From cacd5b04e24c74a813c694ec7b26a1a370b5d666 Mon Sep 17 00:00:00 2001
From: Chunhai Guo <guochunhai@vivo.com>
Date: Mon, 1 Apr 2024 07:55:50 -0600
Subject: erofs: rename utils.c to zutil.c

Currently, utils.c is only useful if CONFIG_EROFS_FS_ZIP is on.
So let's rename it to zutil.c as well as avoid its inclusion if
CONFIG_EROFS_FS_ZIP is explicitly disabled.

Signed-off-by: Chunhai Guo <guochunhai@vivo.com>
Reviewed-by: Gao Xiang <hsiangkao@linux.alibaba.com>
Link: https://lore.kernel.org/r/20240401135550.2550043-1-guochunhai@vivo.com
Signed-off-by: Gao Xiang <hsiangkao@linux.alibaba.com>
---
 fs/erofs/Makefile |   4 +-
 fs/erofs/utils.c  | 287 ------------------------------------------------------
 fs/erofs/zutil.c  | 279 ++++++++++++++++++++++++++++++++++++++++++++++++++++
 3 files changed, 281 insertions(+), 289 deletions(-)
 delete mode 100644 fs/erofs/utils.c
 create mode 100644 fs/erofs/zutil.c

(limited to 'fs')

diff --git a/fs/erofs/Makefile b/fs/erofs/Makefile
index 994d0b9deddf..845eafdcee4a 100644
--- a/fs/erofs/Makefile
+++ b/fs/erofs/Makefile
@@ -1,9 +1,9 @@
 # SPDX-License-Identifier: GPL-2.0-only
 
 obj-$(CONFIG_EROFS_FS) += erofs.o
-erofs-objs := super.o inode.o data.o namei.o dir.o utils.o sysfs.o
+erofs-objs := super.o inode.o data.o namei.o dir.o sysfs.o
 erofs-$(CONFIG_EROFS_FS_XATTR) += xattr.o
-erofs-$(CONFIG_EROFS_FS_ZIP) += decompressor.o zmap.o zdata.o pcpubuf.o
+erofs-$(CONFIG_EROFS_FS_ZIP) += decompressor.o zmap.o zdata.o pcpubuf.o zutil.o
 erofs-$(CONFIG_EROFS_FS_ZIP_LZMA) += decompressor_lzma.o
 erofs-$(CONFIG_EROFS_FS_ZIP_DEFLATE) += decompressor_deflate.o
 erofs-$(CONFIG_EROFS_FS_ONDEMAND) += fscache.o
diff --git a/fs/erofs/utils.c b/fs/erofs/utils.c
deleted file mode 100644
index 518bdd69c823..000000000000
--- a/fs/erofs/utils.c
+++ /dev/null
@@ -1,287 +0,0 @@
-// SPDX-License-Identifier: GPL-2.0-only
-/*
- * Copyright (C) 2018 HUAWEI, Inc.
- *             https://www.huawei.com/
- */
-#include "internal.h"
-
-struct page *erofs_allocpage(struct page **pagepool, gfp_t gfp)
-{
-	struct page *page = *pagepool;
-
-	if (page) {
-		DBG_BUGON(page_ref_count(page) != 1);
-		*pagepool = (struct page *)page_private(page);
-	} else {
-		page = alloc_page(gfp);
-	}
-	return page;
-}
-
-void erofs_release_pages(struct page **pagepool)
-{
-	while (*pagepool) {
-		struct page *page = *pagepool;
-
-		*pagepool = (struct page *)page_private(page);
-		put_page(page);
-	}
-}
-
-#ifdef CONFIG_EROFS_FS_ZIP
-/* global shrink count (for all mounted EROFS instances) */
-static atomic_long_t erofs_global_shrink_cnt;
-
-static bool erofs_workgroup_get(struct erofs_workgroup *grp)
-{
-	if (lockref_get_not_zero(&grp->lockref))
-		return true;
-
-	spin_lock(&grp->lockref.lock);
-	if (__lockref_is_dead(&grp->lockref)) {
-		spin_unlock(&grp->lockref.lock);
-		return false;
-	}
-
-	if (!grp->lockref.count++)
-		atomic_long_dec(&erofs_global_shrink_cnt);
-	spin_unlock(&grp->lockref.lock);
-	return true;
-}
-
-struct erofs_workgroup *erofs_find_workgroup(struct super_block *sb,
-					     pgoff_t index)
-{
-	struct erofs_sb_info *sbi = EROFS_SB(sb);
-	struct erofs_workgroup *grp;
-
-repeat:
-	rcu_read_lock();
-	grp = xa_load(&sbi->managed_pslots, index);
-	if (grp) {
-		if (!erofs_workgroup_get(grp)) {
-			/* prefer to relax rcu read side */
-			rcu_read_unlock();
-			goto repeat;
-		}
-
-		DBG_BUGON(index != grp->index);
-	}
-	rcu_read_unlock();
-	return grp;
-}
-
-struct erofs_workgroup *erofs_insert_workgroup(struct super_block *sb,
-					       struct erofs_workgroup *grp)
-{
-	struct erofs_sb_info *const sbi = EROFS_SB(sb);
-	struct erofs_workgroup *pre;
-
-	DBG_BUGON(grp->lockref.count < 1);
-repeat:
-	xa_lock(&sbi->managed_pslots);
-	pre = __xa_cmpxchg(&sbi->managed_pslots, grp->index,
-			   NULL, grp, GFP_KERNEL);
-	if (pre) {
-		if (xa_is_err(pre)) {
-			pre = ERR_PTR(xa_err(pre));
-		} else if (!erofs_workgroup_get(pre)) {
-			/* try to legitimize the current in-tree one */
-			xa_unlock(&sbi->managed_pslots);
-			cond_resched();
-			goto repeat;
-		}
-		grp = pre;
-	}
-	xa_unlock(&sbi->managed_pslots);
-	return grp;
-}
-
-static void  __erofs_workgroup_free(struct erofs_workgroup *grp)
-{
-	atomic_long_dec(&erofs_global_shrink_cnt);
-	erofs_workgroup_free_rcu(grp);
-}
-
-void erofs_workgroup_put(struct erofs_workgroup *grp)
-{
-	if (lockref_put_or_lock(&grp->lockref))
-		return;
-
-	DBG_BUGON(__lockref_is_dead(&grp->lockref));
-	if (grp->lockref.count == 1)
-		atomic_long_inc(&erofs_global_shrink_cnt);
-	--grp->lockref.count;
-	spin_unlock(&grp->lockref.lock);
-}
-
-static bool erofs_try_to_release_workgroup(struct erofs_sb_info *sbi,
-					   struct erofs_workgroup *grp)
-{
-	int free = false;
-
-	spin_lock(&grp->lockref.lock);
-	if (grp->lockref.count)
-		goto out;
-
-	/*
-	 * Note that all cached pages should be detached before deleted from
-	 * the XArray. Otherwise some cached pages could be still attached to
-	 * the orphan old workgroup when the new one is available in the tree.
-	 */
-	if (erofs_try_to_free_all_cached_folios(sbi, grp))
-		goto out;
-
-	/*
-	 * It's impossible to fail after the workgroup is freezed,
-	 * however in order to avoid some race conditions, add a
-	 * DBG_BUGON to observe this in advance.
-	 */
-	DBG_BUGON(__xa_erase(&sbi->managed_pslots, grp->index) != grp);
-
-	lockref_mark_dead(&grp->lockref);
-	free = true;
-out:
-	spin_unlock(&grp->lockref.lock);
-	if (free)
-		__erofs_workgroup_free(grp);
-	return free;
-}
-
-static unsigned long erofs_shrink_workstation(struct erofs_sb_info *sbi,
-					      unsigned long nr_shrink)
-{
-	struct erofs_workgroup *grp;
-	unsigned int freed = 0;
-	unsigned long index;
-
-	xa_lock(&sbi->managed_pslots);
-	xa_for_each(&sbi->managed_pslots, index, grp) {
-		/* try to shrink each valid workgroup */
-		if (!erofs_try_to_release_workgroup(sbi, grp))
-			continue;
-		xa_unlock(&sbi->managed_pslots);
-
-		++freed;
-		if (!--nr_shrink)
-			return freed;
-		xa_lock(&sbi->managed_pslots);
-	}
-	xa_unlock(&sbi->managed_pslots);
-	return freed;
-}
-
-/* protected by 'erofs_sb_list_lock' */
-static unsigned int shrinker_run_no;
-
-/* protects the mounted 'erofs_sb_list' */
-static DEFINE_SPINLOCK(erofs_sb_list_lock);
-static LIST_HEAD(erofs_sb_list);
-
-void erofs_shrinker_register(struct super_block *sb)
-{
-	struct erofs_sb_info *sbi = EROFS_SB(sb);
-
-	mutex_init(&sbi->umount_mutex);
-
-	spin_lock(&erofs_sb_list_lock);
-	list_add(&sbi->list, &erofs_sb_list);
-	spin_unlock(&erofs_sb_list_lock);
-}
-
-void erofs_shrinker_unregister(struct super_block *sb)
-{
-	struct erofs_sb_info *const sbi = EROFS_SB(sb);
-
-	mutex_lock(&sbi->umount_mutex);
-	/* clean up all remaining workgroups in memory */
-	erofs_shrink_workstation(sbi, ~0UL);
-
-	spin_lock(&erofs_sb_list_lock);
-	list_del(&sbi->list);
-	spin_unlock(&erofs_sb_list_lock);
-	mutex_unlock(&sbi->umount_mutex);
-}
-
-static unsigned long erofs_shrink_count(struct shrinker *shrink,
-					struct shrink_control *sc)
-{
-	return atomic_long_read(&erofs_global_shrink_cnt);
-}
-
-static unsigned long erofs_shrink_scan(struct shrinker *shrink,
-				       struct shrink_control *sc)
-{
-	struct erofs_sb_info *sbi;
-	struct list_head *p;
-
-	unsigned long nr = sc->nr_to_scan;
-	unsigned int run_no;
-	unsigned long freed = 0;
-
-	spin_lock(&erofs_sb_list_lock);
-	do {
-		run_no = ++shrinker_run_no;
-	} while (run_no == 0);
-
-	/* Iterate over all mounted superblocks and try to shrink them */
-	p = erofs_sb_list.next;
-	while (p != &erofs_sb_list) {
-		sbi = list_entry(p, struct erofs_sb_info, list);
-
-		/*
-		 * We move the ones we do to the end of the list, so we stop
-		 * when we see one we have already done.
-		 */
-		if (sbi->shrinker_run_no == run_no)
-			break;
-
-		if (!mutex_trylock(&sbi->umount_mutex)) {
-			p = p->next;
-			continue;
-		}
-
-		spin_unlock(&erofs_sb_list_lock);
-		sbi->shrinker_run_no = run_no;
-
-		freed += erofs_shrink_workstation(sbi, nr - freed);
-
-		spin_lock(&erofs_sb_list_lock);
-		/* Get the next list element before we move this one */
-		p = p->next;
-
-		/*
-		 * Move this one to the end of the list to provide some
-		 * fairness.
-		 */
-		list_move_tail(&sbi->list, &erofs_sb_list);
-		mutex_unlock(&sbi->umount_mutex);
-
-		if (freed >= nr)
-			break;
-	}
-	spin_unlock(&erofs_sb_list_lock);
-	return freed;
-}
-
-static struct shrinker *erofs_shrinker_info;
-
-int __init erofs_init_shrinker(void)
-{
-	erofs_shrinker_info = shrinker_alloc(0, "erofs-shrinker");
-	if (!erofs_shrinker_info)
-		return -ENOMEM;
-
-	erofs_shrinker_info->count_objects = erofs_shrink_count;
-	erofs_shrinker_info->scan_objects = erofs_shrink_scan;
-
-	shrinker_register(erofs_shrinker_info);
-
-	return 0;
-}
-
-void erofs_exit_shrinker(void)
-{
-	shrinker_free(erofs_shrinker_info);
-}
-#endif	/* !CONFIG_EROFS_FS_ZIP */
diff --git a/fs/erofs/zutil.c b/fs/erofs/zutil.c
new file mode 100644
index 000000000000..8cd30ac2091f
--- /dev/null
+++ b/fs/erofs/zutil.c
@@ -0,0 +1,279 @@
+// SPDX-License-Identifier: GPL-2.0-only
+/*
+ * Copyright (C) 2018 HUAWEI, Inc.
+ *             https://www.huawei.com/
+ */
+#include "internal.h"
+
+static atomic_long_t erofs_global_shrink_cnt;	/* for all mounted instances */
+/* protected by 'erofs_sb_list_lock' */
+static unsigned int shrinker_run_no;
+
+/* protects the mounted 'erofs_sb_list' */
+static DEFINE_SPINLOCK(erofs_sb_list_lock);
+static LIST_HEAD(erofs_sb_list);
+static struct shrinker *erofs_shrinker_info;
+
+struct page *erofs_allocpage(struct page **pagepool, gfp_t gfp)
+{
+	struct page *page = *pagepool;
+
+	if (page) {
+		DBG_BUGON(page_ref_count(page) != 1);
+		*pagepool = (struct page *)page_private(page);
+		return page;
+	}
+	return alloc_page(gfp);
+}
+
+void erofs_release_pages(struct page **pagepool)
+{
+	while (*pagepool) {
+		struct page *page = *pagepool;
+
+		*pagepool = (struct page *)page_private(page);
+		put_page(page);
+	}
+}
+
+static bool erofs_workgroup_get(struct erofs_workgroup *grp)
+{
+	if (lockref_get_not_zero(&grp->lockref))
+		return true;
+
+	spin_lock(&grp->lockref.lock);
+	if (__lockref_is_dead(&grp->lockref)) {
+		spin_unlock(&grp->lockref.lock);
+		return false;
+	}
+
+	if (!grp->lockref.count++)
+		atomic_long_dec(&erofs_global_shrink_cnt);
+	spin_unlock(&grp->lockref.lock);
+	return true;
+}
+
+struct erofs_workgroup *erofs_find_workgroup(struct super_block *sb,
+					     pgoff_t index)
+{
+	struct erofs_sb_info *sbi = EROFS_SB(sb);
+	struct erofs_workgroup *grp;
+
+repeat:
+	rcu_read_lock();
+	grp = xa_load(&sbi->managed_pslots, index);
+	if (grp) {
+		if (!erofs_workgroup_get(grp)) {
+			/* prefer to relax rcu read side */
+			rcu_read_unlock();
+			goto repeat;
+		}
+
+		DBG_BUGON(index != grp->index);
+	}
+	rcu_read_unlock();
+	return grp;
+}
+
+struct erofs_workgroup *erofs_insert_workgroup(struct super_block *sb,
+					       struct erofs_workgroup *grp)
+{
+	struct erofs_sb_info *const sbi = EROFS_SB(sb);
+	struct erofs_workgroup *pre;
+
+	DBG_BUGON(grp->lockref.count < 1);
+repeat:
+	xa_lock(&sbi->managed_pslots);
+	pre = __xa_cmpxchg(&sbi->managed_pslots, grp->index,
+			   NULL, grp, GFP_KERNEL);
+	if (pre) {
+		if (xa_is_err(pre)) {
+			pre = ERR_PTR(xa_err(pre));
+		} else if (!erofs_workgroup_get(pre)) {
+			/* try to legitimize the current in-tree one */
+			xa_unlock(&sbi->managed_pslots);
+			cond_resched();
+			goto repeat;
+		}
+		grp = pre;
+	}
+	xa_unlock(&sbi->managed_pslots);
+	return grp;
+}
+
+static void  __erofs_workgroup_free(struct erofs_workgroup *grp)
+{
+	atomic_long_dec(&erofs_global_shrink_cnt);
+	erofs_workgroup_free_rcu(grp);
+}
+
+void erofs_workgroup_put(struct erofs_workgroup *grp)
+{
+	if (lockref_put_or_lock(&grp->lockref))
+		return;
+
+	DBG_BUGON(__lockref_is_dead(&grp->lockref));
+	if (grp->lockref.count == 1)
+		atomic_long_inc(&erofs_global_shrink_cnt);
+	--grp->lockref.count;
+	spin_unlock(&grp->lockref.lock);
+}
+
+static bool erofs_try_to_release_workgroup(struct erofs_sb_info *sbi,
+					   struct erofs_workgroup *grp)
+{
+	int free = false;
+
+	spin_lock(&grp->lockref.lock);
+	if (grp->lockref.count)
+		goto out;
+
+	/*
+	 * Note that all cached pages should be detached before deleted from
+	 * the XArray. Otherwise some cached pages could be still attached to
+	 * the orphan old workgroup when the new one is available in the tree.
+	 */
+	if (erofs_try_to_free_all_cached_folios(sbi, grp))
+		goto out;
+
+	/*
+	 * It's impossible to fail after the workgroup is freezed,
+	 * however in order to avoid some race conditions, add a
+	 * DBG_BUGON to observe this in advance.
+	 */
+	DBG_BUGON(__xa_erase(&sbi->managed_pslots, grp->index) != grp);
+
+	lockref_mark_dead(&grp->lockref);
+	free = true;
+out:
+	spin_unlock(&grp->lockref.lock);
+	if (free)
+		__erofs_workgroup_free(grp);
+	return free;
+}
+
+static unsigned long erofs_shrink_workstation(struct erofs_sb_info *sbi,
+					      unsigned long nr_shrink)
+{
+	struct erofs_workgroup *grp;
+	unsigned int freed = 0;
+	unsigned long index;
+
+	xa_lock(&sbi->managed_pslots);
+	xa_for_each(&sbi->managed_pslots, index, grp) {
+		/* try to shrink each valid workgroup */
+		if (!erofs_try_to_release_workgroup(sbi, grp))
+			continue;
+		xa_unlock(&sbi->managed_pslots);
+
+		++freed;
+		if (!--nr_shrink)
+			return freed;
+		xa_lock(&sbi->managed_pslots);
+	}
+	xa_unlock(&sbi->managed_pslots);
+	return freed;
+}
+
+void erofs_shrinker_register(struct super_block *sb)
+{
+	struct erofs_sb_info *sbi = EROFS_SB(sb);
+
+	mutex_init(&sbi->umount_mutex);
+
+	spin_lock(&erofs_sb_list_lock);
+	list_add(&sbi->list, &erofs_sb_list);
+	spin_unlock(&erofs_sb_list_lock);
+}
+
+void erofs_shrinker_unregister(struct super_block *sb)
+{
+	struct erofs_sb_info *const sbi = EROFS_SB(sb);
+
+	mutex_lock(&sbi->umount_mutex);
+	/* clean up all remaining workgroups in memory */
+	erofs_shrink_workstation(sbi, ~0UL);
+
+	spin_lock(&erofs_sb_list_lock);
+	list_del(&sbi->list);
+	spin_unlock(&erofs_sb_list_lock);
+	mutex_unlock(&sbi->umount_mutex);
+}
+
+static unsigned long erofs_shrink_count(struct shrinker *shrink,
+					struct shrink_control *sc)
+{
+	return atomic_long_read(&erofs_global_shrink_cnt);
+}
+
+static unsigned long erofs_shrink_scan(struct shrinker *shrink,
+				       struct shrink_control *sc)
+{
+	struct erofs_sb_info *sbi;
+	struct list_head *p;
+
+	unsigned long nr = sc->nr_to_scan;
+	unsigned int run_no;
+	unsigned long freed = 0;
+
+	spin_lock(&erofs_sb_list_lock);
+	do {
+		run_no = ++shrinker_run_no;
+	} while (run_no == 0);
+
+	/* Iterate over all mounted superblocks and try to shrink them */
+	p = erofs_sb_list.next;
+	while (p != &erofs_sb_list) {
+		sbi = list_entry(p, struct erofs_sb_info, list);
+
+		/*
+		 * We move the ones we do to the end of the list, so we stop
+		 * when we see one we have already done.
+		 */
+		if (sbi->shrinker_run_no == run_no)
+			break;
+
+		if (!mutex_trylock(&sbi->umount_mutex)) {
+			p = p->next;
+			continue;
+		}
+
+		spin_unlock(&erofs_sb_list_lock);
+		sbi->shrinker_run_no = run_no;
+
+		freed += erofs_shrink_workstation(sbi, nr - freed);
+
+		spin_lock(&erofs_sb_list_lock);
+		/* Get the next list element before we move this one */
+		p = p->next;
+
+		/*
+		 * Move this one to the end of the list to provide some
+		 * fairness.
+		 */
+		list_move_tail(&sbi->list, &erofs_sb_list);
+		mutex_unlock(&sbi->umount_mutex);
+
+		if (freed >= nr)
+			break;
+	}
+	spin_unlock(&erofs_sb_list_lock);
+	return freed;
+}
+
+int __init erofs_init_shrinker(void)
+{
+	erofs_shrinker_info = shrinker_alloc(0, "erofs-shrinker");
+	if (!erofs_shrinker_info)
+		return -ENOMEM;
+
+	erofs_shrinker_info->count_objects = erofs_shrink_count;
+	erofs_shrinker_info->scan_objects = erofs_shrink_scan;
+	shrinker_register(erofs_shrinker_info);
+	return 0;
+}
+
+void erofs_exit_shrinker(void)
+{
+	shrinker_free(erofs_shrinker_info);
+}
-- 
cgit v1.2.3


From f36f3010f67611a45d66e773bc91e4c66a9abab5 Mon Sep 17 00:00:00 2001
From: Chunhai Guo <guochunhai@vivo.com>
Date: Tue, 2 Apr 2024 04:00:36 -0600
Subject: erofs: rename per-CPU buffers to global buffer pool and make it
 configurable

It will cost more time if compressed buffers are allocated on demand for
low-latency algorithms (like lz4) so EROFS uses per-CPU buffers to keep
compressed data if in-place decompression is unfulfilled.  While it is kind
of wasteful of memory for a device with hundreds of CPUs, and only a small
number of CPUs concurrently decompress most of the time.

This patch renames it as 'global buffer pool' and makes it configurable.
This allows two or more CPUs to share a common buffer to reduce memory
occupation.

Suggested-by: Gao Xiang <xiang@kernel.org>
Reviewed-by: Gao Xiang <hsiangkao@linux.alibaba.com>
Signed-off-by: Chunhai Guo <guochunhai@vivo.com>
Link: https://lore.kernel.org/r/20240402100036.2673604-1-guochunhai@vivo.com
Signed-off-by: Sandeep Dhavale <dhavale@google.com>
Link: https://lore.kernel.org/r/20240408215231.3376659-1-dhavale@google.com
Signed-off-by: Gao Xiang <hsiangkao@linux.alibaba.com>
---
 fs/erofs/Makefile       |   2 +-
 fs/erofs/decompressor.c |   6 +-
 fs/erofs/internal.h     |  14 ++---
 fs/erofs/pcpubuf.c      | 148 ------------------------------------------------
 fs/erofs/super.c        |   9 ++-
 fs/erofs/zutil.c        | 148 ++++++++++++++++++++++++++++++++++++++++++++++++
 6 files changed, 166 insertions(+), 161 deletions(-)
 delete mode 100644 fs/erofs/pcpubuf.c

(limited to 'fs')

diff --git a/fs/erofs/Makefile b/fs/erofs/Makefile
index 845eafdcee4a..20d1ec422443 100644
--- a/fs/erofs/Makefile
+++ b/fs/erofs/Makefile
@@ -3,7 +3,7 @@
 obj-$(CONFIG_EROFS_FS) += erofs.o
 erofs-objs := super.o inode.o data.o namei.o dir.o sysfs.o
 erofs-$(CONFIG_EROFS_FS_XATTR) += xattr.o
-erofs-$(CONFIG_EROFS_FS_ZIP) += decompressor.o zmap.o zdata.o pcpubuf.o zutil.o
+erofs-$(CONFIG_EROFS_FS_ZIP) += decompressor.o zmap.o zdata.o zutil.o
 erofs-$(CONFIG_EROFS_FS_ZIP_LZMA) += decompressor_lzma.o
 erofs-$(CONFIG_EROFS_FS_ZIP_DEFLATE) += decompressor_deflate.o
 erofs-$(CONFIG_EROFS_FS_ONDEMAND) += fscache.o
diff --git a/fs/erofs/decompressor.c b/fs/erofs/decompressor.c
index 2ec9b2bb628d..e1239d886984 100644
--- a/fs/erofs/decompressor.c
+++ b/fs/erofs/decompressor.c
@@ -54,7 +54,7 @@ static int z_erofs_load_lz4_config(struct super_block *sb,
 	sbi->lz4.max_distance_pages = distance ?
 					DIV_ROUND_UP(distance, PAGE_SIZE) + 1 :
 					LZ4_MAX_DISTANCE_PAGES;
-	return erofs_pcpubuf_growsize(sbi->lz4.max_pclusterblks);
+	return z_erofs_gbuf_growsize(sbi->lz4.max_pclusterblks);
 }
 
 /*
@@ -159,7 +159,7 @@ static void *z_erofs_lz4_handle_overlap(struct z_erofs_lz4_decompress_ctx *ctx,
 docopy:
 	/* Or copy compressed data which can be overlapped to per-CPU buffer */
 	in = rq->in;
-	src = erofs_get_pcpubuf(ctx->inpages);
+	src = z_erofs_get_gbuf(ctx->inpages);
 	if (!src) {
 		DBG_BUGON(1);
 		kunmap_local(inpage);
@@ -260,7 +260,7 @@ static int z_erofs_lz4_decompress_mem(struct z_erofs_lz4_decompress_ctx *ctx,
 	} else if (maptype == 1) {
 		vm_unmap_ram(src, ctx->inpages);
 	} else if (maptype == 2) {
-		erofs_put_pcpubuf(src);
+		z_erofs_put_gbuf(src);
 	} else if (maptype != 3) {
 		DBG_BUGON(1);
 		return -EFAULT;
diff --git a/fs/erofs/internal.h b/fs/erofs/internal.h
index d28ccfc0352b..ee080d042ab3 100644
--- a/fs/erofs/internal.h
+++ b/fs/erofs/internal.h
@@ -463,11 +463,11 @@ int erofs_try_to_free_all_cached_folios(struct erofs_sb_info *sbi,
 					struct erofs_workgroup *egrp);
 int z_erofs_map_blocks_iter(struct inode *inode, struct erofs_map_blocks *map,
 			    int flags);
-void *erofs_get_pcpubuf(unsigned int requiredpages);
-void erofs_put_pcpubuf(void *ptr);
-int erofs_pcpubuf_growsize(unsigned int nrpages);
-void __init erofs_pcpubuf_init(void);
-void erofs_pcpubuf_exit(void);
+void *z_erofs_get_gbuf(unsigned int requiredpages);
+void z_erofs_put_gbuf(void *ptr);
+int z_erofs_gbuf_growsize(unsigned int nrpages);
+int __init z_erofs_gbuf_init(void);
+void z_erofs_gbuf_exit(void);
 int erofs_init_managed_cache(struct super_block *sb);
 int z_erofs_parse_cfgs(struct super_block *sb, struct erofs_super_block *dsb);
 #else
@@ -477,8 +477,8 @@ static inline int erofs_init_shrinker(void) { return 0; }
 static inline void erofs_exit_shrinker(void) {}
 static inline int z_erofs_init_zip_subsystem(void) { return 0; }
 static inline void z_erofs_exit_zip_subsystem(void) {}
-static inline void erofs_pcpubuf_init(void) {}
-static inline void erofs_pcpubuf_exit(void) {}
+static inline int z_erofs_gbuf_init(void) { return 0; }
+static inline void z_erofs_gbuf_exit(void) {}
 static inline int erofs_init_managed_cache(struct super_block *sb) { return 0; }
 #endif	/* !CONFIG_EROFS_FS_ZIP */
 
diff --git a/fs/erofs/pcpubuf.c b/fs/erofs/pcpubuf.c
deleted file mode 100644
index c7a4b1d77069..000000000000
--- a/fs/erofs/pcpubuf.c
+++ /dev/null
@@ -1,148 +0,0 @@
-// SPDX-License-Identifier: GPL-2.0-only
-/*
- * Copyright (C) Gao Xiang <xiang@kernel.org>
- *
- * For low-latency decompression algorithms (e.g. lz4), reserve consecutive
- * per-CPU virtual memory (in pages) in advance to store such inplace I/O
- * data if inplace decompression is failed (due to unmet inplace margin for
- * example).
- */
-#include "internal.h"
-
-struct erofs_pcpubuf {
-	raw_spinlock_t lock;
-	void *ptr;
-	struct page **pages;
-	unsigned int nrpages;
-};
-
-static DEFINE_PER_CPU(struct erofs_pcpubuf, erofs_pcb);
-
-void *erofs_get_pcpubuf(unsigned int requiredpages)
-	__acquires(pcb->lock)
-{
-	struct erofs_pcpubuf *pcb = &get_cpu_var(erofs_pcb);
-
-	raw_spin_lock(&pcb->lock);
-	/* check if the per-CPU buffer is too small */
-	if (requiredpages > pcb->nrpages) {
-		raw_spin_unlock(&pcb->lock);
-		put_cpu_var(erofs_pcb);
-		/* (for sparse checker) pretend pcb->lock is still taken */
-		__acquire(pcb->lock);
-		return NULL;
-	}
-	return pcb->ptr;
-}
-
-void erofs_put_pcpubuf(void *ptr) __releases(pcb->lock)
-{
-	struct erofs_pcpubuf *pcb = &per_cpu(erofs_pcb, smp_processor_id());
-
-	DBG_BUGON(pcb->ptr != ptr);
-	raw_spin_unlock(&pcb->lock);
-	put_cpu_var(erofs_pcb);
-}
-
-/* the next step: support per-CPU page buffers hotplug */
-int erofs_pcpubuf_growsize(unsigned int nrpages)
-{
-	static DEFINE_MUTEX(pcb_resize_mutex);
-	static unsigned int pcb_nrpages;
-	struct page *pagepool = NULL;
-	int delta, cpu, ret, i;
-
-	mutex_lock(&pcb_resize_mutex);
-	delta = nrpages - pcb_nrpages;
-	ret = 0;
-	/* avoid shrinking pcpubuf, since no idea how many fses rely on */
-	if (delta <= 0)
-		goto out;
-
-	for_each_possible_cpu(cpu) {
-		struct erofs_pcpubuf *pcb = &per_cpu(erofs_pcb, cpu);
-		struct page **pages, **oldpages;
-		void *ptr, *old_ptr;
-
-		pages = kmalloc_array(nrpages, sizeof(*pages), GFP_KERNEL);
-		if (!pages) {
-			ret = -ENOMEM;
-			break;
-		}
-
-		for (i = 0; i < nrpages; ++i) {
-			pages[i] = erofs_allocpage(&pagepool, GFP_KERNEL);
-			if (!pages[i]) {
-				ret = -ENOMEM;
-				oldpages = pages;
-				goto free_pagearray;
-			}
-		}
-		ptr = vmap(pages, nrpages, VM_MAP, PAGE_KERNEL);
-		if (!ptr) {
-			ret = -ENOMEM;
-			oldpages = pages;
-			goto free_pagearray;
-		}
-		raw_spin_lock(&pcb->lock);
-		old_ptr = pcb->ptr;
-		pcb->ptr = ptr;
-		oldpages = pcb->pages;
-		pcb->pages = pages;
-		i = pcb->nrpages;
-		pcb->nrpages = nrpages;
-		raw_spin_unlock(&pcb->lock);
-
-		if (!oldpages) {
-			DBG_BUGON(old_ptr);
-			continue;
-		}
-
-		if (old_ptr)
-			vunmap(old_ptr);
-free_pagearray:
-		while (i)
-			erofs_pagepool_add(&pagepool, oldpages[--i]);
-		kfree(oldpages);
-		if (ret)
-			break;
-	}
-	pcb_nrpages = nrpages;
-	erofs_release_pages(&pagepool);
-out:
-	mutex_unlock(&pcb_resize_mutex);
-	return ret;
-}
-
-void __init erofs_pcpubuf_init(void)
-{
-	int cpu;
-
-	for_each_possible_cpu(cpu) {
-		struct erofs_pcpubuf *pcb = &per_cpu(erofs_pcb, cpu);
-
-		raw_spin_lock_init(&pcb->lock);
-	}
-}
-
-void erofs_pcpubuf_exit(void)
-{
-	int cpu, i;
-
-	for_each_possible_cpu(cpu) {
-		struct erofs_pcpubuf *pcb = &per_cpu(erofs_pcb, cpu);
-
-		if (pcb->ptr) {
-			vunmap(pcb->ptr);
-			pcb->ptr = NULL;
-		}
-		if (!pcb->pages)
-			continue;
-
-		for (i = 0; i < pcb->nrpages; ++i)
-			if (pcb->pages[i])
-				put_page(pcb->pages[i]);
-		kfree(pcb->pages);
-		pcb->pages = NULL;
-	}
-}
diff --git a/fs/erofs/super.c b/fs/erofs/super.c
index 30b49b2eee53..c1dae1fb949b 100644
--- a/fs/erofs/super.c
+++ b/fs/erofs/super.c
@@ -859,7 +859,10 @@ static int __init erofs_module_init(void)
 	if (err)
 		goto deflate_err;
 
-	erofs_pcpubuf_init();
+	err = z_erofs_gbuf_init();
+	if (err)
+		goto gbuf_err;
+
 	err = z_erofs_init_zip_subsystem();
 	if (err)
 		goto zip_err;
@@ -879,6 +882,8 @@ fs_err:
 sysfs_err:
 	z_erofs_exit_zip_subsystem();
 zip_err:
+	z_erofs_gbuf_exit();
+gbuf_err:
 	z_erofs_deflate_exit();
 deflate_err:
 	z_erofs_lzma_exit();
@@ -902,7 +907,7 @@ static void __exit erofs_module_exit(void)
 	z_erofs_lzma_exit();
 	erofs_exit_shrinker();
 	kmem_cache_destroy(erofs_inode_cachep);
-	erofs_pcpubuf_exit();
+	z_erofs_gbuf_exit();
 }
 
 static int erofs_statfs(struct dentry *dentry, struct kstatfs *buf)
diff --git a/fs/erofs/zutil.c b/fs/erofs/zutil.c
index 8cd30ac2091f..2fa90b10b985 100644
--- a/fs/erofs/zutil.c
+++ b/fs/erofs/zutil.c
@@ -5,6 +5,18 @@
  */
 #include "internal.h"
 
+struct z_erofs_gbuf {
+	spinlock_t lock;
+	void *ptr;
+	struct page **pages;
+	unsigned int nrpages;
+};
+
+static struct z_erofs_gbuf *z_erofs_gbufpool;
+static unsigned int z_erofs_gbuf_count, z_erofs_gbuf_nrpages;
+
+module_param_named(global_buffers, z_erofs_gbuf_count, uint, 0444);
+
 static atomic_long_t erofs_global_shrink_cnt;	/* for all mounted instances */
 /* protected by 'erofs_sb_list_lock' */
 static unsigned int shrinker_run_no;
@@ -14,6 +26,142 @@ static DEFINE_SPINLOCK(erofs_sb_list_lock);
 static LIST_HEAD(erofs_sb_list);
 static struct shrinker *erofs_shrinker_info;
 
+static unsigned int z_erofs_gbuf_id(void)
+{
+	return raw_smp_processor_id() % z_erofs_gbuf_count;
+}
+
+void *z_erofs_get_gbuf(unsigned int requiredpages)
+	__acquires(gbuf->lock)
+{
+	struct z_erofs_gbuf *gbuf;
+
+	gbuf = &z_erofs_gbufpool[z_erofs_gbuf_id()];
+	spin_lock(&gbuf->lock);
+	/* check if the buffer is too small */
+	if (requiredpages > gbuf->nrpages) {
+		spin_unlock(&gbuf->lock);
+		/* (for sparse checker) pretend gbuf->lock is still taken */
+		__acquire(gbuf->lock);
+		return NULL;
+	}
+	return gbuf->ptr;
+}
+
+void z_erofs_put_gbuf(void *ptr) __releases(gbuf->lock)
+{
+	struct z_erofs_gbuf *gbuf;
+
+	gbuf = &z_erofs_gbufpool[z_erofs_gbuf_id()];
+	DBG_BUGON(gbuf->ptr != ptr);
+	spin_unlock(&gbuf->lock);
+}
+
+int z_erofs_gbuf_growsize(unsigned int nrpages)
+{
+	static DEFINE_MUTEX(gbuf_resize_mutex);
+	struct page *pagepool = NULL;
+	int delta, ret, i, j;
+
+	mutex_lock(&gbuf_resize_mutex);
+	delta = nrpages - z_erofs_gbuf_nrpages;
+	ret = 0;
+	/* avoid shrinking gbufs, since no idea how many fses rely on */
+	if (delta <= 0)
+		goto out;
+
+	for (i = 0; i < z_erofs_gbuf_count; ++i) {
+		struct z_erofs_gbuf *gbuf = &z_erofs_gbufpool[i];
+		struct page **pages, **tmp_pages;
+		void *ptr, *old_ptr = NULL;
+
+		ret = -ENOMEM;
+		tmp_pages = kcalloc(nrpages, sizeof(*tmp_pages), GFP_KERNEL);
+		if (!tmp_pages)
+			break;
+		for (j = 0; j < nrpages; ++j) {
+			tmp_pages[j] = erofs_allocpage(&pagepool, GFP_KERNEL);
+			if (!tmp_pages[j])
+				goto free_pagearray;
+		}
+		ptr = vmap(tmp_pages, nrpages, VM_MAP, PAGE_KERNEL);
+		if (!ptr)
+			goto free_pagearray;
+
+		pages = tmp_pages;
+		spin_lock(&gbuf->lock);
+		old_ptr = gbuf->ptr;
+		gbuf->ptr = ptr;
+		tmp_pages = gbuf->pages;
+		gbuf->pages = pages;
+		j = gbuf->nrpages;
+		gbuf->nrpages = nrpages;
+		spin_unlock(&gbuf->lock);
+		ret = 0;
+		if (!tmp_pages) {
+			DBG_BUGON(old_ptr);
+			continue;
+		}
+
+		if (old_ptr)
+			vunmap(old_ptr);
+free_pagearray:
+		while (j)
+			erofs_pagepool_add(&pagepool, tmp_pages[--j]);
+		kfree(tmp_pages);
+		if (ret)
+			break;
+	}
+	z_erofs_gbuf_nrpages = nrpages;
+	erofs_release_pages(&pagepool);
+out:
+	mutex_unlock(&gbuf_resize_mutex);
+	return ret;
+}
+
+int __init z_erofs_gbuf_init(void)
+{
+	unsigned int i = num_possible_cpus();
+
+	if (!z_erofs_gbuf_count)
+		z_erofs_gbuf_count = i;
+	else
+		z_erofs_gbuf_count = min(z_erofs_gbuf_count, i);
+
+	z_erofs_gbufpool = kcalloc(z_erofs_gbuf_count,
+			sizeof(*z_erofs_gbufpool), GFP_KERNEL);
+	if (!z_erofs_gbufpool)
+		return -ENOMEM;
+
+	for (i = 0; i < z_erofs_gbuf_count; ++i)
+		spin_lock_init(&z_erofs_gbufpool[i].lock);
+	return 0;
+}
+
+void z_erofs_gbuf_exit(void)
+{
+	int i;
+
+	for (i = 0; i < z_erofs_gbuf_count; ++i) {
+		struct z_erofs_gbuf *gbuf = &z_erofs_gbufpool[i];
+
+		if (gbuf->ptr) {
+			vunmap(gbuf->ptr);
+			gbuf->ptr = NULL;
+		}
+
+		if (!gbuf->pages)
+			continue;
+
+		for (i = 0; i < gbuf->nrpages; ++i)
+			if (gbuf->pages[i])
+				put_page(gbuf->pages[i]);
+		kfree(gbuf->pages);
+		gbuf->pages = NULL;
+	}
+	kfree(z_erofs_gbufpool);
+}
+
 struct page *erofs_allocpage(struct page **pagepool, gfp_t gfp)
 {
 	struct page *page = *pagepool;
-- 
cgit v1.2.3


From d6db47e571dcaecaeaafa8840d00ae849ae3907b Mon Sep 17 00:00:00 2001
From: Chunhai Guo <guochunhai@vivo.com>
Date: Tue, 2 Apr 2024 03:27:57 -0600
Subject: erofs: do not use pagepool in z_erofs_gbuf_growsize()

Let's use alloc_pages_bulk_array() for simplicity and get rid of
unnecessary pagepool.

Signed-off-by: Chunhai Guo <guochunhai@vivo.com>
Reviewed-by: Gao Xiang <hsiangkao@linux.alibaba.com>
Link: https://lore.kernel.org/r/20240402092757.2635257-1-guochunhai@vivo.com
Signed-off-by: Gao Xiang <hsiangkao@linux.alibaba.com>
---
 fs/erofs/zutil.c | 67 ++++++++++++++++++++++++++------------------------------
 1 file changed, 31 insertions(+), 36 deletions(-)

(limited to 'fs')

diff --git a/fs/erofs/zutil.c b/fs/erofs/zutil.c
index 2fa90b10b985..bc88dc4fe8bd 100644
--- a/fs/erofs/zutil.c
+++ b/fs/erofs/zutil.c
@@ -60,63 +60,58 @@ void z_erofs_put_gbuf(void *ptr) __releases(gbuf->lock)
 int z_erofs_gbuf_growsize(unsigned int nrpages)
 {
 	static DEFINE_MUTEX(gbuf_resize_mutex);
-	struct page *pagepool = NULL;
-	int delta, ret, i, j;
+	struct page **tmp_pages = NULL;
+	struct z_erofs_gbuf *gbuf;
+	void *ptr, *old_ptr;
+	int last, i, j;
 
 	mutex_lock(&gbuf_resize_mutex);
-	delta = nrpages - z_erofs_gbuf_nrpages;
-	ret = 0;
 	/* avoid shrinking gbufs, since no idea how many fses rely on */
-	if (delta <= 0)
-		goto out;
+	if (nrpages <= z_erofs_gbuf_nrpages) {
+		mutex_unlock(&gbuf_resize_mutex);
+		return 0;
+	}
 
 	for (i = 0; i < z_erofs_gbuf_count; ++i) {
-		struct z_erofs_gbuf *gbuf = &z_erofs_gbufpool[i];
-		struct page **pages, **tmp_pages;
-		void *ptr, *old_ptr = NULL;
-
-		ret = -ENOMEM;
+		gbuf = &z_erofs_gbufpool[i];
 		tmp_pages = kcalloc(nrpages, sizeof(*tmp_pages), GFP_KERNEL);
 		if (!tmp_pages)
-			break;
-		for (j = 0; j < nrpages; ++j) {
-			tmp_pages[j] = erofs_allocpage(&pagepool, GFP_KERNEL);
-			if (!tmp_pages[j])
-				goto free_pagearray;
-		}
+			goto out;
+
+		for (j = 0; j < gbuf->nrpages; ++j)
+			tmp_pages[j] = gbuf->pages[j];
+		do {
+			last = j;
+			j = alloc_pages_bulk_array(GFP_KERNEL, nrpages,
+						   tmp_pages);
+			if (last == j)
+				goto out;
+		} while (j != nrpages);
+
 		ptr = vmap(tmp_pages, nrpages, VM_MAP, PAGE_KERNEL);
 		if (!ptr)
-			goto free_pagearray;
+			goto out;
 
-		pages = tmp_pages;
 		spin_lock(&gbuf->lock);
+		kfree(gbuf->pages);
+		gbuf->pages = tmp_pages;
 		old_ptr = gbuf->ptr;
 		gbuf->ptr = ptr;
-		tmp_pages = gbuf->pages;
-		gbuf->pages = pages;
-		j = gbuf->nrpages;
 		gbuf->nrpages = nrpages;
 		spin_unlock(&gbuf->lock);
-		ret = 0;
-		if (!tmp_pages) {
-			DBG_BUGON(old_ptr);
-			continue;
-		}
-
 		if (old_ptr)
 			vunmap(old_ptr);
-free_pagearray:
-		while (j)
-			erofs_pagepool_add(&pagepool, tmp_pages[--j]);
-		kfree(tmp_pages);
-		if (ret)
-			break;
 	}
 	z_erofs_gbuf_nrpages = nrpages;
-	erofs_release_pages(&pagepool);
 out:
+	if (i < z_erofs_gbuf_count && tmp_pages) {
+		for (j = 0; j < nrpages; ++j)
+			if (tmp_pages[j] && tmp_pages[j] != gbuf->pages[j])
+				__free_page(tmp_pages[j]);
+		kfree(tmp_pages);
+	}
 	mutex_unlock(&gbuf_resize_mutex);
-	return ret;
+	return i < z_erofs_gbuf_count ? -ENOMEM : 0;
 }
 
 int __init z_erofs_gbuf_init(void)
-- 
cgit v1.2.3


From 0f6273ab46375b62c8dd5c987ce7c15877602831 Mon Sep 17 00:00:00 2001
From: Chunhai Guo <guochunhai@vivo.com>
Date: Tue, 2 Apr 2024 07:15:23 -0600
Subject: erofs: add a reserved buffer pool for lz4 decompression

This adds a special global buffer pool (in the end) for reserved pages.

Using a reserved pool for LZ4 decompression significantly reduces the
time spent on extra temporary page allocation for the extreme cases in
low memory scenarios.

The table below shows the reduction in time spent on page allocation for
LZ4 decompression when using a reserved pool. The results were obtained
from multi-app launch benchmarks on ARM64 Android devices running the
5.15 kernel with an 8-core CPU and 8GB of memory. In the benchmark, we
launched 16 frequently-used apps, and the camera app was the last one in
each round. The data in the table is the average time of camera app for
each round.

After using the reserved pool, there was an average improvement of 150ms
in the overall launch time of our camera app, which was obtained from
the systrace log.

+--------------+---------------+--------------+---------+
|              | w/o page pool | w/ page pool |  diff   |
+--------------+---------------+--------------+---------+
| Average (ms) |     3434      |      21      | -99.38% |
+--------------+---------------+--------------+---------+

Based on the benchmark logs, 64 pages are sufficient for 95% of
scenarios. This value can be adjusted with a module parameter
`reserved_pages`. The default value is 0.

This pool is currently only used for the LZ4 decompressor, but it can be
applied to more decompressors if needed.

Signed-off-by: Chunhai Guo <guochunhai@vivo.com>
Reviewed-by: Gao Xiang <hsiangkao@linux.alibaba.com>
Link: https://lore.kernel.org/r/20240402131523.2703948-1-guochunhai@vivo.com
Signed-off-by: Gao Xiang <hsiangkao@linux.alibaba.com>
---
 fs/erofs/decompressor.c |  2 +-
 fs/erofs/internal.h     |  6 ++++-
 fs/erofs/zutil.c        | 61 +++++++++++++++++++++++++++++++++++++------------
 3 files changed, 52 insertions(+), 17 deletions(-)

(limited to 'fs')

diff --git a/fs/erofs/decompressor.c b/fs/erofs/decompressor.c
index e1239d886984..d2fe8130819e 100644
--- a/fs/erofs/decompressor.c
+++ b/fs/erofs/decompressor.c
@@ -111,7 +111,7 @@ static int z_erofs_lz4_prepare_dstpages(struct z_erofs_lz4_decompress_ctx *ctx,
 			victim = availables[--top];
 			get_page(victim);
 		} else {
-			victim = erofs_allocpage(pagepool, rq->gfp);
+			victim = __erofs_allocpage(pagepool, rq->gfp, true);
 			if (!victim)
 				return -ENOMEM;
 			set_page_private(victim, Z_EROFS_SHORTLIVED_PAGE);
diff --git a/fs/erofs/internal.h b/fs/erofs/internal.h
index ee080d042ab3..53ebba952a2f 100644
--- a/fs/erofs/internal.h
+++ b/fs/erofs/internal.h
@@ -438,7 +438,11 @@ void erofs_unregister_sysfs(struct super_block *sb);
 int __init erofs_init_sysfs(void);
 void erofs_exit_sysfs(void);
 
-struct page *erofs_allocpage(struct page **pagepool, gfp_t gfp);
+struct page *__erofs_allocpage(struct page **pagepool, gfp_t gfp, bool tryrsv);
+static inline struct page *erofs_allocpage(struct page **pagepool, gfp_t gfp)
+{
+	return __erofs_allocpage(pagepool, gfp, false);
+}
 static inline void erofs_pagepool_add(struct page **pagepool, struct page *page)
 {
 	set_page_private(page, (unsigned long)*pagepool);
diff --git a/fs/erofs/zutil.c b/fs/erofs/zutil.c
index bc88dc4fe8bd..036024bce9f7 100644
--- a/fs/erofs/zutil.c
+++ b/fs/erofs/zutil.c
@@ -12,10 +12,12 @@ struct z_erofs_gbuf {
 	unsigned int nrpages;
 };
 
-static struct z_erofs_gbuf *z_erofs_gbufpool;
-static unsigned int z_erofs_gbuf_count, z_erofs_gbuf_nrpages;
+static struct z_erofs_gbuf *z_erofs_gbufpool, *z_erofs_rsvbuf;
+static unsigned int z_erofs_gbuf_count, z_erofs_gbuf_nrpages,
+		z_erofs_rsv_nrpages;
 
 module_param_named(global_buffers, z_erofs_gbuf_count, uint, 0444);
+module_param_named(reserved_pages, z_erofs_rsv_nrpages, uint, 0444);
 
 static atomic_long_t erofs_global_shrink_cnt;	/* for all mounted instances */
 /* protected by 'erofs_sb_list_lock' */
@@ -116,19 +118,30 @@ out:
 
 int __init z_erofs_gbuf_init(void)
 {
-	unsigned int i = num_possible_cpus();
+	unsigned int i, total = num_possible_cpus();
 
-	if (!z_erofs_gbuf_count)
-		z_erofs_gbuf_count = i;
-	else
-		z_erofs_gbuf_count = min(z_erofs_gbuf_count, i);
+	if (z_erofs_gbuf_count)
+		total = min(z_erofs_gbuf_count, total);
+	z_erofs_gbuf_count = total;
 
-	z_erofs_gbufpool = kcalloc(z_erofs_gbuf_count,
-			sizeof(*z_erofs_gbufpool), GFP_KERNEL);
+	/* The last (special) global buffer is the reserved buffer */
+	total += !!z_erofs_rsv_nrpages;
+
+	z_erofs_gbufpool = kcalloc(total, sizeof(*z_erofs_gbufpool),
+				   GFP_KERNEL);
 	if (!z_erofs_gbufpool)
 		return -ENOMEM;
 
-	for (i = 0; i < z_erofs_gbuf_count; ++i)
+	if (z_erofs_rsv_nrpages) {
+		z_erofs_rsvbuf = &z_erofs_gbufpool[total - 1];
+		z_erofs_rsvbuf->pages = kcalloc(z_erofs_rsv_nrpages,
+				sizeof(*z_erofs_rsvbuf->pages), GFP_KERNEL);
+		if (!z_erofs_rsvbuf->pages) {
+			z_erofs_rsvbuf = NULL;
+			z_erofs_rsv_nrpages = 0;
+		}
+	}
+	for (i = 0; i < total; ++i)
 		spin_lock_init(&z_erofs_gbufpool[i].lock);
 	return 0;
 }
@@ -137,7 +150,7 @@ void z_erofs_gbuf_exit(void)
 {
 	int i;
 
-	for (i = 0; i < z_erofs_gbuf_count; ++i) {
+	for (i = 0; i < z_erofs_gbuf_count + (!!z_erofs_rsvbuf); ++i) {
 		struct z_erofs_gbuf *gbuf = &z_erofs_gbufpool[i];
 
 		if (gbuf->ptr) {
@@ -157,16 +170,22 @@ void z_erofs_gbuf_exit(void)
 	kfree(z_erofs_gbufpool);
 }
 
-struct page *erofs_allocpage(struct page **pagepool, gfp_t gfp)
+struct page *__erofs_allocpage(struct page **pagepool, gfp_t gfp, bool tryrsv)
 {
 	struct page *page = *pagepool;
 
 	if (page) {
-		DBG_BUGON(page_ref_count(page) != 1);
 		*pagepool = (struct page *)page_private(page);
-		return page;
+	} else if (tryrsv && z_erofs_rsvbuf && z_erofs_rsvbuf->nrpages) {
+		spin_lock(&z_erofs_rsvbuf->lock);
+		if (z_erofs_rsvbuf->nrpages)
+			page = z_erofs_rsvbuf->pages[--z_erofs_rsvbuf->nrpages];
+		spin_unlock(&z_erofs_rsvbuf->lock);
 	}
-	return alloc_page(gfp);
+	if (!page)
+		page = alloc_page(gfp);
+	DBG_BUGON(page && page_ref_count(page) != 1);
+	return page;
 }
 
 void erofs_release_pages(struct page **pagepool)
@@ -175,6 +194,18 @@ void erofs_release_pages(struct page **pagepool)
 		struct page *page = *pagepool;
 
 		*pagepool = (struct page *)page_private(page);
+		/* try to fill reserved global pool first */
+		if (z_erofs_rsvbuf && z_erofs_rsvbuf->nrpages <
+				z_erofs_rsv_nrpages) {
+			spin_lock(&z_erofs_rsvbuf->lock);
+			if (z_erofs_rsvbuf->nrpages < z_erofs_rsv_nrpages) {
+				z_erofs_rsvbuf->pages[z_erofs_rsvbuf->nrpages++]
+						= page;
+				spin_unlock(&z_erofs_rsvbuf->lock);
+				continue;
+			}
+			spin_unlock(&z_erofs_rsvbuf->lock);
+		}
 		put_page(page);
 	}
 }
-- 
cgit v1.2.3


From 1872df8dcd87bd1e623e1a70076d08636b9c473d Mon Sep 17 00:00:00 2001
From: Hongzhen Luo <hongzhen@linux.alibaba.com>
Date: Tue, 9 Apr 2024 19:30:22 +0800
Subject: erofs: derive fsid from on-disk UUID for .statfs() if possible

Use the superblock's UUID to generate the fsid when it's non-null.

Reviewed-by: Gao Xiang <hsiangkao@linux.alibaba.com>
Reviewed-by: Jingbo Xu <jefflexu@linux.alibaba.com>
Signed-off-by: Hongzhen Luo <hongzhen@linux.alibaba.com>
Link: https://lore.kernel.org/r/20240409113022.74720-1-hongzhen@linux.alibaba.com
Signed-off-by: Gao Xiang <hsiangkao@linux.alibaba.com>
---
 fs/erofs/super.c | 12 +++++-------
 1 file changed, 5 insertions(+), 7 deletions(-)

(limited to 'fs')

diff --git a/fs/erofs/super.c b/fs/erofs/super.c
index c1dae1fb949b..e3438f1a7bac 100644
--- a/fs/erofs/super.c
+++ b/fs/erofs/super.c
@@ -914,22 +914,20 @@ static int erofs_statfs(struct dentry *dentry, struct kstatfs *buf)
 {
 	struct super_block *sb = dentry->d_sb;
 	struct erofs_sb_info *sbi = EROFS_SB(sb);
-	u64 id = 0;
-
-	if (!erofs_is_fscache_mode(sb))
-		id = huge_encode_dev(sb->s_bdev->bd_dev);
 
 	buf->f_type = sb->s_magic;
 	buf->f_bsize = sb->s_blocksize;
 	buf->f_blocks = sbi->total_blocks;
 	buf->f_bfree = buf->f_bavail = 0;
-
 	buf->f_files = ULLONG_MAX;
 	buf->f_ffree = ULLONG_MAX - sbi->inos;
-
 	buf->f_namelen = EROFS_NAME_LEN;
 
-	buf->f_fsid    = u64_to_fsid(id);
+	if (uuid_is_null(&sb->s_uuid))
+		buf->f_fsid = u64_to_fsid(erofs_is_fscache_mode(sb) ? 0 :
+				huge_encode_dev(sb->s_bdev->bd_dev));
+	else
+		buf->f_fsid = uuid_to_fsid(sb->s_uuid.b);
 	return 0;
 }
 
-- 
cgit v1.2.3


From d69189428d50c9a8de6475f3c8c241f86f4e764a Mon Sep 17 00:00:00 2001
From: Gao Xiang <hsiangkao@linux.alibaba.com>
Date: Wed, 8 May 2024 20:33:57 +0800
Subject: erofs: clean up z_erofs_load_full_lcluster()

Only four lcluster types here, remove redundant code.
No real logic changes.

Signed-off-by: Gao Xiang <hsiangkao@linux.alibaba.com>
Link: https://lore.kernel.org/r/20240508123357.3266173-1-hsiangkao@linux.alibaba.com
---
 fs/erofs/erofs_fs.h |  5 +----
 fs/erofs/zmap.c     | 21 +++++----------------
 2 files changed, 6 insertions(+), 20 deletions(-)

(limited to 'fs')

diff --git a/fs/erofs/erofs_fs.h b/fs/erofs/erofs_fs.h
index a03ec70ba6f2..550baf1729d4 100644
--- a/fs/erofs/erofs_fs.h
+++ b/fs/erofs/erofs_fs.h
@@ -396,8 +396,7 @@ enum {
 	Z_EROFS_LCLUSTER_TYPE_MAX
 };
 
-#define Z_EROFS_LI_LCLUSTER_TYPE_BITS        2
-#define Z_EROFS_LI_LCLUSTER_TYPE_BIT         0
+#define Z_EROFS_LI_LCLUSTER_TYPE_MASK	(Z_EROFS_LCLUSTER_TYPE_MAX - 1)
 
 /* (noncompact only, HEAD) This pcluster refers to partial decompressed data */
 #define Z_EROFS_LI_PARTIAL_REF		(1 << 15)
@@ -451,8 +450,6 @@ static inline void erofs_check_ondisk_layout_definitions(void)
 		     sizeof(struct z_erofs_lcluster_index));
 	BUILD_BUG_ON(sizeof(struct erofs_deviceslot) != 128);
 
-	BUILD_BUG_ON(BIT(Z_EROFS_LI_LCLUSTER_TYPE_BITS) <
-		     Z_EROFS_LCLUSTER_TYPE_MAX - 1);
 	/* exclude old compiler versions like gcc 7.5.0 */
 	BUILD_BUG_ON(__builtin_constant_p(fmh) ?
 		     fmh != cpu_to_le64(1ULL << 63) : 0);
diff --git a/fs/erofs/zmap.c b/fs/erofs/zmap.c
index e313c936351d..26637a60eba5 100644
--- a/fs/erofs/zmap.c
+++ b/fs/erofs/zmap.c
@@ -31,7 +31,7 @@ static int z_erofs_load_full_lcluster(struct z_erofs_maprecorder *m,
 			vi->inode_isize + vi->xattr_isize) +
 			lcn * sizeof(struct z_erofs_lcluster_index);
 	struct z_erofs_lcluster_index *di;
-	unsigned int advise, type;
+	unsigned int advise;
 
 	m->kaddr = erofs_read_metabuf(&m->map->buf, inode->i_sb,
 				      erofs_blknr(inode->i_sb, pos), EROFS_KMAP);
@@ -43,10 +43,8 @@ static int z_erofs_load_full_lcluster(struct z_erofs_maprecorder *m,
 	di = m->kaddr + erofs_blkoff(inode->i_sb, pos);
 
 	advise = le16_to_cpu(di->di_advise);
-	type = (advise >> Z_EROFS_LI_LCLUSTER_TYPE_BIT) &
-		((1 << Z_EROFS_LI_LCLUSTER_TYPE_BITS) - 1);
-	switch (type) {
-	case Z_EROFS_LCLUSTER_TYPE_NONHEAD:
+	m->type = advise & Z_EROFS_LI_LCLUSTER_TYPE_MASK;
+	if (m->type == Z_EROFS_LCLUSTER_TYPE_NONHEAD) {
 		m->clusterofs = 1 << vi->z_logical_clusterbits;
 		m->delta[0] = le16_to_cpu(di->di_u.delta[0]);
 		if (m->delta[0] & Z_EROFS_LI_D0_CBLKCNT) {
@@ -60,24 +58,15 @@ static int z_erofs_load_full_lcluster(struct z_erofs_maprecorder *m,
 			m->delta[0] = 1;
 		}
 		m->delta[1] = le16_to_cpu(di->di_u.delta[1]);
-		break;
-	case Z_EROFS_LCLUSTER_TYPE_PLAIN:
-	case Z_EROFS_LCLUSTER_TYPE_HEAD1:
-	case Z_EROFS_LCLUSTER_TYPE_HEAD2:
-		if (advise & Z_EROFS_LI_PARTIAL_REF)
-			m->partialref = true;
+	} else {
+		m->partialref = !!(advise & Z_EROFS_LI_PARTIAL_REF);
 		m->clusterofs = le16_to_cpu(di->di_clusterofs);
 		if (m->clusterofs >= 1 << vi->z_logical_clusterbits) {
 			DBG_BUGON(1);
 			return -EFSCORRUPTED;
 		}
 		m->pblk = le32_to_cpu(di->di_u.blkaddr);
-		break;
-	default:
-		DBG_BUGON(1);
-		return -EOPNOTSUPP;
 	}
-	m->type = type;
 	return 0;
 }
 
-- 
cgit v1.2.3


From 7c35de4df1056a5a1fb4de042197b8f5b1033b61 Mon Sep 17 00:00:00 2001
From: Gao Xiang <hsiangkao@linux.alibaba.com>
Date: Thu, 9 May 2024 07:44:53 +0800
Subject: erofs: Zstandard compression support

Add Zstandard compression as the 4th supported algorithm since it
becomes more popular now and some end users have asked this for
quite a while [1][2].

Each EROFS physical cluster contains only one valid standard
Zstandard frame as described in [3] so that decompression can be
performed on a per-pcluster basis independently.

Currently, it just leverages multi-call stream decompression APIs with
internal sliding window buffers.  One-shot or bufferless decompression
could be implemented later for even better performance if needed.

[1] https://github.com/erofs/erofs-utils/issues/6
[2] https://lore.kernel.org/r/Y08h+z6CZdnS1XBm@B-P7TQMD6M-0146.lan
[3] https://www.rfc-editor.org/rfc/rfc8478.txt

Acked-by: Chao Yu <chao@kernel.org>
Signed-off-by: Gao Xiang <hsiangkao@linux.alibaba.com>
Link: https://lore.kernel.org/r/20240508234453.17896-1-xiang@kernel.org
---
 fs/erofs/Kconfig             |  15 +++
 fs/erofs/Makefile            |   1 +
 fs/erofs/compress.h          |   4 +
 fs/erofs/decompressor.c      |   7 ++
 fs/erofs/decompressor_zstd.c | 279 +++++++++++++++++++++++++++++++++++++++++++
 fs/erofs/erofs_fs.h          |  10 ++
 fs/erofs/internal.h          |   8 ++
 fs/erofs/super.c             |   7 ++
 fs/erofs/zmap.c              |   3 +-
 9 files changed, 333 insertions(+), 1 deletion(-)
 create mode 100644 fs/erofs/decompressor_zstd.c

(limited to 'fs')

diff --git a/fs/erofs/Kconfig b/fs/erofs/Kconfig
index fffd3919343e..7dcdce660cac 100644
--- a/fs/erofs/Kconfig
+++ b/fs/erofs/Kconfig
@@ -112,6 +112,21 @@ config EROFS_FS_ZIP_DEFLATE
 
 	  If unsure, say N.
 
+config EROFS_FS_ZIP_ZSTD
+	bool "EROFS Zstandard compressed data support"
+	depends on EROFS_FS_ZIP
+	select ZSTD_DECOMPRESS
+	help
+	  Saying Y here includes support for reading EROFS file systems
+	  containing Zstandard compressed data.  It gives better compression
+	  ratios than the default LZ4 format, while it costs more CPU
+	  overhead.
+
+	  Zstandard support is an experimental feature for now and so most
+	  file systems will be readable without selecting this option.
+
+	  If unsure, say N.
+
 config EROFS_FS_ONDEMAND
 	bool "EROFS fscache-based on-demand read support"
 	depends on EROFS_FS
diff --git a/fs/erofs/Makefile b/fs/erofs/Makefile
index 20d1ec422443..097d672e6b14 100644
--- a/fs/erofs/Makefile
+++ b/fs/erofs/Makefile
@@ -6,4 +6,5 @@ erofs-$(CONFIG_EROFS_FS_XATTR) += xattr.o
 erofs-$(CONFIG_EROFS_FS_ZIP) += decompressor.o zmap.o zdata.o zutil.o
 erofs-$(CONFIG_EROFS_FS_ZIP_LZMA) += decompressor_lzma.o
 erofs-$(CONFIG_EROFS_FS_ZIP_DEFLATE) += decompressor_deflate.o
+erofs-$(CONFIG_EROFS_FS_ZIP_ZSTD) += decompressor_zstd.o
 erofs-$(CONFIG_EROFS_FS_ONDEMAND) += fscache.o
diff --git a/fs/erofs/compress.h b/fs/erofs/compress.h
index 333587ba6183..19d53c30c8af 100644
--- a/fs/erofs/compress.h
+++ b/fs/erofs/compress.h
@@ -90,8 +90,12 @@ int z_erofs_load_lzma_config(struct super_block *sb,
 			struct erofs_super_block *dsb, void *data, int size);
 int z_erofs_load_deflate_config(struct super_block *sb,
 			struct erofs_super_block *dsb, void *data, int size);
+int z_erofs_load_zstd_config(struct super_block *sb,
+			struct erofs_super_block *dsb, void *data, int size);
 int z_erofs_lzma_decompress(struct z_erofs_decompress_req *rq,
 			    struct page **pagepool);
 int z_erofs_deflate_decompress(struct z_erofs_decompress_req *rq,
 			       struct page **pagepool);
+int z_erofs_zstd_decompress(struct z_erofs_decompress_req *rq,
+			    struct page **pgpl);
 #endif
diff --git a/fs/erofs/decompressor.c b/fs/erofs/decompressor.c
index d2fe8130819e..9d85b6c11c6b 100644
--- a/fs/erofs/decompressor.c
+++ b/fs/erofs/decompressor.c
@@ -399,6 +399,13 @@ const struct z_erofs_decompressor erofs_decompressors[] = {
 		.name = "deflate"
 	},
 #endif
+#ifdef CONFIG_EROFS_FS_ZIP_ZSTD
+	[Z_EROFS_COMPRESSION_ZSTD] = {
+		.config = z_erofs_load_zstd_config,
+		.decompress = z_erofs_zstd_decompress,
+		.name = "zstd"
+	},
+#endif
 };
 
 int z_erofs_parse_cfgs(struct super_block *sb, struct erofs_super_block *dsb)
diff --git a/fs/erofs/decompressor_zstd.c b/fs/erofs/decompressor_zstd.c
new file mode 100644
index 000000000000..63a23cac3af4
--- /dev/null
+++ b/fs/erofs/decompressor_zstd.c
@@ -0,0 +1,279 @@
+// SPDX-License-Identifier: GPL-2.0-or-later
+#include <linux/zstd.h>
+#include "compress.h"
+
+struct z_erofs_zstd {
+	struct z_erofs_zstd *next;
+	u8 bounce[PAGE_SIZE];
+	void *wksp;
+	unsigned int wkspsz;
+};
+
+static DEFINE_SPINLOCK(z_erofs_zstd_lock);
+static unsigned int z_erofs_zstd_max_dictsize;
+static unsigned int z_erofs_zstd_nstrms, z_erofs_zstd_avail_strms;
+static struct z_erofs_zstd *z_erofs_zstd_head;
+static DECLARE_WAIT_QUEUE_HEAD(z_erofs_zstd_wq);
+
+module_param_named(zstd_streams, z_erofs_zstd_nstrms, uint, 0444);
+
+static struct z_erofs_zstd *z_erofs_isolate_strms(bool all)
+{
+	struct z_erofs_zstd *strm;
+
+again:
+	spin_lock(&z_erofs_zstd_lock);
+	strm = z_erofs_zstd_head;
+	if (!strm) {
+		spin_unlock(&z_erofs_zstd_lock);
+		wait_event(z_erofs_zstd_wq, READ_ONCE(z_erofs_zstd_head));
+		goto again;
+	}
+	z_erofs_zstd_head = all ? NULL : strm->next;
+	spin_unlock(&z_erofs_zstd_lock);
+	return strm;
+}
+
+void z_erofs_zstd_exit(void)
+{
+	while (z_erofs_zstd_avail_strms) {
+		struct z_erofs_zstd *strm, *n;
+
+		for (strm = z_erofs_isolate_strms(true); strm; strm = n) {
+			n = strm->next;
+
+			kvfree(strm->wksp);
+			kfree(strm);
+			--z_erofs_zstd_avail_strms;
+		}
+	}
+}
+
+int __init z_erofs_zstd_init(void)
+{
+	/* by default, use # of possible CPUs instead */
+	if (!z_erofs_zstd_nstrms)
+		z_erofs_zstd_nstrms = num_possible_cpus();
+
+	for (; z_erofs_zstd_avail_strms < z_erofs_zstd_nstrms;
+	     ++z_erofs_zstd_avail_strms) {
+		struct z_erofs_zstd *strm;
+
+		strm = kzalloc(sizeof(*strm), GFP_KERNEL);
+		if (!strm) {
+			z_erofs_zstd_exit();
+			return -ENOMEM;
+		}
+		spin_lock(&z_erofs_zstd_lock);
+		strm->next = z_erofs_zstd_head;
+		z_erofs_zstd_head = strm;
+		spin_unlock(&z_erofs_zstd_lock);
+	}
+	return 0;
+}
+
+int z_erofs_load_zstd_config(struct super_block *sb,
+			struct erofs_super_block *dsb, void *data, int size)
+{
+	static DEFINE_MUTEX(zstd_resize_mutex);
+	struct z_erofs_zstd_cfgs *zstd = data;
+	unsigned int dict_size, wkspsz;
+	struct z_erofs_zstd *strm, *head = NULL;
+	void *wksp;
+
+	if (!zstd || size < sizeof(struct z_erofs_zstd_cfgs) || zstd->format) {
+		erofs_err(sb, "unsupported zstd format, size=%u", size);
+		return -EINVAL;
+	}
+
+	if (zstd->windowlog > ilog2(Z_EROFS_ZSTD_MAX_DICT_SIZE) - 10) {
+		erofs_err(sb, "unsupported zstd window log %u", zstd->windowlog);
+		return -EINVAL;
+	}
+	dict_size = 1U << (zstd->windowlog + 10);
+
+	/* in case 2 z_erofs_load_zstd_config() race to avoid deadlock */
+	mutex_lock(&zstd_resize_mutex);
+	if (z_erofs_zstd_max_dictsize >= dict_size) {
+		mutex_unlock(&zstd_resize_mutex);
+		return 0;
+	}
+
+	/* 1. collect/isolate all streams for the following check */
+	while (z_erofs_zstd_avail_strms) {
+		struct z_erofs_zstd *n;
+
+		for (strm = z_erofs_isolate_strms(true); strm; strm = n) {
+			n = strm->next;
+			strm->next = head;
+			head = strm;
+			--z_erofs_zstd_avail_strms;
+		}
+	}
+
+	/* 2. walk each isolated stream and grow max dict_size if needed */
+	wkspsz = zstd_dstream_workspace_bound(dict_size);
+	for (strm = head; strm; strm = strm->next) {
+		wksp = kvmalloc(wkspsz, GFP_KERNEL);
+		if (!wksp)
+			break;
+		kvfree(strm->wksp);
+		strm->wksp = wksp;
+		strm->wkspsz = wkspsz;
+	}
+
+	/* 3. push back all to the global list and update max dict_size */
+	spin_lock(&z_erofs_zstd_lock);
+	DBG_BUGON(z_erofs_zstd_head);
+	z_erofs_zstd_head = head;
+	spin_unlock(&z_erofs_zstd_lock);
+	z_erofs_zstd_avail_strms = z_erofs_zstd_nstrms;
+	wake_up_all(&z_erofs_zstd_wq);
+	if (!strm)
+		z_erofs_zstd_max_dictsize = dict_size;
+	mutex_unlock(&zstd_resize_mutex);
+	return strm ? -ENOMEM : 0;
+}
+
+int z_erofs_zstd_decompress(struct z_erofs_decompress_req *rq,
+			    struct page **pgpl)
+{
+	const unsigned int nrpages_out =
+		PAGE_ALIGN(rq->pageofs_out + rq->outputsize) >> PAGE_SHIFT;
+	const unsigned int nrpages_in =
+		PAGE_ALIGN(rq->inputsize) >> PAGE_SHIFT;
+	zstd_dstream *stream;
+	struct super_block *sb = rq->sb;
+	unsigned int insz, outsz, pofs;
+	struct z_erofs_zstd *strm;
+	zstd_in_buffer in_buf = { NULL, 0, 0 };
+	zstd_out_buffer out_buf = { NULL, 0, 0 };
+	u8 *kin, *kout = NULL;
+	bool bounced = false;
+	int no = -1, ni = 0, j = 0, zerr, err;
+
+	/* 1. get the exact compressed size */
+	kin = kmap_local_page(*rq->in);
+	err = z_erofs_fixup_insize(rq, kin + rq->pageofs_in,
+			min_t(unsigned int, rq->inputsize,
+			      sb->s_blocksize - rq->pageofs_in));
+	if (err) {
+		kunmap_local(kin);
+		return err;
+	}
+
+	/* 2. get an available ZSTD context */
+	strm = z_erofs_isolate_strms(false);
+
+	/* 3. multi-call decompress */
+	insz = rq->inputsize;
+	outsz = rq->outputsize;
+	stream = zstd_init_dstream(z_erofs_zstd_max_dictsize, strm->wksp, strm->wkspsz);
+	if (!stream) {
+		err = -EIO;
+		goto failed_zinit;
+	}
+
+	pofs = rq->pageofs_out;
+	in_buf.size = min_t(u32, insz, PAGE_SIZE - rq->pageofs_in);
+	insz -= in_buf.size;
+	in_buf.src = kin + rq->pageofs_in;
+	do {
+		if (out_buf.size == out_buf.pos) {
+			if (++no >= nrpages_out || !outsz) {
+				erofs_err(sb, "insufficient space for decompressed data");
+				err = -EFSCORRUPTED;
+				break;
+			}
+
+			if (kout)
+				kunmap_local(kout);
+			out_buf.size = min_t(u32, outsz, PAGE_SIZE - pofs);
+			outsz -= out_buf.size;
+			if (!rq->out[no]) {
+				rq->out[no] = erofs_allocpage(pgpl, rq->gfp);
+				if (!rq->out[no]) {
+					kout = NULL;
+					err = -ENOMEM;
+					break;
+				}
+				set_page_private(rq->out[no],
+						 Z_EROFS_SHORTLIVED_PAGE);
+			}
+			kout = kmap_local_page(rq->out[no]);
+			out_buf.dst = kout + pofs;
+			out_buf.pos = 0;
+			pofs = 0;
+		}
+
+		if (in_buf.size == in_buf.pos && insz) {
+			if (++ni >= nrpages_in) {
+				erofs_err(sb, "invalid compressed data");
+				err = -EFSCORRUPTED;
+				break;
+			}
+
+			if (kout) /* unlike kmap(), take care of the orders */
+				kunmap_local(kout);
+			kunmap_local(kin);
+			in_buf.size = min_t(u32, insz, PAGE_SIZE);
+			insz -= in_buf.size;
+			kin = kmap_local_page(rq->in[ni]);
+			in_buf.src = kin;
+			in_buf.pos = 0;
+			bounced = false;
+			if (kout) {
+				j = (u8 *)out_buf.dst - kout;
+				kout = kmap_local_page(rq->out[no]);
+				out_buf.dst = kout + j;
+			}
+		}
+
+		/*
+		 * Handle overlapping: Use bounced buffer if the compressed
+		 * data is under processing; Or use short-lived pages from the
+		 * on-stack pagepool where pages share among the same request
+		 * and not _all_ inplace I/O pages are needed to be doubled.
+		 */
+		if (!bounced && rq->out[no] == rq->in[ni]) {
+			memcpy(strm->bounce, in_buf.src, in_buf.size);
+			in_buf.src = strm->bounce;
+			bounced = true;
+		}
+
+		for (j = ni + 1; j < nrpages_in; ++j) {
+			struct page *tmppage;
+
+			if (rq->out[no] != rq->in[j])
+				continue;
+			tmppage = erofs_allocpage(pgpl, rq->gfp);
+			if (!tmppage) {
+				err = -ENOMEM;
+				goto failed;
+			}
+			set_page_private(tmppage, Z_EROFS_SHORTLIVED_PAGE);
+			copy_highpage(tmppage, rq->in[j]);
+			rq->in[j] = tmppage;
+		}
+		zerr = zstd_decompress_stream(stream, &out_buf, &in_buf);
+		if (zstd_is_error(zerr) || (!zerr && outsz)) {
+			erofs_err(sb, "failed to decompress in[%u] out[%u]: %s",
+				  rq->inputsize, rq->outputsize,
+				  zerr ? zstd_get_error_name(zerr) : "unexpected end of stream");
+			err = -EFSCORRUPTED;
+			break;
+		}
+	} while (outsz || out_buf.pos < out_buf.size);
+failed:
+	if (kout)
+		kunmap_local(kout);
+failed_zinit:
+	kunmap_local(kin);
+	/* 4. push back ZSTD stream context to the global list */
+	spin_lock(&z_erofs_zstd_lock);
+	strm->next = z_erofs_zstd_head;
+	z_erofs_zstd_head = strm;
+	spin_unlock(&z_erofs_zstd_lock);
+	wake_up(&z_erofs_zstd_wq);
+	return err;
+}
diff --git a/fs/erofs/erofs_fs.h b/fs/erofs/erofs_fs.h
index 550baf1729d4..6c0c270c42e1 100644
--- a/fs/erofs/erofs_fs.h
+++ b/fs/erofs/erofs_fs.h
@@ -296,6 +296,7 @@ enum {
 	Z_EROFS_COMPRESSION_LZ4		= 0,
 	Z_EROFS_COMPRESSION_LZMA	= 1,
 	Z_EROFS_COMPRESSION_DEFLATE	= 2,
+	Z_EROFS_COMPRESSION_ZSTD	= 3,
 	Z_EROFS_COMPRESSION_MAX
 };
 #define Z_EROFS_ALL_COMPR_ALGS		((1 << Z_EROFS_COMPRESSION_MAX) - 1)
@@ -322,6 +323,15 @@ struct z_erofs_deflate_cfgs {
 	u8 reserved[5];
 } __packed;
 
+/* 6 bytes (+ length field = 8 bytes) */
+struct z_erofs_zstd_cfgs {
+	u8 format;
+	u8 windowlog;           /* windowLog - ZSTD_WINDOWLOG_ABSOLUTEMIN(10) */
+	u8 reserved[4];
+} __packed;
+
+#define Z_EROFS_ZSTD_MAX_DICT_SIZE      Z_EROFS_PCLUSTER_MAX_SIZE
+
 /*
  * bit 0 : COMPACTED_2B indexes (0 - off; 1 - on)
  *  e.g. for 4k logical cluster size,      4B        if compacted 2B is off;
diff --git a/fs/erofs/internal.h b/fs/erofs/internal.h
index 53ebba952a2f..21def866a482 100644
--- a/fs/erofs/internal.h
+++ b/fs/erofs/internal.h
@@ -502,6 +502,14 @@ static inline int z_erofs_deflate_init(void) { return 0; }
 static inline int z_erofs_deflate_exit(void) { return 0; }
 #endif	/* !CONFIG_EROFS_FS_ZIP_DEFLATE */
 
+#ifdef CONFIG_EROFS_FS_ZIP_ZSTD
+int __init z_erofs_zstd_init(void);
+void z_erofs_zstd_exit(void);
+#else
+static inline int z_erofs_zstd_init(void) { return 0; }
+static inline int z_erofs_zstd_exit(void) { return 0; }
+#endif	/* !CONFIG_EROFS_FS_ZIP_ZSTD */
+
 #ifdef CONFIG_EROFS_FS_ONDEMAND
 int erofs_fscache_register_fs(struct super_block *sb);
 void erofs_fscache_unregister_fs(struct super_block *sb);
diff --git a/fs/erofs/super.c b/fs/erofs/super.c
index e3438f1a7bac..044c79229a78 100644
--- a/fs/erofs/super.c
+++ b/fs/erofs/super.c
@@ -859,6 +859,10 @@ static int __init erofs_module_init(void)
 	if (err)
 		goto deflate_err;
 
+	err = z_erofs_zstd_init();
+	if (err)
+		goto zstd_err;
+
 	err = z_erofs_gbuf_init();
 	if (err)
 		goto gbuf_err;
@@ -884,6 +888,8 @@ sysfs_err:
 zip_err:
 	z_erofs_gbuf_exit();
 gbuf_err:
+	z_erofs_zstd_exit();
+zstd_err:
 	z_erofs_deflate_exit();
 deflate_err:
 	z_erofs_lzma_exit();
@@ -903,6 +909,7 @@ static void __exit erofs_module_exit(void)
 
 	erofs_exit_sysfs();
 	z_erofs_exit_zip_subsystem();
+	z_erofs_zstd_exit();
 	z_erofs_deflate_exit();
 	z_erofs_lzma_exit();
 	erofs_exit_shrinker();
diff --git a/fs/erofs/zmap.c b/fs/erofs/zmap.c
index 26637a60eba5..0a2454d8bcc1 100644
--- a/fs/erofs/zmap.c
+++ b/fs/erofs/zmap.c
@@ -550,7 +550,8 @@ static int z_erofs_do_map_blocks(struct inode *inode,
 	if ((flags & EROFS_GET_BLOCKS_FIEMAP) ||
 	    ((flags & EROFS_GET_BLOCKS_READMORE) &&
 	     (map->m_algorithmformat == Z_EROFS_COMPRESSION_LZMA ||
-	      map->m_algorithmformat == Z_EROFS_COMPRESSION_DEFLATE) &&
+	      map->m_algorithmformat == Z_EROFS_COMPRESSION_DEFLATE ||
+	      map->m_algorithmformat == Z_EROFS_COMPRESSION_ZSTD) &&
 	      map->m_llen >= i_blocksize(inode))) {
 		err = z_erofs_get_extent_decompressedlen(&m);
 		if (!err)
-- 
cgit v1.2.3