Stuart Henderson <s...@spacehopper.org> writes: > That FreeBSD commit prevents using their "hw.ahci.force" tunable on the > device, it's used for attaching as AHCI to certain known chips even if > they're set in legacy IDE mode. > > Does it work to just add the vid/pid to the ahci_devices[] array > without a specific attach function? (like PCI_PRODUCT_ASMEDIA_ASM1061_SATA).
After a little discussion off-list with Jonathan Matthew, I think what I've experienced here is actually an SSD that is on it's way to the great silicon graveyard. What I was experiencing was brief periods (sometimes a couple of seconds) of what seemed like I/O related pauses when trying to access files from that disk (the machine remained responsive the whole time to ssh, tmux etc but whatever process was trying to access that disk would hiccup briefly). However, I experienced it *only* with one particular drive (and it's quite an older one) - other drives did not do this whether regardless of NCQ being on or off. I had not experienced this with that drive in other systems previously. But after doing some testing again today I've now experienced the same hang on a completely different machine with a completely different AHCI controller with the drive getting somewhat warm to the touch when idle - so given that the SSD in question is ~7 years old, I'm going to put it out to pasture. I selected NCQ because IIRC it's the first feature knockout mask listed in the header - it just happened to seem like it worked. Me Googling and misreading the FreeBSD commit wasn't the trigger for selecting NCQ anyway - I was just looking for confirmation I wasn't going crazy but apparently I am. :-) Based on that I *think* attaching it just like the ASMedia device will work, but let me test that a little more thoroughly before I send an updated diff. Thanks for the review and sorry for the confusion, Ash