Skip to contentSkip to conversation

AOSP: erofs-utils: fix sub-directory prefix for canned fs_config

PATCH In review Gao Xiang via Linux-erofs 4 participants 14 messages Dec 26, 2020 – Dec 31, 2020 1 message not in the archive Library code

lore
Share & tools
lore

A message that is not in the archive (sent off-list or not yet received): [email protected]

#1 lore raw

PATCH 1/1 AOSP: erofs-utils: fix sub-directory prefix for canned fs_config

From: Gao Xiang <[email protected]>

"failed to find [%s] in canned fs_config" was observed by using "--fs-config-file" option as reported by Yue Hu [1].

The root cause was that the mountpoint prefix to subdirectories is also needed if "--mount-point" presents. However, such prefix cannot be added by just using erofs_fspath().

One exception is that the root directory itself needs to be handled specially for canned fs_config. For such case, the prefix of the root directory has to be dropped instead.

[1] https://lkml.kernel.org/r/[email protected]

Fixes 8a9e8046f170 ("AOSP: erofs-utils: add fs_config support")

---
Hi Yue, Jianan,

I've verified cuttlefish booting with success, It'd be better to verify this patchset on your sides. Please kingly leave "Tested-by:" if possible.

Hi Guifu,
Could you also review this patch ? This needs to be included in the upcoming v1.2.1 as well...

Thanks,
Gao Xiang

 lib/inode.c | 39 +++++++++++++++++++++++++--------------
 1 file changed, 25 insertions(+), 14 deletions(-)
lib/inode.c +25 -14
index 0c4839dc7152..f88966d26fce 100644
@@ -696,32 +696,43 @@ int erofs_droid_inode_fsconfig(struct erofs_inode *inode,
 	/* filesystem_config does not preserve file type bits */
 	mode_t stat_file_type_mask = st->st_mode & S_IFMT;
 	unsigned int uid = 0, gid = 0, mode = 0;
-	char *fspath;
+	const char *fspath;
+	char *decorated = NULL;
 
 	inode->capabilities = 0;
+	if (!cfg.fs_config_file && !cfg.mount_point)
+		return 0;
+
+	if (!cfg.mount_point ||
+	/* have to drop the mountpoint for rootdir of canned fsconfig */
+	    (cfg.fs_config_file && IS_ROOT(inode))) {
+		fspath = erofs_fspath(path);
+	} else {
+		if (asprintf(&decorated, "%s/%s", cfg.mount_point,
+			     erofs_fspath(path)) <= 0)
+			return -ENOMEM;
+		fspath = decorated;
+	}
+
 	if (cfg.fs_config_file)
-		canned_fs_config(erofs_fspath(path),
+		canned_fs_config(fspath,
 				 S_ISDIR(st->st_mode),
 				 cfg.target_out_path,
 				 &uid, &gid, &mode, &inode->capabilities);
-	else if (cfg.mount_point) {
-		if (asprintf(&fspath, "%s/%s", cfg.mount_point,
-			     erofs_fspath(path)) <= 0)
-			return -ENOMEM;
-
+	else
 		fs_config(fspath, S_ISDIR(st->st_mode),
 			  cfg.target_out_path,
 			  &uid, &gid, &mode, &inode->capabilities);
-		free(fspath);
-	}
-	st->st_uid = uid;
-	st->st_gid = gid;
-	st->st_mode = mode | stat_file_type_mask;
 
 	erofs_dbg("/%s -> mode = 0x%x, uid = 0x%x, gid = 0x%x, "
 		  "capabilities = 0x%" PRIx64 "\n",
-		  erofs_fspath(path),
-		  mode, uid, gid, inode->capabilities);
+		  fspath, mode, uid, gid, inode->capabilities);
+
+	if (decorated)
+		free(decorated);
+	st->st_uid = uid;
+	st->st_gid = gid;
+	st->st_mode = mode | stat_file_type_mask;
 	return 0;
 }
 #else
Signature · 1 line
2.24.0
#2 lore raw
Quoted 19 lines · On Sat, 26 Dec 2020 14:27:36 +0800 Gao Xiang <[email protected]> wrote:

From: Gao Xiang <[email protected]>

"failed to find [%s] in canned fs_config" was observed by using "--fs-config-file" option as reported by Yue Hu [1].

The root cause was that the mountpoint prefix to subdirectories is also needed if "--mount-point" presents. However, such prefix cannot be added by just using erofs_fspath().

One exception is that the root directory itself needs to be handled specially for canned fs_config. For such case, the prefix of the root directory has to be dropped instead.

[1]
https://lkml.kernel.org/r/[email protected]

Fixes 8a9e8046f170 ("AOSP: erofs-utils: add fs_config support")
Quoted 6 lines

---
Hi Yue, Jianan,

I've verified cuttlefish booting with success, It'd be better to verify this patchset on your sides. Please kingly leave "Tested-by:" if possible.

Hi Xiang,

I'm testing now. I will inform once finished.

Thx.

Quoted 75 lines

Hi Guifu,
Could you also review this patch ? This needs to be included in the upcoming v1.2.1 as well...

Thanks,
Gao Xiang

 lib/inode.c | 39 +++++++++++++++++++++++++--------------
 1 file changed, 25 insertions(+), 14 deletions(-)
lib/inode.c +0 -0
index 0c4839dc7152..f88966d26fce 100644
@@ -696,32 +696,43 @@ int erofs_droid_inode_fsconfig(struct

erofs_inode *inode, /* filesystem_config does not preserve file type bits */ mode_t stat_file_type_mask = st->st_mode & S_IFMT;

lib/inode.c +7 -1
 	unsigned int uid = 0, gid = 0, mode = 0;
-	char *fspath;
+	const char *fspath;
+	char *decorated = NULL;
 
 	inode->capabilities = 0;
+	if (!cfg.fs_config_file && !cfg.mount_point)
+		return 0;
+
+	if (!cfg.mount_point ||
+	/* have to drop the mountpoint for rootdir of canned

fsconfig */

lib/inode.c +10 -1
+	    (cfg.fs_config_file && IS_ROOT(inode))) {
+		fspath = erofs_fspath(path);
+	} else {
+		if (asprintf(&decorated, "%s/%s", cfg.mount_point,
+			     erofs_fspath(path)) <= 0)
+			return -ENOMEM;
+		fspath = decorated;
+	}
+
 	if (cfg.fs_config_file)
-		canned_fs_config(erofs_fspath(path),
+		canned_fs_config(fspath,
 				 S_ISDIR(st->st_mode),
 				 cfg.target_out_path,
 				 &uid, &gid, &mode,

&inode->capabilities);

lib/inode.c +8 -12
-	else if (cfg.mount_point) {
-		if (asprintf(&fspath, "%s/%s", cfg.mount_point,
-			     erofs_fspath(path)) <= 0)
-			return -ENOMEM;
-
+	else
 		fs_config(fspath, S_ISDIR(st->st_mode),
 			  cfg.target_out_path,
 			  &uid, &gid, &mode, &inode->capabilities);
-		free(fspath);
-	}
-	st->st_uid = uid;
-	st->st_gid = gid;
-	st->st_mode = mode | stat_file_type_mask;
 
 	erofs_dbg("/%s -> mode = 0x%x, uid = 0x%x, gid = 0x%x, "
 		  "capabilities = 0x%" PRIx64 "\n",
-		  erofs_fspath(path),
-		  mode, uid, gid, inode->capabilities);
+		  fspath, mode, uid, gid, inode->capabilities);
+
+	if (decorated)
+		free(decorated);
+	st->st_uid = uid;
+	st->st_gid = gid;
+	st->st_mode = mode | stat_file_type_mask;
 	return 0;
 }
 #else
#3 lore raw
In reply to Gao Xiang via Linux-erofs
Quoted 44 lines · Gao Xiang

From: Gao Xiang <[email protected]>

"failed to find [%s] in canned fs_config" was observed by using "--fs-config-file" option as reported by Yue Hu [1].

The root cause was that the mountpoint prefix to subdirectories is also needed if "--mount-point" presents. However, such prefix cannot be added by just using erofs_fspath().

One exception is that the root directory itself needs to be handled specially for canned fs_config. For such case, the prefix of the root directory has to be dropped instead.

[1] https://lkml.kernel.org/r/[email protected]

Fixes 8a9e8046f170 ("AOSP: erofs-utils: add fs_config support")

---
Hi Yue, Jianan,

I've verified cuttlefish booting with success, It'd be better to verify this patchset on your sides. Please kingly leave "Tested-by:" if possible.

Hi Guifu,
Could you also review this patch ? This needs to be included in the upcoming v1.2.1 as well...

Thanks,
Gao Xiang

  lib/inode.c | 39 +++++++++++++++++++++++++--------------
  1 file changed, 25 insertions(+), 14 deletions(-)
lib/inode.c +0 -1
index 0c4839dc7152..f88966d26fce 100644
@@ -696,32 +696,43 @@ int erofs_droid_inode_fsconfig(struct erofs_inode *inode,
  	/* filesystem_config does not preserve file type bits */
  	mode_t stat_file_type_mask = st->st_mode & S_IFMT;
  	unsigned int uid = 0, gid = 0, mode = 0;
-	char *fspath;

lib/inode.c int erofs_droid_inode_fsconfig(struct erofs_inode *inode, · line 708

Quoted 10 lines
lib/inode.c +8 -0
+	const char *fspath;
+	char *decorated = NULL;
 
  	inode->capabilities = 0;
+	if (!cfg.fs_config_file && !cfg.mount_point)
+		return 0;
+
+	if (!cfg.mount_point ||
+	/* have to drop the mountpoint for rootdir of canned fsconfig */
+	    (cfg.fs_config_file && IS_ROOT(inode))) {

Hi Xiang,

I have tested this patch with --fs-config-file, and the problem still exists.

IS_ROOT can't be used here to determine the root directory , because inode->i_parent is null at this time. It will be set after this function.

Thanks,

Jianan

Quoted 43 lines
lib/inode.c +17 -13
+		fspath = erofs_fspath(path);
+	} else {
+		if (asprintf(&decorated, "%s/%s", cfg.mount_point,
+			     erofs_fspath(path)) <= 0)
+			return -ENOMEM;
+		fspath = decorated;
+	}
+
  	if (cfg.fs_config_file)
-		canned_fs_config(erofs_fspath(path),
+		canned_fs_config(fspath,
  				 S_ISDIR(st->st_mode),
  				 cfg.target_out_path,
  				 &uid, &gid, &mode, &inode->capabilities);
-	else if (cfg.mount_point) {
-		if (asprintf(&fspath, "%s/%s", cfg.mount_point,
-			     erofs_fspath(path)) <= 0)
-			return -ENOMEM;
-
+	else
  		fs_config(fspath, S_ISDIR(st->st_mode),
  			  cfg.target_out_path,
  			  &uid, &gid, &mode, &inode->capabilities);
-		free(fspath);
-	}
-	st->st_uid = uid;
-	st->st_gid = gid;
-	st->st_mode = mode | stat_file_type_mask;
 
  	erofs_dbg("/%s -> mode = 0x%x, uid = 0x%x, gid = 0x%x, "
  		  "capabilities = 0x%" PRIx64 "\n",
-		  erofs_fspath(path),
-		  mode, uid, gid, inode->capabilities);
+		  fspath, mode, uid, gid, inode->capabilities);
+
+	if (decorated)
+		free(decorated);
+	st->st_uid = uid;
+	st->st_gid = gid;
+	st->st_mode = mode | stat_file_type_mask;
  	return 0;
  }
  #else
#4 lore raw
Quoted 2 lines · Huang Jianan

在 2020/12/26 14:27, Gao Xiang 写道:

...

Quoted 6 lines

lib/inode.c int erofs_droid_inode_fsconfig(struct erofs_inode *inode, · line 708

Quoted 3 lines
lib/inode.c +3 -0
+	if (!cfg.mount_point ||
+	/* have to drop the mountpoint for rootdir of canned fsconfig */
+	    (cfg.fs_config_file && IS_ROOT(inode))) {

Hi Xiang,

Quoted 6 lines

I have tested this patch with --fs-config-file, and the problem still exists.

IS_ROOT can't be used here to determine the root directory , because inode->i_parent is null at this time. It will be set after this function.

Hi Jianan,

Thanks for your report! Sorry for the troublesome.
I need to turn back as the next version.... Very sorry about this!

Thanks,
Gao Xiang

Quoted 47 lines

Thanks,

Jianan

Quoted 42 lines
lib/inode.c +17 -13
+		fspath = erofs_fspath(path);
+	} else {
+		if (asprintf(&decorated, "%s/%s", cfg.mount_point,
+			     erofs_fspath(path)) <= 0)
+			return -ENOMEM;
+		fspath = decorated;
+	}
+
  	if (cfg.fs_config_file)
-		canned_fs_config(erofs_fspath(path),
+		canned_fs_config(fspath,
  				 S_ISDIR(st->st_mode),
  				 cfg.target_out_path,
  				 &uid, &gid, &mode, &inode->capabilities);
-	else if (cfg.mount_point) {
-		if (asprintf(&fspath, "%s/%s", cfg.mount_point,
-			     erofs_fspath(path)) <= 0)
-			return -ENOMEM;
-
+	else
  		fs_config(fspath, S_ISDIR(st->st_mode),
  			  cfg.target_out_path,
  			  &uid, &gid, &mode, &inode->capabilities);
-		free(fspath);
-	}
-	st->st_uid = uid;
-	st->st_gid = gid;
-	st->st_mode = mode | stat_file_type_mask;
  	erofs_dbg("/%s -> mode = 0x%x, uid = 0x%x, gid = 0x%x, "
  		  "capabilities = 0x%" PRIx64 "\n",
-		  erofs_fspath(path),
-		  mode, uid, gid, inode->capabilities);
+		  fspath, mode, uid, gid, inode->capabilities);
+
+	if (decorated)
+		free(decorated);
+	st->st_uid = uid;
+	st->st_gid = gid;
+	st->st_mode = mode | stat_file_type_mask;
  	return 0;
  }
  #else
#5 lore raw

PATCH v2 1/1 AOSP: erofs-utils: fix sub-directory prefix for canned fs_config

In reply to Gao Xiang via Linux-erofs

From: Gao Xiang <[email protected]>

"failed to find [%s] in canned fs_config" was observed by using "--fs-config-file" option as reported by Yue Hu [1].

The root cause was that the mountpoint prefix to subdirectories is also needed if "--mount-point" presents. However, such prefix cannot be added by just using erofs_fspath().

One exception is that the root directory itself needs to be handled specially for canned fs_config. For such case, the prefix of the root directory has to be dropped instead.

[1] https://lkml.kernel.org/r/[email protected]

Fixes 8a9e8046f170 ("AOSP: erofs-utils: add fs_config support")

---
changes since v2:
- fix IS_ROOT misuse reported by Jianan, very sorry about this since I know little about canned fs_config.

(please kindly test again...)

 lib/inode.c | 39 +++++++++++++++++++++++++--------------
 1 file changed, 25 insertions(+), 14 deletions(-)
lib/inode.c +25 -14
index 0c4839d..e6159c9 100644
@@ -696,32 +696,43 @@ int erofs_droid_inode_fsconfig(struct erofs_inode *inode,
 	/* filesystem_config does not preserve file type bits */
 	mode_t stat_file_type_mask = st->st_mode & S_IFMT;
 	unsigned int uid = 0, gid = 0, mode = 0;
-	char *fspath;
+	const char *fspath;
+	char *decorated = NULL;
 
 	inode->capabilities = 0;
+	if (!cfg.fs_config_file && !cfg.mount_point)
+		return 0;
+
+	if (!cfg.mount_point ||
+	/* have to drop the mountpoint for rootdir of canned fsconfig */
+	    (cfg.fs_config_file && erofs_fspath(path)[0] == '\0')) {
+		fspath = erofs_fspath(path);
+	} else {
+		if (asprintf(&decorated, "%s/%s", cfg.mount_point,
+			     erofs_fspath(path)) <= 0)
+			return -ENOMEM;
+		fspath = decorated;
+	}
+
 	if (cfg.fs_config_file)
-		canned_fs_config(erofs_fspath(path),
+		canned_fs_config(fspath,
 				 S_ISDIR(st->st_mode),
 				 cfg.target_out_path,
 				 &uid, &gid, &mode, &inode->capabilities);
-	else if (cfg.mount_point) {
-		if (asprintf(&fspath, "%s/%s", cfg.mount_point,
-			     erofs_fspath(path)) <= 0)
-			return -ENOMEM;
-
+	else
 		fs_config(fspath, S_ISDIR(st->st_mode),
 			  cfg.target_out_path,
 			  &uid, &gid, &mode, &inode->capabilities);
-		free(fspath);
-	}
-	st->st_uid = uid;
-	st->st_gid = gid;
-	st->st_mode = mode | stat_file_type_mask;
 
 	erofs_dbg("/%s -> mode = 0x%x, uid = 0x%x, gid = 0x%x, "
 		  "capabilities = 0x%" PRIx64 "\n",
-		  erofs_fspath(path),
-		  mode, uid, gid, inode->capabilities);
+		  fspath, mode, uid, gid, inode->capabilities);
+
+	if (decorated)
+		free(decorated);
+	st->st_uid = uid;
+	st->st_gid = gid;
+	st->st_mode = mode | stat_file_type_mask;
 	return 0;
 }
 #else
Signature · 1 line
2.27.0
Tested-by on v2 1/1
#6 lore raw

Hi Xiang,

This patch works well with or without --fs-config-file in build.

Thanks,

Jianan

Quoted 91 lines · Gao Xiang

From: Gao Xiang <[email protected]>

"failed to find [%s] in canned fs_config" was observed by using "--fs-config-file" option as reported by Yue Hu [1].

The root cause was that the mountpoint prefix to subdirectories is also needed if "--mount-point" presents. However, such prefix cannot be added by just using erofs_fspath().

One exception is that the root directory itself needs to be handled specially for canned fs_config. For such case, the prefix of the root directory has to be dropped instead.

[1] https://lkml.kernel.org/r/[email protected]

Fixes 8a9e8046f170 ("AOSP: erofs-utils: add fs_config support")

---
changes since v2:
- fix IS_ROOT misuse reported by Jianan, very sorry about this since I know little about canned fs_config.

(please kindly test again...)

  lib/inode.c | 39 +++++++++++++++++++++++++--------------
  1 file changed, 25 insertions(+), 14 deletions(-)
lib/inode.c +25 -14
index 0c4839d..e6159c9 100644
@@ -696,32 +696,43 @@ int erofs_droid_inode_fsconfig(struct erofs_inode *inode,
  	/* filesystem_config does not preserve file type bits */
  	mode_t stat_file_type_mask = st->st_mode & S_IFMT;
  	unsigned int uid = 0, gid = 0, mode = 0;
-	char *fspath;
+	const char *fspath;
+	char *decorated = NULL;
 
  	inode->capabilities = 0;
+	if (!cfg.fs_config_file && !cfg.mount_point)
+		return 0;
+
+	if (!cfg.mount_point ||
+	/* have to drop the mountpoint for rootdir of canned fsconfig */
+	    (cfg.fs_config_file && erofs_fspath(path)[0] == '\0')) {
+		fspath = erofs_fspath(path);
+	} else {
+		if (asprintf(&decorated, "%s/%s", cfg.mount_point,
+			     erofs_fspath(path)) <= 0)
+			return -ENOMEM;
+		fspath = decorated;
+	}
+
  	if (cfg.fs_config_file)
-		canned_fs_config(erofs_fspath(path),
+		canned_fs_config(fspath,
  				 S_ISDIR(st->st_mode),
  				 cfg.target_out_path,
  				 &uid, &gid, &mode, &inode->capabilities);
-	else if (cfg.mount_point) {
-		if (asprintf(&fspath, "%s/%s", cfg.mount_point,
-			     erofs_fspath(path)) <= 0)
-			return -ENOMEM;
-
+	else
  		fs_config(fspath, S_ISDIR(st->st_mode),
  			  cfg.target_out_path,
  			  &uid, &gid, &mode, &inode->capabilities);
-		free(fspath);
-	}
-	st->st_uid = uid;
-	st->st_gid = gid;
-	st->st_mode = mode | stat_file_type_mask;
 
  	erofs_dbg("/%s -> mode = 0x%x, uid = 0x%x, gid = 0x%x, "
  		  "capabilities = 0x%" PRIx64 "\n",
-		  erofs_fspath(path),
-		  mode, uid, gid, inode->capabilities);
+		  fspath, mode, uid, gid, inode->capabilities);
+
+	if (decorated)
+		free(decorated);
+	st->st_uid = uid;
+	st->st_gid = gid;
+	st->st_mode = mode | stat_file_type_mask;
  	return 0;
  }
  #else
#7 lore raw
In reply to Gao Xiang
Quoted 49 lines · On Mon, 28 Dec 2020 18:51:46 +0800 Gao Xiang <[email protected]> wrote:

From: Gao Xiang <[email protected]>

"failed to find [%s] in canned fs_config" was observed by using "--fs-config-file" option as reported by Yue Hu [1].

The root cause was that the mountpoint prefix to subdirectories is also needed if "--mount-point" presents. However, such prefix cannot be added by just using erofs_fspath().

One exception is that the root directory itself needs to be handled specially for canned fs_config. For such case, the prefix of the root directory has to be dropped instead.

[1]
https://lkml.kernel.org/r/[email protected]

Link:
https://lore.kernel.org/r/[email protected]

Fixes 8a9e8046f170 ("AOSP: erofs-utils: add fs_config support")

<[email protected]> ---
changes since v2:
- fix IS_ROOT misuse reported by Jianan, very sorry about this since I know little about canned fs_config.

(please kindly test again...)

 lib/inode.c | 39 +++++++++++++++++++++++++--------------
 1 file changed, 25 insertions(+), 14 deletions(-)
lib/inode.c +0 -0
index 0c4839d..e6159c9 100644
@@ -696,32 +696,43 @@ int erofs_droid_inode_fsconfig(struct

erofs_inode *inode, /* filesystem_config does not preserve file type bits */ mode_t stat_file_type_mask = st->st_mode & S_IFMT;

lib/inode.c +7 -1
 	unsigned int uid = 0, gid = 0, mode = 0;
-	char *fspath;
+	const char *fspath;
+	char *decorated = NULL;
 
 	inode->capabilities = 0;
+	if (!cfg.fs_config_file && !cfg.mount_point)
+		return 0;
+
+	if (!cfg.mount_point ||
+	/* have to drop the mountpoint for rootdir of canned

fsconfig */

lib/inode.c int erofs_droid_inode_fsconfig(struct erofs_inode *inode, · line 717

Quoted 10 lines
lib/inode.c +9 -0
+	    (cfg.fs_config_file && erofs_fspath(path)[0] == '\0')) {
+		fspath = erofs_fspath(path);
+	} else {
+		if (asprintf(&decorated, "%s/%s", cfg.mount_point,
+			     erofs_fspath(path)) <= 0)
+			return -ENOMEM;
+		fspath = decorated;
+	}
+
 	if (cfg.fs_config_file)

Whether we can use the first "cfg.fs_config_file" when loading canned fs-config to reduce/simplify these duplicated calling?

Quoted 35 lines
lib/inode.c +1 -1
-		canned_fs_config(erofs_fspath(path),
+		canned_fs_config(fspath,
 				 S_ISDIR(st->st_mode),
 				 cfg.target_out_path,
 				 &uid, &gid, &mode,

&inode->capabilities);

lib/inode.c +8 -12
-	else if (cfg.mount_point) {
-		if (asprintf(&fspath, "%s/%s", cfg.mount_point,
-			     erofs_fspath(path)) <= 0)
-			return -ENOMEM;
-
+	else
 		fs_config(fspath, S_ISDIR(st->st_mode),
 			  cfg.target_out_path,
 			  &uid, &gid, &mode, &inode->capabilities);
-		free(fspath);
-	}
-	st->st_uid = uid;
-	st->st_gid = gid;
-	st->st_mode = mode | stat_file_type_mask;
 
 	erofs_dbg("/%s -> mode = 0x%x, uid = 0x%x, gid = 0x%x, "
 		  "capabilities = 0x%" PRIx64 "\n",
-		  erofs_fspath(path),
-		  mode, uid, gid, inode->capabilities);
+		  fspath, mode, uid, gid, inode->capabilities);
+
+	if (decorated)
+		free(decorated);
+	st->st_uid = uid;
+	st->st_gid = gid;
+	st->st_mode = mode | stat_file_type_mask;
 	return 0;
 }
 #else
#8 lore raw

Hi Yue,

Quoted 2 lines · Yue Hu

On Mon, 28 Dec 2020 18:51:46 +0800
Gao Xiang <[email protected]> wrote:

...

Quoted 20 lines
Quoted 20 lines
(quoted diff) +0 -0
@@ -696,32 +696,43 @@ int erofs_droid_inode_fsconfig(struct

erofs_inode *inode, /* filesystem_config does not preserve file type bits */ mode_t stat_file_type_mask = st->st_mode & S_IFMT;

lib/inode.c +7 -1
 	unsigned int uid = 0, gid = 0, mode = 0;
-	char *fspath;
+	const char *fspath;
+	char *decorated = NULL;
 
 	inode->capabilities = 0;
+	if (!cfg.fs_config_file && !cfg.mount_point)
+		return 0;
+
+	if (!cfg.mount_point ||
+	/* have to drop the mountpoint for rootdir of canned

fsconfig */

lib/inode.c +5 -0
+	    (cfg.fs_config_file && erofs_fspath(path)[0] == '\0')) {
+		fspath = erofs_fspath(path);
+	} else {
+		if (asprintf(&decorated, "%s/%s", cfg.mount_point,
+			     erofs_fspath(path)) <= 0)
Quoted 8 lines

lib/inode.c int erofs_droid_inode_fsconfig(struct erofs_inode *inode, · line 717

Quoted 5 lines
lib/inode.c +4 -0
+			return -ENOMEM;
+		fspath = decorated;
+	}
+
 	if (cfg.fs_config_file)

Whether we can use the first "cfg.fs_config_file" when loading canned fs-config to reduce/simplify these duplicated calling?

Not sure what you mean... If you mean why not using some "fs_config_func" as squashfs did, that is I'd like to 1) avoid such unneeded indirect function pointers, and 2) no need to introduce such function prototype (I don't want to maintain such function pointer type) since fs_config and canned_fs_config implement differently (and it seems they also behave differently so no need to mix them at all).

Thanks,
Gao Xiang

#9 lore raw
Quoted 39 lines · On Mon, 28 Dec 2020 19:32:47 +0800 Gao Xiang <[email protected]> wrote:

Hi Yue,

Quoted 2 lines · Yue Hu

On Mon, 28 Dec 2020 18:51:46 +0800
Gao Xiang <[email protected]> wrote:

...

Quoted 22 lines
Quoted 22 lines
(quoted diff) +0 -0
@@ -696,32 +696,43 @@ int erofs_droid_inode_fsconfig(struct

erofs_inode *inode, /* filesystem_config does not preserve file type bits */ mode_t stat_file_type_mask = st->st_mode & S_IFMT;

lib/inode.c +7 -1
 	unsigned int uid = 0, gid = 0, mode = 0;
-	char *fspath;
+	const char *fspath;
+	char *decorated = NULL;
 
 	inode->capabilities = 0;
+	if (!cfg.fs_config_file && !cfg.mount_point)
+		return 0;
+
+	if (!cfg.mount_point ||
+	/* have to drop the mountpoint for rootdir of canned

fsconfig */

(quoted diff) +1 -0
+	    (cfg.fs_config_file && erofs_fspath(path)[0] ==

'\0')) {

lib/inode.c +3 -0
+		fspath = erofs_fspath(path);
+	} else {
+		if (asprintf(&decorated, "%s/%s",

cfg.mount_point,

lib/inode.c +1 -0
+			     erofs_fspath(path)) <= 0)
Quoted 8 lines

lib/inode.c int erofs_droid_inode_fsconfig(struct erofs_inode *inode, · line 717

Quoted 5 lines
lib/inode.c +4 -0
+			return -ENOMEM;
+		fspath = decorated;
+	}
+
 	if (cfg.fs_config_file)  

Whether we can use the first "cfg.fs_config_file" when loading canned fs-config to reduce/simplify these duplicated calling?

Quoted 1 line

Not sure what you mean... If you mean why not using some

Sorry for my typo. I mean duplicated 'check' to cfg.fs_config_file.

Quoted 5 lines

"fs_config_func" as squashfs did, that is I'd like to 1) avoid such unneeded indirect function pointers, and 2) no need to introduce such function prototype (I don't want to maintain such function pointer type) since fs_config and canned_fs_config implement differently (and it seems they also behave differently so no need to mix them at all).

Ok.

Quoted 4 lines

Thanks,
Gao Xiang

Tested-by on v2 1/1
#10 lore raw
In reply to Gao Xiang

Hi Xiang,

Works fine to me for canned/non-canned fs_config.

Thx.

Quoted 94 lines · On Mon, 28 Dec 2020 18:51:46 +0800 Gao Xiang <[email protected]> wrote:

From: Gao Xiang <[email protected]>

"failed to find [%s] in canned fs_config" was observed by using "--fs-config-file" option as reported by Yue Hu [1].

The root cause was that the mountpoint prefix to subdirectories is also needed if "--mount-point" presents. However, such prefix cannot be added by just using erofs_fspath().

One exception is that the root directory itself needs to be handled specially for canned fs_config. For such case, the prefix of the root directory has to be dropped instead.

[1]
https://lkml.kernel.org/r/[email protected]

Link:
https://lore.kernel.org/r/[email protected]

Fixes 8a9e8046f170 ("AOSP: erofs-utils: add fs_config support")

<[email protected]> ---
changes since v2:
- fix IS_ROOT misuse reported by Jianan, very sorry about this since I know little about canned fs_config.

(please kindly test again...)

 lib/inode.c | 39 +++++++++++++++++++++++++--------------
 1 file changed, 25 insertions(+), 14 deletions(-)
lib/inode.c +0 -0
index 0c4839d..e6159c9 100644
@@ -696,32 +696,43 @@ int erofs_droid_inode_fsconfig(struct

erofs_inode *inode, /* filesystem_config does not preserve file type bits */ mode_t stat_file_type_mask = st->st_mode & S_IFMT;

lib/inode.c +7 -1
 	unsigned int uid = 0, gid = 0, mode = 0;
-	char *fspath;
+	const char *fspath;
+	char *decorated = NULL;
 
 	inode->capabilities = 0;
+	if (!cfg.fs_config_file && !cfg.mount_point)
+		return 0;
+
+	if (!cfg.mount_point ||
+	/* have to drop the mountpoint for rootdir of canned

fsconfig */

lib/inode.c +10 -1
+	    (cfg.fs_config_file && erofs_fspath(path)[0] == '\0')) {
+		fspath = erofs_fspath(path);
+	} else {
+		if (asprintf(&decorated, "%s/%s", cfg.mount_point,
+			     erofs_fspath(path)) <= 0)
+			return -ENOMEM;
+		fspath = decorated;
+	}
+
 	if (cfg.fs_config_file)
-		canned_fs_config(erofs_fspath(path),
+		canned_fs_config(fspath,
 				 S_ISDIR(st->st_mode),
 				 cfg.target_out_path,
 				 &uid, &gid, &mode,

&inode->capabilities);

lib/inode.c +8 -12
-	else if (cfg.mount_point) {
-		if (asprintf(&fspath, "%s/%s", cfg.mount_point,
-			     erofs_fspath(path)) <= 0)
-			return -ENOMEM;
-
+	else
 		fs_config(fspath, S_ISDIR(st->st_mode),
 			  cfg.target_out_path,
 			  &uid, &gid, &mode, &inode->capabilities);
-		free(fspath);
-	}
-	st->st_uid = uid;
-	st->st_gid = gid;
-	st->st_mode = mode | stat_file_type_mask;
 
 	erofs_dbg("/%s -> mode = 0x%x, uid = 0x%x, gid = 0x%x, "
 		  "capabilities = 0x%" PRIx64 "\n",
-		  erofs_fspath(path),
-		  mode, uid, gid, inode->capabilities);
+		  fspath, mode, uid, gid, inode->capabilities);
+
+	if (decorated)
+		free(decorated);
+	st->st_uid = uid;
+	st->st_gid = gid;
+	st->st_mode = mode | stat_file_type_mask;
 	return 0;
 }
 #else
#12 lore raw
In reply to Gao Xiang

Thank for your works, Yue Hu, Jianan, Gao Xiang

Quoted 46 lines · Gao Xiang

From: Gao Xiang <[email protected]>

"failed to find [%s] in canned fs_config" was observed by using "--fs-config-file" option as reported by Yue Hu [1].

The root cause was that the mountpoint prefix to subdirectories is also needed if "--mount-point" presents. However, such prefix cannot be added by just using erofs_fspath().

One exception is that the root directory itself needs to be handled specially for canned fs_config. For such case, the prefix of the root directory has to be dropped instead.

[1] https://lkml.kernel.org/r/[email protected]

Fixes 8a9e8046f170 ("AOSP: erofs-utils: add fs_config support")

---
changes since v2:
- fix IS_ROOT misuse reported by Jianan, very sorry about this since I know little about canned fs_config.

(please kindly test again...)

 lib/inode.c | 39 +++++++++++++++++++++++++--------------
 1 file changed, 25 insertions(+), 14 deletions(-)
lib/inode.c +6 -1
index 0c4839d..e6159c9 100644
@@ -696,32 +696,43 @@ int erofs_droid_inode_fsconfig(struct erofs_inode *inode,
 	/* filesystem_config does not preserve file type bits */
 	mode_t stat_file_type_mask = st->st_mode & S_IFMT;
 	unsigned int uid = 0, gid = 0, mode = 0;
-	char *fspath;
+	const char *fspath;
+	char *decorated = NULL;
 
 	inode->capabilities = 0;
+	if (!cfg.fs_config_file && !cfg.mount_point)
+		return 0;
+
+	if (!cfg.mount_point ||

lib/inode.c int erofs_droid_inode_fsconfig(struct erofs_inode *inode, · line 716

Quoted 10 lines
lib/inode.c +10 -0
+	/* have to drop the mountpoint for rootdir of canned fsconfig */
+	    (cfg.fs_config_file && erofs_fspath(path)[0] == '\0')) {
+		fspath = erofs_fspath(path);
+	} else {
+		if (asprintf(&decorated, "%s/%s", cfg.mount_point,
+			     erofs_fspath(path)) <= 0)
+			return -ENOMEM;
+		fspath = decorated;
+	}
+

erofs_fspath has been written for three times, and called always. What do you think refact it ? Do it with
fspath = erofs_fspath(path);
if need decorated:
fspath = decorated

#13 lore raw

Hi Guifu,

Quoted 3 lines · Li GuiFu

Thank for your works, Yue Hu, Jianan, Gao Xiang

On 2020/12/28 18:51, Gao Xiang wrote:

...

Quoted 24 lines
Quoted 24 lines
lib/inode.c +13 -1
index 0c4839d..e6159c9 100644
@@ -696,32 +696,43 @@ int erofs_droid_inode_fsconfig(struct erofs_inode *inode,
 	/* filesystem_config does not preserve file type bits */
 	mode_t stat_file_type_mask = st->st_mode & S_IFMT;
 	unsigned int uid = 0, gid = 0, mode = 0;
-	char *fspath;
+	const char *fspath;
+	char *decorated = NULL;
 
 	inode->capabilities = 0;
+	if (!cfg.fs_config_file && !cfg.mount_point)
+		return 0;
+
+	if (!cfg.mount_point ||
+	/* have to drop the mountpoint for rootdir of canned fsconfig */
+	    (cfg.fs_config_file && erofs_fspath(path)[0] == '\0')) {
+		fspath = erofs_fspath(path);
+	} else {
+		if (asprintf(&decorated, "%s/%s", cfg.mount_point,
+			     erofs_fspath(path)) <= 0)
+			return -ENOMEM;
Quoted 8 lines

lib/inode.c int erofs_droid_inode_fsconfig(struct erofs_inode *inode, · line 716

Quoted 3 lines
lib/inode.c +3 -0
+		fspath = decorated;
+	}
+

erofs_fspath has been written for three times, and called always. What do you think refact it ? Do it with
fspath = erofs_fspath(path);
if need decorated:
fspath = decorated

Yeah, it sounds reasonable. Let's refine this as a improvement of an individual patch (if needed) since I have to rebuild AOSP again (that is a bit complex for me now to open my PC...)

Thanks,
Gao Xiang

Quoted 1 line
author Reviewed-by Li Guifu on v2 1/1
#14 lore raw
In reply to Gao Xiang
Quoted 24 lines · Gao Xiang

From: Gao Xiang <[email protected]>

"failed to find [%s] in canned fs_config" was observed by using "--fs-config-file" option as reported by Yue Hu [1].

The root cause was that the mountpoint prefix to subdirectories is also needed if "--mount-point" presents. However, such prefix cannot be added by just using erofs_fspath().

One exception is that the root directory itself needs to be handled specially for canned fs_config. For such case, the prefix of the root directory has to be dropped instead.

[1] https://lkml.kernel.org/r/[email protected]

Fixes 8a9e8046f170 ("AOSP: erofs-utils: add fs_config support")

---
changes since v2:
- fix IS_ROOT misuse reported by Jianan, very sorry about this since I know little about canned fs_config.

Quoted 5 lines

(please kindly test again...)

 lib/inode.c | 39 +++++++++++++++++++++++++--------------
 1 file changed, 25 insertions(+), 14 deletions(-)

It looks good

Thanks,