Skip to contentSkip to conversation

ide.c and compactFlash

DISCUSSION Espen M. Rutger 2 participants 3 messages Aug 28, 2007 – Aug 30, 2007

lore
Share & tools
lore
#0 lore raw

Hi all.

I got problems with the IDE code which causes the kernel to freez after printing out:

hda: status timeout: status=0xd0 { Busy }

kernel used: 2.4.18 crosscompiled with Montavista tools (ppc_82xx-gcc (GCC) 3.2.1 20020930 (MontaVista))

The ide interface chip is a PD6729 configured to ATA mode.

CompactFlash cards: winsys 1GB industrial grade and simpleTech 1GB industrial grade.

I beleive this is a timing issue and have tried to increase a udelay () to 2 microseconds (instead of 1 microsecond) in the ide_wait_stat() function - it took longer time to freez, but it still freezes...

Any clues?

Regards
Espen M. Rutger

#1 lore raw
Quoted 8 lines · Espen M. Rutger

I got problems with the IDE code which causes the kernel to freez after printing out:

hda: status timeout: status=0xd0 { Busy }

kernel used: 2.4.18 crosscompiled with Montavista tools (ppc_82xx-gcc (GCC) 3.2.1 20020930 (MontaVista))

Quoted 8 lines

The ide interface chip is a PD6729 configured to ATA mode.

CompactFlash cards: winsys 1GB industrial grade and simpleTech 1GB industrial grade.

I beleive this is a timing issue and have tried to increase a udelay () to 2 microseconds (instead of 1 microsecond) in the ide_wait_stat() function - it took longer time to freez, but it still freezes...

Try booting with 'ide=nodma'. Some compact flash cards support DMA mode, and of course most IDE controllers support DMA mode, and if both support it the kernel tends to try and enable it but if you don't have the DMA lines connected it will fail and give annoying errors similar to that one. The DMA part of compact flash was a fairly recent addition and the two affected lines used to be reserved on compact flash. The product I work on did not have the DMA lines on the previous board design but we have it on the latest board revision. On the old boards we had to disable DMA support for IDE since otherwise certain types of compact flash would have annoying timeouts trying to enable DMA. On our new board with the DMA lines, everything just works and the DMA access gives about twice the read/write speed to the compact flash cards we are using.

--
Len Sorensen

#2 lore raw
Quoted 29 lines · Lennart Sorensen
Quoted 4 lines · Espen M. Rutger

I got problems with the IDE code which causes the kernel to freez after printing out:

hda: status timeout: status=0xd0 { Busy }

Quoted 8 lines

kernel used: 2.4.18 crosscompiled with Montavista tools (ppc_82xx-gcc (GCC) 3.2.1 20020930 (MontaVista))

The ide interface chip is a PD6729 configured to ATA mode.

CompactFlash cards: winsys 1GB industrial grade and simpleTech 1GB industrial grade.

I beleive this is a timing issue and have tried to increase a udelay () to 2 microseconds (instead of 1 microsecond) in the ide_wait_stat() function - it took longer time to freez, but it still freezes...

Try booting with 'ide=nodma'. Some compact flash cards support DMA mode, and of course most IDE controllers support DMA mode, and if both support it the kernel tends to try and enable it but if you don't have the DMA lines connected it will fail and give annoying errors similar to that one. The DMA part of compact flash was a fairly recent addition and the two affected lines used to be reserved on compact flash. The product I work on did not have the DMA lines on the previous board design but we have it on the latest board revision. On the old boards we had to disable DMA support for IDE since otherwise certain types of compact flash would have annoying timeouts trying to enable DMA. On our new board with the DMA lines, everything just works and the DMA access gives about twice the read/write speed to the compact flash cards we are using.

Quoted 4 lines

--
Len Sorensen

Thank you very much for your response Lennart, but unfortunately this did not solve my problem - the system (got four machines in my test setup) still generate the error and make the kernel wait forever.

regards
Espen M. Rutger