Re: panic in ciss.c

[Jaromír Doleček <jaromir.dolecek%gmail.com@localhost>]

Le ven. 17 avr. 2020 à 22:38, <oskar%fessel.org@localhost> a écrit :
> This was the first time.  this machine survived successful complete source builds and pkgsrc builds with -current up to 9.99.49.  50, 52 and 55 paniced somewhere else. 9.99.56 survived much longer than  these, and then this one showed up.

Looking at the sources for ciss, this seems like a forgotten debugging
KASSERT() - it looks like that normally happens if the polled command
times out.

Having said that, the code handling the condition after the KASSERT()
seems to be wrong anyway, ciss_done() fails to process the result as
error since (ccb_cmd.id & CISS_CMD_ERR) is not set, and also even if
that worked, ciss_error() expects ccb_err.cmd_stat to be set to
little-endian value.

Jaromir