staging: erofs: add compacted compression indexes support

This patch aims at compacted compression indexes:
 1) cleanup z_erofs_map_blocks_iter and move into zmap.c;
 2) add compacted 4/2B decoding support.

On kirin980 platform, sequential read is increased about
6% (725MiB/s -> 770MiB/s) on enwik9 dataset if compacted 2B
feature is enabled.
Reviewed-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Gao Xiang <gaoxiang25@huawei.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

drivers/staging/erofs/Makefile

@@ -9,5 +9,5 @@ obj-$(CONFIG_EROFS_FS) += erofs.o
 ccflags-y += -I $(srctree)/$(src)/include
 erofs-objs := super.o inode.o data.o namei.o dir.o utils.o
 erofs-$(CONFIG_EROFS_FS_XATTR) += xattr.o
-erofs-$(CONFIG_EROFS_FS_ZIP) += unzip_vle.o unzip_vle_lz4.o
+erofs-$(CONFIG_EROFS_FS_ZIP) += unzip_vle.o unzip_vle_lz4.o zmap.o
 

drivers/staging/erofs/inode.c

@@ -210,12 +210,7 @@ static int fill_inode(struct inode *inode, int isdir)
 		}
 
 		if (is_inode_layout_compression(inode)) {
-#ifdef CONFIG_EROFS_FS_ZIP
-			inode->i_mapping->a_ops =
-				&z_erofs_vle_normalaccess_aops;
-#else
-			err = -ENOTSUPP;
-#endif
+			err = z_erofs_fill_inode(inode);
 			goto out_unlock;
 		}
 

drivers/staging/erofs/internal.h

@@ -339,9 +339,11 @@ static inline erofs_off_t iloc(struct erofs_sb_info *sbi, erofs_nid_t nid)
 
 /* atomic flag definitions */
 #define EROFS_V_EA_INITED_BIT	0
+#define EROFS_V_Z_INITED_BIT	1
 
 /* bitlock definitions (arranged in reverse order) */
 #define EROFS_V_BL_XATTR_BIT	(BITS_PER_LONG - 1)
+#define EROFS_V_BL_Z_BIT	(BITS_PER_LONG - 2)
 
 struct erofs_vnode {
 	erofs_nid_t nid;
@@ -356,8 +358,17 @@ struct erofs_vnode {
 	unsigned xattr_shared_count;
 	unsigned *xattr_shared_xattrs;
 
-	erofs_blk_t raw_blkaddr;
-
+	union {
+		erofs_blk_t raw_blkaddr;
+#ifdef CONFIG_EROFS_FS_ZIP
+		struct {
+			unsigned short z_advise;
+			unsigned char  z_algorithmtype[2];
+			unsigned char  z_logical_clusterbits;
+			unsigned char  z_physical_clusterbits[2];
+		};
+#endif
+	};
 	/* the corresponding vfs inode */
 	struct inode vfs_inode;
 };
@@ -447,11 +458,14 @@ struct erofs_map_blocks {
 /* Flags used by erofs_map_blocks() */
 #define EROFS_GET_BLOCKS_RAW    0x0001
 
+/* zmap.c */
 #ifdef CONFIG_EROFS_FS_ZIP
+int z_erofs_fill_inode(struct inode *inode);
 int z_erofs_map_blocks_iter(struct inode *inode,
 			    struct erofs_map_blocks *map,
 			    int flags);
 #else
+static inline int z_erofs_fill_inode(struct inode *inode) { return -ENOTSUPP; }
 static inline int z_erofs_map_blocks_iter(struct inode *inode,
 					  struct erofs_map_blocks *map,
 					  int flags)

drivers/staging/erofs/unzip_vle.c

@@ -1600,289 +1600,3 @@ const struct address_space_operations z_erofs_vle_normalaccess_aops = {
 	.readpages = z_erofs_vle_normalaccess_readpages,
 };
 
-/*
- * Variable-sized Logical Extent (Fixed Physical Cluster) Compression Mode
- * ---
- * VLE compression mode attempts to compress a number of logical data into
- * a physical cluster with a fixed size.
- * VLE compression mode uses "struct z_erofs_vle_decompressed_index".
- */
-#define __vle_cluster_advise(x, bit, bits) \
-	((le16_to_cpu(x) >> (bit)) & ((1 << (bits)) - 1))
-
-#define __vle_cluster_type(advise) __vle_cluster_advise(advise, \
-	Z_EROFS_VLE_DI_CLUSTER_TYPE_BIT, Z_EROFS_VLE_DI_CLUSTER_TYPE_BITS)
-
-#define vle_cluster_type(di)	\
-	__vle_cluster_type((di)->di_advise)
-
-static int
-vle_decompressed_index_clusterofs(unsigned int *clusterofs,
-				  unsigned int clustersize,
-				  struct z_erofs_vle_decompressed_index *di)
-{
-	switch (vle_cluster_type(di)) {
-	case Z_EROFS_VLE_CLUSTER_TYPE_NONHEAD:
-		*clusterofs = clustersize;
-		break;
-	case Z_EROFS_VLE_CLUSTER_TYPE_PLAIN:
-	case Z_EROFS_VLE_CLUSTER_TYPE_HEAD:
-		*clusterofs = le16_to_cpu(di->di_clusterofs);
-		break;
-	default:
-		DBG_BUGON(1);
-		return -EIO;
-	}
-	return 0;
-}
-
-static inline erofs_blk_t
-vle_extent_blkaddr(struct inode *inode, pgoff_t index)
-{
-	struct erofs_sb_info *sbi = EROFS_I_SB(inode);
-	struct erofs_vnode *vi = EROFS_V(inode);
-
-	unsigned int ofs = Z_EROFS_VLE_LEGACY_INDEX_ALIGN(vi->inode_isize +
-							  vi->xattr_isize) +
-		index * sizeof(struct z_erofs_vle_decompressed_index);
-
-	return erofs_blknr(iloc(sbi, vi->nid) + ofs);
-}
-
-static inline unsigned int
-vle_extent_blkoff(struct inode *inode, pgoff_t index)
-{
-	struct erofs_sb_info *sbi = EROFS_I_SB(inode);
-	struct erofs_vnode *vi = EROFS_V(inode);
-
-	unsigned int ofs = Z_EROFS_VLE_LEGACY_INDEX_ALIGN(vi->inode_isize +
-							  vi->xattr_isize) +
-		index * sizeof(struct z_erofs_vle_decompressed_index);
-
-	return erofs_blkoff(iloc(sbi, vi->nid) + ofs);
-}
-
-struct vle_map_blocks_iter_ctx {
-	struct inode *inode;
-	struct super_block *sb;
-	unsigned int clusterbits;
-
-	struct page **mpage_ret;
-	void **kaddr_ret;
-};
-
-static int
-vle_get_logical_extent_head(const struct vle_map_blocks_iter_ctx *ctx,
-			    unsigned int lcn,	/* logical cluster number */
-			    unsigned long long *ofs,
-			    erofs_blk_t *pblk,
-			    unsigned int *flags)
-{
-	const unsigned int clustersize = 1 << ctx->clusterbits;
-	const erofs_blk_t mblk = vle_extent_blkaddr(ctx->inode, lcn);
-	struct page *mpage = *ctx->mpage_ret;	/* extent metapage */
-
-	struct z_erofs_vle_decompressed_index *di;
-	unsigned int cluster_type, delta0;
-
-	if (mpage->index != mblk) {
-		kunmap_atomic(*ctx->kaddr_ret);
-		unlock_page(mpage);
-		put_page(mpage);
-
-		mpage = erofs_get_meta_page(ctx->sb, mblk, false);
-		if (IS_ERR(mpage)) {
-			*ctx->mpage_ret = NULL;
-			return PTR_ERR(mpage);
-		}
-		*ctx->mpage_ret = mpage;
-		*ctx->kaddr_ret = kmap_atomic(mpage);
-	}
-
-	di = *ctx->kaddr_ret + vle_extent_blkoff(ctx->inode, lcn);
-
-	cluster_type = vle_cluster_type(di);
-	switch (cluster_type) {
-	case Z_EROFS_VLE_CLUSTER_TYPE_NONHEAD:
-		delta0 = le16_to_cpu(di->di_u.delta[0]);
-		if (unlikely(!delta0 || delta0 > lcn)) {
-			errln("invalid NONHEAD dl0 %u at lcn %u of nid %llu",
-			      delta0, lcn, EROFS_V(ctx->inode)->nid);
-			DBG_BUGON(1);
-			return -EIO;
-		}
-		return vle_get_logical_extent_head(ctx,
-			lcn - delta0, ofs, pblk, flags);
-	case Z_EROFS_VLE_CLUSTER_TYPE_PLAIN:
-		*flags ^= EROFS_MAP_ZIPPED;
-		/* fallthrough */
-	case Z_EROFS_VLE_CLUSTER_TYPE_HEAD:
-		/* clustersize should be a power of two */
-		*ofs = ((u64)lcn << ctx->clusterbits) +
-			(le16_to_cpu(di->di_clusterofs) & (clustersize - 1));
-		*pblk = le32_to_cpu(di->di_u.blkaddr);
-		break;
-	default:
-		errln("unknown cluster type %u at lcn %u of nid %llu",
-		      cluster_type, lcn, EROFS_V(ctx->inode)->nid);
-		DBG_BUGON(1);
-		return -EIO;
-	}
-	return 0;
-}
-
-int z_erofs_map_blocks_iter(struct inode *inode,
-			    struct erofs_map_blocks *map,
-			    int flags)
-{
-	void *kaddr;
-	const struct vle_map_blocks_iter_ctx ctx = {
-		.inode = inode,
-		.sb = inode->i_sb,
-		.clusterbits = EROFS_I_SB(inode)->clusterbits,
-		.mpage_ret = &map->mpage,
-		.kaddr_ret = &kaddr
-	};
-	const unsigned int clustersize = 1 << ctx.clusterbits;
-	/* if both m_(l,p)len are 0, regularize l_lblk, l_lofs, etc... */
-	const bool initial = !map->m_llen;
-
-	/* logicial extent (start, end) offset */
-	unsigned long long ofs, end;
-	unsigned int lcn;
-	u32 ofs_rem;
-
-	/* initialize `pblk' to keep gcc from printing foolish warnings */
-	erofs_blk_t mblk, pblk = 0;
-	struct page *mpage = map->mpage;
-	struct z_erofs_vle_decompressed_index *di;
-	unsigned int cluster_type, logical_cluster_ofs;
-	int err = 0;
-
-	trace_z_erofs_map_blocks_iter_enter(inode, map, flags);
-
-	/* when trying to read beyond EOF, leave it unmapped */
-	if (unlikely(map->m_la >= inode->i_size)) {
-		DBG_BUGON(!initial);
-		map->m_llen = map->m_la + 1 - inode->i_size;
-		map->m_la = inode->i_size;
-		map->m_flags = 0;
-		goto out;
-	}
-
-	debugln("%s, m_la %llu m_llen %llu --- start", __func__,
-		map->m_la, map->m_llen);
-
-	ofs = map->m_la + map->m_llen;
-
-	/* clustersize should be power of two */
-	lcn = ofs >> ctx.clusterbits;
-	ofs_rem = ofs & (clustersize - 1);
-
-	mblk = vle_extent_blkaddr(inode, lcn);
-
-	if (!mpage || mpage->index != mblk) {
-		if (mpage)
-			put_page(mpage);
-
-		mpage = erofs_get_meta_page(ctx.sb, mblk, false);
-		if (IS_ERR(mpage)) {
-			err = PTR_ERR(mpage);
-			goto out;
-		}
-		map->mpage = mpage;
-	} else {
-		lock_page(mpage);
-		DBG_BUGON(!PageUptodate(mpage));
-	}
-
-	kaddr = kmap_atomic(mpage);
-	di = kaddr + vle_extent_blkoff(inode, lcn);
-
-	debugln("%s, lcn %u mblk %u e_blkoff %u", __func__, lcn,
-		mblk, vle_extent_blkoff(inode, lcn));
-
-	err = vle_decompressed_index_clusterofs(&logical_cluster_ofs,
-						clustersize, di);
-	if (unlikely(err))
-		goto unmap_out;
-
-	if (!initial) {
-		/* [walking mode] 'map' has been already initialized */
-		map->m_llen += logical_cluster_ofs;
-		goto unmap_out;
-	}
-
-	/* by default, compressed */
-	map->m_flags |= EROFS_MAP_ZIPPED;
-
-	end = ((u64)lcn + 1) * clustersize;
-
-	cluster_type = vle_cluster_type(di);
-
-	switch (cluster_type) {
-	case Z_EROFS_VLE_CLUSTER_TYPE_PLAIN:
-		if (ofs_rem >= logical_cluster_ofs)
-			map->m_flags ^= EROFS_MAP_ZIPPED;
-		/* fallthrough */
-	case Z_EROFS_VLE_CLUSTER_TYPE_HEAD:
-		if (ofs_rem == logical_cluster_ofs) {
-			pblk = le32_to_cpu(di->di_u.blkaddr);
-			goto exact_hitted;
-		}
-
-		if (ofs_rem > logical_cluster_ofs) {
-			ofs = (u64)lcn * clustersize | logical_cluster_ofs;
-			pblk = le32_to_cpu(di->di_u.blkaddr);
-			break;
-		}
-
-		/* logical cluster number should be >= 1 */
-		if (unlikely(!lcn)) {
-			errln("invalid logical cluster 0 at nid %llu",
-			      EROFS_V(inode)->nid);
-			err = -EIO;
-			goto unmap_out;
-		}
-		end = ((u64)lcn-- * clustersize) | logical_cluster_ofs;
-		/* fallthrough */
-	case Z_EROFS_VLE_CLUSTER_TYPE_NONHEAD:
-		/* get the correspoinding first chunk */
-		err = vle_get_logical_extent_head(&ctx, lcn, &ofs,
-						  &pblk, &map->m_flags);
-		mpage = map->mpage;
-
-		if (unlikely(err)) {
-			if (mpage)
-				goto unmap_out;
-			goto out;
-		}
-		break;
-	default:
-		errln("unknown cluster type %u at offset %llu of nid %llu",
-		      cluster_type, ofs, EROFS_V(inode)->nid);
-		err = -EIO;
-		goto unmap_out;
-	}
-
-	map->m_la = ofs;
-exact_hitted:
-	map->m_llen = end - ofs;
-	map->m_plen = clustersize;
-	map->m_pa = blknr_to_addr(pblk);
-	map->m_flags |= EROFS_MAP_MAPPED;
-unmap_out:
-	kunmap_atomic(kaddr);
-	unlock_page(mpage);
-out:
-	debugln("%s, m_la %llu m_pa %llu m_llen %llu m_plen %llu m_flags 0%o",
-		__func__, map->m_la, map->m_pa,
-		map->m_llen, map->m_plen, map->m_flags);
-
-	trace_z_erofs_map_blocks_iter_exit(inode, map, flags, err);
-
-	/* aggressively BUG_ON iff CONFIG_EROFS_FS_DEBUG is on */
-	DBG_BUGON(err < 0 && err != -ENOMEM);
-	return err;
-}
-