erofs: introduce physical cluster slab pools

Since multiple pcluster sizes could be used at once, the number of
compressed pages will become a variable factor. It's necessary to
introduce slab pools rather than a single slab cache now.

This limits the pclustersize to 1M (Z_EROFS_PCLUSTER_MAX_SIZE), and
get rid of the obsolete EROFS_FS_CLUSTER_PAGE_LIMIT, which has no
use now.

Link: https://lore.kernel.org/r/20210407043927.10623-4-xiang@kernel.org
Acked-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Gao Xiang <hsiangkao@redhat.com>

5 files changed, 126 insertions, 80 deletions

diff --git a/fs/erofs/Kconfig b/fs/erofs/Kconfig
index 74b0aaa7114c..858b3339f381 100644
--- a/fs/erofs/Kconfig
+++ b/fs/erofs/Kconfig
@@ -76,17 +76,3 @@ config EROFS_FS_ZIP
 
 	  If you don't want to enable compression feature, say N.
 
-config EROFS_FS_CLUSTER_PAGE_LIMIT
-	int "EROFS Cluster Pages Hard Limit"
-	depends on EROFS_FS_ZIP
-	range 1 256
-	default "1"
-	help
-	  Indicates maximum # of pages of a compressed
-	  physical cluster.
-
-	  For example, if files in a image were compressed
-	  into 8k-unit, hard limit should not be configured
-	  less than 2. Otherwise, the image will be refused
-	  to mount on this kernel.
-
diff --git a/fs/erofs/erofs_fs.h b/fs/erofs/erofs_fs.h
index 626b7d3e9ab7..76777673eb63 100644
--- a/fs/erofs/erofs_fs.h
+++ b/fs/erofs/erofs_fs.h
@@ -201,6 +201,9 @@ static inline unsigned int erofs_xattr_entry_size(struct erofs_xattr_entry *e)
 				 e->e_name_len + le16_to_cpu(e->e_value_size));
 }
 
+/* maximum supported size of a physical compression cluster */
+#define Z_EROFS_PCLUSTER_MAX_SIZE	(1024 * 1024)
+
 /* available compression algorithm types (for h_algorithmtype) */
 enum {
 	Z_EROFS_COMPRESSION_LZ4	= 0,
diff --git a/fs/erofs/internal.h b/fs/erofs/internal.h
index 4db085413304..898838e1965e 100644
--- a/fs/erofs/internal.h
+++ b/fs/erofs/internal.h
@@ -194,9 +194,6 @@ static inline int erofs_wait_on_workgroup_freezed(struct erofs_workgroup *grp)
 	return v;
 }
 #endif	/* !CONFIG_SMP */
-
-/* hard limit of pages per compressed cluster */
-#define Z_EROFS_CLUSTER_MAX_PAGES       (CONFIG_EROFS_FS_CLUSTER_PAGE_LIMIT)
 #endif	/* !CONFIG_EROFS_FS_ZIP */
 
 /* we strictly follow PAGE_SIZE and no buffer head yet */
diff --git a/fs/erofs/zdata.c b/fs/erofs/zdata.c
index e3f0100d82d1..db296d324333 100644
--- a/fs/erofs/zdata.c
+++ b/fs/erofs/zdata.c
@@ -11,6 +11,93 @@
 #include <trace/events/erofs.h>
 
 /*
+ * since pclustersize is variable for big pcluster feature, introduce slab
+ * pools implementation for different pcluster sizes.
+ */
+struct z_erofs_pcluster_slab {
+	struct kmem_cache *slab;
+	unsigned int maxpages;
+	char name[48];
+};
+
+#define _PCLP(n) { .maxpages = n }
+
+static struct z_erofs_pcluster_slab pcluster_pool[] __read_mostly = {
+	_PCLP(1), _PCLP(4), _PCLP(16), _PCLP(64), _PCLP(128),
+	_PCLP(Z_EROFS_PCLUSTER_MAX_PAGES)
+};
+
+static void z_erofs_destroy_pcluster_pool(void)
+{
+	int i;
+
+	for (i = 0; i < ARRAY_SIZE(pcluster_pool); ++i) {
+		if (!pcluster_pool[i].slab)
+			continue;
+		kmem_cache_destroy(pcluster_pool[i].slab);
+		pcluster_pool[i].slab = NULL;
+	}
+}
+
+static int z_erofs_create_pcluster_pool(void)
+{
+	struct z_erofs_pcluster_slab *pcs;
+	struct z_erofs_pcluster *a;
+	unsigned int size;
+
+	for (pcs = pcluster_pool;
+	     pcs < pcluster_pool + ARRAY_SIZE(pcluster_pool); ++pcs) {
+		size = struct_size(a, compressed_pages, pcs->maxpages);
+
+		sprintf(pcs->name, "erofs_pcluster-%u", pcs->maxpages);
+		pcs->slab = kmem_cache_create(pcs->name, size, 0,
+					      SLAB_RECLAIM_ACCOUNT, NULL);
+		if (pcs->slab)
+			continue;
+
+		z_erofs_destroy_pcluster_pool();
+		return -ENOMEM;
+	}
+	return 0;
+}
+
+static struct z_erofs_pcluster *z_erofs_alloc_pcluster(unsigned int nrpages)
+{
+	int i;
+
+	for (i = 0; i < ARRAY_SIZE(pcluster_pool); ++i) {
+		struct z_erofs_pcluster_slab *pcs = pcluster_pool + i;
+		struct z_erofs_pcluster *pcl;
+
+		if (nrpages > pcs->maxpages)
+			continue;
+
+		pcl = kmem_cache_zalloc(pcs->slab, GFP_NOFS);
+		if (!pcl)
+			return ERR_PTR(-ENOMEM);
+		pcl->pclusterpages = nrpages;
+		return pcl;
+	}
+	return ERR_PTR(-EINVAL);
+}
+
+static void z_erofs_free_pcluster(struct z_erofs_pcluster *pcl)
+{
+	int i;
+
+	for (i = 0; i < ARRAY_SIZE(pcluster_pool); ++i) {
+		struct z_erofs_pcluster_slab *pcs = pcluster_pool + i;
+
+		if (pcl->pclusterpages > pcs->maxpages)
+			continue;
+
+		kmem_cache_free(pcs->slab, pcl);
+		return;
+	}
+	DBG_BUGON(1);
+}
+
+/*
  * a compressed_pages[] placeholder in order to avoid
  * being filled with file pages for in-place decompression.
  */
@@ -37,12 +124,11 @@ typedef tagptr1_t compressed_page_t;
 	tagptr_fold(compressed_page_t, page, 1)
 
 static struct workqueue_struct *z_erofs_workqueue __read_mostly;
-static struct kmem_cache *pcluster_cachep __read_mostly;
 
 void z_erofs_exit_zip_subsystem(void)
 {
 	destroy_workqueue(z_erofs_workqueue);
-	kmem_cache_destroy(pcluster_cachep);
+	z_erofs_destroy_pcluster_pool();
 }
 
 static inline int z_erofs_init_workqueue(void)
@@ -59,32 +145,16 @@ static inline int z_erofs_init_workqueue(void)
 	return z_erofs_workqueue ? 0 : -ENOMEM;
 }
 
-static void z_erofs_pcluster_init_once(void *ptr)
-{
-	struct z_erofs_pcluster *pcl = ptr;
-	struct z_erofs_collection *cl = z_erofs_primarycollection(pcl);
-	unsigned int i;
-
-	mutex_init(&cl->lock);
-	cl->nr_pages = 0;
-	cl->vcnt = 0;
-	for (i = 0; i < Z_EROFS_CLUSTER_MAX_PAGES; ++i)
-		pcl->compressed_pages[i] = NULL;
-}
-
 int __init z_erofs_init_zip_subsystem(void)
 {
-	pcluster_cachep = kmem_cache_create("erofs_compress",
-					    Z_EROFS_WORKGROUP_SIZE, 0,
-					    SLAB_RECLAIM_ACCOUNT,
-					    z_erofs_pcluster_init_once);
-	if (pcluster_cachep) {
-		if (!z_erofs_init_workqueue())
-			return 0;
-
-		kmem_cache_destroy(pcluster_cachep);
-	}
-	return -ENOMEM;
+	int err = z_erofs_create_pcluster_pool();
+
+	if (err)
+		return err;
+	err = z_erofs_init_workqueue();
+	if (err)
+		z_erofs_destroy_pcluster_pool();
+	return err;
 }
 
 enum z_erofs_collectmode {
@@ -169,7 +239,6 @@ static void preload_compressed_pages(struct z_erofs_collector *clt,
 				     struct list_head *pagepool)
 {
 	const struct z_erofs_pcluster *pcl = clt->pcl;
-	const unsigned int clusterpages = BIT(pcl->clusterbits);
 	struct page **pages = clt->compressedpages;
 	pgoff_t index = pcl->obj.index + (pages - pcl->compressed_pages);
 	bool standalone = true;
@@ -179,7 +248,7 @@ static void preload_compressed_pages(struct z_erofs_collector *clt,
 	if (clt->mode < COLLECT_PRIMARY_FOLLOWED)
 		return;
 
-	for (; pages < pcl->compressed_pages + clusterpages; ++pages) {
+	for (; pages < pcl->compressed_pages + pcl->pclusterpages; ++pages) {
 		struct page *page;
 		compressed_page_t t;
 		struct page *newpage = NULL;
@@ -239,14 +308,13 @@ int erofs_try_to_free_all_cached_pages(struct erofs_sb_info *sbi,
 	struct z_erofs_pcluster *const pcl =
 		container_of(grp, struct z_erofs_pcluster, obj);
 	struct address_space *const mapping = MNGD_MAPPING(sbi);
-	const unsigned int clusterpages = BIT(pcl->clusterbits);
 	int i;
 
 	/*
 	 * refcount of workgroup is now freezed as 1,
 	 * therefore no need to worry about available decompression users.
 	 */
-	for (i = 0; i < clusterpages; ++i) {
+	for (i = 0; i < pcl->pclusterpages; ++i) {
 		struct page *page = pcl->compressed_pages[i];
 
 		if (!page)
@@ -271,13 +339,12 @@ int erofs_try_to_free_cached_page(struct address_space *mapping,
 				  struct page *page)
 {
 	struct z_erofs_pcluster *const pcl = (void *)page_private(page);
-	const unsigned int clusterpages = BIT(pcl->clusterbits);
 	int ret = 0;	/* 0 - busy */
 
 	if (erofs_workgroup_try_to_freeze(&pcl->obj, 1)) {
 		unsigned int i;
 
-		for (i = 0; i < clusterpages; ++i) {
+		for (i = 0; i < pcl->pclusterpages; ++i) {
 			if (pcl->compressed_pages[i] == page) {
 				WRITE_ONCE(pcl->compressed_pages[i], NULL);
 				ret = 1;
@@ -297,9 +364,9 @@ static inline bool z_erofs_try_inplace_io(struct z_erofs_collector *clt,
 					  struct page *page)
 {
 	struct z_erofs_pcluster *const pcl = clt->pcl;
-	const unsigned int clusterpages = BIT(pcl->clusterbits);
 
-	while (clt->compressedpages < pcl->compressed_pages + clusterpages) {
+	while (clt->compressedpages <
+	       pcl->compressed_pages + pcl->pclusterpages) {
 		if (!cmpxchg(clt->compressedpages++, NULL, page))
 			return true;
 	}
@@ -413,10 +480,10 @@ static int z_erofs_register_collection(struct z_erofs_collector *clt,
 	struct erofs_workgroup *grp;
 	int err;
 
-	/* no available workgroup, let's allocate one */
-	pcl = kmem_cache_alloc(pcluster_cachep, GFP_NOFS);
-	if (!pcl)
-		return -ENOMEM;
+	/* no available pcluster, let's allocate one */
+	pcl = z_erofs_alloc_pcluster(map->m_plen >> PAGE_SHIFT);
+	if (IS_ERR(pcl))
+		return PTR_ERR(pcl);
 
 	atomic_set(&pcl->obj.refcount, 1);
 	pcl->obj.index = map->m_pa >> PAGE_SHIFT;
@@ -430,24 +497,18 @@ static int z_erofs_register_collection(struct z_erofs_collector *clt,
 	else
 		pcl->algorithmformat = Z_EROFS_COMPRESSION_SHIFTED;
 
-	pcl->clusterbits = 0;
-
 	/* new pclusters should be claimed as type 1, primary and followed */
 	pcl->next = clt->owned_head;
 	clt->mode = COLLECT_PRIMARY_FOLLOWED;
 
 	cl = z_erofs_primarycollection(pcl);
-
-	/* must be cleaned before freeing to slab */
-	DBG_BUGON(cl->nr_pages);
-	DBG_BUGON(cl->vcnt);
-
 	cl->pageofs = map->m_la & ~PAGE_MASK;
 
 	/*
 	 * lock all primary followed works before visible to others
 	 * and mutex_trylock *never* fails for a new pcluster.
 	 */
+	mutex_init(&cl->lock);
 	DBG_BUGON(!mutex_trylock(&cl->lock));
 
 	grp = erofs_insert_workgroup(inode->i_sb, &pcl->obj);
@@ -471,7 +532,7 @@ static int z_erofs_register_collection(struct z_erofs_collector *clt,
 
 err_out:
 	mutex_unlock(&cl->lock);
-	kmem_cache_free(pcluster_cachep, pcl);
+	z_erofs_free_pcluster(pcl);
 	return err;
 }
 
@@ -517,7 +578,7 @@ out:
 
 	clt->compressedpages = clt->pcl->compressed_pages;
 	if (clt->mode <= COLLECT_PRIMARY) /* cannot do in-place I/O */
-		clt->compressedpages += Z_EROFS_CLUSTER_MAX_PAGES;
+		clt->compressedpages += clt->pcl->pclusterpages;
 	return 0;
 }
 
@@ -530,9 +591,8 @@ static void z_erofs_rcu_callback(struct rcu_head *head)
 	struct z_erofs_collection *const cl =
 		container_of(head, struct z_erofs_collection, rcu);
 
-	kmem_cache_free(pcluster_cachep,
-			container_of(cl, struct z_erofs_pcluster,
-				     primary_collection));
+	z_erofs_free_pcluster(container_of(cl, struct z_erofs_pcluster,
+					   primary_collection));
 }
 
 void erofs_workgroup_free_rcu(struct erofs_workgroup *grp)
@@ -784,9 +844,8 @@ static int z_erofs_decompress_pcluster(struct super_block *sb,
 				       struct list_head *pagepool)
 {
 	struct erofs_sb_info *const sbi = EROFS_SB(sb);
-	const unsigned int clusterpages = BIT(pcl->clusterbits);
 	struct z_erofs_pagevec_ctor ctor;
-	unsigned int i, outputsize, llen, nr_pages;
+	unsigned int i, inputsize, outputsize, llen, nr_pages;
 	struct page *pages_onstack[Z_EROFS_VMAP_ONSTACK_PAGES];
 	struct page **pages, **compressed_pages, *page;
 
@@ -866,7 +925,7 @@ static int z_erofs_decompress_pcluster(struct super_block *sb,
 	overlapped = false;
 	compressed_pages = pcl->compressed_pages;
 
-	for (i = 0; i < clusterpages; ++i) {
+	for (i = 0; i < pcl->pclusterpages; ++i) {
 		unsigned int pagenr;
 
 		page = compressed_pages[i];
@@ -919,12 +978,13 @@ static int z_erofs_decompress_pcluster(struct super_block *sb,
 		partial = true;
 	}
 
+	inputsize = pcl->pclusterpages * PAGE_SIZE;
 	err = z_erofs_decompress(&(struct z_erofs_decompress_req) {
 					.sb = sb,
 					.in = compressed_pages,
 					.out = pages,
 					.pageofs_out = cl->pageofs,
-					.inputsize = PAGE_SIZE,
+					.inputsize = inputsize,
 					.outputsize = outputsize,
 					.alg = pcl->algorithmformat,
 					.inplace_io = overlapped,
@@ -933,7 +993,7 @@ static int z_erofs_decompress_pcluster(struct super_block *sb,
 
 out:
 	/* must handle all compressed pages before ending pages */
-	for (i = 0; i < clusterpages; ++i) {
+	for (i = 0; i < pcl->pclusterpages; ++i) {
 		page = compressed_pages[i];
 
 		if (erofs_page_is_managed(sbi, page))
@@ -1236,7 +1296,7 @@ static void z_erofs_submit_queue(struct super_block *sb,
 		pcl = container_of(owned_head, struct z_erofs_pcluster, next);
 
 		cur = pcl->obj.index;
-		end = cur + BIT(pcl->clusterbits);
+		end = cur + pcl->pclusterpages;
 
 		/* close the main owned chain at first */
 		owned_head = cmpxchg(&pcl->next, Z_EROFS_PCLUSTER_TAIL,
diff --git a/fs/erofs/zdata.h b/fs/erofs/zdata.h
index b503b353d4ab..942ee69dff6a 100644
--- a/fs/erofs/zdata.h
+++ b/fs/erofs/zdata.h
@@ -10,6 +10,7 @@
 #include "internal.h"
 #include "zpvec.h"
 
+#define Z_EROFS_PCLUSTER_MAX_PAGES	(Z_EROFS_PCLUSTER_MAX_SIZE / PAGE_SIZE)
 #define Z_EROFS_NR_INLINE_PAGEVECS      3
 
 /*
@@ -59,16 +60,17 @@ struct z_erofs_pcluster {
 	/* A: point to next chained pcluster or TAILs */
 	z_erofs_next_pcluster_t next;
 
-	/* A: compressed pages (including multi-usage pages) */
-	struct page *compressed_pages[Z_EROFS_CLUSTER_MAX_PAGES];
-
 	/* A: lower limit of decompressed length and if full length or not */
 	unsigned int length;
 
+	/* I: physical cluster size in pages */
+	unsigned short pclusterpages;
+
 	/* I: compression algorithm format */
 	unsigned char algorithmformat;
-	/* I: bit shift of physical cluster size */
-	unsigned char clusterbits;
+
+	/* A: compressed pages (can be cached or inplaced pages) */
+	struct page *compressed_pages[];
 };
 
 #define z_erofs_primarycollection(pcluster) (&(pcluster)->primary_collection)
@@ -82,8 +84,6 @@ struct z_erofs_pcluster {
 
 #define Z_EROFS_PCLUSTER_NIL            (NULL)
 
-#define Z_EROFS_WORKGROUP_SIZE  sizeof(struct z_erofs_pcluster)
-
 struct z_erofs_decompressqueue {
 	struct super_block *sb;
 	atomic_t pending_bios;