The problem is that since OpenSolaris commit
l2ad_hand is kept aligned based on ashift (which is derived from
the cache device's logical and physical block sizes).
So, the hand could be advanced by more than the sum of b_asize-s
of the written L2ARC buffers. This is because b_asize is a misnomer
at the moment as it does not always represent the allocated size:
if a buffer is compressed, then the compressed size is properly rounded,
but if the compression fails or it is not applied, then the original
size is kept and it could be smaller than what ashift requires.
For the same reasons arcstat_l2_asize and the reported used space
on the cache device could be smaller than the actual allocated size
if ashift > 9. That problem is not fixed by this change.
This change only ensures that l2ad_hand is not advanced by more
than target_sz. Otherwise we would overwrite active (unevicted)
L2ARC buffers. That problem is manifested as growing l2_cksum_bad
and l2_io_error counters.
This change also changes 'p' prefix to 'a' prefix in a few places
where variables represent allocated rather than physical size.
The resolved problem may also result in the reported allocated size
being greater than the cache device's capacity, because of the
overwritten buffers (more than one buffer claiming the same disk
PR: 195746 (possibly related)