Re: [autofs] Re: autofs4 looks up wrong path element when ghostingis enabled

[Jeff Moyer]

==> Regarding Re: [autofs] Re: autofs4 looks up wrong path element when ghosting is enabled; Ian Kent <raven@xxxxxxxxxx> adds:

raven> On Mon, 31 Oct 2005, Ian Kent wrote:
>> On Mon, 31 Oct 2005, Jeff Moyer wrote:
>> 
>> > ==> Regarding Re: autofs4 looks up wrong path element when ghosting is
>> enabled; Ian Kent <raven@xxxxxxxxxx> adds:
>> > 
>> > raven> So to resolve this we need to ignore negative and unhashed
>> dentries > raven> when checking if directory dentry is empty.
>> > >>
>> > raven> Please test this patch and let me know how you go.  > >> OK,
>> I've finally got 'round to testing your patch.  It does fix the test >
>> >> case I was using.  My only concern is the potential for regressions.
>> > >> I'll try making sure all of my various maps still work as
>> advertised.
>> > 
>> > raven> I've spotted a regression with this patch.  It doesn't stop
>> autofs > raven> from appearing to function correctly. It causes mount
>> callbacks when > raven> they shouldn't made. So it seems that there is a
>> case when an autofs > raven> directory is, as yet unhashed, but should
>> be used.
>> > 
>> > I'm not sure I follow.  What do you mean it doesn't stop autofs from >
>> *appearing* to function correctly?  Do you have a reproducer that fails?
>> 
>> Any pseudo direct map will produce the behaviour.
>> 
>> It behaves as if the ghost option was not specified even when it
>> has. This is because the altered test for an empty directory is always
>> returning true even though the directory isn't empty.
>> 
>> I'm still trying to understand why this happens. In theory it's just not
>> the expected behaviour. I must be missing something in the directory
>> creation. I've been here before and looked several times and I just
>> can't see why it happens this way.

raven> I couldn't work out why this patch shouldn't work so I tried to
raven> duplicate the problem I saw before and I can't.

raven> I've tested the patch against autofs-4.1.3-123 (latest source rpm I
raven> could find) and autofs-4.1.4-8 with a couple of my patches (only on
raven> the 4.1.4 version) that shouldn't affect it and I can't seem to
raven> break it.

raven> I must have got my kernel modules mixed up. It probably means
raven> there's some backward compatibilty work to be done on the version 5
raven> module.

raven> Anyway, some more testing, as you suggested, would be great.

raven> The only other question is what to do about the cacheing of mount
raven> failures which has never worked by the look of it. We can remove it
raven> or fix it.

OK, good to know that it works.  I'm really not sure why we're caching
mount failures, either.  Perhaps it was thought to be a performance
optimization?  I see no reason to keep it around, to be honest.

Note that the patch you posted has been in Fedora Core 5's devel
repository since October 20th, and no bugs have been filed.

-Jeff
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/