Re: kern/38057: ffs makes assuptions about devvp file system

[Andrew Doran <ad%netbsd.org@localhost>]

The following reply was made to PR kern/38057; it has been noted by GNATS.

From: Andrew Doran <ad%netbsd.org@localhost>
To: gnats-bugs%NetBSD.org@localhost
Cc: 
Subject: Re: kern/38057: ffs makes assuptions about devvp file system
Date: Thu, 21 Feb 2008 14:35:07 +0000

 On Wed, Feb 20, 2008 at 02:40:01PM +0000, YAMAMOTO Takashi wrote:
 > The following reply was made to PR kern/38057; it has been noted by GNATS.
 > 
 > From: yamt%mwd.biglobe.ne.jp@localhost (YAMAMOTO Takashi)
 > To: gnats-bugs%NetBSD.org@localhost
 > Cc: kern-bug-people%netbsd.org@localhost, gnats-admin%netbsd.org@localhost, 
 > netbsd-bugs%netbsd.org@localhost,
 >         pooka%iki.fi@localhost
 > Subject: Re: kern/38057: ffs makes assuptions about devvp file system
 > Date: Wed, 20 Feb 2008 23:38:35 +0900 (JST)
 > 
 >  >  This is an even more obvious problem for some code yet to be checked in,
 >  >  that it is for softdep.  I was thinking if it could be fixed in a general
 >  >  manner by requiring that VBLK devvp has the properies of the file system
 >  >  it is used for instead of the file system that the node resides on.
 >  
 >  how about:
 >  
 >      flush_buffers_for_fsync(vp)
 >      {
 >              mp = vp->v_specmountpoint;
 >              if (mp != NULL) {
 >                      VFS_foo(mp, vp);
 >              } else {
 >                      vflushbuf();
 >              }
 >      }
 
 In another direction but just as ugly, we could record it in the spec_node
 and have something like spec_synchook_set() and spec_synchook_clear() to
 be used on devvp.
 
 Andrew