Messages in this thread | | | From | Aaron Carroll <> | Date | Thu, 24 Apr 2008 12:07:06 +1000 | Subject | Re: [RFC][PATCH 3/3] Do not use rqhash when merges disabled |
| |
Alan D. Brunelle wrote: > Good catch - that was a last minute addition, the ELV_ON_HASH should be > sufficient without the check for blk_queue_nomerges, right?
Yes. With that line reverted it seems to work fine. FWIW, I'm testing it like so:
=================================x8=======================================
#!/bin/bash
DEV=sdb
echo deadline > /sys/block/${DEV}/queue/scheduler
~/bin-ia64/fio - <<! & [reader] filename=/dev/${DEV} ioengine=sync direct=1 norandommap randrepeat=0 zero_buffers bs=16k rw=randread:16 numjobs=128 time_based runtime=600 !
while /bin/true ; do echo 0 > /sys/block/${DEV}/queue/nomerges sleep 0.1 echo 1 > /sys/block/${DEV}/queue/nomerges sleep 0.1 done
| |