[PATCH v1-changed] erofs-utils: add hybrid source support for local metadata and gzran
ChengyuZhu6
hudson at cyzhu.com
Sun Sep 28 13:39:57 AEST 2025
From: Chengyu Zhu <hudsonzhu at tencent.com>
Add support for combining local metadata files with remote OCI blobs
through a new hybrid source mechanism. This enables local metadata
storage while keeping blob data in remote registries.
Signed-off-by: Chengyu Zhu <hudsonzhu at tencent.com>
---
lib/liberofs_oci.h | 2 +
lib/remotes/oci.c | 20 ---
mount/main.c | 374 ++++++++++++++++++++++++++++++++++++++++++++-
3 files changed, 369 insertions(+), 27 deletions(-)
diff --git a/lib/liberofs_oci.h b/lib/liberofs_oci.h
index 71c8879..a818003 100644
--- a/lib/liberofs_oci.h
+++ b/lib/liberofs_oci.h
@@ -35,6 +35,8 @@ struct ocierofs_config {
char *password;
char *blob_digest;
int layer_index;
+ char *local_meta_path;
+ char *zinfo_path;
};
struct ocierofs_layer_info {
diff --git a/lib/remotes/oci.c b/lib/remotes/oci.c
index b2f1f59..b25e0b2 100644
--- a/lib/remotes/oci.c
+++ b/lib/remotes/oci.c
@@ -1461,19 +1461,6 @@ static void ocierofs_io_close(struct erofs_vfile *vfile)
*(struct ocierofs_iostream **)vfile->payload = NULL;
}
-static int ocierofs_is_erofs_native_image(struct ocierofs_ctx *ctx)
-{
- if (ctx->layer_count > 0 && ctx->layers[0] &&
- ctx->layers[0]->media_type) {
- const char *media_type = ctx->layers[0]->media_type;
- size_t len = strlen(media_type);
-
- if (len >= 6 && strcmp(media_type + len - 6, ".erofs") == 0)
- return 0;
- }
- return -ENOENT;
-}
-
static struct erofs_vfops ocierofs_io_vfops = {
.pread = ocierofs_io_pread,
.read = ocierofs_io_read,
@@ -1497,13 +1484,6 @@ int ocierofs_io_open(struct erofs_vfile *vfile, const struct ocierofs_config *cf
return err;
}
- err = ocierofs_is_erofs_native_image(ctx);
- if (err) {
- ocierofs_ctx_cleanup(ctx);
- free(ctx);
- return err;
- }
-
oci_iostream = calloc(1, sizeof(*oci_iostream));
if (!oci_iostream) {
ocierofs_ctx_cleanup(ctx);
diff --git a/mount/main.c b/mount/main.c
index eb0dd01..811732a 100644
--- a/mount/main.c
+++ b/mount/main.c
@@ -16,6 +16,7 @@
#include "erofs/io.h"
#include "../lib/liberofs_nbd.h"
#include "../lib/liberofs_oci.h"
+#include "../lib/liberofs_gzran.h"
#ifdef HAVE_LINUX_LOOP_H
#include <linux/loop.h>
#else
@@ -141,7 +142,25 @@ static int erofsmount_parse_oci_option(const char *option)
if (!oci_cfg->password)
return -ENOMEM;
} else {
- return -EINVAL;
+ p = strstr(option, "oci.local_meta=");
+ if (p != NULL) {
+ p += strlen("oci.local_meta=");
+ free(oci_cfg->local_meta_path);
+ oci_cfg->local_meta_path = strdup(p);
+ if (!oci_cfg->local_meta_path)
+ return -ENOMEM;
+ } else {
+ p = strstr(option, "oci.zinfo=");
+ if (p != NULL) {
+ p += strlen("oci.zinfo=");
+ free(oci_cfg->zinfo_path);
+ oci_cfg->zinfo_path = strdup(p);
+ if (!oci_cfg->zinfo_path)
+ return -ENOMEM;
+ } else {
+ return -EINVAL;
+ }
+ }
}
}
}
@@ -332,11 +351,265 @@ static int erofsmount_fuse(const char *source, const char *mountpoint,
return 0;
}
+struct erofs_hybrid_source {
+ struct erofs_vfile local_vf;
+ struct erofs_vfile *gzran_vf;
+ u64 local_size;
+};
+
struct erofsmount_nbd_ctx {
struct erofs_vfile vd; /* virtual device */
struct erofs_vfile sk; /* socket file */
};
+static ssize_t erofs_hybrid_pread(struct erofs_vfile *vf, void *buf,
+ size_t count, u64 offset)
+{
+ struct erofs_hybrid_source *hs;
+ ssize_t local_read, remote_read;
+
+ hs = *(struct erofs_hybrid_source **)vf->payload;
+ if (!hs)
+ return -EINVAL;
+
+ /* Handle device boundary probe requests */
+ if (offset >= (INT64_MAX >> 9))
+ return 0;
+
+ if (hs->local_size == 0)
+ return hs->gzran_vf->ops->pread(hs->gzran_vf, buf, count, offset);
+
+ if (offset >= hs->local_size) {
+ u64 remote_offset = offset - hs->local_size;
+
+ return hs->gzran_vf->ops->pread(hs->gzran_vf, buf, count, remote_offset);
+ }
+
+ if (offset + count <= hs->local_size)
+ return erofs_io_pread(&hs->local_vf, buf, count, offset);
+
+ u64 local_part = hs->local_size - offset;
+ u64 remote_part = count - local_part;
+
+ local_read = erofs_io_pread(&hs->local_vf, buf, local_part, offset);
+ if (local_read < 0)
+ return local_read;
+
+ remote_read = hs->gzran_vf->ops->pread(hs->gzran_vf,
+ (char *)buf + local_read,
+ remote_part, 0);
+ if (remote_read < 0)
+ return remote_read;
+ return local_read + remote_read;
+}
+
+static void erofs_hybrid_close(struct erofs_vfile *vf)
+{
+ struct erofs_hybrid_source *hs;
+
+ if (!vf)
+ return;
+
+ hs = *(struct erofs_hybrid_source **)vf->payload;
+ if (!hs)
+ return;
+
+ if (hs->local_size > 0)
+ erofs_io_close(&hs->local_vf);
+
+ if (hs->gzran_vf)
+ erofs_io_close(hs->gzran_vf);
+
+ free(hs);
+}
+
+static int load_file_to_buf(const char *path, void **out, unsigned int *out_len)
+{
+ FILE *fp = NULL;
+ void *buf = NULL;
+ int ret = 0;
+ long sz;
+ size_t num;
+
+ fp = fopen(path, "rb");
+ if (!fp)
+ return -errno;
+
+ if (fseek(fp, 0, SEEK_END) != 0) {
+ ret = -errno;
+ goto out;
+ }
+ sz = ftell(fp);
+ if (sz < 0) {
+ ret = -errno;
+ goto out;
+ }
+ if (fseek(fp, 0, SEEK_SET) != 0) {
+ ret = -errno;
+ goto out;
+ }
+ if (sz == 0) {
+ ret = -EINVAL;
+ goto out;
+ }
+
+ buf = malloc((size_t)sz);
+ if (!buf) {
+ ret = -ENOMEM;
+ goto out;
+ }
+
+ num = fread(buf, 1, (size_t)sz, fp);
+ if (num != (size_t)sz) {
+ ret = -EIO;
+ goto out;
+ }
+
+ *out = buf;
+ *out_len = (unsigned int)sz;
+ buf = NULL;
+
+out:
+ if (fp)
+ fclose(fp);
+ if (ret < 0 && buf)
+ free(buf);
+ return ret;
+}
+
+static int erofsmount_init_gzran(struct erofs_vfile **gzran_vf,
+ const struct ocierofs_config *oci_cfg,
+ const char *zinfo_path)
+{
+ int err = 0;
+ void *zinfo_data = NULL;
+ unsigned int zinfo_len = 0;
+ struct erofs_vfile *oci_vf = NULL;
+
+ err = load_file_to_buf(zinfo_path, &zinfo_data, &zinfo_len);
+ if (err)
+ return err;
+
+ oci_vf = malloc(sizeof(*oci_vf));
+ if (!oci_vf) {
+ err = -ENOMEM;
+ goto cleanup;
+ }
+
+ err = ocierofs_io_open(oci_vf, oci_cfg);
+ if (err) {
+ free(oci_vf);
+ goto cleanup;
+ }
+
+ *gzran_vf = erofs_gzran_zinfo_open(oci_vf, zinfo_data, zinfo_len);
+ if (IS_ERR(*gzran_vf)) {
+ err = PTR_ERR(*gzran_vf);
+ *gzran_vf = NULL;
+ erofs_io_close(oci_vf);
+ free(oci_vf);
+ goto cleanup;
+ }
+
+ free(zinfo_data);
+ return 0;
+
+cleanup:
+ if (zinfo_data)
+ free(zinfo_data);
+ return err;
+}
+
+static ssize_t erofs_hybrid_sendfile(struct erofs_vfile *vout, struct erofs_vfile *vin,
+ off_t *pos, size_t count)
+{
+ static char buf[32768];
+ ssize_t total_written = 0, ret = 0, written;
+ size_t to_read;
+ u64 read_offset;
+
+ while (count > 0) {
+ to_read = min_t(size_t, count, sizeof(buf));
+ read_offset = pos ? *pos : 0;
+
+ ret = erofs_hybrid_pread(vin, buf, to_read, read_offset);
+ if (ret <= 0) {
+ if (ret < 0 && total_written == 0)
+ return ret;
+ break;
+ }
+
+ written = __erofs_io_write(vout->fd, buf, ret);
+ if (written < 0) {
+ ret = -errno;
+ break;
+ }
+ if (written != ret)
+ ret = written;
+
+ total_written += ret;
+ count -= ret;
+ if (pos)
+ *pos += ret;
+ }
+ return count;
+}
+
+static struct erofs_vfops hybrid_vfile_ops = {
+ .pread = erofs_hybrid_pread,
+ .sendfile = erofs_hybrid_sendfile,
+ .close = erofs_hybrid_close,
+};
+
+static int erofs_create_hybrid_source(struct erofs_vfile *out_vf,
+ const struct ocierofs_config *oci_cfg,
+ const char *local_meta_path,
+ const char *zinfo_path)
+{
+ struct erofs_hybrid_source *hs;
+ int err;
+ struct stat st;
+
+ hs = calloc(1, sizeof(*hs));
+ if (!hs)
+ return -ENOMEM;
+
+ if (local_meta_path) {
+ hs->local_vf.fd = open(local_meta_path, O_RDONLY);
+ if (hs->local_vf.fd < 0) {
+ err = -errno;
+ goto cleanup;
+ }
+
+ hs->local_vf.ops = NULL;
+ hs->local_vf.offset = 0;
+
+ if (fstat(hs->local_vf.fd, &st) < 0) {
+ err = -errno;
+ goto cleanup;
+ }
+ hs->local_size = st.st_size;
+ }
+
+ if (zinfo_path) {
+ err = erofsmount_init_gzran(&hs->gzran_vf, oci_cfg, zinfo_path);
+ if (err)
+ goto cleanup;
+ }
+
+ out_vf->ops = &hybrid_vfile_ops;
+ out_vf->fd = 0;
+ out_vf->offset = 0;
+ *(struct erofs_hybrid_source **)out_vf->payload = hs;
+
+ return 0;
+cleanup:
+ if (local_meta_path && hs->local_vf.fd >= 0)
+ close(hs->local_vf.fd);
+ free(hs);
+ return err;
+}
+
static void *erofsmount_nbd_loopfn(void *arg)
{
struct erofsmount_nbd_ctx *ctx = arg;
@@ -388,9 +661,17 @@ static int erofsmount_startnbd(int nbdfd, struct erofs_nbd_source *source)
int err, err2;
if (source->type == EROFSNBD_SOURCE_OCI) {
- err = ocierofs_io_open(&ctx.vd, &source->ocicfg);
- if (err)
- goto out_closefd;
+ if (source->ocicfg.local_meta_path || source->ocicfg.zinfo_path) {
+ err = erofs_create_hybrid_source(&ctx.vd, &source->ocicfg,
+ source->ocicfg.local_meta_path,
+ source->ocicfg.zinfo_path);
+ if (err)
+ goto out_closefd;
+ } else {
+ err = ocierofs_io_open(&ctx.vd, &source->ocicfg);
+ if (err)
+ goto out_closefd;
+ }
} else {
err = open(source->device_path, O_RDONLY);
if (err < 0) {
@@ -440,6 +721,19 @@ static int erofsmount_write_recovery_oci(FILE *f, struct erofs_nbd_source *sourc
return PTR_ERR(b64cred);
}
+ if ((source->ocicfg.local_meta_path || source->ocicfg.zinfo_path) &&
+ source->ocicfg.blob_digest && *source->ocicfg.blob_digest) {
+ ret = fprintf(f, "GZRAN_OCI_BLOB %s %s %s %s %s %s\n",
+ source->ocicfg.image_ref ?: "",
+ source->ocicfg.platform ?: "",
+ source->ocicfg.blob_digest,
+ b64cred ?: "",
+ source->ocicfg.local_meta_path ?: "",
+ source->ocicfg.zinfo_path ?: "");
+ free(b64cred);
+ return ret < 0 ? -ENOMEM : 0;
+ }
+
if (source->ocicfg.blob_digest && *source->ocicfg.blob_digest) {
ret = fprintf(f, "OCI_NATIVE_BLOB %s %s %s %s\n",
source->ocicfg.image_ref ?: "",
@@ -635,6 +929,60 @@ static int erofsmount_reattach_oci(struct erofs_vfile *vf,
}
#endif
+static int erofsmount_reattach_gzran_oci(struct erofsmount_nbd_ctx *ctx,
+ char *source)
+{
+ char *tokens[6] = {0}, *p = source, *space, *oci_source, *meta_path, *zinfo_path;
+ int token_count = 0, err;
+ const char *b64cred;
+ struct erofs_vfile temp_vd;
+ struct ocierofs_config oci_cfg = {};
+
+ while (token_count < 5) {
+ space = strchr(p, ' ');
+ if (!space)
+ break;
+
+ *space = '\0';
+ p = space + 1;
+ tokens[token_count++] = p;
+ }
+
+ if (token_count < 5)
+ return -EINVAL;
+
+ b64cred = (token_count > 2 && tokens[2]) ? tokens[2] : "";
+
+ err = asprintf(&oci_source, "%s %s %s %s",
+ source, tokens[0], tokens[1], b64cred);
+ if (err < 0)
+ return -ENOMEM;
+
+ err = erofsmount_reattach_oci(&ctx->vd, "OCI_NATIVE_BLOB", oci_source);
+ free(oci_source);
+ if (err)
+ return err;
+
+ temp_vd = ctx->vd;
+ oci_cfg.image_ref = strdup(source);
+ if (!oci_cfg.image_ref) {
+ erofs_io_close(&temp_vd);
+ return -ENOMEM;
+ }
+
+ if (token_count > 3 && tokens[3] && strlen(tokens[3]) > 0)
+ meta_path = tokens[3];
+ if (token_count > 4 && tokens[4] && strlen(tokens[4]) > 0)
+ zinfo_path = tokens[4];
+
+ err = erofs_create_hybrid_source(&ctx->vd, &oci_cfg,
+ meta_path, zinfo_path);
+ free(oci_cfg.image_ref);
+ erofs_io_close(&temp_vd);
+
+ return err;
+}
+
static int erofsmount_nbd_fix_backend_linkage(int num, char **recp)
{
char *newrecp;
@@ -679,9 +1027,17 @@ static int erofsmount_startnbd_nl(pid_t *pid, struct erofs_nbd_source *source)
exit(EXIT_FAILURE);
if (source->type == EROFSNBD_SOURCE_OCI) {
- err = ocierofs_io_open(&ctx.vd, &source->ocicfg);
- if (err)
- exit(EXIT_FAILURE);
+ if (source->ocicfg.local_meta_path || source->ocicfg.zinfo_path) {
+ err = erofs_create_hybrid_source(&ctx.vd, &source->ocicfg,
+ source->ocicfg.local_meta_path,
+ source->ocicfg.zinfo_path);
+ if (err)
+ exit(EXIT_FAILURE);
+ } else {
+ err = ocierofs_io_open(&ctx.vd, &source->ocicfg);
+ if (err)
+ exit(EXIT_FAILURE);
+ }
} else {
err = open(source->device_path, O_RDONLY);
if (err < 0)
@@ -794,6 +1150,10 @@ static int erofsmount_reattach(const char *target)
goto err_line;
}
ctx.vd.fd = err;
+ } else if (!strcmp(line, "GZRAN_OCI_BLOB")) {
+ err = erofsmount_reattach_gzran_oci(&ctx, source);
+ if (err)
+ goto err_line;
} else if (!strcmp(line, "OCI_LAYER") || !strcmp(line, "OCI_NATIVE_BLOB")) {
err = erofsmount_reattach_oci(&ctx.vd, line, source);
if (err)
--
2.51.0
More information about the Linux-erofs
mailing list