From: Matthias Andree (no email)
Date: Fri Aug 01 2003 - 09:43:05 EDT
On Fri, 01 Aug 2003, Wietse Venema wrote:
> Linux XFS breaks both link() AND rename(); therefore patching for
> the link() problems alone is pointless.
Virtually all Linux file systems "break" these, i. e. do these
operations asynchronously. The difficulty the mail administrator has is
finding out which file system will heed chattr +S or mount -o sync at
all -- IF he is aware of the problem at all, not many are. :-/
Synchronous operations are only handled stepmotherly by most Linux file
systems, as mentioned before; today, a reiserfs mount -o sync filesystem
still writes astonishingly fast when compared to XFS, even if I don't
use data journalling. Scepticism is highly advised.
> For qmail, someone did an LD_PRELOAD solution that intercepts all
> the relevant file system operations; such an implementation can
> actually be tested and verified. Lots of little changes all over
> Postfix are not.
There are some concerns about the implementation tough.
I think using cscope to figure the full list of rename() and link() is
safe enough, ultimately, there'd be one function sync_rename() and one
sync_link() and these functions would be the only ones that call
rename() or link().
-- Matthias Andree
|
|
|