From: Dan Williams <dan.j.williams@intel.com>
To: neilb@suse.de, akpm@linux-foundation.org, christopher.leech@intel.com
Cc: linux-kernel@vger.kernel.org, linux-raid@vger.kernel.org
Subject: [PATCH 08/16] md: move raid5 compute block operations to raid5_run_ops
Date: Tue, 01 May 2007 23:18:41 -0700 [thread overview]
Message-ID: <20070502061841.7066.3911.stgit@dwillia2-linux.ch.intel.com> (raw)
In-Reply-To: <20070502060949.7066.357.stgit@dwillia2-linux.ch.intel.com>
handle_stripe sets STRIPE_OP_COMPUTE_BLK to request servicing from
raid5_run_ops. It also sets a flag for the block being computed to let
other parts of handle_stripe submit dependent operations. raid5_run_ops
guarantees that the compute operation completes before any dependent
operation starts.
Changelog:
* remove the req_compute BUG_ON
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
---
drivers/md/raid5.c | 126 +++++++++++++++++++++++++++++++++++++++-------------
1 files changed, 94 insertions(+), 32 deletions(-)
diff --git a/drivers/md/raid5.c b/drivers/md/raid5.c
index 03a435d..844bd9b 100644
--- a/drivers/md/raid5.c
+++ b/drivers/md/raid5.c
@@ -1998,7 +1998,7 @@ static void handle_stripe5(struct stripe_head *sh)
int i;
int syncing, expanding, expanded;
int locked=0, uptodate=0, to_read=0, to_write=0, failed=0, written=0;
- int non_overwrite = 0;
+ int compute=0, req_compute=0, non_overwrite=0;
int failed_num=0;
struct r5dev *dev;
unsigned long pending=0;
@@ -2050,8 +2050,8 @@ static void handle_stripe5(struct stripe_head *sh)
/* now count some things */
if (test_bit(R5_LOCKED, &dev->flags)) locked++;
if (test_bit(R5_UPTODATE, &dev->flags)) uptodate++;
+ if (test_bit(R5_Wantcompute, &dev->flags)) BUG_ON(++compute > 1);
-
if (dev->toread) to_read++;
if (dev->towrite) {
to_write++;
@@ -2206,31 +2206,83 @@ static void handle_stripe5(struct stripe_head *sh)
* parity, or to satisfy requests
* or to load a block that is being partially written.
*/
- if (to_read || non_overwrite || (syncing && (uptodate < disks)) || expanding) {
- for (i=disks; i--;) {
- dev = &sh->dev[i];
- if (!test_bit(R5_LOCKED, &dev->flags) && !test_bit(R5_UPTODATE, &dev->flags) &&
- (dev->toread ||
- (dev->towrite && !test_bit(R5_OVERWRITE, &dev->flags)) ||
- syncing ||
- expanding ||
- (failed && (sh->dev[failed_num].toread ||
- (sh->dev[failed_num].towrite && !test_bit(R5_OVERWRITE, &sh->dev[failed_num].flags))))
- )
- ) {
- /* we would like to get this block, possibly
- * by computing it, but we might not be able to
+ if (to_read || non_overwrite || (syncing && (uptodate + compute < disks)) || expanding ||
+ test_bit(STRIPE_OP_COMPUTE_BLK, &sh->ops.pending)) {
+
+ /* Clear completed compute operations. Parity recovery
+ * (STRIPE_OP_MOD_REPAIR_PD) implies a write-back which is handled
+ * later on in this routine
+ */
+ if (test_bit(STRIPE_OP_COMPUTE_BLK, &sh->ops.complete) &&
+ !test_bit(STRIPE_OP_MOD_REPAIR_PD, &sh->ops.pending)) {
+ clear_bit(STRIPE_OP_COMPUTE_BLK, &sh->ops.complete);
+ clear_bit(STRIPE_OP_COMPUTE_BLK, &sh->ops.ack);
+ clear_bit(STRIPE_OP_COMPUTE_BLK, &sh->ops.pending);
+ }
+
+ /* look for blocks to read/compute, skip this if a compute
+ * is already in flight, or if the stripe contents are in the
+ * midst of changing due to a write
+ */
+ if (!test_bit(STRIPE_OP_COMPUTE_BLK, &sh->ops.pending) &&
+ !test_bit(STRIPE_OP_PREXOR, &sh->ops.pending) &&
+ !test_bit(STRIPE_OP_POSTXOR, &sh->ops.pending)) {
+ for (i=disks; i--;) {
+ dev = &sh->dev[i];
+
+ /* don't schedule compute operations or reads on
+ * the parity block while a check is in flight
*/
- if (uptodate == disks-1) {
- PRINTK("Computing block %d\n", i);
- compute_block(sh, i);
- uptodate++;
- } else if (test_bit(R5_Insync, &dev->flags)) {
- set_bit(R5_LOCKED, &dev->flags);
- set_bit(R5_Wantread, &dev->flags);
- locked++;
- PRINTK("Reading block %d (sync=%d)\n",
- i, syncing);
+ if ((i == sh->pd_idx) && test_bit(STRIPE_OP_CHECK, &sh->ops.pending))
+ continue;
+
+ if (!test_bit(R5_LOCKED, &dev->flags) && !test_bit(R5_UPTODATE, &dev->flags) &&
+ (dev->toread ||
+ (dev->towrite && !test_bit(R5_OVERWRITE, &dev->flags)) ||
+ syncing ||
+ expanding ||
+ (failed && (sh->dev[failed_num].toread ||
+ (sh->dev[failed_num].towrite &&
+ !test_bit(R5_OVERWRITE, &sh->dev[failed_num].flags))))
+ )
+ ) {
+ /* 1/ We would like to get this block, possibly
+ * by computing it, but we might not be able to.
+ *
+ * 2/ Since parity check operations potentially
+ * make the parity block !uptodate it will need
+ * to be refreshed before any compute operations
+ * on data disks are scheduled.
+ *
+ * 3/ We hold off parity block re-reads until check
+ * operations have quiesced.
+ */
+ if ((uptodate == disks-1) && !test_bit(STRIPE_OP_CHECK, &sh->ops.pending)) {
+ set_bit(STRIPE_OP_COMPUTE_BLK, &sh->ops.pending);
+ set_bit(R5_Wantcompute, &dev->flags);
+ sh->ops.target = i;
+ req_compute = 1;
+ sh->ops.count++;
+ /* Careful: from this point on 'uptodate' is in the eye of
+ * raid5_run_ops which services 'compute' operations before
+ * writes. R5_Wantcompute flags a block that will be R5_UPTODATE
+ * by the time it is needed for a subsequent operation.
+ */
+ uptodate++;
+ break; /* uptodate + compute == disks */
+ } else if ((uptodate < disks-1) && test_bit(R5_Insync, &dev->flags)) {
+ /* Note: we hold off compute operations while checks are in flight,
+ * but we still prefer 'compute' over 'read' hence we only read if
+ * (uptodate < disks-1)
+ */
+ set_bit(R5_LOCKED, &dev->flags);
+ set_bit(R5_Wantread, &dev->flags);
+ if (!test_and_set_bit(STRIPE_OP_IO, &sh->ops.pending))
+ sh->ops.count++;
+ locked++;
+ PRINTK("Reading block %d (sync=%d)\n",
+ i, syncing);
+ }
}
}
}
@@ -2305,7 +2357,7 @@ static void handle_stripe5(struct stripe_head *sh)
if ((dev->towrite || i == sh->pd_idx) &&
(!test_bit(R5_LOCKED, &dev->flags)
) &&
- !test_bit(R5_UPTODATE, &dev->flags)) {
+ !(test_bit(R5_UPTODATE, &dev->flags) || test_bit(R5_Wantcompute, &dev->flags))) {
if (test_bit(R5_Insync, &dev->flags)
/* && !(!mddev->insync && i == sh->pd_idx) */
)
@@ -2316,7 +2368,7 @@ static void handle_stripe5(struct stripe_head *sh)
if (!test_bit(R5_OVERWRITE, &dev->flags) && i != sh->pd_idx &&
(!test_bit(R5_LOCKED, &dev->flags)
) &&
- !test_bit(R5_UPTODATE, &dev->flags)) {
+ !(test_bit(R5_UPTODATE, &dev->flags) || test_bit(R5_Wantcompute, &dev->flags))) {
if (test_bit(R5_Insync, &dev->flags)) rcw++;
else rcw += 2*disks;
}
@@ -2329,7 +2381,8 @@ static void handle_stripe5(struct stripe_head *sh)
for (i=disks; i--;) {
dev = &sh->dev[i];
if ((dev->towrite || i == sh->pd_idx) &&
- !test_bit(R5_LOCKED, &dev->flags) && !test_bit(R5_UPTODATE, &dev->flags) &&
+ !test_bit(R5_LOCKED, &dev->flags) &&
+ !(test_bit(R5_UPTODATE, &dev->flags) || test_bit(R5_Wantcompute, &dev->flags)) &&
test_bit(R5_Insync, &dev->flags)) {
if (test_bit(STRIPE_PREREAD_ACTIVE, &sh->state))
{
@@ -2348,7 +2401,8 @@ static void handle_stripe5(struct stripe_head *sh)
for (i=disks; i--;) {
dev = &sh->dev[i];
if (!test_bit(R5_OVERWRITE, &dev->flags) && i != sh->pd_idx &&
- !test_bit(R5_LOCKED, &dev->flags) && !test_bit(R5_UPTODATE, &dev->flags) &&
+ !test_bit(R5_LOCKED, &dev->flags) &&
+ !(test_bit(R5_UPTODATE, &dev->flags) || test_bit(R5_Wantcompute, &dev->flags)) &&
test_bit(R5_Insync, &dev->flags)) {
if (test_bit(STRIPE_PREREAD_ACTIVE, &sh->state))
{
@@ -2363,8 +2417,16 @@ static void handle_stripe5(struct stripe_head *sh)
}
}
/* now if nothing is locked, and if we have enough data, we can start a write request */
- if (locked == 0 && (rcw == 0 ||rmw == 0) &&
- !test_bit(STRIPE_BIT_DELAY, &sh->state))
+ /* since handle_stripe can be called at any time we need to handle the case
+ * where a compute block operation has been submitted and then a subsequent
+ * call wants to start a write request. raid5_run_ops only handles the case where
+ * compute block and postxor are requested simultaneously. If this
+ * is not the case then new writes need to be held off until the compute
+ * completes.
+ */
+ if ((req_compute || !test_bit(STRIPE_OP_COMPUTE_BLK, &sh->ops.pending)) &&
+ (locked == 0 && (rcw == 0 ||rmw == 0) &&
+ !test_bit(STRIPE_BIT_DELAY, &sh->state)))
locked += handle_write_operations5(sh, rcw == 0, 0);
}
next prev parent reply other threads:[~2007-05-02 6:18 UTC|newest]
Thread overview: 24+ messages / expand[flat|nested] mbox.gz Atom feed top
2007-05-02 6:14 [PATCH 00/16] raid acceleration and asynchronous offload api for 2.6.22 Dan Williams
2007-05-02 6:15 ` [PATCH 01/16] dmaengine: add base support for the async_tx api Dan Williams
2007-05-02 6:15 ` [PATCH 02/16] dmaengine: move channel management to the client Dan Williams
2007-05-02 6:16 ` [PATCH 03/16] ARM: Add drivers/dma to arch/arm/Kconfig Dan Williams
2007-05-02 6:16 ` [PATCH 04/16] dmaengine: add the async_tx api Dan Williams
2007-05-02 6:17 ` [PATCH 05/16] md: add raid5_run_ops and support routines Dan Williams
2007-05-02 6:17 ` [PATCH 06/16] md: use raid5_run_ops for stripe cache operations Dan Williams
2007-05-02 6:18 ` [PATCH 07/16] md: move write operations to raid5_run_ops Dan Williams
2007-05-02 6:18 ` Dan Williams [this message]
2007-05-02 6:19 ` [PATCH 09/16] md: move raid5 parity checks " Dan Williams
2007-05-02 6:19 ` [PATCH 10/16] md: satisfy raid5 read requests via raid5_run_ops Dan Williams
2007-05-02 6:20 ` [PATCH 11/16] md: use async_tx and raid5_run_ops for raid5 expansion operations Dan Williams
2007-05-02 6:20 ` [PATCH 12/16] md: move raid5 io requests to raid5_run_ops Dan Williams
2007-05-02 6:21 ` [PATCH 13/16] md: remove raid5 compute_block and compute_parity5 Dan Williams
2007-05-02 6:21 ` [PATCH 14/16] dmaengine: driver for the iop32x, iop33x, and iop13xx raid engines Dan Williams
2007-05-02 6:22 ` [PATCH 15/16] iop13xx: Surface the iop13xx adma units to the iop-adma driver Dan Williams
2007-05-02 6:22 ` [PATCH 16/16] iop3xx: Surface the iop3xx DMA and AAU " Dan Williams
2007-05-02 6:55 ` [PATCH 00/16] raid acceleration and asynchronous offload api for 2.6.22 Nick Piggin
2007-05-02 15:45 ` Williams, Dan J
2007-05-02 15:55 ` Justin Piszcz
2007-05-02 16:17 ` Williams, Dan J
2007-05-02 16:19 ` Justin Piszcz
2007-05-02 16:36 ` Dagfinn Ilmari Mannsåker
2007-05-02 16:42 ` Williams, Dan J
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20070502061841.7066.3911.stgit@dwillia2-linux.ch.intel.com \
--to=dan.j.williams@intel.com \
--cc=akpm@linux-foundation.org \
--cc=christopher.leech@intel.com \
--cc=linux-kernel@vger.kernel.org \
--cc=linux-raid@vger.kernel.org \
--cc=neilb@suse.de \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).