aic7xxx v 6.0.8 BETA looses device

[Dr. Michael Weller]

Dear people,
thx for this beta.

This is about 6.0.8beta on vanilla linux 2.2.18 running a
2940u2w interacting with two compaq disks and a:

  Vendor: FUJITSU   Model: M2512A            Rev: 1403
  Type:   Direct-Access                      ANSI SCSI revision: 01

MO disk (which can also do scsi-2).



With all driver version up to the Adaptec supported ones, I had
massive problems with this removable disk.

The driver apparently 'gobbled' a few CheckConditions and the Media change
mode sense with unsurprising disastrous results. If it happened once the
device didn't report any media changes anymore. A remove and add single
device to the kernel did repair this.

I can no longer reproduce this with the 6.0.8beta. (which is good).

>From paging through the patch I got the impression the controller now
reads mode sense on its own and store it with the SCBs (if so, it's good).
(maybe that's the point with the MO?)

However two other issues:

Reading /proc/scsi/aic7xxx/0 give segmentation fault and kernel internal
divide by zero:

----
Jan 13 18:16:52 lorena kernel: divide error: 0000 Jan 13 18:16:52 lorena
kernel: CPU:    0
Jan 13 18:16:52 lorena kernel: EIP:    0010:[scsi_calc_syncsrate+77/88]
Jan 13 18:16:52 lorena kernel: EFLAGS: 00010246
Jan 13 18:16:52 lorena kernel: eax: 00989680   ebx: c0268c18   ecx:
00989680   edx: 00000000
Jan 13 18:16:52 lorena kernel: esi: 00000000   edi: c0268c14   ebp:
00000018   esp: d4b13d90
Jan 13 18:16:52 lorena kernel: ds: 0018   es: 0018   ss: 0018
Jan 13 18:16:52 lorena kernel: Process cat (pid: 286, process nr: 40,
stackpage=d4b13000)
Jan 13 18:16:52 lorena kernel: Stack: d7fef559 d4b13f20 00000000 c01ea59f
00000000 d7fef54c 00000004 d4b13f20
Jan 13 18:16:52 lorena kernel:        d7fd1c00 00009c40 c01ea713 d4b13f20
d7fef559 d4b13f20 c0240e6b d4b13f20
Jan 13 18:16:52 lorena kernel:        c0240e40 00000041 00000004 00000004
d7fd1c00 0000000f d4b13f20 00000007
Jan 13 18:16:52 lorena kernel: Call Trace: [ahc_format_transinfo+39/300]
[ahc_dump_target_state+111/212] [min_sb_len.500+5931/6684]
[min_sb_len.500+5888/6684] [aic7xxx_proc_info+281/332]
[get_new_inode+173/292] [iget4+110/120]
Jan 13 18:16:52 lorena kernel:        [iget+19/24] [proc_get_inode+62/240]
[proc_lookup+172/188] [dispatch_scsi_info+142/164] [proc_readscsi+151/296]
[sys_read+174/196] [system_call+52/56]
Jan 13 18:16:52 lorena kernel: Code: f7 f6 89 c1 5b 5e 5f 5d 59 c3 90 83
ec 04 55 57 56 31 f6 53
-----

I assume it's a simple printing/formatting error.


However, causing this bug and especially provoking error or timeout
on the removable device (attempt to access w/o a disk or mount while
disk is still spinning up) renders the device completely unaccessible.

Unaccessible means any mount attempt yields:
/dev/sdc is not a valid block device.

I can no longer remove it with remove/add-single device on
/proc/scsi/scsi.

The device remains unaccessible until reboot. No errors (even timeout or
disk busy) are printed to the logs. Please advice how to debug this.

Regards,
Michael.


--

Michael Weller: eowmob at exp-math.uni-essen.de, eowmob at ms.exp-math.uni-essen.de,
or even mat42b at spi.power.uni-essen.de. If you encounter an eowmob account on
any machine in the net, it's very likely it's me.



To Unsubscribe: send mail to majordomo at FreeBSD.org
with "unsubscribe aic7xxx" in the body of the message