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: Sat, 31 Dec 2016 23:53:05 +0000

 On Mon, Dec 26, 2016 at 03:15:00PM +0000, Martin Husemann wrote:
 
  >  Interesting - I was creating a FFSv1 file system. With newfs -O2 I
  >  still get the same discard failures, but not the file system
  >  corruption. The same happens if I do not use "async" (wd_trim
  >  errors, but no corruption)
 
 Curious; I would have expected it to be the other way around, since
 it's ffsv2 that does lazy initialization of on-disk inodes. (And that
 code has had races/issues in the past.)
 
 Can you try it with the VOP_FDISCARD call commented out? That would
 tell us if the problem is the discard/trim causing corruption or if
 the corruption is a consequence of the delayed asynchronous block
 freeing that the -o discard logic uses.
 
 (also, what's freeing blocks while untarring? that itself seems odd
 since untarring should really only be writing out new files and
 directories... especially on a fresh volume)
 
 -- 
 David A. Holland
 dholland%netbsd.org@localhost