Re: kern/51740: mount -o discard considered harmfull

[David Holland <dholland-bugs%netbsd.org@localhost>]

The following reply was made to PR kern/51740; it has been noted by GNATS.

From: David Holland <dholland-bugs%netbsd.org@localhost>
To: gnats-bugs%NetBSD.org@localhost
Cc: 
Subject: Re: kern/51740: mount -o discard considered harmfull
Date: Sun, 25 Dec 2016 18:21:52 +0000

 On Sat, Dec 24, 2016 at 01:35:00PM +0000, martin%NetBSD.org@localhost wrote:
  > For some currently unknown reasons discard operations fail on a new
  > ssd drive I am trying to populate. This leads to catastrophic file
  > system inconsistencies.
 
 odd...
 
 some questions:
    - is newfs generating an ffsv1 or ffsv2 volume?
    - does the corruption still happen if you comment out the actual
      discarding (it's the VOP_FDISCARD call on/near line 1597 of
      ffs_alloc.c)?
    - based on memory size and how long it takes before it croaks, is
      it probably all in memory, or does it die after reading
      previously written (and maybe mangled) stuff back from the disk?
    - is it linked to -o async?
 
 Because -o async is faster, being linked to -o async might either be
 an async bug or a race. It might be worth trying to rig out to test
 this on mfs.
 
 -- 
 David A. Holland
 dholland%netbsd.org@localhost