bad inode causes xfs_bulkstat to loop

[Roger Willcocks <willcor@xxxxxxxxx>]

xfs_bulkstat can get into a tight loop if there are bad inodes. (Seen
running xfs_fsr - it took 100% of one cpu on a dual cpu machine.
Worked fine after xfs_repair.)

It looks like the 'break' caused by the xfs_itobp call in xfs_bulkstat
returning an error means no terminating condition is set for the outer
loop.

The break looks slightly suspicious anyway - shouldn't the inode
simply be ignored? (Ref. the comment at the top of
xfs_bulkstat_single).

Perhaps add 'ubleft = 0; rval = error;' before the break (as happens
if the formatter returns 'bulkstat_rv_giveup') ?

--
Rger