I think I've found a limitation in the badblocks implementation
(block/badblocks.c), which we now also use for nvdimm badblocks.
Consider the following operations:
badblocks_set(bb, 32, 1);
badblocks_set(bb, 34, 1);
badblocks_set(bb, 36, 1);
badblocks_show will now correctly report:
32 1
34 1
36 1
Now if I do:
badblocks_set(bb, 32, 12);
Ideally, this should collapse all ranges into a single (32, 12).. But
looks like badblocks_set only merges one set of adjacent mergeable
ranges, and then returns, resulting in:
32 3
36 1
Now if I add the same (32, 12) range again, I get:
32 5
as one more set is merged. And finally after adding (32, 12) one final
time, I get the expected
32 12.
Neil - do you think it will be relatively straightforward to extend
badblocks_set to be aware of multi-overlap ranges like above? I haven't
really looked at the internals of it (I'll take a look tomorrow), but I
wanted to get this out in the meantime.
Thanks,
-Vishal
Show replies by date