All of lore.kernel.org
 help / color / mirror / Atom feed
* Re: Adaptec driver crashes (1/3 and 2/3)
@ 2009-05-10 10:09 ` Norman Diamond
  0 siblings, 0 replies; 7+ messages in thread
From: Norman Diamond @ 2009-05-10 10:09 UTC (permalink / raw)
  To: James Bottomley; +Cc: linux-kernel, linux-scsi

Norman Diamond (I) wrote:
> James Bottomley wrote:
>> On Wed, 2009-05-06 at 16:13 +0900, Norman Diamond wrote:
>>> The easiest 100% reproducible way to crash a Linux system is as follows.
>>>
>>> Insert either an Adaptec 1460 PCMCIA card or Adaptec 1480 CardBus card. 
>>> Remove the card.  Insert the card again.  Even if you knew to set the 
>>> console to a text mode terminal first (since Linux doesn't give Blue 
>>> Screens of Death otherwise), you still won't get a dump.  Not much to go 
>>> on without a dump, but at least it's 100% repro.
>>
>> Sounds like something's still pinned.  When you remove it the first time, 
>> can you remove the module? (that's aha152x_cs right?).
>
> With the 1460, after removing the first time, no I couldn't rmmod the 
> module.
> However, yesterday with the 1460 I couldn't repro the kernel crash. 
> Yesterday with the 1460 I could only repro less severe lossage.

Today I tried Knoppix 6.0.1, kernel 2.6.28.4.  aha152x_cs is a module here 
too.  When inserting a 1460, nothing happened.  Removing a 1460, nothing 
happened.  Inserting a 1460, nothing happened.  I modprobed aha152x_cs.  It 
modprobed successfully.  Removing a 1460, inserting, removing, inserting, 
nothing happened, even though the module remained loaded.  The older 
partially working kernel was better than this.

> Yesterday with the 1480 I did repro the kernel crash, the same as before. 
> The aic7xxx driver was compiled in, not a module.

Today with Knoppix 6.0.1, kernel 2.6.28.4, this one is worse too.  Here 
aic7xxx is a module, not compiled in.  I don't remember if I had to modprobe 
it manually but I think not.  I think that inserting a 1480 automatically 
modprobed the module and it worked for the time being.  However, removing 
the card started the time bomb.  Reinsertion was no longer necessary. 
Around 20 seconds after removal, if the console was in text mode then there 
would be a 1-line crash message with no dump.  Attempts to rmmod aic7xxx 
failed.  It didn't matter if I tried 0, 1, or multiple times while the time 
bomb was running, rmmod failed and the crash was still coming.

One time I shut down Knoppix 6.0.1 from whatever KDE's Start button is 
called, the usual way of doing a shutdown, instead of letting a crash do it. 
When the CD-ROM tray popped open, I guessed that Knoppix was nearly done 
shutting down, so I hit the Enter key, and the power went off a few seconds 
later.  Older versions had some text mode messages on the screen during 
shutdown, and told us at the same time as it popped open the CD tray.  OK, 
this one isn't the fault of a SCSI driver.

Sometimes there are reasons to stick to older partially working versions. 

--------------------------------------
Power up the Internet with Yahoo! Toolbar.
http://pr.mail.yahoo.co.jp/toolbar/

^ permalink raw reply	[flat|nested] 7+ messages in thread

* Re: Adaptec driver crashes (1/3 and 2/3)
@ 2009-05-10 10:09 ` Norman Diamond
  0 siblings, 0 replies; 7+ messages in thread
From: Norman Diamond @ 2009-05-10 10:09 UTC (permalink / raw)
  To: James Bottomley; +Cc: linux-kernel, linux-scsi

Norman Diamond (I) wrote:
> James Bottomley wrote:
>> On Wed, 2009-05-06 at 16:13 +0900, Norman Diamond wrote:
>>> The easiest 100% reproducible way to crash a Linux system is as follows.
>>>
>>> Insert either an Adaptec 1460 PCMCIA card or Adaptec 1480 CardBus card. 
>>> Remove the card.  Insert the card again.  Even if you knew to set the 
>>> console to a text mode terminal first (since Linux doesn't give Blue 
>>> Screens of Death otherwise), you still won't get a dump.  Not much to go 
>>> on without a dump, but at least it's 100% repro.
>>
>> Sounds like something's still pinned.  When you remove it the first time, 
>> can you remove the module? (that's aha152x_cs right?).
>
> With the 1460, after removing the first time, no I couldn't rmmod the 
> module.
> However, yesterday with the 1460 I couldn't repro the kernel crash. 
> Yesterday with the 1460 I could only repro less severe lossage.

Today I tried Knoppix 6.0.1, kernel 2.6.28.4.  aha152x_cs is a module here 
too.  When inserting a 1460, nothing happened.  Removing a 1460, nothing 
happened.  Inserting a 1460, nothing happened.  I modprobed aha152x_cs.  It 
modprobed successfully.  Removing a 1460, inserting, removing, inserting, 
nothing happened, even though the module remained loaded.  The older 
partially working kernel was better than this.

> Yesterday with the 1480 I did repro the kernel crash, the same as before. 
> The aic7xxx driver was compiled in, not a module.

Today with Knoppix 6.0.1, kernel 2.6.28.4, this one is worse too.  Here 
aic7xxx is a module, not compiled in.  I don't remember if I had to modprobe 
it manually but I think not.  I think that inserting a 1480 automatically 
modprobed the module and it worked for the time being.  However, removing 
the card started the time bomb.  Reinsertion was no longer necessary. 
Around 20 seconds after removal, if the console was in text mode then there 
would be a 1-line crash message with no dump.  Attempts to rmmod aic7xxx 
failed.  It didn't matter if I tried 0, 1, or multiple times while the time 
bomb was running, rmmod failed and the crash was still coming.

One time I shut down Knoppix 6.0.1 from whatever KDE's Start button is 
called, the usual way of doing a shutdown, instead of letting a crash do it. 
When the CD-ROM tray popped open, I guessed that Knoppix was nearly done 
shutting down, so I hit the Enter key, and the power went off a few seconds 
later.  Older versions had some text mode messages on the screen during 
shutdown, and told us at the same time as it popped open the CD tray.  OK, 
this one isn't the fault of a SCSI driver.

Sometimes there are reasons to stick to older partially working versions. 

--------------------------------------
Power up the Internet with Yahoo! Toolbar.
http://pr.mail.yahoo.co.jp/toolbar/

^ permalink raw reply	[flat|nested] 7+ messages in thread

* Re: Adaptec driver crashes (1/3 and 2/3)
  2009-05-08 19:38 ` James Bottomley
@ 2009-05-09 23:41     ` Norman Diamond
  0 siblings, 0 replies; 7+ messages in thread
From: Norman Diamond @ 2009-05-09 23:41 UTC (permalink / raw)
  To: James Bottomley; +Cc: linux-kernel, linux-scsi

James Bottomley wrote:
> On Wed, 2009-05-06 at 16:13 +0900, Norman Diamond wrote:
>> The easiest 100% reproducible way to crash a Linux system is as follows.
>>
>> Insert either an Adaptec 1460 PCMCIA card or Adaptec 1480 CardBus card. 
>> Remove the card.  Insert the card again.  Even if you knew to set the 
>> console to a text mode terminal first (since Linux doesn't give Blue 
>> Screens of Death otherwise), you still won't get a dump.  Not much to go 
>> on without a dump, but at least it's 100% repro.
>
> Sounds like something's still pinned.  When you remove it the first time, 
> can you remove the module? (that's aha152x_cs right?).

With the 1460, after removing the first time, no I couldn't rmmod the 
module.  You're either right or nearly right about the module name since 
some were renamed to use hyphens instead of underbars, but I don't remember 
if that was included.  Anyway, the module was in use by 4 unknown users, and 
couldn't be rmmodded.

However, yesterday with the 1460 I couldn't repro the kernel crash. 
Yesterday with the 1460 I could only repro less severe lossage.  Removal of 
the card showed up in dmesg.  Reinsertions of the card were completely 
ignored, not showing up in dmesg, not recreating the special file /dev/sdb, 
and if I mknodded /dev/sdb it was still inaccessible (no such device).  Also 
shutdown hanged and I had to hold the power switch to shut down.

Yesterday with the 1480 I did repro the kernel crash, the same as before. 
The aic7xxx driver was compiled in, not a module.  I don't recall if removal 
of the card showed up in dmesg.  Reinsertion didn't, but the system lived 
long enough for dmesg to run and not show a reinsertion.  After that came a 
1 line kernel crash message with no dump.  If I recall correctly, 
Ctrl+Alt+Del accomplished a reboot so I didn't have to hold the power 
switch. 

--------------------------------------
Power up the Internet with Yahoo! Toolbar.
http://pr.mail.yahoo.co.jp/toolbar/

^ permalink raw reply	[flat|nested] 7+ messages in thread

* Re: Adaptec driver crashes (1/3 and 2/3)
@ 2009-05-09 23:41     ` Norman Diamond
  0 siblings, 0 replies; 7+ messages in thread
From: Norman Diamond @ 2009-05-09 23:41 UTC (permalink / raw)
  To: James Bottomley; +Cc: linux-kernel, linux-scsi

James Bottomley wrote:
> On Wed, 2009-05-06 at 16:13 +0900, Norman Diamond wrote:
>> The easiest 100% reproducible way to crash a Linux system is as follows.
>>
>> Insert either an Adaptec 1460 PCMCIA card or Adaptec 1480 CardBus card. 
>> Remove the card.  Insert the card again.  Even if you knew to set the 
>> console to a text mode terminal first (since Linux doesn't give Blue 
>> Screens of Death otherwise), you still won't get a dump.  Not much to go 
>> on without a dump, but at least it's 100% repro.
>
> Sounds like something's still pinned.  When you remove it the first time, 
> can you remove the module? (that's aha152x_cs right?).

With the 1460, after removing the first time, no I couldn't rmmod the 
module.  You're either right or nearly right about the module name since 
some were renamed to use hyphens instead of underbars, but I don't remember 
if that was included.  Anyway, the module was in use by 4 unknown users, and 
couldn't be rmmodded.

However, yesterday with the 1460 I couldn't repro the kernel crash. 
Yesterday with the 1460 I could only repro less severe lossage.  Removal of 
the card showed up in dmesg.  Reinsertions of the card were completely 
ignored, not showing up in dmesg, not recreating the special file /dev/sdb, 
and if I mknodded /dev/sdb it was still inaccessible (no such device).  Also 
shutdown hanged and I had to hold the power switch to shut down.

Yesterday with the 1480 I did repro the kernel crash, the same as before. 
The aic7xxx driver was compiled in, not a module.  I don't recall if removal 
of the card showed up in dmesg.  Reinsertion didn't, but the system lived 
long enough for dmesg to run and not show a reinsertion.  After that came a 
1 line kernel crash message with no dump.  If I recall correctly, 
Ctrl+Alt+Del accomplished a reboot so I didn't have to hold the power 
switch. 

--------------------------------------
Power up the Internet with Yahoo! Toolbar.
http://pr.mail.yahoo.co.jp/toolbar/

^ permalink raw reply	[flat|nested] 7+ messages in thread

* Re: Adaptec driver crashes (1/3 and 2/3)
  2009-05-06  7:13 ` Norman Diamond
  (?)
@ 2009-05-08 19:38 ` James Bottomley
  2009-05-09 23:41     ` Norman Diamond
  -1 siblings, 1 reply; 7+ messages in thread
From: James Bottomley @ 2009-05-08 19:38 UTC (permalink / raw)
  To: Norman Diamond; +Cc: linux-kernel, linux-scsi

On Wed, 2009-05-06 at 16:13 +0900, Norman Diamond wrote:
> The easiest 100% reproducible way to crash a Linux system is as follows.
> 
> Insert either an Adaptec 1460 PCMCIA card or Adaptec 1480 CardBus card. 
> Remove the card.  Insert the card again.  Even if you knew to set the 
> console to a text mode terminal first (since Linux doesn't give Blue Screens 
> of Death otherwise), you still won't get a dump.  Not much to go on without 
> a dump, but at least it's 100% repro.

Sounds like something's still pinned.  When you remove it the first
time, can you remove the module? (that's aha152x_cs right?).

James



^ permalink raw reply	[flat|nested] 7+ messages in thread

* Adaptec driver crashes (1/3 and 2/3)
@ 2009-05-06  7:13 ` Norman Diamond
  0 siblings, 0 replies; 7+ messages in thread
From: Norman Diamond @ 2009-05-06  7:13 UTC (permalink / raw)
  To: linux-kernel, linux-scsi

The easiest 100% reproducible way to crash a Linux system is as follows.

Insert either an Adaptec 1460 PCMCIA card or Adaptec 1480 CardBus card. 
Remove the card.  Insert the card again.  Even if you knew to set the 
console to a text mode terminal first (since Linux doesn't give Blue Screens 
of Death otherwise), you still won't get a dump.  Not much to go on without 
a dump, but at least it's 100% repro.

It doesn't happen with a NinjaSCSI card.  Insert card, access SCSI disk, 
remove card, insert card again, access SCSI disk again, remove card, insert 
again, no problem.  So there's something about two Adaptec drivers. 

--------------------------------------
Power up the Internet with Yahoo! Toolbar.
http://pr.mail.yahoo.co.jp/toolbar/

^ permalink raw reply	[flat|nested] 7+ messages in thread

* Adaptec driver crashes (1/3 and 2/3)
@ 2009-05-06  7:13 ` Norman Diamond
  0 siblings, 0 replies; 7+ messages in thread
From: Norman Diamond @ 2009-05-06  7:13 UTC (permalink / raw)
  To: linux-kernel, linux-scsi

The easiest 100% reproducible way to crash a Linux system is as follows.

Insert either an Adaptec 1460 PCMCIA card or Adaptec 1480 CardBus card. 
Remove the card.  Insert the card again.  Even if you knew to set the 
console to a text mode terminal first (since Linux doesn't give Blue Screens 
of Death otherwise), you still won't get a dump.  Not much to go on without 
a dump, but at least it's 100% repro.

It doesn't happen with a NinjaSCSI card.  Insert card, access SCSI disk, 
remove card, insert card again, access SCSI disk again, remove card, insert 
again, no problem.  So there's something about two Adaptec drivers. 

--------------------------------------
Power up the Internet with Yahoo! Toolbar.
http://pr.mail.yahoo.co.jp/toolbar/

^ permalink raw reply	[flat|nested] 7+ messages in thread

end of thread, other threads:[~2009-05-10 10:10 UTC | newest]

Thread overview: 7+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2009-05-10 10:09 Adaptec driver crashes (1/3 and 2/3) Norman Diamond
2009-05-10 10:09 ` Norman Diamond
  -- strict thread matches above, loose matches on Subject: below --
2009-05-06  7:13 Norman Diamond
2009-05-06  7:13 ` Norman Diamond
2009-05-08 19:38 ` James Bottomley
2009-05-09 23:41   ` Norman Diamond
2009-05-09 23:41     ` Norman Diamond

This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.