Hello, Matthew.

On Mon, May 11, 2015 at 01:24:18PM -0700, Matthew Garrett wrote:

Ok, I've had one person testing this report the following:

[ 1896.902726] ata1.00: exception Emask 0x0 SAct 0x800 SErr 0x40000 action 0x0
[ 1896.902735] ata1.00: irq_stat 0x40000008
[ 1896.902741] ata1: SError: { CommWake }
[ 1896.902748] ata1.00: failed command: WRITE FPDMA QUEUED
[ 1896.902758] ata1.00: cmd 61/18:58:88:d5:5d/00:00:0d:00:00/40 tag 11
ncq 12288 out
         res 41/04:00:9f:d5:5d/00:00:0d:00:00/40 Emask 0x401 (device error) <F>
[ 1896.902764] ata1.00: status: { DRDY ERR }
[ 1896.902768] ata1.00: error: { ABRT }
[ 1896.912725] ata1.00: supports DRM functions and may not be fully accessible
[ 1896.932716] ata1.00: supports DRM functions and may not be fully accessible
[ 1896.942783] ata1.00: configured for UDMA/133
[ 1896.942815] ata1: EH complete

which seems to occur in batches. I'm a little confused - isn't it
legitimate for the phy to report comwake here?

CommWake isn't the problem here.  SError is being dumped just for
information.  The disk is reporting failure on a write command which
is diagnosed as "device error" and thus the link is not reset.  It's
really the device actively reporting command failure.

Thanks.

-- 
tejun