cvs commit: src/sbin/fsck_ffs fsck.h pass5.c src/sys/ufs/ffs ffs_alloc.c ffs_softdep.c fs.h

[David Schultz]

On Sun, Feb 20, 2005, Xin LI wrote:
> delphij     2005-02-20 08:02:16 UTC
> 
>   FreeBSD src repository
> 
>   Modified files:
>     sbin/fsck_ffs        fsck.h pass5.c 
>     sys/ufs/ffs          ffs_alloc.c ffs_softdep.c fs.h 
>   Log:
>   The recomputation of file system summary at mount time can be a
>   very slow process, especially for large file systems that is just
>   recovered from a crash.
>   
>   Since the summary is already re-sync'ed every 30 second, we will
>   not lag behind too much after a crash.  With this consideration
>   in mind, it is more reasonable to transfer the responsibility to
>   background fsck, to reduce the delay after a crash.

I'm not sure that I completely buy this explanation.  If an
application has a 1 GB temporary file open and unlinked at the
time of the crash, then upon reboot, this change will make it seem
as though I have 1 GB less space than I really do.  This could
lead to spurious disk full errors.  (Or will that happen anyway if
bgfsck hasn't recomputed all the free block bitmaps yet?)

I don't mean to suggest that this is a bad idea; to the contrary,
I think it's a great idea.  But unless I'm missing something, it
has larger adverse effects than claimed in the commit message.

FWIW, I run bgfsck on my development box once a month from a cron
job, rather than after every crash.  As long as there's free space
and no bugs in the filesystem or I/O system (okay, a big assumption),
this doesn't hurt anything and saves me lots of time.