svn commit: r328013 - head/sbin/fsck_ffs

[Mark Johnston]

On Fri, Mar 09, 2018 at 09:36:25PM -0500, David Bright wrote:
> On Mar 9, 2018, at 17:31, Ian Lepore <ian at freebsd.org> wrote:
> > 
> > On Fri, 2018-03-09 at 17:09 -0500, Mark Johnston wrote:
> >> 
> >> etc/rc.d/fsck doesn't know how to interpret the new exit code and now
> >> just drops to a single-user shell when it is encountered. […]
> >> 
> >> Is there any reason etc/rc.d/fsck shouldn't automatically retry (up to
> 
> This is, in fact, the reason that I made the change I did. I was trying to put in a retry loop to rc.d/fsck, but found that I couldn’t get it to work because fsck and fsck_ffs were not exiting with non-zero status. The drop to single user is not really due to the specific (new) error code of 16, it is due to the fact that fsck_ffs is now exiting with a non-zero status when it hasn’t completely cleaned the file system;

Sure, but that's a regression IMO: before, I believe we'd successfully
mount the FS even without retrying fsck, and continue booting.

> /any/ non-zero status would cause the current rc.d/fsck script to go to single user. Prior to my change, fsck_ffs was exiting with a zero status even though it had not completely cleaned the filesystem and told the user to run it again.
> 
> > 
> > fsck_ffs already has a -R flag to automatically retry, wouldn't that be
> > a better mechanism for handling this new type of retry?
> 
> That’s true; however, there is currently no way to pass that flag through the filesystem-agnostic fsck wrapper called from rc.d/fsck to the filesystem-specific fsck_ffs program that it calls. One could implement a similar flag on the fsck wrapper to be passed along to the filesystem-specific checker, but I think fsck_ffs is the only one that currently implements such a flag. 

As was pointed out by others, this isn't true. In my experience it's
fsck -p that is exiting with status 16. It thus seems like it would be
desirable to add "-T ffs:-R" to the initial fsck invocation in
rc.d/fsck.