mm: zswap: global shrinker fix and proactive shrink

This patch fixes an issue where the zswap global shrinker stopped iterating through the memcg tree.

The problem was that shrink_worker() would stop iterating when a memcg was being offlined and restart from the tree root. Now, it properly handles the offlie memcg and continues shrinking with the next memcg.

Note that, to avoid a refcount leak of offline memcg encountered during the memcg tree walking, shrink_worker() must continue iterating to find the next online memcg.

The following minor issues in the existing code are also resolved by the change in the iteration logic:

- A rare temporary refcount leak in the offline memcg cleaner, where the next memcg of the offlined memcg is also offline. The leaked memcg cannot be freed until the next shrink_worker() releases the reference.

- One memcg was skipped from shrinking when the offline memcg cleaner advanced the cursor of memcg tree. It is addressed by a flag to indicate that the cursor has already been advanced.

Fixes a65b0e7607cc ("zswap: make shrinking memcg-aware")

Signed-off-by Takero Funaki <[email protected]>

---
  mm/zswap.c | 94 ++++++++++++++++++++++++++++++++++++++++++------------
  1 file changed, 73 insertions(+), 21 deletions(-)

mm/zswap.c +73 -21

index a50e2986cd2f..29944d8145af 100644
@@ -171,6 +171,7 @@ static struct list_lru zswap_list_lru;
  /* The lock protects zswap_next_shrink updates. */
  static DEFINE_SPINLOCK(zswap_shrink_lock);
  static struct mem_cgroup *zswap_next_shrink;
+static bool zswap_next_shrink_changed;
  static struct work_struct zswap_shrink_work;
  static struct shrinker *zswap_shrinker;
 

@@ -775,12 +776,39 @@ void zswap_folio_swapin(struct folio *folio)
  	}
  }
 
+/*
+ * This function should be called when a memcg is being offlined.
+ *
+ * Since the global shrinker shrink_worker() may hold a reference
+ * of the memcg, we must check and release the reference in
+ * zswap_next_shrink.
+ *
+ * shrink_worker() must handle the case where this function releases
+ * the reference of memcg being shrunk.
+ */
  void zswap_memcg_offline_cleanup(struct mem_cgroup *memcg)
  {
  	/* lock out zswap shrinker walking memcg tree */
  	spin_lock(&zswap_shrink_lock);
-	if (zswap_next_shrink == memcg)
-		zswap_next_shrink = mem_cgroup_iter(NULL, zswap_next_shrink, NULL);
+	if (zswap_next_shrink == memcg) {
+		/*
+		 * We advances the cursor to put back the offlined memcg.
+		 * shrink_worker() should not advance the cursor again.
+		 */
+		zswap_next_shrink_changed = true;
+
+		do {
+			zswap_next_shrink = mem_cgroup_iter(NULL,
+					zswap_next_shrink, NULL);
+		} while (zswap_next_shrink &&
+				!mem_cgroup_online(zswap_next_shrink));
+		/*
+		 * We verified the next memcg is online.  Even if the next
+		 * memcg is being offlined here, another cleaner must be
+		 * waiting for our lock.  We can leave the online memcg
+		 * reference.
+		 */
+	}
  	spin_unlock(&zswap_shrink_lock);
  }
 

@@ -1319,18 +1347,42 @@ static void shrink_worker(struct work_struct *w)
  	/* Reclaim down to the accept threshold */
  	thr = zswap_accept_thr_pages();
 
-	/* global reclaim will select cgroup in a round-robin fashion. */
+	/* global reclaim will select cgroup in a round-robin fashion.
+	 *
+	 * We save iteration cursor memcg into zswap_next_shrink,
+	 * which can be modified by the offline memcg cleaner
+	 * zswap_memcg_offline_cleanup().
+	 *
+	 * Since the offline cleaner is called only once, we cannot leave an
+	 * offline memcg reference in zswap_next_shrink.
+	 * We can rely on the cleaner only if we get online memcg under lock.
+	 *
+	 * If we get an offline memcg, we cannot determine the cleaner has
+	 * already been called or will be called later. We must put back the
+	 * reference before returning from this function. Otherwise, the
+	 * offline memcg left in zswap_next_shrink will hold the reference
+	 * until the next run of shrink_worker().
+	 */
  	do {
  		spin_lock(&zswap_shrink_lock);
-		zswap_next_shrink = mem_cgroup_iter(NULL, zswap_next_shrink, NULL);
-		memcg = zswap_next_shrink;
 
  		/*
-		 * We need to retry if we have gone through a full round trip, or if we
-		 * got an offline memcg (or else we risk undoing the effect of the
-		 * zswap memcg offlining cleanup callback). This is not catastrophic
-		 * per se, but it will keep the now offlined memcg hostage for a while.
-		 *
+		 * Start shrinking from the next memcg after zswap_next_shrink.
+		 * To not skip a memcg, do not advance the cursor when it has
+		 * already been advanced by the offline cleaner.
+		 */
+		do {
+			if (zswap_next_shrink_changed) {
+				/* cleaner advanced the cursor */
+				zswap_next_shrink_changed = false;
+			} else {
+				zswap_next_shrink = mem_cgroup_iter(NULL,
+						zswap_next_shrink, NULL);
+			}
+			memcg = zswap_next_shrink;
+		} while (memcg && !mem_cgroup_tryget_online(memcg));
+
+		/*
  		 * Note that if we got an online memcg, we will keep the extra
  		 * reference in case the original reference obtained by mem_cgroup_iter
  		 * is dropped by the zswap memcg offlining callback, ensuring that the

@@ -1344,17 +1396,11 @@ static void shrink_worker(struct work_struct *w)
  			goto resched;
  		}
 
-		if (!mem_cgroup_tryget_online(memcg)) {
-			/* drop the reference from mem_cgroup_iter() */
-			mem_cgroup_iter_break(NULL, memcg);
-			zswap_next_shrink = NULL;
-			spin_unlock(&zswap_shrink_lock);
-
-			if (++failures == MAX_RECLAIM_RETRIES)
-				break;
-
-			goto resched;
-		}
+		/*
+		 * We verified the memcg is online and got an extra memcg
+		 * reference.  Our memcg might be offlined concurrently but the
+		 * respective offline cleaner must be waiting for our lock.
+		 */
  		spin_unlock(&zswap_shrink_lock);
 
  		ret = shrink_memcg(memcg);

@@ -1368,6 +1414,12 @@ static void shrink_worker(struct work_struct *w)
  resched:
  		cond_resched();
  	} while (zswap_total_pages() > thr);
+
+	/*
+	 * We can still hold the original memcg reference.
+	 * The reference is stored in zswap_next_shrink, and then reused
+	 * by the next shrink_worker().
+	 */
  }
 
  /*********************************

This patch fixes an issue where the zswap global shrinker stopped iterating through the memcg tree.

The problem was that shrink_worker() would stop iterating when a memcg was being offlined and restart from the tree root. Now, it properly handles the offlie memcg and continues shrinking with the next memcg.

Note that, to avoid a refcount leak of offline memcg encountered during the memcg tree walking, shrink_worker() must continue iterating to find the next online memcg.

The following minor issues in the existing code are also resolved by the change in the iteration logic:

- A rare temporary refcount leak in the offline memcg cleaner, where the next memcg of the offlined memcg is also offline. The leaked memcg cannot be freed until the next shrink_worker() releases the reference.

- One memcg was skipped from shrinking when the offline memcg cleaner advanced the cursor of memcg tree. It is addressed by a flag to indicate that the cursor has already been advanced.

Fixes a65b0e7607cc ("zswap: make shrinking memcg-aware")

Signed-off-by Takero Funaki <[email protected]>

---
 mm/zswap.c | 94 ++++++++++++++++++++++++++++++++++++++++++------------
 1 file changed, 73 insertions(+), 21 deletions(-)

mm/zswap.c +11 -0

index a50e2986cd2f..29944d8145af 100644
@@ -171,6 +171,7 @@ static struct list_lru zswap_list_lru;
 /* The lock protects zswap_next_shrink updates. */
 static DEFINE_SPINLOCK(zswap_shrink_lock);
 static struct mem_cgroup *zswap_next_shrink;
+static bool zswap_next_shrink_changed;
 static struct work_struct zswap_shrink_work;
 static struct shrinker *zswap_shrinker;
 

@@ -775,12 +776,39 @@ void zswap_folio_swapin(struct folio *folio)
        }
 }
 
+/*
+ * This function should be called when a memcg is being offlined.
+ *
+ * Since the global shrinker shrink_worker() may hold a reference
+ * of the memcg, we must check and release the reference in
+ * zswap_next_shrink.
+ *
+ * shrink_worker() must handle the case where this function releases
+ * the reference of memcg being shrunk.
+ */
 void zswap_memcg_offline_cleanup(struct mem_cgroup *memcg)
 {

mm/zswap.c +14 -1

+
+               do {
+                       zswap_next_shrink = mem_cgroup_iter(NULL,
+                                       zswap_next_shrink, NULL);
+               } while (zswap_next_shrink &&
+                               !mem_cgroup_online(zswap_next_shrink));
+               /*
+                * We verified the next memcg is online.  Even if the next
+                * memcg is being offlined here, another cleaner must be
+                * waiting for our lock.  We can leave the online memcg
+                * reference.
+                */
+       }
        spin_unlock(&zswap_shrink_lock);
 }
 

@@ -1319,18 +1347,42 @@ static void shrink_worker(struct work_struct *w)
        /* Reclaim down to the accept threshold */
        thr = zswap_accept_thr_pages();
 
-       /* global reclaim will select cgroup in a round-robin fashion. */
+       /* global reclaim will select cgroup in a round-robin fashion.

mm/zswap.c +26 -18

+        * already been called or will be called later. We must put back the
+        * reference before returning from this function. Otherwise, the
+        * offline memcg left in zswap_next_shrink will hold the reference
+        * until the next run of shrink_worker().
+        */
        do {
                spin_lock(&zswap_shrink_lock);
-               zswap_next_shrink = mem_cgroup_iter(NULL, zswap_next_shrink, NULL);
-               memcg = zswap_next_shrink;
 
                /*
-                * We need to retry if we have gone through a full round trip, or if we
-                * got an offline memcg (or else we risk undoing the effect of the
-                * zswap memcg offlining cleanup callback). This is not catastrophic
-                * per se, but it will keep the now offlined memcg hostage for a while.
-                *
+                * Start shrinking from the next memcg after zswap_next_shrink.
+                * To not skip a memcg, do not advance the cursor when it has
+                * already been advanced by the offline cleaner.
+                */
+               do {
+                       if (zswap_next_shrink_changed) {
+                               /* cleaner advanced the cursor */
+                               zswap_next_shrink_changed = false;
+                       } else {
+                               zswap_next_shrink = mem_cgroup_iter(NULL,
+                                               zswap_next_shrink, NULL);
+                       }
+                       memcg = zswap_next_shrink;
+               } while (memcg && !mem_cgroup_tryget_online(memcg));
+
+               /*
                 * Note that if we got an online memcg, we will keep the extra
                 * reference in case the original reference obtained by mem_cgroup_iter
                 * is dropped by the zswap memcg offlining callback, ensuring that the

@@ -1344,17 +1396,11 @@ static void shrink_worker(struct work_struct *w)
                        goto resched;
                }
 
-               if (!mem_cgroup_tryget_online(memcg)) {
-                       /* drop the reference from mem_cgroup_iter() */
-                       mem_cgroup_iter_break(NULL, memcg);
-                       zswap_next_shrink = NULL;
-                       spin_unlock(&zswap_shrink_lock);
-
-                       if (++failures == MAX_RECLAIM_RETRIES)
-                               break;
-
-                       goto resched;
-               }
+               /*
+                * We verified the memcg is online and got an extra memcg
+                * reference.  Our memcg might be offlined concurrently but the
+                * respective offline cleaner must be waiting for our lock.
+                */
                spin_unlock(&zswap_shrink_lock);
 
                ret = shrink_memcg(memcg);

mm/zswap.c +0 -1

 }
 
 /*********************************
--

2.43.0

This patch fixes zswap global shrinker that did not shrink zpool as expected.

The issue it addresses is that shrink_worker() did not distinguish between unexpected errors and expected error codes that should be skipped, such as when there is no stored page in a memcg. This led to the shrinking process being aborted on the expected error codes.

The shrinker should ignore these cases and skip to the next memcg. However, skipping all memcgs presents another problem. To address this, this patch tracks progress while walking the memcg tree and checks for progress once the tree walk is completed.

To handle the empty memcg case, the helper function shrink_memcg() is modified to check if the memcg is empty and then return -ENOENT.

Fixes a65b0e7607cc ("zswap: make shrinking memcg-aware")

Signed-off-by Takero Funaki <[email protected]>

---
 mm/zswap.c | 31 +++++++++++++++++++++++++------
 1 file changed, 25 insertions(+), 6 deletions(-)

mm/zswap.c +8 -3

index 29944d8145af..f092932e652b 100644
@@ -1317,10 +1317,10 @@ static struct shrinker *zswap_alloc_shrinker(void)
 
 static int shrink_memcg(struct mem_cgroup *memcg)
 {
-       int nid, shrunk = 0;
+       int nid, shrunk = 0, scanned = 0;
 
        if (!mem_cgroup_zswap_writeback_enabled(memcg))
-               return -EINVAL;
+               return -ENOENT;
 
        /*
         * Skip zombies because their LRUs are reparented and we would be

@@ -1334,19 +1334,30 @@ static int shrink_memcg(struct mem_cgroup *memcg)
 
                shrunk += list_lru_walk_one(&zswap_list_lru, nid, memcg,
                                            &shrink_memcg_cb, NULL, &nr_to_walk);
+               scanned += 1 - nr_to_walk;
        }
+
+       if (!scanned)
+               return -ENOENT;
+
        return shrunk ? 0 : -EAGAIN;
 }
 
 static void shrink_worker(struct work_struct *w)
 {
        struct mem_cgroup *memcg;
-       int ret, failures = 0;
+       int ret, failures = 0, progress;
        unsigned long thr;

mm/zswap.c +4 -1

        /* global reclaim will select cgroup in a round-robin fashion.
         *
         * We save iteration cursor memcg into zswap_next_shrink,

@@ -1390,9 +1401,12 @@ static void shrink_worker(struct work_struct *w)
                 */
                if (!memcg) {
                        spin_unlock(&zswap_shrink_lock);
-                       if (++failures == MAX_RECLAIM_RETRIES)
+
+                       /* tree walk completed but no progress */
+                       if (!progress && ++failures == MAX_RECLAIM_RETRIES)
                                break;
 
+                       progress = 0;
                        goto resched;
                }
 

mm/zswap.c +0 -1

 resched:
                cond_resched();
        } while (zswap_total_pages() > thr);
--

2.43.0

This patch implements proactive shrinking of zswap pool before the max pool size limit is reached. This also changes zswap to accept new pages while the shrinker is running.

To prevent zswap from rejecting new pages and incurring latency when zswap is full, this patch queues the global shrinker by a pool usage threshold between 100% and accept_thr_percent, instead of the max pool size. The pool size will be controlled between 90% to 91% for the default accept_thr_percent=90. Since the current global shrinker continues to shrink until accept_thr_percent, we do not need to maintain the hysteresis variable tracking the pool limit overage in zswap_store().

Before this patch, zswap rejected pages while the shrinker is running without incrementing zswap_pool_limit_hit counter. It could be a reason why zswap writethrough new pages before writeback old pages. With this patch, zswap accepts new pages while shrinking, and zswap increments the counter when and only when zswap rejects pages by the max pool size.

Now, reclaims smaller than the proactive shrinking amount finish instantly and trigger background shrinking. Admins can check if new pages are buffered by zswap by monitoring the pool_limit_hit counter.

The name of sysfs tunable accept_thr_percent is unchanged as it is still the stop condition of the shrinker.
The respective documentation is updated to describe the new behavior.

Signed-off-by Takero Funaki <[email protected]>

This patch allows zswap to accept incompressible pages and store them into zpool if possible.

This change is required to achieve zero rejection on zswap_store(). With proper amount of proactive shrinking, swapout can be buffered by zswap without IO latency. Storing incompressible pages may seem costly, but it can reduce latency. A rare incompressible page in a large batch of compressive pages can delay the entire batch during swapping.

The memory overhead is negligible because the underlying zsmalloc already accepts nearly incompressible pages. zsmalloc stores data close to PAGE_SIZE to a dedicated page. Thus storing as-is saves decompression cycles without allocation overhead. zswap itself has not rejected pages in these cases.

To store the page as-is, use the compressed data size field length in struct zswap_entry. The length == PAGE_SIZE indicates incompressible data.

If a zpool backend does not support allocating PAGE_SIZE (zbud), the behavior remains unchanged. The allocation failure reported by the zpool blocks accepting the page as before.

Signed-off-by Takero Funaki <[email protected]>

I tried to propose something similar in the past. Please read the following discussion:

https://lore.kernel.org/all/CAJD7tka6XRyzYndRNEFZmi0Zj4DD2KnVzt=vMGhfF4iN2B4VKw@mail.gmail.com/

But, the TLDR is Yosry was (rightly) concerned that with this approach, memory reclaiming could end up increasing memory usage rather than reducing (since we do not free up the page that fail to zswap-out, and we need extra memory for the zswap metadata of that page).

So my vote on this patch would be NACK, until we get around this issue somehow :)

2024年7月7日(日) 8:53 Nhat Pham <[email protected]>:

Quoted 6 lines

I tried to propose something similar in the past. Please read the following discussion:

https://lore.kernel.org/all/CAJD7tka6XRyzYndRNEFZmi0Zj4DD2KnVzt=vMGhfF4iN2B4VKw@mail.gmail.com/

Quoted 8 lines

But, the TLDR is Yosry was (rightly) concerned that with this approach, memory reclaiming could end up increasing memory usage rather than reducing (since we do not free up the page that fail to zswap-out, and we need extra memory for the zswap metadata of that page).

So my vote on this patch would be NACK, until we get around this issue somehow :)

It seems the discussion on the thread mixed up memory allocation failure (system runs out of memory reserve) and incompressible pages (compression algorithm successfully compressed but the result is equal to or larger than PAGE_SIZE).

zswap has been storing pages into dedicated pages 1:1 when compressed to near PAGE_SIZE. Using zsmalloc, current zswap stores pages compressed to between 3633 bytes (=hugeclass+1) to 4095 bytes (=PAGE_SIZE-1) into 1 page. This patch changes the range to 3633 to 4096 by treating PAGE_SIZE as a special case. I could not find a reason to reject only PAGE_SIZE while accepting PAGE_SIZE-1.

zswap wastes memory for metadata for all accepted pages but reduces IO

amount and latency by compressed buffer memory. For pages between 3633 to 4096 bytes, zswap reduces the latency only. This is still beneficial because the rare incompressible pages trigger urgent pageout IO and incur a head-of-line blocking on the subsequent pages. It also keeps LRU priority for pagein latency.

In the worst case or with a malicious dataset, zswap will waste a significant amount of memory, but this patch does not affect nor resolve the scenario. For example, if a user allocates pages compressed to 3633 bytes, current zswap using zsmalloc cannot gain memory as the compression ratio, including zsmalloc overhead, becomes 1:1. This also applies to zbud. The compression ratio will be 1:1 as zbud cannot find buddies smaller than 463 bytes. zswap will be less efficient but still work in this situation since the max pool percent and background writeback ensure the pool size does not overwhelm usable memory.

I suppose the current zswap has accepted the possible waste of memory, at least since the current zswap_compress() logic was implemented. If zswap had to ensure the compression ratio is better than 1:1, and only prefers reducing IO amount (not latency), there would have been a compression ratio threshold to reject pages not compressible to under 2048 bytes. I think accepting nearly incompressible pages is beneficial and changing the range to 4096 does not negatively affect the current behavior.

This patch allows zswap to accept incompressible pages and store them into zpool if possible.

This change is required to achieve zero rejection on zswap_store(). With proper amount of proactive shrinking, swapout can be buffered by zswap without IO latency. Storing incompressible pages may seem costly, but it can reduce latency. A rare incompressible page in a large batch of compressive pages can delay the entire batch during swapping.

The memory overhead is negligible because the underlying zsmalloc already accepts nearly incompressible pages. zsmalloc stores data close to PAGE_SIZE to a dedicated page. Thus storing as-is saves decompression cycles without allocation overhead. zswap itself has not rejected pages in these cases.

To store the page as-is, use the compressed data size field length in struct zswap_entry. The length == PAGE_SIZE indicates incompressible data.

If a zpool backend does not support allocating PAGE_SIZE (zbud), the behavior remains unchanged. The allocation failure reported by the zpool blocks accepting the page as before.

Signed-off-by Takero Funaki <[email protected]>

---
  mm/zswap.c | 36 +++++++++++++++++++++++++++++++++---
  1 file changed, 33 insertions(+), 3 deletions(-)

mm/zswap.c +2 -0

index 76691ca7b6a7..def0f948a4ab 100644
@@ -186,6 +186,8 @@ static struct shrinker *zswap_shrinker;
   * length - the length in bytes of the compressed page data.  Needed during
   *          decompression. For a same value filled page length is 0, and both
   *          pool and lru are invalid and must be ignored.
+ *          If length is equal to PAGE_SIZE, the data stored in handle is
+ *          not compressed. The data must be copied to page as-is.
   * pool - the zswap_pool the entry's data is in
   * handle - zpool allocation handle that stores the compressed page data
   * value - value of the same-value filled pages which have same content

mm/zswap.c +12 -0

+	/*
+	 * If the data cannot be compressed well, store the data as-is.
+	 * Switching by a threshold at
+	 * PAGE_SIZE - (allocation granularity)
+	 * zbud and z3fold use 64B granularity.
+	 * zsmalloc stores >3632B in one page for 4K page arch.
+	 */
+	if (comp_ret || dlen > PAGE_SIZE - 64) {
+		/* we do not use compressed result anymore */
+		comp_ret = 0;
+		dlen = PAGE_SIZE;
+	}
  	zpool = zswap_find_zpool(entry);
  	gfp = __GFP_NORETRY | __GFP_NOWARN | __GFP_KSWAPD_RECLAIM;
  	if (zpool_malloc_support_movable(zpool))

mm/zswap.c +12 -1

+	else
+		memcpy(buf, dst, dlen);
+
  	zpool_unmap_handle(zpool, handle);
 
  	entry->handle = handle;
  	entry->length = dlen;
 
  unlock:
-	if (comp_ret == -ENOSPC || alloc_ret == -ENOSPC)
+	if (alloc_ret == -ENOSPC)
  		zswap_reject_compress_poor++;
  	else if (comp_ret)
  		zswap_reject_compress_fail++;

@@ -1006,6 +1028,14 @@ static void zswap_decompress(struct zswap_entry *entry, struct page *page)
  	struct crypto_acomp_ctx *acomp_ctx;
  	u8 *src;
 
+	if (entry->length == PAGE_SIZE) {
+		/* the content is not compressed. copy back as-is.  */
+		src = zpool_map_handle(zpool, entry->handle, ZPOOL_MM_RO);
+		memcpy_to_page(page, 0, src, entry->length);
+		zpool_unmap_handle(zpool, entry->handle);
+		return;
+	}
+
  	acomp_ctx = raw_cpu_ptr(entry->pool->acomp_ctx);
  	mutex_lock(&acomp_ctx->mutex);
 

mm/zswap.c static bool zswap_compress(struct folio *folio, struct zswap_entry *entry) · line 977

Quoted 7 lines

      comp_ret = crypto_wait_req(crypto_acomp_compress(acomp_ctx->req), &acomp_ctx->wait);
      dlen = acomp_ctx->req->dlen;

mm/zswap.c +3 -1

-     if (comp_ret)
+
+     /* coa_compress returns -EINVAL for errors including insufficient dlen */
+     if (comp_ret && comp_ret != -EINVAL)
              goto unlock;

Seems we don't need to care about? "comp_ret" is useless anymore.

Just:

if (comp_ret || dlen > PAGE_SIZE - 64)

        dlen = PAGE_SIZE;

And remove the checkings of comp_ret at the end.

We actually don't need to hold mutex if we are just copying folio.

Thanks.

2024年7月8日(月) 12:56 Chengming Zhou <[email protected]>:

Quoted 21 lines

mm/zswap.c static bool zswap_compress(struct folio *folio, struct zswap_entry *entry) · line 977

Quoted 7 lines

       comp_ret = crypto_wait_req(crypto_acomp_compress(acomp_ctx->req), &acomp_ctx->wait);
       dlen = acomp_ctx->req->dlen;

mm/zswap.c +3 -1

-     if (comp_ret)
+
+     /* coa_compress returns -EINVAL for errors including insufficient dlen */
+     if (comp_ret && comp_ret != -EINVAL)
               goto unlock;

Seems we don't need to care about? "comp_ret" is useless anymore.

Just:

if (comp_ret || dlen > PAGE_SIZE - 64)

         dlen = PAGE_SIZE;

And remove the checkings of comp_ret at the end.

We actually don't need to hold mutex if we are just copying folio.

Quoted 2 lines

Thanks.

Thanks for reviewing.

For comp_ret, can we consolidate all possible error codes as incompressible data?

if we do not need to distinguish -EINVAL and the others, diff v2..v3 can be like:

(quoted diff) +0 -3

@@ -62,8 +62,6 @@ static u64 zswap_pool_limit_hit;
  static u64 zswap_written_back_pages;
  /* Store failed due to a reclaim failure after pool limit was reached */
  static u64 zswap_reject_reclaim_fail;
-/* Store failed due to compression algorithm failure */
-static u64 zswap_reject_compress_fail;
  /* Compressed page was too big for the allocator to (optimally) store */
  static u64 zswap_reject_compress_poor;
  /* Store failed because underlying allocator could not get memory */

@@ -1043,10 +1041,6 @@ static bool zswap_compress(struct folio *folio,
 struct zswap_entry *entry)
         comp_ret =
 crypto_wait_req(crypto_acomp_compress(acomp_ctx->req),
 &acomp_ctx->wait);
         dlen = acomp_ctx->req->dlen;
 
-       /* coa_compress returns -EINVAL for errors including

insufficient dlen */

(quoted diff) +2 -4

-       if (comp_ret && comp_ret != -EINVAL)
-               goto unlock;
-
         /*
          * If the data cannot be compressed well, store the data as-is.
          * Switching by a threshold at

@@ -1056,7 +1050,8 @@ static bool zswap_compress(struct folio *folio,
 struct zswap_entry *entry)
          */
         if (comp_ret || dlen > PAGE_SIZE - 64) {
                 /* we do not use compressed result anymore */
-               comp_ret = 0;
+               mutex_unlock(&acomp_ctx->mutex);
+               acomp_ctx = NULL;
                 dlen = PAGE_SIZE;
         }

         zpool = zswap_find_zpool(entry);

(quoted diff) +0 -2

  }

@@ -1886,8 +1880,6 @@ static int zswap_debugfs_init(void)
                            zswap_debugfs_root, &zswap_reject_alloc_fail);
         debugfs_create_u64("reject_kmemcache_fail", 0444,
                            zswap_debugfs_root, &zswap_reject_kmemcache_fail);
-       debugfs_create_u64("reject_compress_fail", 0444,
-                          zswap_debugfs_root, &zswap_reject_compress_fail);
         debugfs_create_u64("reject_compress_poor", 0444,
                            zswap_debugfs_root, &zswap_reject_compress_poor);
         debugfs_create_u64("written_back_pages", 0444,

Quoted 26 lines · Takero Funaki

2024年7月8日(月) 12:56 Chengming Zhou <[email protected]>:

Quoted 23 lines

mm/zswap.c static bool zswap_compress(struct folio *folio, struct zswap_entry *entry) · line 977

Quoted 7 lines

       comp_ret = crypto_wait_req(crypto_acomp_compress(acomp_ctx->req), &acomp_ctx->wait);
       dlen = acomp_ctx->req->dlen;

mm/zswap.c +3 -1

-     if (comp_ret)
+
+     /* coa_compress returns -EINVAL for errors including insufficient dlen */
+     if (comp_ret && comp_ret != -EINVAL)
               goto unlock;

Seems we don't need to care about? "comp_ret" is useless anymore.

Just:

if (comp_ret || dlen > PAGE_SIZE - 64)

         dlen = PAGE_SIZE;

And remove the checkings of comp_ret at the end.

We actually don't need to hold mutex if we are just copying folio.

Thanks.

Quoted 4 lines

Thanks for reviewing.

For comp_ret, can we consolidate all possible error codes as incompressible data?

Maybe we still want these debug counters? I'm not sure.

Quoted 33 lines · Chengming Zhou

Quoted 30 lines · Takero Funaki

2024年7月8日(月) 12:56 Chengming Zhou <[email protected]>:

Quoted 15 lines

mm/zswap.c static bool zswap_compress(struct folio *folio, struct zswap_entry *entry) · line 977

Quoted 7 lines

        comp_ret = crypto_wait_req(crypto_acomp_compress(acomp_ctx->req), &acomp_ctx->wait);
        dlen = acomp_ctx->req->dlen;

mm/zswap.c +3 -1

-     if (comp_ret)
+
+     /* coa_compress returns -EINVAL for errors including insufficient dlen */
+     if (comp_ret && comp_ret != -EINVAL)
                goto unlock;

Seems we don't need to care about? "comp_ret" is useless anymore.

Just:

if (comp_ret || dlen > PAGE_SIZE - 64)

          dlen = PAGE_SIZE;

Quoted 8 lines

And remove the checkings of comp_ret at the end.

We actually don't need to hold mutex if we are just copying folio.

Thanks.

Thanks for reviewing.

For comp_ret, can we consolidate all possible error codes as incompressible data?

Quoted 1 line

Maybe we still want these debug counters? I'm not sure.

I'm a bit torn, but ATM I have no strong opinions on these two error codes. If you do decide to consolidate these two, may I ask you to separate it into its own patch so that we can review + discuss it separately?

To prevent the zswap global shrinker from writing back pages simultaneously with IO performed for memory reclaim and faults, delay the writeback when zswap_store() rejects pages or zswap_load() cannot find entry in pool.

When the zswap shrinker is running and zswap rejects an incoming page, simulatenous zswap writeback and the rejected page lead to IO contention on swap device. In this case, the writeback of the rejected page must be higher priority as it is necessary for actual memory reclaim progress. The zswap global shrinker can run in the background and should not interfere with memory reclaim.

The same logic applies to zswap_load(). When zswap cannot find requested page from pool and read IO is performed, shrinker should be interrupted.

To avoid IO contention, save the timestamp jiffies when zswap cannot buffer the pagein/out IO and interrupt the global shrinker. The shrinker resumes the writeback in 500 msec since the saved timestamp.

Signed-off-by Takero Funaki <[email protected]>

---
 mm/zswap.c | 47 +++++++++++++++++++++++++++++++++++++++++++++--
 1 file changed, 45 insertions(+), 2 deletions(-)

mm/zswap.c +2 -0

index def0f948a4ab..59ba4663c74f 100644
@@ -35,6 +35,8 @@
 #include <linux/pagemap.h>
 #include <linux/workqueue.h>
 #include <linux/list_lru.h>
+#include <linux/delay.h>
+#include <linux/jiffies.h>
 
 #include "swap.h"
 #include "internal.h"

mm/zswap.c +4 -0

+ */
+#define ZSWAP_GLOBAL_SHRINKER_DELAY_MS 500
+static unsigned long zswap_shrinker_delay_start;
+
 /*
  * struct zswap_entry
  *

mm/zswap.c +3 -0

+
 /*********************************
 * pool functions
 **********************************/

@@ -1378,6 +1396,8 @@ static void shrink_worker(struct work_struct *w)
        struct mem_cgroup *memcg;
        int ret, failures = 0, progress;
        unsigned long thr;
+       unsigned long now, sleepuntil;
+       const unsigned long delay = msecs_to_jiffies(ZSWAP_GLOBAL_SHRINKER_DELAY_MS);
 
        /* Reclaim down to the accept threshold */
        thr = zswap_accept_thr_pages();

mm/zswap.c +20 -3

+               now = jiffies;
+               /*
+                * If zswap did not reject pages for long, sleepuntil-now may
+                * underflow.  We assume the timestamp is valid only if
+                * now < sleepuntil < now + delay + 1
+                */
+               if (time_before(now, sleepuntil) &&
+                               time_before(sleepuntil, now + delay + 1))
+                       fsleep(jiffies_to_usecs(sleepuntil - now));
+
                spin_lock(&zswap_shrink_lock);
 
                /*

@@ -1526,8 +1561,10 @@ bool zswap_store(struct folio *folio)
        VM_WARN_ON_ONCE(!folio_test_swapcache(folio));
 
        /* Large folios aren't supported */
-       if (folio_test_large(folio))
+       if (folio_test_large(folio)) {
+               zswap_shrinker_delay_update();
                return false;
+       }
 
        if (!zswap_enabled)
                goto check_old;

@@ -1648,6 +1685,8 @@ bool zswap_store(struct folio *folio)
        zswap_entry_cache_free(entry);
 reject:
        obj_cgroup_put(objcg);
+       zswap_shrinker_delay_update();
+
        if (need_global_shrink)
                queue_work(shrink_wq, &zswap_shrink_work);
 check_old:

@@ -1691,8 +1730,10 @@ bool zswap_load(struct folio *folio)
        else
                entry = xa_load(tree, offset);
 
-       if (!entry)
+       if (!entry) {
+               zswap_shrinker_delay_update();
                return false;
+       }
 
        if (entry->length)
                zswap_decompress(entry, page);

@@ -1835,6 +1876,8 @@ static int zswap_setup(void)
        if (ret)
                goto hp_fail;
 
+       zswap_shrinker_delay_update();
+
        shrink_wq = alloc_workqueue("zswap-shrink",
                        WQ_UNBOUND, 1);
        if (!shrink_wq)
--

2.43.0

To prevent the zswap global shrinker from writing back pages simultaneously with IO performed for memory reclaim and faults, delay the writeback when zswap_store() rejects pages or zswap_load() cannot find entry in pool.

When the zswap shrinker is running and zswap rejects an incoming page, simulatenous zswap writeback and the rejected page lead to IO contention on swap device. In this case, the writeback of the rejected page must be higher priority as it is necessary for actual memory reclaim progress. The zswap global shrinker can run in the background and should not interfere with memory reclaim.

The same logic applies to zswap_load(). When zswap cannot find requested page from pool and read IO is performed, shrinker should be interrupted.

Do you see this problem actually manifesting in real life? Does not sound infeasible to me, but I wonder how likely this is the case.

Do you have any userspace-visible metrics, or any tracing logs etc. that proves that it actually happens?

This might also affect the dynamic shrinker as well FWIW.

Quoted 4 lines

+/*
+ * To avoid IO contention between pagein/out and global shrinker writeback,
+ * track the last jiffies of pagein/out and delay the writeback.
+ * Default to 500msec in alignment with mq-deadline read timeout.

If there is a future version, could you include the reason why you select 500msec in the patch's changelog as well?

Hmmm is there a reason why we do not just do:

zswap_shrinker_delay_start = jiffies;

or, more unnecessarily:

unsigned long now = jiffies;

zswap_shrinker_delay_start = now;

IOW, why the branching here? Does not seem necessary to me, but perhaps there is a correctness/compiler reason I'm not seeing?

In fact, if it's the first version, then we could manually inline it.

Additionally/alternatively, I wonder if it is more convenient to do it at the mm/page_io.c zswap callsites, i.e whenever zswap_store() and zswap_load() returns false, then delay the shrinker before proceeding with the IO steps.

mm/zswap.c static void shrink_worker(struct work_struct *w) · line 1432

Quoted 6 lines

        do {

mm/zswap.c +5 -0

+               /*
+                * delay shrinking to allow the last rejected page completes
+                * its writeback
+                */
+               sleepuntil = delay + READ_ONCE(zswap_shrinker_delay_start);

I assume we do not care about racy access here right? Same goes for updates - I don't see any locks protecting these operations (but I could be missing something).

Hmm what about this scenario: when we disable zswap writeback on a cgroup, if zswap_store() fails, we are delaying the global shrinker for no gain essentially. There is no subsequent IO. I don't think we are currently handling this, right?

Quoted 4 lines

The same logic applies to zswap_load(). When zswap cannot find requested page from pool and read IO is performed, shrinker should be interrupted.

Yet another (less concerning IMHO) scenario is when a cgroup disables zswap by setting zswap.max = 0 (for instance, if the sysadmin knows that this cgroup's data are really cold, and/or that the workload is latency-tolerant, and do not want it to take up valuable memory resources of other cgroups). Every time this cgroup reclaims memory, it would disable the global shrinker (including the new proactive behavior) for other cgroup, correct? And, when they do need to swap in, it would further delay the global shrinker. Would this break of isolation be a problem?

There are other concerns I raised in the cover letter's response as well - please take a look :)

2024年7月9日(火) 4:17 Nhat Pham <[email protected]>:

Quoted 9 lines

Do you see this problem actually manifesting in real life? Does not sound infeasible to me, but I wonder how likely this is the case.

Do you have any userspace-visible metrics, or any tracing logs etc. that proves that it actually happens?

This might also affect the dynamic shrinker as well FWIW.

Although it is rare, on a small VM with 0.5GB RAM, performing apt upgrade for ubuntu kernel update degrades system responsiveness. Since kernel upgrade is memory consuming for zstd compressed initramfs, there is heavy memory pressure like the benchmark.

Unfortunately I could not get evidence that clearly indicates the contention. Perhaps IO latency can be a metric?

While allocating large memory, perf showed that __swap_writepage() was consuming time and was called mostly from kswapd and some fraction from user faults of python script and from shrink_worker. CPU was mostly idling even in a single CPU system, so lock contention and compression should not be the reason. I believe these behaviors suggest contention on writeback IO.

As shown in the benchmark, reducing shrinker writeback by patches 3 to 6 reduced elapsed time, which also indicates IO contention.

Quoted 8 lines

Quoted 4 lines

+/*
+ * To avoid IO contention between pagein/out and global shrinker writeback,
+ * track the last jiffies of pagein/out and delay the writeback.
+ * Default to 500msec in alignment with mq-deadline read timeout.

If there is a future version, could you include the reason why you select 500msec in the patch's changelog as well?

The 500ms can be any value longer than the average interval of each pageout/in and is not significant for behavior. If subsequent pageout rejection occurs while the shrinker is sleeping, writeback will be delayed again by 500ms from the last timestamp update. If pageout occurs at a 1ms interval on average, the minimal delay should be 1+ms.

I chose 500ms from the mq-deadline scheduler that tries to perform read IO in a 500ms timeframe by default (bfq for HDD uses a shorter timeout).
When the shrinker performs writeback IO with a 500ms delay from the last pagein, the write IO will be of lower priority than the read IO waiting in the queue, as the pagein read becomes the highest priority by the deadline. This logic emulates low-priority write IO by voluntarily delaying IO.

Quoted 11 lines

Hmmm is there a reason why we do not just do:

zswap_shrinker_delay_start = jiffies;

or, more unnecessarily:

unsigned long now = jiffies;

zswap_shrinker_delay_start = now;

Quoted 5 lines

IOW, why the branching here? Does not seem necessary to me, but perhaps there is a correctness/compiler reason I'm not seeing?

In fact, if it's the first version, then we could manually inline it.

That was to avoid invalidating the CPU cache of the shared variable unnecessarily. Removing the branch and manually inlining it for v3.

Quoted 7 lines

Additionally/alternatively, I wonder if it is more convenient to do it at the mm/page_io.c zswap callsites, i.e whenever zswap_store() and zswap_load() returns false, then delay the shrinker before proceeding with the IO steps.

Should we expose the timestamp variable? It is only used in zswap internally, and the timestamp is not required when zswap is disabled.

Quoted 8 lines

mm/zswap.c static void shrink_worker(struct work_struct *w) · line 1432

Quoted 7 lines

        do {

mm/zswap.c +5 -0

+               /*
+                * delay shrinking to allow the last rejected page completes
+                * its writeback
+                */
+               sleepuntil = delay + READ_ONCE(zswap_shrinker_delay_start);

Quoted 4 lines

I assume we do not care about racy access here right? Same goes for updates - I don't see any locks protecting these operations (but I could be missing something).

Right. Do we need atomic or spinlock for safety?
I think the bare store/load of unsigned long is sufficient here. The
possible deviation by concurrent updates is mostly +/-1 jiffy. Sleep does not need ms accuracy.

Ah, I found a mistake here. v2 missed continue statement in the loop. The delay should be extended if zswap_store() rejects another page. In v2, one writeback was allowed per 500ms, which was not my intended behavior.
The corrected logic for v3 should be:

               if (time_before(now, sleepuntil) &&
                               time_before(sleepuntil, now + delay + 1)) {
                       fsleep(jiffies_to_usecs(sleepuntil - now));
                       /* check if subsequent pageout/in extended delay */
                       continue;
               }

2024年7月9日(火) 9:57 Nhat Pham <[email protected]>:

Quoted 21 lines

Hmm what about this scenario: when we disable zswap writeback on a cgroup, if zswap_store() fails, we are delaying the global shrinker for no gain essentially. There is no subsequent IO. I don't think we are currently handling this, right?

Quoted 4 lines

The same logic applies to zswap_load(). When zswap cannot find requested page from pool and read IO is performed, shrinker should be interrupted.

Yet another (less concerning IMHO) scenario is when a cgroup disables zswap by setting zswap.max = 0 (for instance, if the sysadmin knows that this cgroup's data are really cold, and/or that the workload is latency-tolerant, and do not want it to take up valuable memory resources of other cgroups). Every time this cgroup reclaims memory, it would disable the global shrinker (including the new proactive behavior) for other cgroup, correct? And, when they do need to swap in, it would further delay the global shrinker. Would this break of isolation be a problem?

Quoted 2 lines

There are other concerns I raised in the cover letter's response as well - please take a look :)

I haven't considered these cases much, but I suppose the global shrinker should be delayed in both cases as well. In general, any pagein/out should be prefered over shrinker writeback throughput.

When zswap writeback was disabled for a memcg (memcg.zswap.writeback=0), I suppose disabling/delaying writeback is harmless.
If the rejection incurs no IO, there is no more memory pressure and shrinking is not urgent. We can postpone the shrinker writeback. If the rejection incurs IO (i.e. mm choose another page from a memcg with writeback enabled), again we should delay the shrinker.

For pageout from latency-tolerant memcg (zswap.max=0), I think pageout latency may affect other memcgs.
For example, when a latency-sensitive workload tries to allocate memory, mm might choose to swap out pages from zswap-disabled memcg. The slow direct pageout may indirectly delay allocation of the latency-sensitive workload. IOW, we cannot determine which workload would be blocked by a slow pageout based on which memcg owns the page. In this case, it would be better to delay shrinker writeback even if the owner is latency tolerant memcg.
Also for pagein, we cannot determine how urgent the pagein is.

Delaying shrinker on any pagein/out diminishes proactive shrinking progress, but that is still better than the existing shrinker that cannot shrink.

Quoted 17 lines

Quoted 10 lines

Yet another (less concerning IMHO) scenario is when a cgroup disables zswap by setting zswap.max = 0 (for instance, if the sysadmin knows that this cgroup's data are really cold, and/or that the workload is latency-tolerant, and do not want it to take up valuable memory resources of other cgroups). Every time this cgroup reclaims memory, it would disable the global shrinker (including the new proactive behavior) for other cgroup, correct? And, when they do need to swap in, it would further delay the global shrinker. Would this break of isolation be a problem?

Quoted 2 lines

There are other concerns I raised in the cover letter's response as well - please take a look :)

I haven't considered these cases much, but I suppose the global shrinker should be delayed in both cases as well. In general, any pagein/out should be prefered over shrinker writeback throughput.

Quoted 7 lines

When zswap writeback was disabled for a memcg (memcg.zswap.writeback=0), I suppose disabling/delaying writeback is harmless.
If the rejection incurs no IO, there is no more memory pressure and shrinking is not urgent. We can postpone the shrinker writeback. If the rejection incurs IO (i.e. mm choose another page from a memcg with writeback enabled), again we should delay the shrinker.

You are delaying writeback globally right? IOW, other cgroup is also affected?

Say we are under memory pressure, with two cgroups under reclaim - one with zswap writeback disabled. The one with writeback disabled will constantly fail at zswap_store(), and delay the global shrinking action, which could have targeted the other cgroup (which should proceed fully because there is no contention here since the first cgroup is not swapping either).

This series addresses some issues and introduces a minor improvement in the zswap global shrinker.

Seems that patches 1 & 2 might be worthy of backporting into earlier kernels? Could you please provide a description of the userspace-visible effects of the bugs so the desirability of such an action can be better understood?

Visible issue to resolve
-------------------------------
The visible issue with the current global shrinker is that pageout/in operations from active processes are slow when zswap is near its max pool size. This is particularly significant on small memory systems where total swap usage exceeds what zswap can store. This results in old pages occupying most of the zswap pool space, with recent pages using the swap disk directly.

Root cause of the issue
-------------------------------
This issue is caused by zswap maintaining the pool size near 100%. Since the shrinker fails to shrink the pool to accept_threshold_percent, zswap rejects incoming pages more frequently than it should. The rejected pages are directly written to disk while zswap protects old pages from eviction, leading to slow pageout/in performance for recent pages.

This series addresses some issues and introduces a minor improvement in the zswap global shrinker.

This version addresses issues discovered during the review of v1:
https://lore.kernel.org/linux-mm/[email protected]/
and includes three additional patches to fix another issue uncovered by applying v1.

Changes in v2:
- Added 3 patches to reduce resource contention.
mm: zswap: fix global shrinker memcg iteration:
- Change the loop style (Yosry, Nhat, Shakeel)
- To not skip online memcg, save zswap_last_shrink to detect cursor change by cleaner. (Yosry, Nhat, Shakeel)
mm: zswap: fix global shrinker error handling logic:
- Change error code for no-writeback memcg. (Yosry)
- Use nr_scanned to check if lru is empty. (Yosry)

Changes in v1:
mm: zswap: fix global shrinker memcg iteration:
- Drop and reacquire spinlock before skipping a memcg.
- Add some comment to clarify the locking mechanism.
mm: zswap: proactive shrinking before pool size limit is hit:
- Remove unneeded check before scheduling work.
- Change shrink start threshold to accept_thr_percent + 1%.

Patches
===============================

1. Fix the memcg iteration logic that abort iteration on offline memcgs.
2. Fix the error path that aborts on expected error codes.
3. Add proactive shrinking at accept threshold + 1%.
4. Drop the global shrinker workqueue WQ_MEM_RECLAIM flag to not block pageout.
5. Store incompressible pages as-is to accept all pages.
6. Interrupt the shrinker to avoid blocking page-in/out.

Patches 1 to 3 should be applied in this order to avoid potential loops caused by the first issue. Patch 3 and later can be applied independently, but the first two issues must be resolved to ensure the shrinker can evict pages. Patches 4 to 6 address resource contention issues in the current shrinker uncovered by applying patch 3.

With this series applied, the shrinker will continue to evict pages until the accept_threshold_percent proactively, as documented in patch
3. As a side effect of changes in the hysteresis logic, zswap will no longer reject pages under the max pool limit.

The series is rebased on mainline 6.10-rc6.

The counters below are from vmstat and zswap debugfs stats. The times are average seconds using /usr/bin/time -v.

pre-patch, 6.10-rc4

|                    | Start       | End         | Delta      |
|--------------------|-------------|-------------|------------|
| pool_limit_hit     | 845         | 845         | 0          |
| pool_total_size    | 201539584   | 201539584   | 0          |
| stored_pages       | 63138       | 63138       | 0          |
| written_back_pages | 12          | 12          | 0          |
| pswpin             | 387         | 32412       | 32025      |
| pswpout            | 153078      | 197138      | 44060      |
| zswpin             | 0           | 0           | 0          |
| zswpout            | 63150       | 63150       | 0          |
| zswpwb             | 12          | 12          | 0          |

| Time              |              |
|-------------------|--------------|
| elapsed           | 8.473        |
| user              | 1.881        |
| system            | 0.814        |

post-patch, 6.10-rc4 with patch 1 to 5

|                    | Start       | End         | Delta      |
|--------------------|-------------|-------------|------------|
| pool_limit_hit     | 81861       | 81861       | 0          |
| pool_total_size    | 75001856    | 87117824    | 12115968   |
| reject_reclaim_fail| 0           | 32          | 32         |
| same_filled_pages  | 135         | 135         | 0          |
| stored_pages       | 23919       | 27549       | 3630       |
| written_back_pages | 97665       | 106994      | 10329      |
| pswpin             | 4981        | 8223        | 3242       |
| pswpout            | 179554      | 188883      | 9329       |

| zswpin             | 293863      | 590014      | 296151     |
| zswpout            | 455338      | 764882      | 309544     |
| zswpwb             | 97665       | 106994      | 10329      |

| Time              |              |
|-------------------|--------------|
| elapsed           | 4.525        |
| user              | 1.839        |
| system            | 1.467        |

Although the pool_limit_hit is not increased in both cases, zswap_store() rejected pages before this patch. Note that, before this series, zswap_store() did not increment pool_limit_hit on rejection by limit hit hysteresis (only the first few hits were counted).

From the pre-patch result, the existing zswap global shrinker cannot write back effectively and locks the old pages in the pool. The pswpin/out indicates the active process uses the swap device directly.

From the post-patch result, zswpin/out/wb are increased as expected, indicating the active process uses zswap and the old pages of the background services are evicted from the pool. pswpin/out are significantly reduced from pre-patch results.

System responsiveness issue (patch 4 to 6)
===============================
After applying patches 1 to 3, I encountered severe responsiveness degradation while zswap global shrinker is running under heavy memory pressure.

Visible issue to resolve
-------------------------------
The visible issue happens with patches 1 to 3 applied when a large amount of memory allocation happens and zswap cannot store the incoming pages.
While global shrinker is writing back pages, system stops responding as if under heavy memory thrashing.

This issue is less likely to happen without patches 1 to 3 or zswap is disabled. I believe this is because the global shrinker could not write back a meaningful amount of pages, as described in patch 2.

Root cause and changes to resolve the issue
-------------------------------
It seems that zswap shrinker blocking IO for memory reclaim and faults is the root cause of this responsiveness issue. I introduced three patches to reduce possible blocking in the following problematic situations:

1. Contention on workqueue thread pools by shrink_worker() using WQ_MEM_RECLAIM unnecessarily.

Although the shrinker runs simultaneously with memory reclaim, shrinking is not required to reclaim memory since zswap_store() can reject pages without interfering with memory reclaim progress. shrink_worker() should not use WQ_MEM_RECLAIM and should be delayed when another work in WQ_MEM_RECLAIM is reclaiming memory. The existing code requires allocating memory inside shrink_worker(), potentially blocking other latency-sensitive reclaim work.

2. Contention on swap IO.

Since zswap_writeback_entry() performs write IO in 4KB pages, it consumes a lot of IOPS, increasing the IO latency of swapout/in. We should not perform IO for background shrinking while zswap_store() is rejecting pages or zswap_load() is failing to find stored pages. This series implements two mitigation logics to reduce the IO contention:

2-a. Do not reject pages in zswap_store().
This is mostly achieved by patch 3. With patch 3, zswap can prepare space proactively and accept pages while the global shrinker is running.

To avoid rejection further, patch 5 (store incompressible pages) is added. This reduces rejection by storing incompressible pages. When zsmalloc is used, we can accept incompressible pages with small memory overhead. It is a minor optimization, but I think it is worth implementing. This does not improve performance on current zbud but does not incur a performance penalty.

2-b. Interrupt writeback while pagein/out.
Once zswap runs out of prepared space, we cannot accept incoming pages, incurring direct writes to the swap disk. At this moment, the shrinker is proactively evicting pages, leading to IO contention with memory reclaim.

Performing low-priority IO is straightforward but requires reimplementing a low-priority version of __swap_writepage(). Instead, in patch 6, I implemented a heuristic, delaying the next zswap writeback based on the elapsed time since zswap_store() rejected a page.

When zswap_store() hits the max pool size and rejects pages, swap_writepage() immediately performs the writeback to disk. The time jiffies is saved to tell shrink_worker() to sleep up to ZSWAP_GLOBAL_SHRINK_DELAY msec.

The same logic applied to zswap_load(). When zswap cannot find a page in the stored pool, pagein requires read IO from the swap device. The global shrinker should be interrupted here.

This patch proposes a constant delay of 500 milliseconds, aligning with the mq-deadline target latency.

Visible change
-------------------------------
With patches 4 to 6, the global shrinker pauses the writeback while pagein/out operations are using the swap device. This change reduces resource contention and makes memory reclaim/faults complete faster, thereby reducing system responsiveness degradation.

Intended scenario for memory reclaim:
1. zswap pool < accept_threshold as the initial state. This is achieved by patch 3, proactive shrinking.
2. Active processes start allocating pages. Pageout is buffered by zswap without IO.
3. zswap reaches shrink_start_threshold. zswap continues to buffer incoming pages and starts writeback immediately in the background.
4. zswap reaches max pool size. zswap interrupts the global shrinker and starts rejecting pages. Write IO for the rejected page will consume all IO resources.

5. Active processes stop allocating pages. After the delay, the shrinker resumes writeback until the accept threshold.

Benchmark
-------------------------------
To demonstrate that the shrinker writeback is not interfering with pagein/out operations, I measured the elapsed time of allocating 2GB of 3/4 compressible data by a Python script, averaged over 10 runs:

|                      | elapsed| user  | sys   |
|----------------------|--------|-------|-------|
| With patches 1 to 3  | 13.10  | 0.183 | 2.049 |
| With all patches     | 11.17  | 0.116 | 1.490 |
| zswap off (baseline) | 11.81  | 0.149 | 1.381 |

Although this test cannot distinguish responsiveness issues caused by zswap writeback from normal memory thrashing between plain pagein/out, the difference from the baseline indicates that the patches reduced performance degradation on pageout caused by zswap writeback.

Quoted 1 line

post-patch, 6.10-rc4 with patch 1 to 5

You mean 1 to 6? There are 6 patches, no?

Just out of pure curiosity, could you include the stats from patch 1-3 only?

Ah this is interesting. Did you actually see improvement in your real deployment (i.e not the benchmark) with patch 4-6 in?

Quoted 11 lines

Intended scenario for memory reclaim:
1. zswap pool < accept_threshold as the initial state. This is achieved by patch 3, proactive shrinking.
2. Active processes start allocating pages. Pageout is buffered by zswap without IO.
3. zswap reaches shrink_start_threshold. zswap continues to buffer incoming pages and starts writeback immediately in the background.
4. zswap reaches max pool size. zswap interrupts the global shrinker and starts rejecting pages. Write IO for the rejected page will consume all IO resources.

This sounds like the proactive shrinker is still not aggressive enough, and/or there are some sort of misspecifications of the zswap setting... Correct me if I'm wrong, but the new proactive global shrinker begins 1% after the acceptance threshold, and shrinks down to acceptance threshold, right? How are we still hitting the pool limit...

My concern is that we are knowingly (and perhaps unnecessarily) creating an LRU inversion here - preferring swapping out the rejected pages over the colder pages in the zswap pool. Shouldn't it be the other way around? For instance, can we spiral into the following scenario:

1. zswap pool becomes full.
2. Memory is still tight, so anonymous memory will be reclaimed. zswap keeps rejecting incoming pages, and putting a hold on the global shrinker.
3. The pages that are swapped out are warmer than the ones stored in the zswap pool, so they will be more likely to be swapped in (which, IIUC, will also further delay the global shrinker).

and the cycle keeps going on and on?

Have you experimented with synchronous reclaim in the case the pool is full? All the way to the acceptance threshold is too aggressive of course - you might need to find something in between :)

I wonder if this contention would show up in PSI metrics (/proc/pressure/io, or the cgroup variants if you use them ). Maybe correlate reclaim counters (pgscan, zswpout, pswpout, zswpwb etc.) with IO pressure to show the pattern, i.e the contention problem was there before, and is now resolved? :)

2024年7月9日(火) 9:53 Nhat Pham <[email protected]>:

Quoted 3 lines

Quoted 1 line

post-patch, 6.10-rc4 with patch 1 to 5

You mean 1 to 6? There are 6 patches, no?

oops. with patches 1 to 6.

Quoted 3 lines

Just out of pure curiosity, could you include the stats from patch 1-3 only?

I will rerun the bench in v3. I assume this bench does not reflect patches 4 to 6, as delta pool_limit_hit=0 means no rejection from zswap.

Quoted 3 lines

Ah this is interesting. Did you actually see improvement in your real deployment (i.e not the benchmark) with patch 4-6 in?

As replied in patch 6, memory consuming tasks like apt upgrade for instance.

Quoted 12 lines

Quoted 11 lines

Intended scenario for memory reclaim:
1. zswap pool < accept_threshold as the initial state. This is achieved by patch 3, proactive shrinking.
2. Active processes start allocating pages. Pageout is buffered by zswap without IO.
3. zswap reaches shrink_start_threshold. zswap continues to buffer incoming pages and starts writeback immediately in the background.
4. zswap reaches max pool size. zswap interrupts the global shrinker and starts rejecting pages. Write IO for the rejected page will consume all IO resources.

Quoted 7 lines

This sounds like the proactive shrinker is still not aggressive enough, and/or there are some sort of misspecifications of the zswap setting... Correct me if I'm wrong, but the new proactive global shrinker begins 1% after the acceptance threshold, and shrinks down to acceptance threshold, right? How are we still hitting the pool limit...

Proactive shrinking should not be aggressive. With patches 4 and 6, I modified the global shrinker to be less aggressive against pagein/out. Shrinking proactively cannot avoid hitting the pool limit when memory pressure grows faster.

Quoted 14 lines

My concern is that we are knowingly (and perhaps unnecessarily) creating an LRU inversion here - preferring swapping out the rejected pages over the colder pages in the zswap pool. Shouldn't it be the other way around? For instance, can we spiral into the following scenario:

1. zswap pool becomes full.
2. Memory is still tight, so anonymous memory will be reclaimed. zswap keeps rejecting incoming pages, and putting a hold on the global shrinker.
3. The pages that are swapped out are warmer than the ones stored in the zswap pool, so they will be more likely to be swapped in (which, IIUC, will also further delay the global shrinker).

Quoted 1 line

and the cycle keeps going on and on?

I agree this does not follow LRU, but I think the LRU priority inversion is unavoidable once the pool limit is hit.
The accept_thr_percent should be lowered to reduce the probability of LRU inversion if it matters. (it is why I implemented proactive shrinker.)

When the writeback throughput is slower than memory usage grows, zswap_store() will have to reject pages sooner or later.
If we evict the oldest stored pages synchronously before rejecting a new page (rotating pool to keep LRU), it will affect latency depending how much writeback is required to store the new page. If the oldest pages were compressed well, we would have to evict too many pages to store a warmer page, which blocks the reclaim progress. Fragmentation in the zspool may also increase the required writeback amount. We cannot accomplish both maintaining LRU priority and maintaining pageout latency.

Additionally, zswap_writeback_entry() is slower than direct pageout. I assume this is because shrinker performs 4KB IO synchronously. I am seeing shrinking throughput is limited by disk IOPS * 4KB while much higher throughput can be achieved by disabling zswap. direct pageout can be faster than zswap writeback, possibly because of bio optimization or sequential allocation of swap.

Quoted 6 lines

Have you experimented with synchronous reclaim in the case the pool is full? All the way to the acceptance threshold is too aggressive of course - you might need to find something in between :)

I don't get what the expected situation is.
The benchmark of patch 6 is performing synchronous reclaim in the case the pool is full, since bulk memory allocation (write to mmapped space) is much faster than writeback throughput. The zswap pool is filled instantly at the beginning of benchmark runs. The accept_thr_percent is not significant for the benchmark, I think.

Quoted 8 lines

I wonder if this contention would show up in PSI metrics (/proc/pressure/io, or the cgroup variants if you use them ). Maybe correlate reclaim counters (pgscan, zswpout, pswpout, zswpwb etc.) with IO pressure to show the pattern, i.e the contention problem was there before, and is now resolved? :)

Unfortunately, I could not find a reliable metric other than elapsed time. It seems PSI does not distinguish stalls for rejected pageout from stalls for shrinker writeback.
For counters, this issue affects latency but does not increase the number of pagein/out. Is there any better way to observe the origin of contention?

Thanks.

Quoted 6 lines

I agree this does not follow LRU, but I think the LRU priority inversion is unavoidable once the pool limit is hit.
The accept_thr_percent should be lowered to reduce the probability of LRU inversion if it matters. (it is why I implemented proactive shrinker.)

And yet, in your own benchmark it fails to prevent that, no? I think you lower it all the way down to 50%.

Quoted 11 lines

When the writeback throughput is slower than memory usage grows, zswap_store() will have to reject pages sooner or later.
If we evict the oldest stored pages synchronously before rejecting a new page (rotating pool to keep LRU), it will affect latency depending how much writeback is required to store the new page. If the oldest pages were compressed well, we would have to evict too many pages to store a warmer page, which blocks the reclaim progress. Fragmentation in the zspool may also increase the required writeback amount. We cannot accomplish both maintaining LRU priority and maintaining pageout latency.

Hmm yeah, I guess this is fair. Looks like there is not a lot of choice, if you want to maintain decent pageout latency...

I could suggest that you have a budgeted zswap writeback on zswap store - i.e if the pool is full, then try to zswap writeback until we have enough space or if the budget is reached. But that feels like even more engineering - the IO priority approach might even be easier at that point LOL.

Oh well, global shrinker delay it is :)

Quoted 7 lines

Additionally, zswap_writeback_entry() is slower than direct pageout. I assume this is because shrinker performs 4KB IO synchronously. I am seeing shrinking throughput is limited by disk IOPS * 4KB while much higher throughput can be achieved by disabling zswap. direct pageout can be faster than zswap writeback, possibly because of bio optimization or sequential allocation of swap.

Hah, this is interesting!

I wonder though, if the solution here is to perform some sort of batching for zswap writeback.

BTW, what is the type of the storage device you are using for swap? Is it SSD or HDD etc?

Quoted 7 lines

Quoted 6 lines

Have you experimented with synchronous reclaim in the case the pool is full? All the way to the acceptance threshold is too aggressive of course - you might need to find something in between :)

Quoted 6 lines

I don't get what the expected situation is.
The benchmark of patch 6 is performing synchronous reclaim in the case the pool is full, since bulk memory allocation (write to mmapped space) is much faster than writeback throughput. The zswap pool is filled instantly at the beginning of benchmark runs. The accept_thr_percent is not significant for the benchmark, I think.

No. I meant synchronous reclaim as in triggering zswap writeback within the zswap store path, to make space for the incoming new zswap pages. But you already addressed it above :)

Quoted 17 lines

Quoted 8 lines

I wonder if this contention would show up in PSI metrics (/proc/pressure/io, or the cgroup variants if you use them ). Maybe correlate reclaim counters (pgscan, zswpout, pswpout, zswpwb etc.) with IO pressure to show the pattern, i.e the contention problem was there before, and is now resolved? :)

Unfortunately, I could not find a reliable metric other than elapsed time. It seems PSI does not distinguish stalls for rejected pageout from stalls for shrinker writeback.
For counters, this issue affects latency but does not increase the number of pagein/out. Is there any better way to observe the origin of contention?

Thanks.

2024年7月13日(土) 8:02 Nhat Pham <[email protected]>:

It was tested on an Azure VM with SSD-backed storage. The total IOPS was capped at 4K IOPS by the VM host. The max throughput of the global shrinker was around 16 MB/s. Proactive shrinking cannot prevent pool_limit_hit since memory allocation can be on the order of GB/s. (The benchmark script allocates 2 GB sequentially, which was compressed to 1.3 GB, while the zswap pool was limited to 200 MB.)

__read_swap_cache_async(), or in shrink_lruvec()), we are doing block

Quoted 15 lines

My concern is that we are knowingly (and perhaps unnecessarily) creating an LRU inversion here - preferring swapping out the rejected pages over the colder pages in the zswap pool. Shouldn't it be the other way around? For instance, can we spiral into the following scenario:

1. zswap pool becomes full.
2. Memory is still tight, so anonymous memory will be reclaimed. zswap keeps rejecting incoming pages, and putting a hold on the global shrinker.
3. The pages that are swapped out are warmer than the ones stored in the zswap pool, so they will be more likely to be swapped in (which, IIUC, will also further delay the global shrinker).

Quoted 1 line

and the cycle keeps going on and on?

I agree this does not follow LRU, but I think the LRU priority inversion is unavoidable once the pool limit is hit.
The accept_thr_percent should be lowered to reduce the probability of LRU inversion if it matters. (it is why I implemented proactive shrinker.)

[..]

Quoted 15 lines

Quoted 15 lines

My concern is that we are knowingly (and perhaps unnecessarily) creating an LRU inversion here - preferring swapping out the rejected pages over the colder pages in the zswap pool. Shouldn't it be the other way around? For instance, can we spiral into the following scenario:

1. zswap pool becomes full.
2. Memory is still tight, so anonymous memory will be reclaimed. zswap keeps rejecting incoming pages, and putting a hold on the global shrinker.
3. The pages that are swapped out are warmer than the ones stored in the zswap pool, so they will be more likely to be swapped in (which, IIUC, will also further delay the global shrinker).

Quoted 7 lines

Quoted 1 line

and the cycle keeps going on and on?

I agree this does not follow LRU, but I think the LRU priority inversion is unavoidable once the pool limit is hit.
The accept_thr_percent should be lowered to reduce the probability of LRU inversion if it matters. (it is why I implemented proactive shrinker.)

Why?

Let's take a step back. You are suggesting that we throttle zswap writeback to allow reclaim to swapout warmer pages to swap device. As Nhat said, we are proliferating LRU inversion instead of fixing it.

I think I had a similar discussion with Johannes about this before, and we discussed that if zswap becomes full, we should instead throttle reclaim and allow zswap writeback to proceed (i.e. the opposite of what this series is doing). This would be similar to how we throttle reclaim today to wait for dirty pages to be written back.

This should reduce/fix the LRU inversion instead of proliferating it, and it should reduce the total amout of IO as colder pages should go to disk while warmer pages go to zswap. I am wondering if we can reuse the reclaim_throttle() mechanism here.

One concern I have is that we will also throttle file pages if we use reclaim_throttle(), since I don't see per-type throttling there. This could be fine, since we similarly throttle zswap reclaim if there are too many dirty file pages. I am not super familiar with reclaim throttling, so maybe I missed something obvious or there is a better way, but I believe that from a high level this should be the right way to go.

I actually think if we do this properly, and throttle reclaim when zswap becomes full, we may be able to drop the acceptance hysteresis and rely on the throttling mechanism to make sure we stop reclaim until we free up enough space in zswap to avoid consistently hitting the limit, but this could be a future extension.

Johannes, any thoughts here?

Anyway, since patches 1-2 are independent of the rest of the series, feel free to send them separately, and we can continue the discussion on the best way forward for the rest of the series.

Quoted 31 lines · Yosry Ahmed

[..]

Quoted 15 lines

Quoted 15 lines

My concern is that we are knowingly (and perhaps unnecessarily) creating an LRU inversion here - preferring swapping out the rejected pages over the colder pages in the zswap pool. Shouldn't it be the other way around? For instance, can we spiral into the following scenario:

1. zswap pool becomes full.
2. Memory is still tight, so anonymous memory will be reclaimed. zswap keeps rejecting incoming pages, and putting a hold on the global shrinker.
3. The pages that are swapped out are warmer than the ones stored in the zswap pool, so they will be more likely to be swapped in (which, IIUC, will also further delay the global shrinker).

Quoted 7 lines

Quoted 1 line

and the cycle keeps going on and on?

I agree this does not follow LRU, but I think the LRU priority inversion is unavoidable once the pool limit is hit.
The accept_thr_percent should be lowered to reduce the probability of LRU inversion if it matters. (it is why I implemented proactive shrinker.)

Why?

Let's take a step back. You are suggesting that we throttle zswap writeback to allow reclaim to swapout warmer pages to swap device. As Nhat said, we are proliferating LRU inversion instead of fixing it.

Quoted 6 lines

I think I had a similar discussion with Johannes about this before, and we discussed that if zswap becomes full, we should instead throttle reclaim and allow zswap writeback to proceed (i.e. the opposite of what this series is doing). This would be similar to how we throttle reclaim today to wait for dirty pages to be written back.

I completely agree with this analysis and proposal - it's somewhat similar to what I have in mind, but more fleshed out :)

Quoted 5 lines

This should reduce/fix the LRU inversion instead of proliferating it, and it should reduce the total amout of IO as colder pages should go to disk while warmer pages go to zswap. I am wondering if we can reuse the reclaim_throttle() mechanism here.

Quoted 7 lines

One concern I have is that we will also throttle file pages if we use reclaim_throttle(), since I don't see per-type throttling there. This could be fine, since we similarly throttle zswap reclaim if there are too many dirty file pages. I am not super familiar with reclaim throttling, so maybe I missed something obvious or there is a better way, but I believe that from a high level this should be the right way to go.

I don't think we have any infrastructure for anon-only throttling in vmscan logic, but it sounds trivial to implement if necessary :)

Quoted 6 lines

I actually think if we do this properly, and throttle reclaim when zswap becomes full, we may be able to drop the acceptance hysteresis and rely on the throttling mechanism to make sure we stop reclaim until we free up enough space in zswap to avoid consistently hitting the limit, but this could be a future extension.

Agree - this hysteresis heuristics needs to die.

IMHO, I think we should still have the proactive global shrinking action that Takero is proposing in patch 3. The throttling is nice, but it'd be even nicer if we can get ahead of that :)

I have always thought that the shrinker should play this role in one way or another. Instead of an arbitrary watermark and asynchronous work, it incrementally pushes the zswap LRU toward disk as reclaim activity increases.

Is the point behind proactive shrinking is to reduce the latency in the reclaim path?

Quoted 66 lines · Yosry Ahmed

Quoted 31 lines

[..]

Quoted 15 lines

Quoted 15 lines

My concern is that we are knowingly (and perhaps unnecessarily) creating an LRU inversion here - preferring swapping out the rejected pages over the colder pages in the zswap pool. Shouldn't it be the other way around? For instance, can we spiral into the following scenario:

1. zswap pool becomes full.
2. Memory is still tight, so anonymous memory will be reclaimed. zswap keeps rejecting incoming pages, and putting a hold on the global shrinker.
3. The pages that are swapped out are warmer than the ones stored in the zswap pool, so they will be more likely to be swapped in (which, IIUC, will also further delay the global shrinker).

Quoted 7 lines

Quoted 1 line

and the cycle keeps going on and on?

I agree this does not follow LRU, but I think the LRU priority inversion is unavoidable once the pool limit is hit.
The accept_thr_percent should be lowered to reduce the probability of LRU inversion if it matters. (it is why I implemented proactive shrinker.)

Why?

Let's take a step back. You are suggesting that we throttle zswap writeback to allow reclaim to swapout warmer pages to swap device. As Nhat said, we are proliferating LRU inversion instead of fixing it.

Quoted 6 lines

I think I had a similar discussion with Johannes about this before, and we discussed that if zswap becomes full, we should instead throttle reclaim and allow zswap writeback to proceed (i.e. the opposite of what this series is doing). This would be similar to how we throttle reclaim today to wait for dirty pages to be written back.

I completely agree with this analysis and proposal - it's somewhat similar to what I have in mind, but more fleshed out :)

Quoted 5 lines

This should reduce/fix the LRU inversion instead of proliferating it, and it should reduce the total amout of IO as colder pages should go to disk while warmer pages go to zswap. I am wondering if we can reuse the reclaim_throttle() mechanism here.

Quoted 7 lines

One concern I have is that we will also throttle file pages if we use reclaim_throttle(), since I don't see per-type throttling there. This could be fine, since we similarly throttle zswap reclaim if there are too many dirty file pages. I am not super familiar with reclaim throttling, so maybe I missed something obvious or there is a better way, but I believe that from a high level this should be the right way to go.

I don't think we have any infrastructure for anon-only throttling in vmscan logic, but it sounds trivial to implement if necessary :)

Quoted 6 lines

I actually think if we do this properly, and throttle reclaim when zswap becomes full, we may be able to drop the acceptance hysteresis and rely on the throttling mechanism to make sure we stop reclaim until we free up enough space in zswap to avoid consistently hitting the limit, but this could be a future extension.

Quoted 5 lines

Agree - this hysteresis heuristics needs to die.

IMHO, I think we should still have the proactive global shrinking action that Takero is proposing in patch 3. The throttling is nice, but it'd be even nicer if we can get ahead of that :)

I have always thought that the shrinker should play this role in one way or another. Instead of an arbitrary watermark and asynchronous work, it incrementally pushes the zswap LRU toward disk as reclaim activity increases.

Is the point behind proactive shrinking is to reduce the latency in the reclaim path?