Problem in 2.6.0-test9-mm1 with siimage+hdparm

Problem in 2.6.0-test9-mm1 with siimage+hdparm

[Krisztian VASAS]

Hello all...

I've got an Abit NF7-S v2.0 mainboard with an Athlon XP 2000+ CPU.

Yesterday I've compiled the 2.6.0-test9 kernel with -mm1 patch. My
system is on the first sata device (/dev/hde).

After reboot I've noticed an Oops after setting dma and other settings
with hdparm. Without -mm1 patch the machine works well. 

After the oops I've tried to find what the problem was. 

hdparm -c1 /dev/hde -> nothing was happened, switched off...
hdparm -d1 /dev/hde -> Oops...

I've got this messages:

hde: DMA timeout retry status error: status=0x58 { DriveReady
SeekComplete DataRequest }

ide2: reset phy, status=0x00000113 siimage_reset

.
.
.

Buffer I/O error on device hde2, logical block 97865
lost page write due to I/O error on hde2

.
.
.

EXT3-fs error (device hde2) in start_transaction: Journal has aborted



The rootfs is /dev/hde2.

Please cc the answers to my address, because I'm not lkml member...

Thanks:


IroNiQ
-- 
Web: http://ironiq.hu
E-mail: iron@ironiq.hu

Re: Problem in 2.6.0-test9-mm1 with siimage+hdparm

[Prakash K. Cheemplavam]

Krisztian VASAS wrote:
> Hello all...
> 
> I've got an Abit NF7-S v2.0 mainboard with an Athlon XP 2000+ CPU.
> 
> Yesterday I've compiled the 2.6.0-test9 kernel with -mm1 patch. My
> system is on the first sata device (/dev/hde).
> 
> After reboot I've noticed an Oops after setting dma and other settings
> with hdparm. Without -mm1 patch the machine works well. 
> 
> After the oops I've tried to find what the problem was. 
> 
> hdparm -c1 /dev/hde -> nothing was happened, switched off...
> hdparm -d1 /dev/hde -> Oops...
> 
> I've got this messages:
> 
> hde: DMA timeout retry status error: status=0x58 { DriveReady
> SeekComplete DataRequest }
> 
> ide2: reset phy, status=0x00000113 siimage_reset

Strange enough I have an Abit NF7-s Rev2.0 as well, but for me the 
driver works - somehow. I havbe latest bios 1.9 (modified with latest 
silicon image bios). Since siimage 1.06 drvier. I have DMA without using 
hdparm, but it is not as fast as in windows. With kernel 2.6 tranfer 
dropped very much to about 20mb/s (playing with read-aheed didn't help). 
But I am getting this error messages in dmesg and at boot up (that's why 
I commented it out in siimage.c):

hde: sata_error = 0x00000000, watchdog = 0, siimage_mmio_ide_dma_test_irq

Other than that it runs fine - forgetting that performance it bad right 
now. Tried various shedulers, but didn't help. With kernel 2.4.22-ac4 I 
had about 37mb/sec, now only 20mb/sec...

Prakash

Re: Problem in 2.6.0-test9-mm1 with siimage+hdparm

[Krisztian VASAS]

On 2003-11-05 Prakash K. Cheemplavam wrote:
> Strange enough I have an Abit NF7-s Rev2.0 as well, but for me the 
> driver works - somehow. I havbe latest bios 1.9 (modified with latest 
> silicon image bios). Since siimage 1.06 drvier. I have DMA without using 
> hdparm, but it is not as fast as in windows. With kernel 2.6 tranfer 
> dropped very much to about 20mb/s (playing with read-aheed didn't help). 
> But I am getting this error messages in dmesg and at boot up (that's why 
> I commented it out in siimage.c):
> 
> hde: sata_error = 0x00000000, watchdog = 0, siimage_mmio_ide_dma_test_irq
> 
> Other than that it runs fine - forgetting that performance it bad right 
> now. Tried various shedulers, but didn't help. With kernel 2.4.22-ac4 I 
> had about 37mb/sec, now only 20mb/sec...

I haven't problem with the speed. I'd like to set up DMA to the disk and
the -d switch (using_dma) causes the panic. With vanilla 2.6.0-test9
everything works fine, no speed problem. But if I boot up with -mm1, I
have to work without DMA on hde. And another interesting thing: hdparm
-d1 /dev/hdd works fine, so the problem is somewhere in the siimage
driver. I don't know exactly what changed, but something has been
corrupted around the driver. 


IroNiQ
-- 
Krisztian VASAS <iron@ironiq.hu>