Hi Paul,
I've been looking through the bitmap code again and found the
handling of resync a bit hard to wrap my brain around. The handling
of the counter here seems a bit non-obvious and possibly fragile.
I would like to make a change, but thought I should bounce it off you
first.
I would like to change the "RESYNC_MASK" bit to mean:
At least one block in this chunk is out-of-sync.
Then:
- When we read the bitmap from disk we set this bit and the
SHADOW_MASK, but leave the counter at zero.
- When we get a failed write, we set this bit, but still decrement
the counter.
- When we are performing a resync, we periodically clear the bit on
recently completed chunks.
- We only clear the SHADOW_MASK and on-disk bit when the counter
hits zero *and* this bit is clear.
I would find this approach a lot easier to understand. Are you OK
with it?
Also, I would like to move the bitmap_testbit in md_do_sync down into
the personality. This should make life easier for other
personalities like raid10 which use a very different approach for
resync than for recovery.
NeilBrown
-
To unsubscribe from this list: send the line "unsubscribe linux-raid" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
