Re: autofs4 looks up wrong path element when ghosting is enabled

[Jeff Moyer <jmoyer@redhat.com>]

==> Regarding Re: autofs4 looks up wrong path element when ghosting is enabled; Ian Kent <raven@themaw.net> adds:

raven> On Tue, 20 Sep 2005, Jeff Moyer wrote:
>> Hi, Ian, list,
>> 
>> I have a bug filed against autofs when ghosting is enabled.  The best way
>> to describe the bug is to walk through the reproducer, I guess.
>> 
>> Take the following maps, for example:
>> 
>> auto.master
>> /sbox	auto.sbox
>> 
>> auto.sbox:
>> src	segfault:/sbox/src/
>> 
>> Let's say that there is a file, id3_0.12.orig.tar.gz, in segfault:/sbox/src/.
>> 
>> To reproduce the problem, stop the nfs service on the server.
>> 
>> On the client, do an 'ls /sbox/src/id3_012.orig.tar.gz'.  This will fail,
>> as well it should.  However, if we look in the logs, we find this:
>> 
>> automount[1182]: handle_packet_missing: token 1, name src 
>> automount[1182]: attempting to mount entry /sbox/src
>> ...
>> automount[1481]: mount(nfs): calling mkdir_path /sbox/src
>> automount[1481]: mount(nfs): calling mount -t nfs -s-o tcp,intr,timeo=600,rsize=8192,wsize=8192,retrans=5 segfault:/sbox/src /sbox/src
>> automount[1481]: >> mount: RPC: Program not registered
>> automount[1481]: mount(nfs): add_bad_host: segfault:/sbox/src
>> automount[1481]: mount(nfs): nfs: mount failure segfault:/sbox/src on /sbox/src
>> automount[1481]: failed to mount /sbox/src
>> ...
>> automount[1182]: send_fail: token=1 
>> automount[1182]: handle_packet: type = 0 
>> automount[1182]: handle_packet_missing: token 2, name src/id3_0.12.orig.tar.gz 
>> automount[1182]: attempting to mount entry /sbox/src/id3_0.12.orig.tar.gz
>> 
>> Noteworthy are these last two lines!  Even though the mount failed, we are
>> continuing the lookup.  The culprit is here, in cached_lookup:
>> 
>> if (!dentry->d_op->d_revalidate(dentry, flags) && !d_invalidate(dentry)) { 
>> dput(dentry); 
>> dentry = NULL; 
>> } 
>> 
>> d_revalidate points to autofs4_revalidate, which calls try_to_fill_dentry,
>> which will return a status of 0.  Since ghosting is enabled,
>> d_invalidate(dentry) will return -EBUSY, and so we return the dentry to 
raven> the
>> caller, which then continues the lookup.
>> 
>> Ian, I'm not really sure how we can address this issue without VFS
>> changes.  Any ideas?
>> 

raven> I'm aware of this problem.
raven> I'm not sure how to deal with it yet.
raven> The case above is probably not that difficult to solve but if the last 
raven> component is a directory it's hard to work out it's a problem.

Ugh.  If you're thinking what I think you're thinking, that's an ugly hack.

raven> There's more information here than I've gathhered so far.

>> Oh, also note that, once the nfs service is started up again on the server,
>> the lookup of a specific file name will still fail!  In this case, the
>> daemon won't even be called.

raven> I'll have to check this out.
raven> It could be helpful.

Well, I've provided a reproducer.  If you'd like log output from the kernel
side, let me know.  I can certainly provide that.

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