Re: dm-io: reject unsupported DISCARD/WRITE SAME requests with EOPNOTSUPP

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



On Fri, Feb 13, 2015 at 02:21:01PM -0500, Martin K. Petersen wrote:
> >>>>> "Mike" == Mike Snitzer <snitzer@xxxxxxxxxx> writes:
> 
> Mike> When I implemented dm_table_supports_discards() I consciously
> Mike> allowed a DM table to contain a mix of discard support.  I'm now
> Mike> wondering where it is we benefit from that?  Seems like more of a
> Mike> liability than anything -- so a bigger hammer approach to fixing
> Mike> this would be to require all targets and all devices in a DM table
> Mike> support discard.
> 
> I think our original rationale was that since discard is only a hint it
> would be fine to mix and match. And at the time there seemed to be value
> in supporting a heterogeneous setups with say a disk drive and an SSD.
> 
> Back then the SSD vendors were all busy telling us how crucial discard
> would be going forward. However, that turned out not to be the case and
> discard often causes more problems than it solves. So I'm perfectly OK
> with requiring all devices in a table to have the same capabilities. In
> many ways I think that's a cleaner approach.

I agree that imposing that extra requirement would be cleaner from a software
engineering POV.

That said, given that discard is advisory and most of the callers seem to
anticipate that it might not work, why not allow heterogeneous mixes?  It seems
unfortunate to remove that ability just because there are kernel bugs.  If
you're implementing thin provisioning and are unmapping storage when discard
requests come through, wouldn't it be an antifeature that this suddenly stops
just because something got in the way?  fstrim ought to be able to talk to
the parts of the compound device that can be trimmed.

Second question: Can a dm device detect that q->limits.max_discard_sectors has
changed in one of the devices it's sitting on?  Say I have this:

X -> Y -> SSD1
 \-> Z -> SSD2

SSD*, Z, Y, and X all advertise discard.

Then I change Y to point to a spinny disk:

X -> Y -> HDD
 \-> Z -> SSD2

Even if Y notices that it no longer supports discard, will X figure that out
too?

--D

> 
> -- 
> Martin K. Petersen	Oracle Linux Engineering

--
dm-devel mailing list
dm-devel@xxxxxxxxxx
https://www.redhat.com/mailman/listinfo/dm-devel




[Index of Archives]     [DM Crypt]     [Fedora Desktop]     [ATA RAID]     [Fedora Marketing]     [Fedora Packaging]     [Fedora SELinux]     [Yosemite Discussion]     [KDE Users]     [Fedora Docs]

  Powered by Linux