lkml.org 
[lkml]   [2008]   [Jun]   [27]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
Patches in this message
/
From
Date
Subject[PATCH 019 of 29] md: md: kill STRIPE_OP_MOD_DMA in raid5 offload

From: Dan Williams <dan.j.williams@intel.com>

This micro-optimization allowed the raid code to skip a re-read of the
parity block after checking parity. It took advantage of the fact that
xor-offload-engines have their own internal result buffer and can check
parity without writing to memory. Remove it for the following reasons:

1/ It is a layering violation for MD to need to manage the DMA and
non-DMA paths within async_xor_zero_sum
2/ Bad precedent to toggle the 'ops' flags outside the lock
3/ Hard to realize a performance gain as reads will not need an updated
parity block and writes will dirty it anyways.

Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Signed-off-by: Neil Brown <neilb@suse.de>

### Diffstat output
./drivers/md/raid5.c | 10 ----------
./include/linux/raid/raid5.h | 2 --
2 files changed, 12 deletions(-)
diff .prev/drivers/md/raid5.c ./drivers/md/raid5.c
--- .prev/drivers/md/raid5.c 2008-06-27 16:20:58.000000000 +1000
+++ ./drivers/md/raid5.c 2008-06-27 16:22:03.000000000 +1000
@@ -837,15 +837,10 @@ ops_run_postxor(struct stripe_head *sh,
static void ops_complete_check(void *stripe_head_ref)
{
struct stripe_head *sh = stripe_head_ref;
- int pd_idx = sh->pd_idx;

pr_debug("%s: stripe %llu\n", __func__,
(unsigned long long)sh->sector);

- if (test_and_clear_bit(STRIPE_OP_MOD_DMA_CHECK, &sh->ops.pending) &&
- sh->ops.zero_sum_result == 0)
- set_bit(R5_UPTODATE, &sh->dev[pd_idx].flags);
-
set_bit(STRIPE_OP_CHECK, &sh->ops.complete);
set_bit(STRIPE_HANDLE, &sh->state);
release_stripe(sh);
@@ -873,11 +868,6 @@ static void ops_run_check(struct stripe_
tx = async_xor_zero_sum(xor_dest, xor_srcs, 0, count, STRIPE_SIZE,
&sh->ops.zero_sum_result, 0, NULL, NULL, NULL);

- if (tx)
- set_bit(STRIPE_OP_MOD_DMA_CHECK, &sh->ops.pending);
- else
- clear_bit(STRIPE_OP_MOD_DMA_CHECK, &sh->ops.pending);
-
atomic_inc(&sh->count);
tx = async_trigger_callback(ASYNC_TX_DEP_ACK | ASYNC_TX_ACK, tx,
ops_complete_check, sh);
diff .prev/include/linux/raid/raid5.h ./include/linux/raid/raid5.h
--- .prev/include/linux/raid/raid5.h 2008-06-27 16:20:33.000000000 +1000
+++ ./include/linux/raid/raid5.h 2008-06-27 16:22:03.000000000 +1000
@@ -267,10 +267,8 @@ struct r6_state {

/* modifiers to the base operations
* STRIPE_OP_MOD_REPAIR_PD - compute the parity block and write it back
- * STRIPE_OP_MOD_DMA_CHECK - parity is not corrupted by the check
*/
#define STRIPE_OP_MOD_REPAIR_PD 7
-#define STRIPE_OP_MOD_DMA_CHECK 8

/*
* Plugging:

\
 
 \ /
  Last update: 2008-06-27 09:01    [from the cache]
©2003-2011 Jasper Spaans