This fixes a small typo from when the code was copied from raid1 and removes a
loop break when a read is atomic on only 'near' layouts. 'Far' layouts never
triggered this break and so it would spread all read I/Os to all available
devices. This removes that break since it is contrary to the comments previous
to it and works as intended in the 'far' layout. After this change random I/O
for 'near' layout is distributed to all disks instead of being all serviced by
the first disk. Sequential I/O is not affected as it would not trigger the
break.
i.e. Before patch
# mdadm --create /dev/md14 --level 10 --raid-devices 4 -p n4 /dev/loop{11..14}
# fio -rw=randread --size=5G --name=mdadm_test
<snip>
Disk stats (read/write):
md14: ios=1304718/0, merge=0/0, ticks=0/0, in_queue=0, util=0.00%,
aggrios=327680/0, aggrmerge=0/0, aggrticks=18719/0, aggrin_queue=18689,
aggrutil=88.37%
loop13: ios=0/0, merge=0/0, ticks=0/0, in_queue=0, util=0.00%
loop11: ios=1310108/0, merge=0/0, ticks=74856/0, in_queue=74736, util=88.37%
loop14: ios=0/0, merge=0/0, ticks=0/0, in_queue=0, util=0.00%
loop12: ios=612/0, merge=0/0, ticks=20/0, in_queue=20, util=0.02%
After patch:
<snip>
Disk stats (read/write):
md14: ios=1309172/1, merge=0/0, ticks=0/0, in_queue=0, util=0.00%,
aggrios=327680/3, aggrmerge=0/0, aggrticks=19866/29, aggrin_queue=19871,
aggrutil=37.54%
loop13: ios=331904/3, merge=0/0, ticks=21184/24, in_queue=21180, util=24.64%
loop11: ios=330075/3, merge=0/0, ticks=32252/24, in_queue=32248, util=37.54%
loop14: ios=327612/3, merge=0/0, ticks=3204/24, in_queue=3200, util=3.73%
loop12: ios=321129/3, merge=0/0, ticks=22824/44, in_queue=22856, util=26.59%
--
To unsubscribe from this list: send the line "unsubscribe linux-raid" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
