On Saturday 14 March 2009, Mark Lord wrote: > > It seems that with libata-based drivers disabling dma is only > > possible at boot time and not dynamically, so this isn't possible > > anyway. I don't know if access errors could cause the driver to fall > > back to PIO, but I don't really want to test that :-) > > Yeah, tricky to see that. With sata_mv, though, DMA is disabled > for any irregular (non block layer R/W) commands. > > So the code path is being exercised, just too quickly to see if > the LED actually blinks much in non-DMA modes. > I think you could test the non-DMA mode with this: > > #!/bin/bash > i=0 > while ( sleep 0.3 ) ; do hdparm --read-sector $i --direct /dev/sda ; > i=$((i + 1)) ; done Tried that. The blinking behavior is the same as with 2.6.25, so blink mode does get disabled correctly. It's funny though that, as you expected, with that particular command you don't really get any blinking (varying sleep times don't make any real difference), so for this particular case it would actually be better to have blink mode enabled (I tested that also) even when not using DMA/NCQ. But I suggest not to go there. > I'll queue up the patch for 2.6.29 now, but if you could try the above > then that would also be cool. Somehow I doubt you'll manage that, but .30 would be nice :-) Cheers, FJP -- To unsubscribe from this list: send the line "unsubscribe linux-ide" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html