ref: https://lore.kernel.org/linux-nvme/CAMof9cU5s0ck20fcSAYSvK-xpcP5_Hv2BmN20Bh3btpLum7THg@mail.gmail.com/.

Thanks for this. I should have found that, but was coming to the idea
that it might be related to having an older tlshd anyway.

> If the tests fail with ktls-utils 1.5.0 as well, please share further
> details about the blktest failure logs and kernel logs so I can also
> take a look.

The initial environment I was bisecting in still seemed to be having
issues, but a clean Fedora 44 + ktls-utils 1.5.0 does look to be working
as expected.  I'll spend some time trying to sort out if I've made a
mess of this system in some other way.

- Chris