Bug in the sg driver

Bug in the sg driver

[Dag Nygren]

Hi,

just got back from a customer with major problems
with a LTO-drive and an Adaptec 19160.

The problems was ones that HP claimed a firmware
upgrade would fix and even gave a tool for doing this:
hp_ltt.

Running this the first time would always segfault and trying
a second time would consistently panic the system.

We traced the segfault to sg_ioctl trying to do something.
After finding a vague hint with google I tried to boot with
mem=512M (The machine has 2GB of memory) and voila:
The update worked without any crashes.
We don't know yet if the firmware update fixed the original problem,
but the conclusion is:

sg_ioctl seems to address illegal parts of memory when used with
kernels where highmem is enabled.

Any sg-driver maintainers out there?

Best Regards

-- 
Dag Nygren                               email: dag@newtech.fi
Oy Espoon NewTech Ab                     phone: +358 9 8024910
Träsktorpet 3                              fax: +358 9 8024916
02360 ESBO                              Mobile: +358 400 426312
FINLAND

Re: Bug in the sg driver

[Randy.Dunlap]

On Mon, 06 Oct 2003 21:06:36 +0300 Dag Nygren <dag@newtech.fi> wrote:

| 
| Hi,
| 
| just got back from a customer with major problems
| with a LTO-drive and an Adaptec 19160.
| 
| The problems was ones that HP claimed a firmware
| upgrade would fix and even gave a tool for doing this:
| hp_ltt.
| 
| Running this the first time would always segfault and trying
| a second time would consistently panic the system.
| 
| We traced the segfault to sg_ioctl trying to do something.
| After finding a vague hint with google I tried to boot with
| mem=512M (The machine has 2GB of memory) and voila:
| The update worked without any crashes.
| We don't know yet if the firmware update fixed the original problem,
| but the conclusion is:
| 
| sg_ioctl seems to address illegal parts of memory when used with
| kernels where highmem is enabled.
| 
| Any sg-driver maintainers out there?


from MAINTAINERS file:
SCSI SG DRIVER
P:	Doug Gilbert
M:	dgilbert@interlog.com
L:	linux-scsi@vger.kernel.org
W:	http://www.torque.net/sg
S:	Maintained


Any details, like kernel version, oops or panic logs, etc.?

--
~Randy

Re: Bug in the sg driver

[Dag Nygren]

> 
> from MAINTAINERS file:
> SCSI SG DRIVER
> P:	Doug Gilbert
> M:	dgilbert@interlog.com
> L:	linux-scsi@vger.kernel.org
> W:	http://www.torque.net/sg
> S:	Maintained

Should I send a bug report to him or is he online here?

> Any details, like kernel version, oops or panic logs, etc.?

Kernel version is 2.4.20 (Redhat 9.0 + newest upgrade)

The oops and the panic logs were not written down as the major focus
was getting this (production) system up and running, sorry for that

 Dag

Re: Bug in the sg driver

[Randy.Dunlap]

On Mon, 06 Oct 2003 21:47:42 +0300 Dag Nygren <dag@newtech.fi> wrote:

| > 
| > from MAINTAINERS file:
| > SCSI SG DRIVER
| > P:	Doug Gilbert
| > M:	dgilbert@interlog.com
| > L:	linux-scsi@vger.kernel.org
| > W:	http://www.torque.net/sg
| > S:	Maintained
| 
| Should I send a bug report to him or is he online here?

Um, if he's around, he'll see it, although ISTR that he's away
for awhile (like 6 weeks ?).

| > Any details, like kernel version, oops or panic logs, etc.?
| 
| Kernel version is 2.4.20 (Redhat 9.0 + newest upgrade)
| 
| The oops and the panic logs were not written down as the major focus
| was getting this (production) system up and running, sorry for that

Well, unless someone happens to know something about your exact
reported problem....
in general, you will get better help/responses the better your
problem reports are (IMHO).

--
~Randy

Re: Bug in the sg driver

[Dag Nygren]

> | > Any details, like kernel version, oops or panic logs, etc.?
> | 
> | Kernel version is 2.4.20 (Redhat 9.0 + newest upgrade)
> | 
> | The oops and the panic logs were not written down as the major focus
> | was getting this (production) system up and running, sorry for that
> 
> Well, unless someone happens to know something about your exact
> reported problem....
> in general, you will get better help/responses the better your
> problem reports are (IMHO).

Perfectly realize that. Found one report of the exact same problem
on Google. The URL is: 

http://groups.google.fi/groups?q=hp_ltt+linux&hl=sv&lr=&ie=UTF-8&selm=zDJq.3O5.
11%40gated-at.bofh.it&rnum=1

The Oops trace for that is:
kernel BUG in header file at line 162
kernel BUG at panic.c:141!
invalid operand: 0000
CPU:    0
EIP:    0010:[__out_of_line_bug+15/36] Tainted: P
EFLAGS: 00010086
eax: 00000026   ebx: f7928a00   ecx: 00000002 edx: 02000000
esi: c3e8207c   edi: f79ba1f0   ebp: f79ba1d8 esp: f6fabc68
ds: 0018   es: 0018   ss: 0018
Process hp_ltt (pid: 590, stackpage=f6fab000)
Stack: c026fa80 000000a2 c0202f38 000000a2 f7928a00 c3e8207c f79ba1c0 f79ba1d8
       c02c2934 c0132070 00000000 00000000 ffffffff 00000000 0000000e 00000060
       00000000 00000002 c0204831 c3e8207c f7928a00 f79ba1c0 0000005a 00000293
Call Trace:    [mptscsih_AddSGE+200/832] [__alloc_pages+64/352]
 [mptscsih_qcmd+621/1288]  [scsi_dispatch_cmd+649/904] [scsi_old_done+0/1500]
 [scsi_request_fn+826/892]  [__scsi_insert_special+110/128]
 [scsi_insert_special_req+26/32] [scsi_do_req+328/368]
 [sg_common_write+587/604] [sg_cmd_done_bh+0/912] [sg_new_write+539/576]
 [sg_ioctl+616/3004] [journal_dirty_metadata+356/396] [__alloc_pages+64/352]
 [do_wp_page+112/668] [handle_mm_fault+135/184] [do_page_fault+380/1178]
 [do_page_fault+0/1178] [sys_rt_sigaction+159/324] [sys_ioctl+685/746]
 [error_code+52/60] [system_call+51/56]

Code: 0f 0b 8d 00 a6 fa 26 c0 eb fe 8d 74 26 00 8d bc 27 00 00 00

-- Dag