The 2.6.10-gentoo-r4 kernel still has the LSI logic SCSI regression that has been present since 2.6.9 the problem exists when there are Single Ended devices on an LVD bus (which is allowed, all devices on that channel will run in SE mode) Messages are being reported by syslog anytime there is access to an SE device. Jan 10 01:54:16 shrapnel sym1:1:0:phase change 6-7 11@2fd27f84 resid=6. Jan 10 01:54:16 shrapnel sym1:1:0:phase change 6-7 11@2fd27f84 resid=6. Jan 10 01:54:16 shrapnel sym1:1:0:phase change 6-7 11@2fd27f84 resid=6. Jan 10 01:54:16 shrapnel sym1:1:0:phase change 6-7 11@2fd27f84 resid=6. A run of hdparm -t of al lmy scsi disks shows that the LVD devices (on channel 1) all are running at their max speed but the SE devices are running at 1/4 speed (I'm only getting 3MB/sec whereas with the 2.6.8-gentoo-r10 kernel they are getting 14-18MB/sec) Reproducible: Always Steps to Reproduce: 1. install 2.6.10-r4 kernel 2. reboot 3. Actual Results: scsi performance for SE devices on an LVD bus goes to snot... Expected Results: SE devices should have performed normally and hteerror messages should not have happened... SCSI controller: 0000:02:04.0 SCSI storage controller: LSI Logic / Symbios Logic 53c1010 Ultra3 SCSI Adapter (rev 01) Flags: bus master, medium devsel, latency 72, IRQ 16 I/O ports at 2000 Memory at e9009800 (64-bit, non-prefetchable) [size=1K] Memory at e9004000 (64-bit, non-prefetchable) [size=8K] Capabilities: <available only to root> 0000:02:04.1 SCSI storage controller: LSI Logic / Symbios Logic 53c1010 Ultra3 SCSI Adapter (rev 01) Flags: bus master, medium devsel, latency 72, IRQ 16 I/O ports at 2400 Memory at e9009c00 (64-bit, non-prefetchable) [size=1K] Memory at e9006000 (64-bit, non-prefetchable) [size=8K] Capabilities: <available only to root> boot messages for kernel 2.6.10-r4 for scsi controller Jan 10 01:54:16 shrapnel sym0: <1010-33> rev 0x1 at pci 0000:02:04.0 irq 16 Jan 10 01:54:16 shrapnel sym0: Symbios NVRAM, ID 7, Fast-80, LVD, parity checking Jan 10 01:54:16 shrapnel sym0: open drain IRQ line driver, using on-chip SRAM Jan 10 01:54:16 shrapnel sym0: using LOAD/STORE-based firmware. Jan 10 01:54:16 shrapnel sym0: handling phase mismatch from SCRIPTS. Jan 10 01:54:16 shrapnel sym0: SCSI BUS has been reset. Jan 10 01:54:16 shrapnel scsi0 : sym-2.1.18m SCSI disks: Attached devices: Host: scsi0 Channel: 00 Id: 00 Lun: 00 Vendor: FUJITSU Model: MAG3182L SUN18G Rev: 1111 Type: Direct-Access ANSI SCSI revision: 02 Host: scsi0 Channel: 00 Id: 02 Lun: 00 Vendor: FUJITSU Model: MAG3182L SUN18G Rev: 1111 Type: Direct-Access ANSI SCSI revision: 02 Host: scsi1 Channel: 00 Id: 01 Lun: 00 Vendor: IBM Model: DXHS18Y Rev: 0430 Type: Direct-Access ANSI SCSI revision: 03 Host: scsi1 Channel: 00 Id: 02 Lun: 00 Vendor: SEAGATE Model: SX118202LS Rev: B808 Type: Direct-Access ANSI SCSI revision: 02 Host: scsi1 Channel: 00 Id: 04 Lun: 00 Vendor: GENERIC Model: CRD-BP3 Rev: 1.03 Type: CD-ROM ANSI SCSI revision: 02 SCSI1, id's 1,2 and 4 are all SE devices these are the ones where performance goes to shit with a 2.6.10 kernel
There is a patch out at: http://marc.theaimsgroup.com/?l=linux-scsi&m=110435131708275&w=2 That does fix it, but it doesn't apply cleanly to the already patched kernel. I did it by hand to my own and it works and solves the problem... This should already be on it's way into mainline acording to the LSI Logic scsi developer...
Get this patch into upstream first please.