dm integrity: implement fair range locks
authorMikulas Patocka <mpatocka@redhat.com>
Tue, 3 Jul 2018 18:13:27 +0000 (20:13 +0200)
committerMike Snitzer <snitzer@redhat.com>
Fri, 27 Jul 2018 19:24:22 +0000 (15:24 -0400)
commit724376a04d1a63d145fdfe7b24c0b13128a3ffd6
tree517522cfdf4adc4dbc3c4d2487e1d68a592aac23
parent518748b1a744c496a657a5a7923e49e002a6f259
dm integrity: implement fair range locks

dm-integrity locks a range of sectors to prevent concurrent I/O or journal
writeback.  These locks were not fair - so that many small overlapping I/Os
could starve a large I/O indefinitely.

Fix this by making the range locks fair.  The ranges that are waiting are
added to the list "wait_list".  If a new I/O overlaps some of the waiting
I/Os, it is not dispatched, but it is also added to that wait list.
Entries on the wait list are processed in first-in-first-out order, so
that an I/O can't starve indefinitely.

Signed-off-by: Mikulas Patocka <mpatocka@redhat.com>
Signed-off-by: Mike Snitzer <snitzer@redhat.com>
drivers/md/dm-integrity.c