On Sun, 2019-10-06 at 19:23 -0600, Theo de Raadt wrote: > Kurt Miller <[email protected]> wrote: > > > > > On Sun, 2019-10-06 at 20:01 -0400, Kurt Miller wrote: > > > > > > On Sun, 2019-10-06 at 17:47 -0600, Theo de Raadt wrote: > > > > > > > > > > > > > > > > > > > > > > > > > > > > I applied your diff and built RAMDISK to test it. However, the > > > > > boot failed with this (full boot message later on): > > > > > > > > > > root on rd0a swap on rd0b dump on rd0b > > > > > panic: cannot open disk, 0x1100/0x2f02, error 2 > > > > I believe your bsd.rd has no ramdisk filesystem inside it. > > > > > > > You are correct, thank you. bsd.mp booted fine. I'll make > > > a proper bsd.rd and report back results of testing install > > > with it. > > I completed 5 bsd.rd upgrades off /mnt/home/_sysupgrade > > without issue and performed a series of fio tests also > > without issue. The bsd.rd upgrades would not have survived > > 2 in a row prior to your change. > > > > Disabling command queueing is looking like it stabilized > > the card for me. I haven't noticed a drop in performance > > either in an unscientific comparison to numbers I jotted > > down on Sep 23. > > > > Thank you for the quick fix. > I think that cynical workaround is fine for release, it probably > affects other people with this ahci block (which has a bad reputation > in other forums) >
I hit the issue again using the latest snapshot which includes the work-around. ahci0: log page read failed, slot 31 was still active. ahci0: stopping the port, softreset slot 31 was still active. ahci0: failed to reset port during timeout handling, disabling it No panic this time. The work-around helped with stability in the RAMDISK env where it was very unstable. Perhaps it is still a good idea. -Kurt
