dma_timer_expiry when using hard disk

Electro · 06-17-2006, 02:27 AM

Did you say yes to CONFIG_IDEDISK_MULTI_MODE and CONFIG_IDEDMA_IVB before compiling the kernel. If you have not, recompile the kernel with these settings. If it still gives you problems and you think it is the drive, I am not stopping you from buying a new hard drives but buy hard drives from either Hitachi or Western Digital.

I suggest either HighPoint Rocket133 (PATA), HighPoint RocketRAID 454 (PATA), HighPoint RocketRAID 1520 (SATA), or HighPoint RocketRAID 1640 (SATA).

eohrnberger · 06-18-2006, 10:59 AM

Been digging in google groups, and have found a large number of reports with the same errors. Seems to me like some sort of bug has crawled into the kernel code for this particluar hardware combination.

eohrnberger · 06-19-2006, 05:35 PM

Set the data drives to UDMA2, seems to be helping. Running 24 hours now without trouble so far.

eohrnberger · 06-21-2006, 11:16 PM

I've tracked this down further now, and have found in google groups, the linux kernel group and the linux ide group, that there is in fact a problem with the driver for the later model Promise controllers.

eohrnberger · 06-22-2006, 06:48 PM

OK, so I was in touch with the kernel developers that are developing the libata kernel package. Essentailly, they are wrapping IDE drives with a SCSI layer. With this installed and configured, it seems that the DMA_Expiry problems go away. However, this would be bleeding edge kernel code stuff. Not for the fait of heart.