Skip to contentSkip to conversation

mm: zswap: increase reject_compress_poor but not reject_compress_fail if compression returns ENOSPC

PATCH Superseded Barry Song 5 participants 7 messages Feb 17, 2024 – Feb 18, 2024

Subsystems (2)

v2 of 3 Superseded Latest is v3 →

lore
Share & tools
lore

Barry Song

author
#0 lore raw

PATCH v2 1/1 mm: zswap: increase reject_compress_poor but not reject_compress_fail if compression returns ENOSPC

From: Barry Song <[email protected]>

We used to rely on the returned -ENOSPC of zpool_malloc() to increase reject_compress_poor. But the code wouldn't get to there after commit 744e1885922a ("crypto: scomp - fix req->dst buffer overflow") as the new code will goto out immediately after the special compression case happens. So there might be no longer a chance to execute zpool_malloc now. We are incorrectly increasing zswap_reject_compress_fail instead. Thus, we need to fix the counters handling right after compressions return ENOSPC. This patch also centralizes the counters handling for all of compress_poor, compress_fail and alloc_fail.

Fixes 744e1885922a ("crypto: scomp - fix req->dst buffer overflow")
Cc Chengming Zhou <[email protected]>
Cc Nhat Pham <[email protected]>
Cc Sergey Senozhatsky <[email protected]>

---
-v2:
* correct the fixes target according to Yosry, Chengming, Nhat's comments;
* centralize the counters handling according to Yosry's comment

 mm/zswap.c | 25 ++++++++++++-------------
 1 file changed, 12 insertions(+), 13 deletions(-)
mm/zswap.c +12 -13
index 350dd2fc8159..47cf07d56362 100644
@@ -1498,6 +1498,7 @@ bool zswap_store(struct folio *folio)
 	struct zswap_tree *tree = zswap_trees[type];
 	struct zswap_entry *entry, *dupentry;
 	struct scatterlist input, output;
+	int comp_ret = 0, alloc_ret = 0;
 	struct crypto_acomp_ctx *acomp_ctx;
 	struct obj_cgroup *objcg = NULL;
 	struct mem_cgroup *memcg = NULL;
@@ -1508,7 +1509,6 @@ bool zswap_store(struct folio *folio)
 	char *buf;
 	u8 *src, *dst;
 	gfp_t gfp;
-	int ret;
 
 	VM_WARN_ON_ONCE(!folio_test_locked(folio));
 	VM_WARN_ON_ONCE(!folio_test_swapcache(folio));
@@ -1621,28 +1621,20 @@ bool zswap_store(struct folio *folio)
 	 * but in different threads running on different cpu, we have different
 	 * acomp instance, so multiple threads can do (de)compression in parallel.
 	 */
-	ret = crypto_wait_req(crypto_acomp_compress(acomp_ctx->req), &acomp_ctx->wait);
+	comp_ret = crypto_wait_req(crypto_acomp_compress(acomp_ctx->req), &acomp_ctx->wait);
 	dlen = acomp_ctx->req->dlen;
 
-	if (ret) {
-		zswap_reject_compress_fail++;
+	if (comp_ret)
 		goto put_dstmem;
-	}
 
 	/* store */
 	zpool = zswap_find_zpool(entry);
 	gfp = __GFP_NORETRY | __GFP_NOWARN | __GFP_KSWAPD_RECLAIM;
 	if (zpool_malloc_support_movable(zpool))
 		gfp |= __GFP_HIGHMEM | __GFP_MOVABLE;
-	ret = zpool_malloc(zpool, dlen, gfp, &handle);
-	if (ret == -ENOSPC) {
-		zswap_reject_compress_poor++;
-		goto put_dstmem;
-	}
-	if (ret) {
-		zswap_reject_alloc_fail++;
+	alloc_ret = zpool_malloc(zpool, dlen, gfp, &handle);
+	if (alloc_ret)
 		goto put_dstmem;
-	}
 	buf = zpool_map_handle(zpool, handle, ZPOOL_MM_WO);
 	memcpy(buf, dst, dlen);
 	zpool_unmap_handle(zpool, handle);
@@ -1689,6 +1681,13 @@ bool zswap_store(struct folio *folio)
 	return true;
 
 put_dstmem:
+	if (comp_ret == -ENOSPC || alloc_ret == -ENOSPC)
+		zswap_reject_compress_poor++;
+	else if (comp_ret)
+		zswap_reject_compress_fail++;
+	else if (alloc_ret)
+		zswap_reject_alloc_fail++;
+
 	mutex_unlock(&acomp_ctx->mutex);
 put_pool:
 	zswap_pool_put(entry->pool);
Signature · 1 line
2.34.1
maintainer Reviewed-by on v2 1/1
#1 lore raw
Quoted 13 lines · Barry Song

From: Barry Song <[email protected]>

We used to rely on the returned -ENOSPC of zpool_malloc() to increase reject_compress_poor. But the code wouldn't get to there after commit 744e1885922a ("crypto: scomp - fix req->dst buffer overflow") as the new code will goto out immediately after the special compression case happens. So there might be no longer a chance to execute zpool_malloc now. We are incorrectly increasing zswap_reject_compress_fail instead. Thus, we need to fix the counters handling right after compressions return ENOSPC. This patch also centralizes the counters handling for all of compress_poor, compress_fail and alloc_fail.

Quoted 5 lines
Fixes 744e1885922a ("crypto: scomp - fix req->dst buffer overflow")
Cc Chengming Zhou <[email protected]>
Cc Nhat Pham <[email protected]>
Cc Sergey Senozhatsky <[email protected]>

LGTM.

Acked-by on v2 1/1
#2 lore raw
In reply to Barry Song
Quoted 17 lines · Barry Song

From: Barry Song <[email protected]>

We used to rely on the returned -ENOSPC of zpool_malloc() to increase reject_compress_poor. But the code wouldn't get to there after commit 744e1885922a ("crypto: scomp - fix req->dst buffer overflow") as the new code will goto out immediately after the special compression case happens. So there might be no longer a chance to execute zpool_malloc now. We are incorrectly increasing zswap_reject_compress_fail instead. Thus, we need to fix the counters handling right after compressions return ENOSPC. This patch also centralizes the counters handling for all of compress_poor, compress_fail and alloc_fail.

Fixes 744e1885922a ("crypto: scomp - fix req->dst buffer overflow")
Cc Chengming Zhou <[email protected]>
Cc Nhat Pham <[email protected]>
Cc Sergey Senozhatsky <[email protected]>
Quoted 5 lines

---
-v2:
* correct the fixes target according to Yosry, Chengming, Nhat's comments;
* centralize the counters handling according to Yosry's comment

Yet Yosry is not CC'd :P

The patch LGTM, but it won't apply on top of mm-unstable given the amount of zswap refactoring there. I would rebase on top of mm-unstable if I were you (and if you did, add mm-unstable in the subject prefix).

Thanks!

author
#3 lore raw
Quoted 25 lines · Yosry Ahmed
Quoted 17 lines · Barry Song

From: Barry Song <[email protected]>

We used to rely on the returned -ENOSPC of zpool_malloc() to increase reject_compress_poor. But the code wouldn't get to there after commit 744e1885922a ("crypto: scomp - fix req->dst buffer overflow") as the new code will goto out immediately after the special compression case happens. So there might be no longer a chance to execute zpool_malloc now. We are incorrectly increasing zswap_reject_compress_fail instead. Thus, we need to fix the counters handling right after compressions return ENOSPC. This patch also centralizes the counters handling for all of compress_poor, compress_fail and alloc_fail.

Fixes 744e1885922a ("crypto: scomp - fix req->dst buffer overflow")
Cc Chengming Zhou <[email protected]>
Cc Nhat Pham <[email protected]>
Cc Sergey Senozhatsky <[email protected]>
Quoted 5 lines

---
-v2:
* correct the fixes target according to Yosry, Chengming, Nhat's comments;
* centralize the counters handling according to Yosry's comment

Quoted 1 line

Yet Yosry is not CC'd :P

terribly sorry. I thought you were in my git send-email list ... but you were not...

Quoted 4 lines

The patch LGTM, but it won't apply on top of mm-unstable given the amount of zswap refactoring there. I would rebase on top of mm-unstable if I were you (and if you did, add mm-unstable in the subject prefix).

This patch has a "fixes" tag, so I assume it should be also in 6.8?

Quoted 2 lines

thanks!

Quoted 2 lines

Thanks!

#4 lore raw
Quoted 27 lines · Barry Song
Quoted 25 lines · Yosry Ahmed
Quoted 22 lines · Barry Song

From: Barry Song <[email protected]>

We used to rely on the returned -ENOSPC of zpool_malloc() to increase reject_compress_poor. But the code wouldn't get to there after commit 744e1885922a ("crypto: scomp - fix req->dst buffer overflow") as the new code will goto out immediately after the special compression case happens. So there might be no longer a chance to execute zpool_malloc now. We are incorrectly increasing zswap_reject_compress_fail instead. Thus, we need to fix the counters handling right after compressions return ENOSPC. This patch also centralizes the counters handling for all of compress_poor, compress_fail and alloc_fail.

Fixes 744e1885922a ("crypto: scomp - fix req->dst buffer overflow")
Cc Chengming Zhou <[email protected]>
Cc Nhat Pham <[email protected]>
Cc Sergey Senozhatsky <[email protected]>

---
-v2:
* correct the fixes target according to Yosry, Chengming, Nhat's comments;
* centralize the counters handling according to Yosry's comment

Quoted 4 lines
Quoted 1 line

Yet Yosry is not CC'd :P

terribly sorry. I thought you were in my git send-email list ... but you were not...

No problem, I caught it on linux-mm anyway :)

Quoted 7 lines
Quoted 5 lines

The patch LGTM, but it won't apply on top of mm-unstable given the amount of zswap refactoring there. I would rebase on top of mm-unstable if I were you (and if you did, add mm-unstable in the subject prefix).

This patch has a "fixes" tag, so I assume it should be also in 6.8?

Hmm that's up to Andrew. This fixes debug counters so it's not critical. On the other hand, it will conflict with the cleanup series in his tree and he'll have to rebase and fix the conflicts (which aren't a lot, but could still be annoying). Personally I think this can wait till v6.9, but if Andrew doesn't have a problem taking it for v6.8 that's fine too.

maintainer
#5 lore raw
Quoted 6 lines · On Sat, 17 Feb 2024 15:14:34 -0800 Yosry Ahmed <[email protected]> wrote:
Quoted 6 lines
Quoted 5 lines

The patch LGTM, but it won't apply on top of mm-unstable given the amount of zswap refactoring there. I would rebase on top of mm-unstable if I were you (and if you did, add mm-unstable in the subject prefix).

Quoted 8 lines
Quoted 1 line

This patch has a "fixes" tag, so I assume it should be also in 6.8?

Hmm that's up to Andrew. This fixes debug counters so it's not critical. On the other hand, it will conflict with the cleanup series in his tree and he'll have to rebase and fix the conflicts (which aren't a lot, but could still be annoying). Personally I think this can wait till v6.9, but if Andrew doesn't have a problem taking it for v6.8 that's fine too.

Yes, there are some pretty extensive repairs needed after this change. I'd prefer not to because lazy, and there are risks involved.

So against mm-unstable would be preferred please.

Reviewed-by on v2 1/1
#6 lore raw
In reply to Barry Song
Quoted 12 lines · Barry Song

From: Barry Song <[email protected]>

We used to rely on the returned -ENOSPC of zpool_malloc() to increase reject_compress_poor. But the code wouldn't get to there after commit 744e1885922a ("crypto: scomp - fix req->dst buffer overflow") as the new code will goto out immediately after the special compression case happens. So there might be no longer a chance to execute zpool_malloc now. We are incorrectly increasing zswap_reject_compress_fail instead. Thus, we need to fix the counters handling right after compressions return ENOSPC. This patch also centralizes the counters handling for all of compress_poor, compress_fail and alloc_fail.

Quoted 5 lines
Fixes 744e1885922a ("crypto: scomp - fix req->dst buffer overflow")
Cc Chengming Zhou <[email protected]>
Cc Nhat Pham <[email protected]>
Cc Sergey Senozhatsky <[email protected]>

LGTM, thanks!

Quoted 76 lines

---
-v2:
* correct the fixes target according to Yosry, Chengming, Nhat's comments;
* centralize the counters handling according to Yosry's comment

 mm/zswap.c | 25 ++++++++++++-------------
 1 file changed, 12 insertions(+), 13 deletions(-)
mm/zswap.c +12 -13
index 350dd2fc8159..47cf07d56362 100644
@@ -1498,6 +1498,7 @@ bool zswap_store(struct folio *folio)
 	struct zswap_tree *tree = zswap_trees[type];
 	struct zswap_entry *entry, *dupentry;
 	struct scatterlist input, output;
+	int comp_ret = 0, alloc_ret = 0;
 	struct crypto_acomp_ctx *acomp_ctx;
 	struct obj_cgroup *objcg = NULL;
 	struct mem_cgroup *memcg = NULL;
@@ -1508,7 +1509,6 @@ bool zswap_store(struct folio *folio)
 	char *buf;
 	u8 *src, *dst;
 	gfp_t gfp;
-	int ret;
 
 	VM_WARN_ON_ONCE(!folio_test_locked(folio));
 	VM_WARN_ON_ONCE(!folio_test_swapcache(folio));
@@ -1621,28 +1621,20 @@ bool zswap_store(struct folio *folio)
 	 * but in different threads running on different cpu, we have different
 	 * acomp instance, so multiple threads can do (de)compression in parallel.
 	 */
-	ret = crypto_wait_req(crypto_acomp_compress(acomp_ctx->req), &acomp_ctx->wait);
+	comp_ret = crypto_wait_req(crypto_acomp_compress(acomp_ctx->req), &acomp_ctx->wait);
 	dlen = acomp_ctx->req->dlen;
 
-	if (ret) {
-		zswap_reject_compress_fail++;
+	if (comp_ret)
 		goto put_dstmem;
-	}
 
 	/* store */
 	zpool = zswap_find_zpool(entry);
 	gfp = __GFP_NORETRY | __GFP_NOWARN | __GFP_KSWAPD_RECLAIM;
 	if (zpool_malloc_support_movable(zpool))
 		gfp |= __GFP_HIGHMEM | __GFP_MOVABLE;
-	ret = zpool_malloc(zpool, dlen, gfp, &handle);
-	if (ret == -ENOSPC) {
-		zswap_reject_compress_poor++;
-		goto put_dstmem;
-	}
-	if (ret) {
-		zswap_reject_alloc_fail++;
+	alloc_ret = zpool_malloc(zpool, dlen, gfp, &handle);
+	if (alloc_ret)
 		goto put_dstmem;
-	}
 	buf = zpool_map_handle(zpool, handle, ZPOOL_MM_WO);
 	memcpy(buf, dst, dlen);
 	zpool_unmap_handle(zpool, handle);
@@ -1689,6 +1681,13 @@ bool zswap_store(struct folio *folio)
 	return true;
 
 put_dstmem:
+	if (comp_ret == -ENOSPC || alloc_ret == -ENOSPC)
+		zswap_reject_compress_poor++;
+	else if (comp_ret)
+		zswap_reject_compress_fail++;
+	else if (alloc_ret)
+		zswap_reject_alloc_fail++;
+
 	mutex_unlock(&acomp_ctx->mutex);
 put_pool:
 	zswap_pool_put(entry->pool);