UFS performance oddities
I've tried installing kFreeBSD on an eee 701 laptop I had spare. While
it runs fine and the installation went flawlessly, it took about three
hours to do the bare-bones installation...
Investigation shows that something very odd is going on with the
filesystem. While dd to files shows respectable performance to the eee's
internal SSD, doing anything with metadata is painful.
$ time mkdir {1,2,3,4,5,6,7,8,9,0}{1,2,3,4,5,6,7,8,9,0}{1,2,3,4,5,6,7,8,9,0}
real 1m44.084s
$ time rmdir {1,2,3,4,5,6,7,8,9,0}{1,2,3,4,5,6,7,8,9,0}{1,2,3,4,5,6,7,8,9,0}
real 0m30.655s
My suspicion is that there's something cache-related with the SSD, in
that it's doing all metadata writes synchronously. Perhaps the SSD
doesn't support write barriers?
On Linux I'd diagnose this with hdparm and kernel messages, but FreeBSD
doesn't have hdparm and the kernel is much less wordy that Linux is, so
I don't see any indications of what's going on. All I see in the system
logs is:
kernel: ada0 at ata1 bus 0 scbus1 target 0 lun 0
kernel: ada0: <SILICONMOTION SM223AC > ATA-4 device
kernel: ada0: 66.700MB/s transfers (UDMA4, PIO 512bytes)
kernel: ada0: 3815MB (7815024 512 byte sectors: 16H 63S/T
7753C)
kernel: ada0: Previously was known as ad2
[...]
kernel: Trying to mount root from ufs:/dev/ad2s2 [rw]...
The next stage is to use a vanilla FreeBSD live image on the eee to try
and figure out whether this is a generic FreeBSD problem or something
specific to the kFreeBSD port; but in the mean time, any advice as to
how to figure this out?
--
┌─── ｄｇ＠ｃｏｗｌａｒｋ．ｃｏｍ ───── http://www.cowlark.com ─────
│ "Parents let children ride bicycles on the street. But parents do not
│ allow children to hear vulgar words. Therefore we can deduce that
│ cursing is more dangerous than being hit by a car." --- Scott Adams
Reply to: