On Thu, Oct 13, 2011 at 8:21 AM, David Howells <dhowells@xxxxxxxxxx> wrote: > Mark Moseley <moseleymark@xxxxxxxxx> wrote: > >> So on a cleared cache with SLAB, it took a while but this finally came >> up. One interesting thing is that at some point, it logged this: >> >> [13461.605871] [httpd ] <== __fscache_read_or_alloc_pages() = -ENOBUFS >> [invalidating] > > That's okay. Basically, a read-from-cache operation was rejected because the > cache object was in the early phase of being invalidated. I kept it simple > here - the read might complete next time it is tried, but it's just a cache so > that shouldn't matter. Ok, noted >> It was a while from when it logged that until when I happened to check >> on the box again, but when I did (shortly before this traceback), >> despite constant NFS activity, nothing in the fscache cache was >> getting written out (i.e. the used bytes on the partition stopped >> changing), and without any messages about withdrawing the cache or >> anythin. > > Did you look at /proc/fs/fscache/stats at all? I didn't but I can repeat it. Which of the stats in /proc/fs/fscache/stats would be best to track? >> [20839.802118] kernel BUG at fs/fscache/object-list.c:83! >> [20839.802733] invalid opcode: 0000 [#1] SMP > > That fits with the previous BUG elsewhere in object-list.c. It sounds like > there's a refcounting problem somewhere. Any sys or proc settings I should turn on to track that? -- Linux-cachefs mailing list Linux-cachefs@xxxxxxxxxx https://www.redhat.com/mailman/listinfo/linux-cachefs
