Re: dm: retain stacked max_sectors when setting queue_limits

[Mike Snitzer <snitzer@kernel.org>]

On Thu, May 23, 2024 at 04:49:38PM +0200, Christoph Hellwig wrote:
> On Thu, May 23, 2024 at 10:12:24AM -0400, Mike Snitzer wrote:
> > Not sure what is sketchy about the max_sectors == 0 check, the large
> > comment block explains that check quite well.  We want to avoid EIO
> > for unsupported operations (otherwise we'll get spurious path failures
> > in the context of dm-multipath).  Could be we can remove this check
> > after an audit of how LLD handle servicing IO for unsupported
> > operations -- so best to work through it during a devel cycle.
> 
> Think of what happens if you create a dm device, and then reduce
> max_sectors using sysfs on the lower device after the dm device
> was created: you'll trivially trigger this check.
> 
> > Not sure why scsi_debug based testing with mptest isn't triggering it
> > for you. Are you seeing these limits for the underlying scsi_debug
> > devices?
> > 
> > ./max_hw_sectors_kb:2147483647
> > ./max_sectors_kb:512
> 
> root@testvm:~/mptest# cat /sys/block/sdc/queue/max_hw_sectors_kb 
> 2147483647
> 
> root@testvm:~/mptest# cat /sys/block/sdd/queue/max_sectors_kb 
> 512
> 
> root@testvm:~/mptest# cat /sys/block/dm-0/queue/max_hw_sectors_kb 
> 2147483647
> 
> root@testvm:~/mptest# cat /sys/block/dm-0/queue/max_sectors_kb 
> 1280
> 
> so they don't match, but for some reason bigger bios never get built.

Weird... I'm running in a VMware guest but I don't see why that'd make
a difference on larger IOs being formed (given it is virtual
scsi_debug devices).

In any case, we know I can reproduce with this scsi_debug-based mptest
test and Marco has verified my fix resolves the issue on his FC
multipath testbed.

But I've just floated a patch to elevate the fix to block core (based
on Ming's suggestion):
https://patchwork.kernel.org/project/dm-devel/patch/Zk9i7V2GRoHxBPRu@kernel.org/

Let me know, thanks.