The message is meant as a safety check to make sure that we properly account for all inodes that the kernel module allocates (which is a good thing). However, it keeps the counters globally rather than per super block. This means that you can trigger the message by just doing this:
mount -t pvfs2 tcp://localhost:3334/pvfs2-fs /mnt/pvfs2
mount -t pvfs2 tcp://localhost:3334/pvfs2-fs /mnt/pvfs2-a
mount -t pvfs2 tcp://localhost:3334/pvfs2-fs /mnt/pvfs2-b
mount -t pvfs2 tcp://localhost:3334/pvfs2-fs /mnt/pvfs2-c
umount /mnt/pvfs2-c
umount /mnt/pvfs2-b
umount /mnt/pvfs2-a
umount /mnt/pvfs2
dmesg
[92701.204000] pvfs2_kill_sb: (WARNING) number of inode allocs (135) != number of inode deallocs (132)
[92702.704000] pvfs2_kill_sb: (WARNING) number of inode allocs (135) != number of inode deallocs (133)
[92703.960000] pvfs2_kill_sb: (WARNING) number of inode allocs (135) != number of inode deallocs (134)
I think we probably can just move "pvfs2_inode_alloc_count" and "pvfs2_inode_dealloc_count" to be tracked within the sb struct.
-Phil