[PATCH v10] erofs-utils: introduce segment compression

Li Guifu bluce.lee at aliyun.com
Wed Jul 1 03:27:58 AEST 2020


Support segment compression which seperates files in several logic
units (segments) and each segment is compressed independently.

Advantages:
 - more friendly for data differencing;
 - it can also be used for parallel compression in the same file later.

Signed-off-by: Li Guifu <bluce.lee at aliyun.com>
---
 include/erofs/config.h |  1 +
 lib/compress.c         | 29 +++++++++++++++++++++++------
 lib/config.c           |  1 +
 man/mkfs.erofs.1       |  4 ++++
 mkfs/main.c            | 16 +++++++++++++++-
 5 files changed, 44 insertions(+), 7 deletions(-)

diff --git a/include/erofs/config.h b/include/erofs/config.h
index 2f09749..e5f1bfb 100644
--- a/include/erofs/config.h
+++ b/include/erofs/config.h
@@ -35,6 +35,7 @@ struct erofs_configure {
 	char *c_img_path;
 	char *c_src_path;
 	char *c_compr_alg_master;
+	u64 c_compr_seg_size;
 	int c_compr_level_master;
 	int c_force_inodeversion;
 	/* < 0, xattr disabled and INT_MAX, always use inline xattrs */
diff --git a/lib/compress.c b/lib/compress.c
index 6cc68ed..2ea5809 100644
--- a/lib/compress.c
+++ b/lib/compress.c
@@ -32,6 +32,7 @@ struct z_erofs_vle_compress_ctx {
 
 	erofs_blk_t blkaddr;	/* pointing to the next blkaddr */
 	u16 clusterofs;
+	u64 segavail;
 };
 
 #define Z_EROFS_LEGACY_MAP_HEADER_SIZE	\
@@ -116,7 +117,7 @@ static void vle_write_indexes(struct z_erofs_vle_compress_ctx *ctx,
 }
 
 static int write_uncompressed_block(struct z_erofs_vle_compress_ctx *ctx,
-				    unsigned int *len,
+				    unsigned int *len, unsigned int *ucomproft,
 				    char *dst)
 {
 	int ret;
@@ -125,14 +126,19 @@ static int write_uncompressed_block(struct z_erofs_vle_compress_ctx *ctx,
 	/* reset clusterofs to 0 if permitted */
 	if (!erofs_sb_has_lz4_0padding() &&
 	    ctx->head >= ctx->clusterofs) {
+		*ucomproft = ctx->clusterofs;
 		ctx->head -= ctx->clusterofs;
 		*len += ctx->clusterofs;
 		ctx->clusterofs = 0;
+		count = min(EROFS_BLKSIZ, *len);
+	} else {
+		*ucomproft = 0;
+		count = min_t(u64, ctx->segavail, *len);
+		if (count > EROFS_BLKSIZ)
+			count = EROFS_BLKSIZ;
 	}
 
 	/* write uncompressed data */
-	count = min(EROFS_BLKSIZ, *len);
-
 	memcpy(dst, ctx->queue + ctx->head, count);
 	memset(dst + count, 0, EROFS_BLKSIZ - count);
 
@@ -157,14 +163,21 @@ static int vle_compress_one(struct erofs_inode *inode,
 
 	while (len) {
 		bool raw;
+		unsigned int ucomproft = 0;
+
+		if (ctx->segavail <= EROFS_BLKSIZ) {
+			if (len < ctx->segavail && !final)
+				break;
+
+			goto nocompression;
+		}
 
 		if (len <= EROFS_BLKSIZ) {
 			if (final)
 				goto nocompression;
 			break;
 		}
-
-		count = len;
+		count = min_t(u64, len, ctx->segavail);
 		ret = erofs_compress_destsize(h, compressionlevel,
 					      ctx->queue + ctx->head,
 					      &count, dst, EROFS_BLKSIZ);
@@ -175,7 +188,7 @@ static int vle_compress_one(struct erofs_inode *inode,
 					  erofs_strerror(ret));
 			}
 nocompression:
-			ret = write_uncompressed_block(ctx, &len, dst);
+			ret = write_uncompressed_block(ctx, &len, &ucomproft, dst);
 			if (ret < 0)
 				return ret;
 			count = ret;
@@ -202,6 +215,9 @@ nocompression:
 
 		++ctx->blkaddr;
 		len -= count;
+		ctx->segavail -= count - ucomproft;
+		if (!ctx->segavail)
+			ctx->segavail = cfg.c_compr_seg_size;
 
 		if (!final && ctx->head >= EROFS_CONFIG_COMPR_MAX_SZ) {
 			const unsigned int qh_aligned =
@@ -422,6 +438,7 @@ int erofs_write_compressed_file(struct erofs_inode *inode)
 	ctx.head = ctx.tail = 0;
 	ctx.clusterofs = 0;
 	remaining = inode->i_size;
+	ctx.segavail = cfg.c_compr_seg_size;
 
 	while (remaining) {
 		const u64 readcount = min_t(u64, remaining,
diff --git a/lib/config.c b/lib/config.c
index da0c260..721ff61 100644
--- a/lib/config.c
+++ b/lib/config.c
@@ -23,6 +23,7 @@ void erofs_init_configure(void)
 	cfg.c_force_inodeversion = 0;
 	cfg.c_inline_xattr_tolerance = 2;
 	cfg.c_unix_timestamp = -1;
+	cfg.c_compr_seg_size = -1;
 }
 
 void erofs_show_config(void)
diff --git a/man/mkfs.erofs.1 b/man/mkfs.erofs.1
index 891c5a8..2a4ef71 100644
--- a/man/mkfs.erofs.1
+++ b/man/mkfs.erofs.1
@@ -52,6 +52,10 @@ Forcely generate extended inodes (64-byte inodes) to output.
 Set all files to the given UNIX timestamp. Reproducible builds requires setting
 all to a specific one.
 .TP
+.BI "\-S " #
+Set max input stream size for each individual segment (disabled if 0).
+The default value is 0. It has be aligned with blocksize.
+.TP
 .BI "\-\-exclude-path=" path
 Ignore file that matches the exact literal path.
 You may give multiple `--exclude-path' options.
diff --git a/mkfs/main.c b/mkfs/main.c
index 94bf1e6..cded973 100644
--- a/mkfs/main.c
+++ b/mkfs/main.c
@@ -61,6 +61,7 @@ static void usage(void)
 	      " -x#               set xattr tolerance to # (< 0, disable xattrs; default 2)\n"
 	      " -EX[,...]         X=extended options\n"
 	      " -T#               set a fixed UNIX timestamp # to all files\n"
+	      " -S#               Set max input stream size # for each individual segment\n"
 	      " --exclude-path=X  avoid including file X (X = exact literal path)\n"
 	      " --exclude-regex=X avoid including files that match X (X = regular expression)\n"
 #ifdef HAVE_LIBSELINUX
@@ -138,7 +139,7 @@ static int mkfs_parse_options_cfg(int argc, char *argv[])
 	char *endptr;
 	int opt, i;
 
-	while((opt = getopt_long(argc, argv, "d:x:z:E:T:",
+	while((opt = getopt_long(argc, argv, "d:x:z:E:T:S:",
 				 long_options, NULL)) != -1) {
 		switch (opt) {
 		case 'z':
@@ -188,6 +189,19 @@ static int mkfs_parse_options_cfg(int argc, char *argv[])
 				return -EINVAL;
 			}
 			break;
+		case 'S':
+			i = strtoll(optarg, &endptr, 0);
+			if (*endptr != '\0') {
+				erofs_err("invalid compress segment size %s",
+					  optarg);
+				return -EINVAL;
+			}
+			if (!i) {
+				cfg.c_compr_seg_size = -1;
+			} else {
+				cfg.c_compr_seg_size = i * EROFS_BLKSIZ;
+			}
+			break;
 		case 2:
 			opt = erofs_parse_exclude_path(optarg, false);
 			if (opt) {
-- 
2.17.1



More information about the Linux-erofs mailing list