bioq_disksort() on SSDs

[Bryan Venteicher]

Hi,

At $JOB, we've noticed bioq_disksort() creep up in the profiling
output when the bio queue gets backed up at our caching SSD drives.
Thoughts on the attached patch which does away with the disksort
for SCSI da* and ATA ada* SSD drives?

Support for quirks and sysctl/tunables to force the behavior either
way are desirable but not implemented in this patch. Nor is some
mechanism to pass this information up to GEOM (to say gsched). And
perhaps this should done fairer to writes as DragonflyBSD's
bioqdisksort() does.

(I've only been able to test an similar patch adapted to 8.x and this
patch on a virtual machine.)

http://www.daemoninthecloset.org/~bryanv/patches/freebsd/ssd_disksort.patch
-------------- next part --------------
A non-text attachment was scrubbed...
Name: ssd_disksort.patch
Type: text/x-patch
Size: 10144 bytes
Desc: not available
URL: <http://lists.freebsd.org/pipermail/freebsd-scsi/attachments/20121017/c07cc9ed/attachment.bin>