Bug 6118

Summary: 2.6.15.4 freez with the io sheduler
Product: IO/Storage Reporter: Szabolcs Toth (totya)
Component: Block LayerAssignee: Jens Axboe (axboe)
Status: REJECTED DUPLICATE    
Severity: high CC: akpm, greg
Priority: P2    
Hardware: i386   
OS: Linux   
Kernel Version: 2.6.15.4 Subsystem:
Regression: --- Bisected commit-id:

Description Szabolcs Toth 2006-02-22 01:21:44 UTC
Distribution:  Debian Sarge 3r1
Hardware Environment: 
MSI 875P mainboard
I. Pentium4 2.6GHz.
Nvidia MX400 VGA card.
No extra hardware

Problem Description:
During the boot process when the io sheduler(s) were registered the system 
freez without any error messsage.
I tried with the elevator= parameter in the lilo.conf  but the result is the 
same.
Comment 1 Andrew Morton 2006-02-22 01:34:16 UTC
This isn't related to the IO scheduler, but I forget what it is.

Can you add `initcall_debug' to the kernel boot command line then send the last
few lines of output?
Comment 2 Szabolcs Toth 2006-02-22 03:07:18 UTC
Hi !
I added the initcall_debug to the kernel boot command line but the output is 
the same. The last lines are:

io sheduler noop registered
io sheduler anticipatory registered
io sheduler deadline registered
io sheduler cfq registered

and nothing else.  The system freez. I can reboot with only the reset button.
Comment 3 Andrew Morton 2006-02-22 03:22:25 UTC
Please show us the last ten lines.

You should get hundreds of lines of the form "Calling initcall 0x....".  If
you don't get that, you didn't set initcall_debug properly.

We'd need to see the final such message.

Comment 4 Szabolcs Toth 2006-02-22 04:17:11 UTC
Dear Jens Axboe,

I tried now with the 'debug'  command line parameter and I get the following 
few lines:
Calling initcall 0xc0416b51 : noop_init+0x0/0x13()
io sheduler noop registered
Calling initcall 0xc0416b64 : as_init+0x0/0x78()
io sheduler anticipatory registered
Calling initcall 0xc0416bdc : deadline_init0x0/0x76()
io sheduler deadline registered
Calling initcall 0xc0416d24 : cfq_init+0x0/0x5e()
io sheduler cfq registered
Calling initcall 0xc020d8f6 : pci_init+0x0/0x3d()

regards,
Szabolcs Toth
Comment 5 Andrew Morton 2006-02-22 04:28:23 UTC
argh, everyone's reporting this, and it's a USB problem but I
forget what the workaround was.   Greg, do you remember?
Comment 6 Greg Kroah-Hartman 2006-02-22 09:50:49 UTC
It was due to EHCI bios issues, I think the scheduler just got blamed for being
the last thing that printed out.

I'll go dig up that bug number...
Comment 7 Greg Kroah-Hartman 2006-02-22 09:51:44 UTC

*** This bug has been marked as a duplicate of 6011 ***