system hang on HDIO_DRIVE_RESET! help!
Hi,
I did HDIO_DRIVE_RESET ioctl, but system hung without any response,
only printed some mesages from kernel(v2.4.20):
hda: DMA disabled
hda: ide_set_handler: handler not null; old=c01ce300, new=c01d4400
bug: kernel timer added twice at c01ce102
would you please help me with it?
Regards
rain.w
Quoted 9 lines · rain.wang
Hi,
I didHDIO_DRIVE_RESETioctl, but system hung without any response,
only printed some mesages from kernel(v2.4.20):hda: DMA disabled
hda:ide_set_handler: handler not null; old=c01ce300, new=c01d4400
bug: kernel timer added twice atc01ce102would you please help me with it?
Does this still occur on 2.4.21pre. It should be fixed now
Quoted 5 lines · Alan Cox
Quoted 4 lines · rain.wang
Hi,
I didHDIO_DRIVE_RESETioctl, but system hung without any response,
only printed some mesages from kernel(v2.4.20):
Quoted 7 lines
Quoted 5 lines
hda: DMA disabled
hda:ide_set_handler: handler not null; old=c01ce300, new=c01d4400
bug: kernel timer added twice atc01ce102would you please help me with it?
Does this still occur on 2.4.21pre. It should be fixed now
I had tested 'hdparm -w /dev/hda' under 2.4.21-pre4, but problem sill exist,
just same message as in 2.4.20.
rain.w
Quoted 5 lines · rain.wang
Quoted 1 line
Does this still occur on 2.4.21pre. It should be fixed now
I had tested 'hdparm -w
/dev/hda' under 2.4.21-pre4, but problem sill exist,just same message as in 2.4.20.
What controller are you using and I'll look into it a bit further
Quoted 8 lines · Alan Cox
Quoted 5 lines · rain.wang
Quoted 1 line
Does this still occur on 2.4.21pre. It should be fixed now
I had tested 'hdparm -w
/dev/hda' under 2.4.21-pre4, but problem sill exist,just same message as in 2.4.20.
What controller are you using and I'll look into it a bit further
Intel 82801AA host controller, and I found when I disabled DMA before doing drive reset, system wouldn't hang at most time. It seemed not tight related with
host chip, does it?
rain.w
Quoted 13 lines · rain.wang
Quoted 11 lines · Alan Cox
Quoted 9 lines · rain.wang
Hi,
I didHDIO_DRIVE_RESETioctl, but system hung without any response,
only printed some mesages from kernel(v2.4.20):hda: DMA disabled
hda:ide_set_handler: handler not null; old=c01ce300, new=c01d4400
bug: kernel timer added twice atc01ce102would you please help me with it?
Quoted 7 lines
Quoted 1 line
Does this still occur on 2.4.21pre. It should be fixed now
I had tested 'hdparm -w
/dev/hda' under 2.4.21-pre4, but problem sill exist,just same message as in 2.4.20.
rain.w
Hi Alan,
I had tested 'hdparm -w /dev/hda' under 2.4.25-pre5-ac1, system
crashed
with
kernel oops message:
kernel BUG at ide-iops:1046!
...
can this be resolved?
rain.w
Quoted 8 lines · rain.wang
I had tested 'hdparm -w /dev/hda' under 2.4.25-pre5-ac1, systemcrashed
with
kernel oops message:
kernel BUG at ide-iops:1046!
...can this be resolved?
Once I understand what the problems all are yes. The BUG() is good, it confirms that what we are both seeing is the same thing - the reset is managing to issue two commands to the controller at the same time.
Quoted 8 lines · Alan Cox
Quoted 7 lines · rain.wang
I had tested 'hdparm -w /dev/hda' under 2.4.25-pre5-ac1, systemcrashed
with
kernel oops message:
kernel BUG at ide-iops:1046!
...
Quoted 5 lines
Quoted 1 line
can this be resolved?
Once I understand what the problems all are yes. The
BUG()is good, it confirms that what we are both seeing is the same thing - the reset is managing to issue two commands to the controller at the same time.
Hi,
thank you, Alan. I tested pre5-ac2 patch and that seems all ok.
rain.w
Quoted 6 lines · rain.wang
Quoted 3 lines
Once I understand what the problems all are yes. The
BUG()is good, it confirms that what we are both seeing is the same thing - the reset is managing to issue two commands to the controller at the same time.Hi,
thank you, Alan. I tested pre5-ac2 patch and that seems all ok.
Thanks for the confirmation it is fixed
Quoted 9 lines · Alan Cox
Quoted 6 lines · rain.wang
Quoted 3 lines
Once I understand what the problems all are yes. The
BUG()is good, it confirms that what we are both seeing is the same thing - the reset is managing to issue two commands to the controller at the same time.Hi,
thank you, Alan. I tested pre5-ac2 patch and that seems all ok.Thanks for the confirmation it is fixed
Hi Alan,
for 2.4.21-pre5-ac2 and -ac3 patch also.
there's still problem on reset. when I do 'hdparm -w /dev/hda' once
after another, all seems ok. but when I make a shell script and let 'hdparm -w' run in several times loop, system would always crashed at the second time and left oops messages:
kernel BUG at ide.c:1700!
...so, if any bugs still locking there?
rain.w
Rain,
The only way to deal with this is to treat the operations a failed and punch them back out to block for clean up. Now we failed the a command. However, I think I need to set a default block hook during the reset process for the drive, channel, hba ... depending on the magnitude of the wrecking ball generated. I need to offline Alan for this core dump.
The hang is in the clean ups after the reset.
I suspect the driver/hba is in DMA and drive is not.
Cheers,
Andre Hedrick
LAD Storage Consulting Group
------------------------------------
Pokemon (n), A Jamaican proctologist
------------------------------------
Quoted 31 lines · rain.wang
Quoted 9 lines · Alan Cox
Quoted 6 lines · rain.wang
Quoted 3 lines
Once I understand what the problems all are yes. The
BUG()is good, it confirms that what we are both seeing is the same thing - the reset is managing to issue two commands to the controller at the same time.Hi,
thank you, Alan. I tested pre5-ac2 patch and that seems all ok.Thanks for the confirmation it is fixed
Hi Alan,
for 2.4.21-pre5-ac2 and -ac3 patch also.
there's still problem on reset. when I do 'hdparm -w/dev/hda' once
after another, all seems ok. but when I make a shell script and let 'hdparm -w' run in several times loop, system would always crashed at the second time and left oops messages:kernel BUG at ide.c:1700! ...so, if any bugs still locking there?
rain.w
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
Quoted 7 lines · Andre Hedrick
Rain,
The only way to deal with this is to treat the operations a failed and punch them back out to block for clean up. Now we failed the a command. However, I think I need to set a default block hook during the reset process for the drive, channel, hba ... depending on the magnitude of the wrecking ball generated. I need to offline Alan for this core dump.
I fixed one set of races with resets and it doesnt suprise me there is another right now.