raid5: adjust order of some operations in handle_stripe

This is to revert ef5b7c69b7a1b8b8744a6168b6f. handle_stripe_clean_event() handles finished stripes, which really should be the first thing to do. The original changelog says checking reconstruct_state should be the first as handle_stripe_clean_event can clear some dev->flags and impact checking reconstruct_state code path. It's unclear to me why this happens, because I thought written finish and reconstruct_state equals to *_result can't happen in the same time.

I also moved checking reconstruct_state code path after handle_stripe_dirtying. If that code sets reconstruct_state to reconstruct_state_idle, the order change will make us miss one handle_stripe_dirtying. But the stripe will be eventually handled again when written is finished.

Signed-off-by Shaohua Li <[email protected]>

---
 drivers/md/raid5.c |   74 ++++++++++++++++++++++++++---------------------------
 1 file changed, 37 insertions(+), 37 deletions(-)

Index: linux/drivers/md/raid5.c
===================================================================

drivers/md/raid5.c +37 -37

@@ -3747,43 +3747,6 @@ static void handle_stripe(struct stripe_
 			handle_failed_sync(conf, sh, &s);
 	}
 
-	/* Now we check to see if any write operations have recently
-	 * completed
-	 */
-	prexor = 0;
-	if (sh->reconstruct_state == reconstruct_state_prexor_drain_result)
-		prexor = 1;
-	if (sh->reconstruct_state == reconstruct_state_drain_result ||
-	    sh->reconstruct_state == reconstruct_state_prexor_drain_result) {
-		sh->reconstruct_state = reconstruct_state_idle;
-
-		/* All the 'written' buffers and the parity block are ready to
-		 * be written back to disk
-		 */
-		BUG_ON(!test_bit(R5_UPTODATE, &sh->dev[sh->pd_idx].flags) &&
-		       !test_bit(R5_Discard, &sh->dev[sh->pd_idx].flags));
-		BUG_ON(sh->qd_idx >= 0 &&
-		       !test_bit(R5_UPTODATE, &sh->dev[sh->qd_idx].flags) &&
-		       !test_bit(R5_Discard, &sh->dev[sh->qd_idx].flags));
-		for (i = disks; i--; ) {
-			struct r5dev *dev = &sh->dev[i];
-			if (test_bit(R5_LOCKED, &dev->flags) &&
-				(i == sh->pd_idx || i == sh->qd_idx ||
-				 dev->written)) {
-				pr_debug("Writing block %d\n", i);
-				set_bit(R5_Wantwrite, &dev->flags);
-				if (prexor)
-					continue;
-				if (!test_bit(R5_Insync, &dev->flags) ||
-				    ((i == sh->pd_idx || i == sh->qd_idx)  &&
-				     s.failed == 0))
-					set_bit(STRIPE_INSYNC, &sh->state);
-			}
-		}
-		if (test_and_clear_bit(STRIPE_PREREAD_ACTIVE, &sh->state))
-			s.dec_preread_active = 1;
-	}
-
 	/*
 	 * might be able to return some write requests if the parity blocks
 	 * are safe, or on a failed drive

@@ -3827,6 +3790,43 @@ static void handle_stripe(struct stripe_
 	if (s.to_write && !sh->reconstruct_state && !sh->check_state)
 		handle_stripe_dirtying(conf, sh, &s, disks);
 
+	/* Now we check to see if any write operations have recently
+	 * completed
+	 */
+	prexor = 0;
+	if (sh->reconstruct_state == reconstruct_state_prexor_drain_result)
+		prexor = 1;
+	if (sh->reconstruct_state == reconstruct_state_drain_result ||
+	    sh->reconstruct_state == reconstruct_state_prexor_drain_result) {
+		sh->reconstruct_state = reconstruct_state_idle;
+
+		/* All the 'written' buffers and the parity block are ready to
+		 * be written back to disk
+		 */
+		BUG_ON(!test_bit(R5_UPTODATE, &sh->dev[sh->pd_idx].flags) &&
+		       !test_bit(R5_Discard, &sh->dev[sh->pd_idx].flags));
+		BUG_ON(sh->qd_idx >= 0 &&
+		       !test_bit(R5_UPTODATE, &sh->dev[sh->qd_idx].flags) &&
+		       !test_bit(R5_Discard, &sh->dev[sh->qd_idx].flags));
+		for (i = disks; i--; ) {
+			struct r5dev *dev = &sh->dev[i];
+			if (test_bit(R5_LOCKED, &dev->flags) &&
+				(i == sh->pd_idx || i == sh->qd_idx ||
+				 dev->written)) {
+				pr_debug("Writing block %d\n", i);
+				set_bit(R5_Wantwrite, &dev->flags);
+				if (prexor)
+					continue;
+				if (!test_bit(R5_Insync, &dev->flags) ||
+				    ((i == sh->pd_idx || i == sh->qd_idx)  &&
+				     s.failed == 0))
+					set_bit(STRIPE_INSYNC, &sh->state);
+			}
+		}
+		if (test_and_clear_bit(STRIPE_PREREAD_ACTIVE, &sh->state))
+			s.dec_preread_active = 1;
+	}
+
 	/* maybe we need to check and possibly fix the parity for this stripe
 	 * Any reads will already have been scheduled, so we just see if enough
 	 * data is available.  The parity check is held off while parity

Quoted 8 lines · On Thu, 22 May 2014 19:24:31 +0800 Shaohua Li <[email protected]> wrote:

This is to revert ef5b7c69b7a1b8b8744a6168b6f. handle_stripe_clean_event() handles finished stripes, which really should be the first thing to do. The original changelog says checking reconstruct_state should be the first as handle_stripe_clean_event can clear some dev->flags and impact checking reconstruct_state code path. It's unclear to me why this happens, because I thought written finish and reconstruct_state equals to *_result can't happen in the same time.

"unclear to me" "I thought" are sufficient to justify a change, though they are certainly sufficient to ask a question.

Are you asking a question or submitting a change?

You may well be correct that if reconstruct_state is not reconstruct_state_idle, then handle_stripe_clean_event cannot possible be called. In that case, maybe we should change the code flow to make that more obvious, but certainly the changelog comment should be clear about exactly why.

Quoted 5 lines

I also moved checking reconstruct_state code path after handle_stripe_dirtying. If that code sets reconstruct_state to reconstruct_state_idle, the order change will make us miss one handle_stripe_dirtying. But the stripe will be eventually handled again when written is finished.

You haven't said here why this patch is a good thing, only why it isn't obviously bad. I really need some justification to make a change and you haven't provided any, at least not in this changelog comment.

Maybe we need a completely different approach.
Instead of repeatedly shuffling code inside handle_stripe(), how about we put all of handle_stripe inside a loop which runs as long as STRIPE_HANDLE is set and sh->count == 1.
ie.

	if (test_and_set_bit_lock(STRIPE_ACTIVE, &sh->state)) {
		/* already being handled, ensure it gets handled
		 * again when current action finishes */
		set_bit(STRIPE_HANDLE, &sh->state);
		return;
	}

        do {
	        clear_bit(STRIPE_HANDLE, &sh->state);
                __handle_stripe(sh);
        } while (test_bit(STRIPE_HANDLE, &sh->state)
                 && atomic_read(&sh->count) == 1);
	clear_bit_unlock(STRIPE_ACTIVE, &sh->state);

where the rest of the current handle_stripe() goes in to __handle_stripe().

Would that address your performance concerns, or is there still too much overhead?

Quoted 14 lines · NeilBrown

Quoted 8 lines · On Thu, 22 May 2014 19:24:31 +0800 Shaohua Li <[email protected]> wrote:

This is to revert ef5b7c69b7a1b8b8744a6168b6f. handle_stripe_clean_event() handles finished stripes, which really should be the first thing to do. The original changelog says checking reconstruct_state should be the first as handle_stripe_clean_event can clear some dev->flags and impact checking reconstruct_state code path. It's unclear to me why this happens, because I thought written finish and reconstruct_state equals to *_result can't happen in the same time.

"unclear to me" "I thought" are sufficient to justify a change, though they are certainly sufficient to ask a question.

Quoted 7 lines

Are you asking a question or submitting a change?

You may well be correct that if reconstruct_state is not reconstruct_state_idle, then handle_stripe_clean_event cannot possible be called. In that case, maybe we should change the code flow to make that more obvious, but certainly the changelog comment should be clear about exactly why.

I'm sorry, it's more like a question. I really didn't understand why we have ef5b7c69b7a1b8b8744a6168b6f, so I'm not 100% sure about. It would be great you can help share a hint.

Quoted 10 lines

Quoted 6 lines

I also moved checking reconstruct_state code path after handle_stripe_dirtying. If that code sets reconstruct_state to reconstruct_state_idle, the order change will make us miss one handle_stripe_dirtying. But the stripe will be eventually handled again when written is finished.

You haven't said here why this patch is a good thing, only why it isn't obviously bad. I really need some justification to make a change and you haven't provided any, at least not in this changelog comment.

ok, I'll add more about this.

Quoted 21 lines

Maybe we need a completely different approach.
Instead of repeatedly shuffling code inside handle_stripe(), how about we put all of handle_stripe inside a loop which runs as long as STRIPE_HANDLE is set and sh->count == 1.
ie.

	if (test_and_set_bit_lock(STRIPE_ACTIVE, &sh->state)) {
		/* already being handled, ensure it gets handled
		 * again when current action finishes */
		set_bit(STRIPE_HANDLE, &sh->state);
		return;
	}

        do {
	        clear_bit(STRIPE_HANDLE, &sh->state);
                __handle_stripe(sh);
        } while (test_bit(STRIPE_HANDLE, &sh->state)
                 && atomic_read(&sh->count) == 1);
	clear_bit_unlock(STRIPE_ACTIVE, &sh->state);

Quoted 4 lines

where the rest of the current handle_stripe() goes in to __handle_stripe().

Would that address your performance concerns, or is there still too much overhead?

Let me try. One issue here is we still have massive cache miss when checking stripe/dev state. I suppose this doesn't help but data should prove.

Quoted 23 lines · On Wed, 28 May 2014 11:45:07 +0800 Shaohua Li <[email protected]> wrote:

Quoted 14 lines · NeilBrown

Quoted 8 lines · On Thu, 22 May 2014 19:24:31 +0800 Shaohua Li <[email protected]> wrote:

This is to revert ef5b7c69b7a1b8b8744a6168b6f. handle_stripe_clean_event() handles finished stripes, which really should be the first thing to do. The original changelog says checking reconstruct_state should be the first as handle_stripe_clean_event can clear some dev->flags and impact checking reconstruct_state code path. It's unclear to me why this happens, because I thought written finish and reconstruct_state equals to *_result can't happen in the same time.

"unclear to me" "I thought" are sufficient to justify a change, though they are certainly sufficient to ask a question.

Quoted 7 lines

Are you asking a question or submitting a change?

You may well be correct that if reconstruct_state is not reconstruct_state_idle, then handle_stripe_clean_event cannot possible be called. In that case, maybe we should change the code flow to make that more obvious, but certainly the changelog comment should be clear about exactly why.

Quoted 3 lines

I'm sorry, it's more like a question. I really didn't understand why we have ef5b7c69b7a1b8b8744a6168b6f, so I'm not 100% sure about. It would be great you can help share a hint.

It's a while ago and I don't remember, but I suspect that I added that patch because handle_stripe_clean_event was about to change to clear R5_UPTODATE, and this code which was previously *after* handle_stripe_clean_event tested R5_UPTODATE (and could BUG if it wasn't set).

You may well be right that the two pieces of code cannot both run in the one invocation of handle_stripe(). I haven't analysed the code closely to be sure, but on casual reflection it seems likely. However we always need to be careful of races in unusual situations.

If that is correct, and if there are two (or more) different situations in which handle_stripe runs, maybe one after IO has completed and one after reconstruction has completed, and one when new devices have been added, then there might be value in clearly delineating these so we don't bother testing for cases that cannot happen.

If it is not correct, then your proposed change might be dangerous.

Quoted 39 lines

Quoted 10 lines

Quoted 6 lines

I also moved checking reconstruct_state code path after handle_stripe_dirtying. If that code sets reconstruct_state to reconstruct_state_idle, the order change will make us miss one handle_stripe_dirtying. But the stripe will be eventually handled again when written is finished.

You haven't said here why this patch is a good thing, only why it isn't obviously bad. I really need some justification to make a change and you haven't provided any, at least not in this changelog comment.

ok, I'll add more about this.

Quoted 25 lines

Maybe we need a completely different approach.
Instead of repeatedly shuffling code inside handle_stripe(), how about we put all of handle_stripe inside a loop which runs as long as STRIPE_HANDLE is set and sh->count == 1.
ie.

	if (test_and_set_bit_lock(STRIPE_ACTIVE, &sh->state)) {
		/* already being handled, ensure it gets handled
		 * again when current action finishes */
		set_bit(STRIPE_HANDLE, &sh->state);
		return;
	}

        do {
	        clear_bit(STRIPE_HANDLE, &sh->state);
                __handle_stripe(sh);
        } while (test_bit(STRIPE_HANDLE, &sh->state)
                 && atomic_read(&sh->count) == 1);
	clear_bit_unlock(STRIPE_ACTIVE, &sh->state);

where the rest of the current handle_stripe() goes in to __handle_stripe().

Would that address your performance concerns, or is there still too much overhead?

Quoted 2 lines

Let me try. One issue here is we still have massive cache miss when checking stripe/dev state. I suppose this doesn't help but data should prove.

That would be great - thanks.
If you can identify exactly where the cache misses are causing a problem, we might be able to optimise around that.