Fedora Rawhide Flips On New SATA Power Management Policy

[phoronix]

Phoronix: Fedora Rawhide Flips On New SATA Power Management Policy

If you are running Fedora Rawhide (their daily/development packages) and using an Intel mobile chipset, be forewarned that they are enabling the SATA link power change that runs the slight risk of potentially causing disk corruption...

http://www.phoronix.com/scan.php?page=news_item&px=Rawhide-Flips-On-SATA-Med-Power

[darkbasic]

Is it safe with btrfs? Btrfs used to be particularly annoying with sata power management...

[starshipeleven]

Is it safe with btrfs? Btrfs used to be particularly annoying with sata power management...

btrfs still lacks the logic to deal with disappearing devices (the patches are maturing, and many prerequsites are getting merged, but the feature is not merged yet).

So I'd say it depends from how this feature works, if it does get in the way then btrfs will trample.

[hansdegoede]

One of the ways this can theoretically cause issues is for the disk to drop of the bus. If btrfs does not handle that well then that would be an additional risk with brtfs, but TBH the disk dropping of the bus scenario (which has been reported with min_power in some cases 2 year ago) is not my biggest worry and likely fixed already.

Some disks show data corruption with min_power without clear errors in dmesg, so you don't notice it until e.g. ext4 trips over invalid inode entries. 2 users have tested such disks already and reported that the problem does not happen with med_power_with_dipm. So I'm hopeful this will all just work.

[starshipeleven]

One of the ways this can theoretically cause issues is for the disk to drop of the bus. If btrfs does not handle that well

Yeah, btrfs does not handle that at all.

Some disks show data corruption with min_power without clear errors in dmesg, so you don't notice it until e.g. ext4 trips over invalid inode entries. 2 users have tested such disks already and reported that the problem does not happen with med_power_with_dipm. So I'm hopeful this will all just work.

buggy firmware in hard drives? Shocked, I'm shocked, I tell you.

[R00KIE]

Some disks show data corruption with min_power without clear errors in dmesg, so you don't notice it until e.g. ext4 trips over invalid inode entries. 2 users have tested such disks already and reported that the problem does not happen with med_power_with_dipm. So I'm hopeful this will all just work.

Any links you can provide to the reports? It would be good to know just in case I or someone else has to debug such a problem later on as that is something to take into account.