Well. We got timeouts and it seems that the behaviour of the driver has
not been as nice as the middle scsi driver was expecting.
Note that I am expecting a middle scsi driver more kind with low-level
drivers about timeout detection and management.
I am interested in the results with tagged queue disabled.
Just for the test and in order to avoid timeouts, you can decrease
SCSI_NCR_MAX_TAGS to 2. The max number of command per lun "cmd_per_lun"
is defined from this constants (even if you do not enable tagged queue).
Under linux, it seems not possible to change this field dynamically at
run time and drivers must choose the max value at initialization time.
(You can too increase the hard-coded timeout value in sd.c).
On Sun, 7 Jul 1996, Alexander Sanda wrote:
> On Sun, 7 Jul 1996, Gerard Roudier wrote:
>
> Allright, I just finished re-testing, here are the results:
>
> 53c7,8xx: 10MHz FAST SCSI-2, disconnect enabled
>
> -------Sequential Output-------- ---Sequential Input-- --Random--
> -Per Char- --Block--- -Rewrite-- -Per Char- --Block--- --Seeks---
> Machine MB K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU /sec %CPU
> 53C7xx 210 3628 67.6 3404 19.6 1549 17.4 3362 53.9 3819 18.1 33.7 1.9
>
>
> 53C8xx: 10MHz, disconnect enabled, TQ4 (see note below)
>
> -------Sequential Output-------- ---Sequential Input-- --Random--
> -Per Char- --Block--- -Rewrite-- -Per Char- --Block--- --Seeks---
> Machine MB K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU /sec %CPU
> 53C8xx 210 3529 69.7 3328 20.8 1481 17.7 3150 51.6 3559 16.4 36.9 2.2
>
>
> I tried to re-test the 53C8xx driver with tagged queuing set to 0, but
> unfortunately, Bonnie crashed during the test. I was able to switch VC,
> but couldn't kill the Bonnie task, which was consuming 99.9% CPU and
> totally hung.
>
> I browsed my kernel-log and found the following:
>
>
> kernel: scsi : aborting command due to timeout : pid 1876, scsi0, channel 0, id 0, lun 0 0x0a 06 62 df 46 00
> kernel: ncr53c8xx_abort : abort call
> kernel: scsi : aborting command due to timeout : pid 1979, scsi0, channel 0, id 0, lun 0 0x0a 06 af 07 f4 00
> kernel: ncr53c8xx_abort : abort call
> kernel: scsi : aborting command due to timeout : pid 1980, scsi0, channel 0, id 0, lun 0 0x0a 06 af fb f4 00
> kernel: ncr53c8xx_abort : abort call
> kernel: scsi : aborting command due to timeout : pid 1876, scsi0, channel 0, id 0, lun 0 0x0a 06 62 df 46 00
> kernel: ncr53c8xx_abort : abort call
> kernel: scsi : aborting command due to timeout : pid 1979, scsi0, channel 0, id 0, lun 0 0x0a 06 af 07 f4 00
> kernel: ncr53c8xx_abort : abort call
> kernel: scsi : aborting command due to timeout : pid 1980, scsi0, channel 0, id 0, lun 0 0x0a 06 af fb f4 00
> kernel: ncr53c8xx_abort : abort call
> kernel: scsi0 channel 0 : resetting for second half of retries.
> kernel: SCSI bus is being reset for host 0 channel 0.
> kernel: ncr53c810-0: restart (scsi bus reset).
> kernel: ncr53c810-0-<target 0, lun 0>: FAST SCSI-2 100ns (10 Mb/sec) offset 8.
> kernel: scsi : aborting command due to timeout : pid 3393, scsi0, channel 0, id 0, lun 0 0x08 09 d3 63 02 00
> kernel: ncr53c8xx_abort : abort call
> kernel: scsi : aborting command due to timeout : pid 1876, scsi0, channel 0, id 0, lun 0 0x0a 06 62 df 46 00
> kernel: ncr53c8xx_abort : abort call
> kernel: scsi : aborting command due to timeout : pid 1876, scsi0, channel 0, id 0, lun 0 0x0a 06 62 df 46 00
> kernel: ncr53c8xx_abort : abort call
> kernel: scsi : aborting command due to timeout : pid 4203, scsi0, channel 0, id 0, lun 0 0x08 09 68 cb 02 00
> kernel: ncr53c8xx_abort : abort call
> kernel: scsi : aborting command due to timeout : pid 4203, scsi0, channel 0, id 0, lun 0 0x08 09 68 cb 02 00
> kernel: ncr53c8xx_abort : abort call
> kernel: scsi0 channel 0 : resetting for second half of retries.
> kernel: SCSI bus is being reset for host 0 channel 0.
> kernel: ncr53c810-0: restart (scsi bus reset).
> kernel: ncr53c810-0-<target 0, lun 0>: FAST SCSI-2 100ns (10 Mb/sec) offset 8.
> kernel: scsi : aborting command due to timeout : pid 4260, scsi0, channel 0, id 0, lun 0 0x08 04 b8 f9 02 00
> kernel: ncr53c8xx_abort : abort call
> kernel: scsi : aborting command due to timeout : pid 4407, scsi0, channel 0, id 0, lun 0 0x08 09 d3 63 02 00
> kernel: ncr53c8xx_abort : abort call
> kernel: scsi : aborting command due to timeout : pid 4444, scsi0, channel 0, id 0, lun 0 0x08 09 d7 d5 02 00
> kernel: ncr53c8xx_abort : abort call
> kernel: scsi : aborting command due to timeout : pid 4515, scsi0, channel 0, id 0, lun 0 0x08 05 13 45 02 00
> kernel: ncr53c8xx_abort : abort call
> kernel: scsi : aborting command due to timeout : pid 4518, scsi0, channel 0, id 0, lun 0 0x0a 04 f2 35 02 00
> kernel: ncr53c8xx_abort : abort call
>
> [continued with some more lines like the above]
>
>
> # /AS/ powered by #
> # http://www.ping.at/members/alexa/ Linux 2.0 #
> -----------------------------------------------------------------------
>
Gerard.