diff options
| author | Jens Axboe <jens.axboe@oracle.com> | 2008-09-25 11:37:50 +0200 |
|---|---|---|
| committer | Jens Axboe <jens.axboe@oracle.com> | 2008-10-09 08:56:19 +0200 |
| commit | f7d7b7a7a3db6526a84ea755c1c54a051e9a52de (patch) | |
| tree | 4d02c3abfca1c6175527473dff12769b09b630bb /block | |
| parent | 8bff7c6b0f63c7ee9c5e3a076338d74125b8debb (diff) | |
| download | kernel-common-f7d7b7a7a3db6526a84ea755c1c54a051e9a52de.tar.gz kernel-common-f7d7b7a7a3db6526a84ea755c1c54a051e9a52de.tar.bz2 kernel-common-f7d7b7a7a3db6526a84ea755c1c54a051e9a52de.zip | |
block: as/cfq ssd idle check update
We really need to know about the hardware tagging support as well,
since if the SSD does not do tagging then we still want to idle.
Otherwise have the same dependent sync IO vs flooding async IO
problem as on rotational media.
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Diffstat (limited to 'block')
| -rw-r--r-- | block/as-iosched.c | 4 | ||||
| -rw-r--r-- | block/cfq-iosched.c | 6 |
2 files changed, 7 insertions, 3 deletions
diff --git a/block/as-iosched.c b/block/as-iosched.c index 4c6fafbba933..71f0abb219ee 100644 --- a/block/as-iosched.c +++ b/block/as-iosched.c @@ -745,11 +745,13 @@ static int as_can_break_anticipation(struct as_data *ad, struct request *rq) */ static int as_can_anticipate(struct as_data *ad, struct request *rq) { +#if 0 /* disable for now, we need to check tag level as well */ /* * SSD device without seek penalty, disable idling */ - if (blk_queue_nonrot(ad->q)) + if (blk_queue_nonrot(ad->q)) axman return 0; +#endif if (!ad->io_context) /* diff --git a/block/cfq-iosched.c b/block/cfq-iosched.c index 03a5953bb5df..6a062eebbd15 100644 --- a/block/cfq-iosched.c +++ b/block/cfq-iosched.c @@ -879,9 +879,11 @@ static void cfq_arm_slice_timer(struct cfq_data *cfqd) unsigned long sl; /* - * SSD device without seek penalty, disable idling + * SSD device without seek penalty, disable idling. But only do so + * for devices that support queuing, otherwise we still have a problem + * with sync vs async workloads. */ - if (blk_queue_nonrot(cfqd->queue)) + if (blk_queue_nonrot(cfqd->queue) && cfqd->hw_tag) return; WARN_ON(!RB_EMPTY_ROOT(&cfqq->sort_list)); |