On Tue, 6 Sep 2011 13:21:01 +0900 Tejun Heo <tj@xxxxxxxxxx> wrote: > Some configurations have problems with using 32bit PIO on ATAPI cdb > transfers. The most likely cause is different division of payload > across FISes but hasn't been confirmed. There isn't much to be gained > by using 32bit PIO for small transfers. Play it safe and use 16bit > PIO for transfers smaller than 512 bytes. This seems to me to be the wrong approach as it hurts lots of other commands unnecessarily and may cause regressions on other setups because their behaviour is unnecessarily changed and performance reduced. The problem report is specific to Sandybridge in IDE mode, which is a recent controller so it should be chased up with the Intel folks who submitted the device ID as an apparent hardware problem - that way if it is a chipset problem (and I have no idea if it is) it might get a proper workaround, or fixed in future. Secondly it's a SATA specific problem. Thirdly we've been doing 32bit PIO on old IDE since it was invented. This seems an extreme reaction for a report of a single specific problem on a single specific device and controller. So NACK. Add a piix_sata_data_xfer32_maybe() function to the ata_piix driver specifically for this chipset. I've added Seth who submitted the relevant PCI identifers to the email. I'm a bit at a loss to understand why none of the previous discussion or patch was Cc'd in his direction ? Seth - the bug report is https://bugzilla.kernel.org/show_bug.cgi?id=40592 and I guess the underlying question is "Are there any known 32bit PIO errata that would explain this ?" Before the changes go in it might also be worth testing CD burning and audio ripping etc to see if the problem is one dependent on transfer size or the fact the transfer is the cdb on a SATA transfer, so that any fixing is done right. Alan -- 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
