/proc/scsi/aic7xxx/? considered harmful (2.4.19-pre9)

[Neil Brown]

Hi,
 I have 3 NFS servers with ext3 on raid5 on scsi with assorted
aic7xxx scsi controllers, all running 2.4.19-pre9 (plus some ext3 and
raid and nfs patches) using the "new" aic7xxx drivers.

While trying to diagnose some problems I ran a little script which
extracts the "Commands Queued" value for each drive and prints out
differences every second, so I can watch traffic.

One our newer machine, which reports
Jun  3 17:33:21 eno kernel: scsi0 : Adaptec AIC7XXX EISA/VLB/PCI SCSI HBA DRIVER, Rev 6.2.6
Jun  3 17:33:21 eno kernel:         <Adaptec aic7899 Ultra160 SCSI adapter>
Jun  3 17:33:21 eno kernel:         aic7899: Ultra160 Wide Channel B, SCSI Id=7, 32/253 SCBs
Jun  3 17:33:21 eno kernel: 
Jun  3 17:33:21 eno kernel: scsi1 : Adaptec AIC7XXX EISA/VLB/PCI SCSI HBA DRIVER, Rev 6.2.6
Jun  3 17:33:21 eno kernel:         <Adaptec aic7899 Ultra160 SCSI adapter>
Jun  3 17:33:21 eno kernel:         aic7899: Ultra160 Wide Channel A, SCSI Id=7, 32/253 SCBs
Jun  3 17:33:21 eno kernel: 
Jun  3 17:33:21 eno kernel: scsi2 : Adaptec AIC7XXX EISA/VLB/PCI SCSI HBA DRIVER, Rev 6.2.6
Jun  3 17:33:21 eno kernel:         <Adaptec 29160B Ultra160 SCSI adapter>
Jun  3 17:33:21 eno kernel:         aic7892: Ultra160 Wide Channel A, SCSI Id=7, 32/253 SCBs

This works fine.

On an older machine, which reports

Jan  1 11:01:39 cage kernel: scsi0 : Adaptec AIC7XXX EISA/VLB/PCI SCSI HBA DRIVER, Rev 6.2.6
Jan  1 11:01:39 cage kernel:         <Adaptec 3950B Ultra2 SCSI adapter>
Jan  1 11:01:39 cage kernel:         aic7896/97: Ultra2 Wide Channel A, SCSI Id=7, 32/253 SCBs
Jan  1 11:01:39 cage kernel: 
Jan  1 11:01:39 cage kernel: scsi1 : Adaptec AIC7XXX EISA/VLB/PCI SCSI HBA DRIVER, Rev 6.2.6
Jan  1 11:01:39 cage kernel:         <Adaptec 3950B Ultra2 SCSI adapter>
Jan  1 11:01:39 cage kernel:         aic7896/97: Ultra2 Wide Channel B, SCSI Id=7, 32/253 SCBs
Jan  1 11:01:39 cage kernel: 

I get lots of errors:
Jun  6 09:38:01 cage kernel: scsi1: Signaled a Target Abort
Jun  6 09:38:02 cage kernel: scsi0: PCI error Interrupt at seqaddr = 0x8
Jun  6 09:38:02 cage kernel: scsi0: Signaled a Target Abort
Jun  6 09:38:02 cage kernel: scsi0: PCI error Interrupt at seqaddr = 0x9
Jun  6 09:38:02 cage kernel: scsi0: Signaled a Target Abort
Jun  6 09:38:02 cage kernel: scsi0: PCI error Interrupt at seqaddr = 0x9
Jun  6 09:38:02 cage kernel: scsi0: Signaled a Target Abort
Jun  6 09:38:02 cage kernel: scsi0: PCI error Interrupt at seqaddr = 0x8
Jun  6 09:38:02 cage kernel: scsi0: Signaled a Target Abort
Jun  6 09:38:02 cage kernel: scsi1: PCI error Interrupt at seqaddr = 0x8
Jun  6 09:38:02 cage kernel: scsi1: Signaled a Target Abort
Jun  6 09:38:02 cage kernel: scsi1: PCI error Interrupt at seqaddr = 0x9

but it seems to keep working..

On the last machine, which is similar to the second but only has one
even-older scsi card:

Jan  1 11:10:35 glass kernel: scsi0 : Adaptec AIC7XXX EISA/VLB/PCI SCSI HBA DRIVER, Rev 6.2.6
Jan  1 11:10:35 glass kernel:         <Adaptec 2940 Ultra2 SCSI adapter>
Jan  1 11:10:35 glass kernel:         aic7890/91: Ultra2 Wide Channel A, SCSI Id=7, 32/253 SCBs
Jan  1 11:10:35 glass kernel: 

I get the above errors for about 10 seconds, and then the machine
freezes solid.

I guess I will re-write my script to use /proc/partitions to monitor
disc traffic.

NeilBrown