On Thu, Nov 5, 2009 at 9:51 AM, Philip Pokorny
<[email protected]> wrote:
>>
>> [ ... ]
>>
>> [ 8864.326702] <6>[0]: scst_queue_retry_cmd:1099:TGT QUEUE FULL:
>> incrementing retry_cmds 1
>> [ 8864.326709] <6>[0]: scst_queue_retry_cmd:1106:Some command(s) finished,
>> direct retry (finished_cmds=2023031, tgt->finished_cmds=2023137,
>> retry_cmds=0)
>> [ ... ]

(replying to an e-mail of one month ago)

The above issue should be fixed in SCST revision 1391, together with
the following changes:
* Support for OFED 1.5.
* The maximum supported value for the ib_srp kernel module parameter
srp_sg_tablesize has been increased from 58 to 128, at least when
ib_srpt is loaded with default parameters. Modifying the new ib_srpt
kernel module parameter srp_max_rdma_size makes it possible to support
even higher values of srp_sg_tablesize.
* Various small performance improvements.

The results I obtained with QDR HCA's, a RAM disk as target and a
vanilla 2.6.27.39 kernel + CFQ on the initiator system are:
* for asynchronous (buffered) I/O, writing speeds up to 1070 MB/s
(srp_sg_tablesize = 128; block size = 16 KB).
* for asynchronous (buffered) I/O, reading speeds up to 930 MB/s
(srp_sg_tablesize = 12; block size = 16 KB).
* for direct (non-buffered) I/O, writing speeds up to 1000 MB/s
(srp_sg_tablesize = 32; block size = 2 MB).
* for direct (non-buffered) I/O, reading speeds up to 1710 MB/s
(srp_sg_tablesize = 16; block size = 64 MB).

Bart.
--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to [email protected]
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Reply via email to