linux-kernel - Re: FS-Cache: Duplicate cookie detected

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <alpine.DEB.2.21.999.1903112235120.21644@trent.utfs.org>
Date:   Mon, 11 Mar 2019 23:07:36 -0700 (PDT)
From:   Christian Kujau <lists@...dbynature.de>
To:     David Howells <dhowells@...hat.com>
cc:     linux-kernel@...r.kernel.org,
        Anna Schumaker <anna.schumaker@...app.com>,
        Steve Dickson <steved@...hat.com>
Subject: Re: FS-Cache: Duplicate cookie detected

On Mon, 11 Mar 2019, David Howells wrote:
> I've a couple more patches for you - one a bugfix and one that will print more
> information.  They don't actually affect the problem you're seeing.  I'll post
> them as replies to this message.

Thanks for the patches. I've applied all three to v5.0 and ran 
"nfstest_cache" and was able to reproduce the messages. Please note that 
I'm only running "nfstest_cache" because it's somehow able to reproduce 
the message reliably - otherwise the message just shows up once or twice 
in syslog, but I didn't know how to reproduce it.

But I noticed something else this time, and I did not notice that before: 
while running nfstest_cache, the "duplicate cookie" messages were only 
triggered when my other, non-test mount was also mounted during the test. 
Let me describe my F29 test VM again:

* VM boots, and /usr/local/src gets mounted via NFS, read-only, and
  with w/o fsc options. cachefilesd isn't even installed here.

* I run nfstest_cache and apparently it's mounting the same NFS export
  from the server to /mnt/t, as a readonly mount.

So two mounts, one in /usr/local/src, the other in /mnt/t, both readonly 
and both w/o "fsc", but the "duplicate cookie" message is only printed 
when /usr/local/src was mounted. If /usr/local/src wasn't mounted, the 
test would complete[0] and no "duplicate message was printed. And then I 
noticed:

----------------------------------------------------------
$ mount | tail -2 | fold
horus:/usr/local/src on /usr/local/src type nfs4 
(ro,relatime,vers=4.2,rsize=1048576,wsize=1048576,namlen=255, 
hard,proto=tcp,timeo=600,retrans=2,sec=sys,clientaddr=192.168.56.139, 
local_lock=none,addr=192.168.0.115)

horus:/ on /mnt/t type nfs4 
(rw,relatime,vers=4.1,rsize=4096,wsize=4096,namlen=255, 
hard,proto=tcp,timeo=600,retrans=2,sec=sys,clientaddr=192.168.56.139, 
local_lock=none,addr=192.168.0.115) 
----------------------------------------------------------

My /usr/local/src mount was mounted with vers=4.2 (default), while 
nfstest_cache was mounting its test-mount with vers=4.1! Apart from the 
different rsize/wsize values, the version number stood out. And indeed, 
when I mount my regular NFS mount /usr/local/src with vers=4.1, the 
"duplicate cookie" is no longer printed.

For simplicity, I've attached two logs to this email:

* nfs_no-mount.txt.xz - showing /proc/fs/nfsfs/volumes and
  /proc/fs/fscache/stats every 0.01 seconds, while running nfstest_cache
  in another terminal. Note that no duplicate "cookie messages" were 
  triggered, as /usr/local/src was not mounted.

* nfs_with-mount.txt.xz - same, but here /usr/local/src was mounted (and 
  defaulted to vers=4.2), and thus "duplicate cookie" messages were 
  printed.

I fear that all this may complicate this strange behaviour, and now we're 
examining NFS mount versions, but I only noticed that now, not earlier :-\

I can't comment on the patches much, as you mentioned they won't make the 
message go away, but I hope it printed more details now.

Thanks,
Christian.

[0] Again, I'm using nfstest_cache only to trigger the message. Everytime
    I execute it, the test fails, because I think it expects a rw-mount:

 $ nfstest_cache --server horus --client fedora0 --runtest=acregmin_attr
 *** Verify consistency of attribute caching with NFSv4.1 on a file 
 acregmin = 10
     TEST: Running test 'acregmin_attr'
     FAIL: Traceback (most recent call last):
            File "/usr/bin/nfstest_cache", line 199, in do_file_test
              fdw = open(self.absfile, "w")
            IOError: [Errno 30] Read-only file system: 
 '/mnt/t/nfstest_cache_20190311223404_f_1'
    TIME: 4.497078s
 1 tests (0 passed, 1 failed)
 Total time: 5.529826s

-- 
BOFH excuse #209:

Only people with names beginning with 'A' are getting mail this week (a la Microsoft)
Download attachment "nfs_with-mount.txt.xz" of type "application/x-xz" (5284 bytes)

Download attachment "nfs_no-mount.txt.xz" of type "application/x-xz" (4820 bytes)