Hello,
On Fri, 10 Aug 2012, Shaohua Li wrote:
This is the patchset to support MD discard, which I refresh against latest
kernel. It's pretty straightforward for raid linear/0/1/10. The raid456 discard
support is tricky, please see the log in the patches for details.
Previously when I posted the patches out, people complain SATA SSD discard
request merge error. Now a patch to disable discard request merge is in Jens's
tree. We don't have this issue any more.
First many thanks for sending those updated patches. An also many thanks
to Neil that you still maintain MD for linux! Thank you for all the great
work!
I have tried the raid linear/0/1/10 patches on 3.6.0-rc3 and I sometimes
get this in /var/log/messages:
Aug 27 21:44:53 yoda kernel: [10576.964134] request botched: dev sda: type=1, flags=916c081
Aug 27 21:44:53 yoda kernel: [10576.964138] sector 316197889, nr/cnr 0/256
Aug 27 21:44:53 yoda kernel: [10576.964140] bio ffff8803741b8600, biotail ffff8803741b8600, buffer (null), len 0
Aug 27 21:44:53 yoda kernel: [10576.971023] request botched: dev sdb: type=1, flags=917c081
Aug 27 21:44:53 yoda kernel: [10576.971026] sector 320574465, nr/cnr 0/384
Aug 27 21:44:53 yoda kernel: [10576.971029] bio ffff8803741b9e00, biotail ffff8803741b9e00, buffer (null), len 0
Aug 27 21:44:53 yoda kernel: [10576.972736] request botched: dev sdb: type=1, flags=916c081
Aug 27 21:44:53 yoda kernel: [10576.972740] sector 320575489, nr/cnr 0/288
Aug 27 21:44:53 yoda kernel: [10576.972742] bio ffff8803ff80e840, biotail ffff8803ff80e840, buffer (null), len 0
Aug 27 21:44:53 yoda kernel: [10576.973851] request botched: dev sda: type=1, flags=917c081
Aug 27 21:44:53 yoda kernel: [10576.973854] sector 320575105, nr/cnr 0/640
Aug 27 21:44:53 yoda kernel: [10576.973857] bio ffff8803741b9e80, biotail ffff8803741b9e80, buffer (null), len 0
This happened on a Raid0 when I deleted lots of files (make clean in
kernel directory). The Raid0 is accross two SSD's:
/dev/sda Crucial M4-CT512M4SSD2 Firmware Revision: 000F
/dev/sdb Crucial M4-CT256M4SSD2 Firmware Revision: 000F
cat /proc/mounts for the partition looks as follows:
/dev/md1 / ext4 rw,noatime,discard,journal_checksum,journal_async_commit,commit=600,stripe=256,data=ordered 0 0
And here cat /proc/mdstat:
Personalities : [raid0]
md1 : active raid0 sdb3[1] sda3[0]
413851648 blocks super 1.2 512k chunks
unused devices: <none>
However, when I apply this patch which Shaohua Li send to me on 13th March
(http://lkml.indiana.edu/hypermail/linux/kernel/1203.1/03323.html) these
messages do NOT show up:
--- linux-3.6-rc3/include/linux/blkdev.h.original 2012-08-22 22:29:06.000000000 +0200
+++ linux-3.6-rc3/include/linux/blkdev.h 2012-08-27 21:42:19.979566802 +0200
@@ -601,7 +601,7 @@
* it already be started by driver.
*/
#define RQ_NOMERGE_FLAGS \
- (REQ_NOMERGE | REQ_STARTED | REQ_SOFTBARRIER | REQ_FLUSH | REQ_FUA)
+ (REQ_NOMERGE | REQ_STARTED | REQ_SOFTBARRIER | REQ_FLUSH | REQ_FUA | REQ_DISCARD)
#define rq_mergeable(rq) \
(!((rq)->cmd_flags & RQ_NOMERGE_FLAGS) && \
(((rq)->cmd_flags & REQ_DISCARD) || \
Regards,
Holger
--
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
