Re: problem booting 5.10

[John Garry]

On 08/12/2020 19:19, Linus Torvalds wrote:
> On Tue, Dec 8, 2020 at 10:59 AM Martin K. Petersen
> <martin.petersen@xxxxxxxxxx> wrote:
> > > So I'm adding SCSI people to the cc, just in case they go "Hmm..".
> >
> > Only change in this department was:
> >
> > 831e3405c2a3 scsi: core: Don't start concurrent async scan on same host
>
> Yeah, I found that one too, and dismissed it for the same reason you
> did - it wasn't in rc1. Plus it looked very simple.
>
> That said, maybe Julia might have misspoken, and rc1 was ok, so I
> guess it's possible. The scan_mutex does show up in that "locks held"
> list, although I can't see why it would matter. But it does
> potentially change timing (so it could expose some existing race), if
> nothing else.
>
> But let's make sure Jens is aware of this too, in case it's some ATA
> issue. Not that any of those handful of 5.10 changes look remotely
> likely _either_.
>
> Jens, see
>
>     https://lore.kernel.org/lkml/alpine.DEB.2.22.394.2012081813310.2680@hadrien/
>
> if you don't already have the lkml thread locally.. There's not enough
> of the dmesg to even really guess what Julia's actual hardware is,
> apart from it being a Seagate SATA disk. Julia? What controllers and
> disks do you have show up when things work?
>
>               Linus
> .

JFYI, About "scsi: megaraid_sas: Added support for shared host tagset 
for cpuhotplug", we did have an issue reported here already from Qian 
about a boot hang:

https://lore.kernel.org/linux-scsi/fe3dff7dae4494e5a88caffbb4d877bbf472dceb.camel@xxxxxxxxxx/

And the solution to that specific problem is in:
https://lore.kernel.org/linux-block/20201203012638.543321-1-ming.lei@xxxxxxxxxx/

This issue may be related, so you could test by reverting that megaraid 
sas commit or setting the driver module param "host_tagset_enable=0" 
just to see.

Thanks,
John