[LTP] [PATCH v2 0/3] tst_device.c: Handle Android path for backing device

[LTP] [PATCH v2 0/3] tst_device.c: Handle Android path for backing device

[Edward Liaw via ltp]

e1b1ae66b240 ("tst_find_backing_dev: Get dev name from
/sys/dev/block/*/uevent") added a hardcoded path to check for the
backing device in /dev.  On Android, it needs to check /dev/block.

The set_dev_path function was renamed to set_dev_loop_path and its
return value was changed from 1 on success to 0 on success to be more
consistent with other functions.  A check was added to
tst_find_free_loopdev in the event that the loop device is not ready
yet, which appears to happen occasionally on Android.

v1->v2:
Changed the function signature of tst_find_backing_dev to add the length
of the path string.  Updated all references for this function to include
the added parameter.

Edward Liaw (3):
  tst_device.c: Use PATH_MAX more consistently
  set_dev_loop_path: Refactor set_dev_path and check return value
  tst_find_backing_dev: Also check /dev/block/ for backing device

 doc/c-test-api.txt                            |  2 +-
 include/tst_device.h                          |  3 +-
 lib/newlib_tests/tst_device.c                 |  2 +-
 lib/tst_device.c                              | 59 ++++++++++++-------
 .../kernel/syscalls/ioctl/ioctl_loop05.c      |  2 +-
 5 files changed, 42 insertions(+), 26 deletions(-)

-- 
2.40.0.rc1.284.g88254d51c5-goog


-- 
Mailing list info: https://lists.linux.it/listinfo/ltp

[LTP] [PATCH v2 1/3] tst_device.c: Use PATH_MAX more consistently

[Edward Liaw via ltp]

Replace hardcoded 1024 path sizes with PATH_MAX.

Signed-off-by: Edward Liaw <edliaw@google.com>
---
 lib/tst_device.c | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)

diff --git a/lib/tst_device.c b/lib/tst_device.c
index 48d7e3ab6..a61c5a748 100644
--- a/lib/tst_device.c
+++ b/lib/tst_device.c
@@ -50,7 +50,7 @@
 #define UUID_STR_SZ 37
 #define UUID_FMT "%02x%02x%02x%02x-%02x%02x-%02x%02x-%02x%02x-%02x%02x%02x%02x%02x%02x"
 
-static char dev_path[1024];
+static char dev_path[PATH_MAX];
 static int device_acquired;
 static unsigned long prev_dev_sec_write;
 
@@ -79,7 +79,7 @@ int tst_find_free_loopdev(char *path, size_t path_len)
 {
 	int ctl_fd, dev_fd, rc, i;
 	struct loop_info loopinfo;
-	char buf[1024];
+	char buf[PATH_MAX];
 
 	/* since Linux 3.1 */
 	ctl_fd = open(LOOP_CONTROL_FILE, O_RDWR);
@@ -489,7 +489,7 @@ int find_stat_file(const char *dev, char *path, size_t path_len)
 unsigned long tst_dev_bytes_written(const char *dev)
 {
 	unsigned long dev_sec_write = 0, dev_bytes_written, io_ticks = 0;
-	char dev_stat_path[1024];
+	char dev_stat_path[PATH_MAX];
 
 	if (!find_stat_file(dev, dev_stat_path, sizeof(dev_stat_path)))
 		tst_brkm(TCONF, NULL, "Test device stat file: %s not found",
@@ -641,7 +641,7 @@ int tst_dev_block_size(const char *path)
 {
 	int fd;
 	int size;
-	char dev_name[1024];
+	char dev_name[PATH_MAX];
 
 	tst_find_backing_dev(path, dev_name);
 
-- 
2.40.0.rc1.284.g88254d51c5-goog


-- 
Mailing list info: https://lists.linux.it/listinfo/ltp

Re: [LTP] [PATCH v2 1/3] tst_device.c: Use PATH_MAX more consistently

[Petr Vorel]

Hi Edward,

obviously correct, merged.

Kind regards,
Petr

-- 
Mailing list info: https://lists.linux.it/listinfo/ltp

[LTP] [PATCH v2 2/3] set_dev_loop_path: Refactor set_dev_path and check return value

[Edward Liaw via ltp]

tst_find_free_loopdev does not check the return value of set_dev_path
and will return the last attempted path even if it does not pass a stat
check.  set_dev_path also has a return value that is not consistent with
the other functions in this file.

Renames the function to set_dev_loop_path, the const array to
dev_loop_variants and changes the return value to 0 on success and 1 on
failure.  Check the return value when called in tst_find_free_loopdev
for failure to acquire a loop device.

Signed-off-by: Edward Liaw <edliaw@google.com>
---
 lib/tst_device.c | 18 +++++++++---------
 1 file changed, 9 insertions(+), 9 deletions(-)

diff --git a/lib/tst_device.c b/lib/tst_device.c
index a61c5a748..ae665f7b6 100644
--- a/lib/tst_device.c
+++ b/lib/tst_device.c
@@ -54,25 +54,25 @@ static char dev_path[PATH_MAX];
 static int device_acquired;
 static unsigned long prev_dev_sec_write;
 
-static const char *dev_variants[] = {
+static const char *dev_loop_variants[] = {
 	"/dev/loop%i",
 	"/dev/loop/%i",
 	"/dev/block/loop%i"
 };
 
-static int set_dev_path(int dev, char *path, size_t path_len)
+static int set_dev_loop_path(int dev, char *path, size_t path_len)
 {
 	unsigned int i;
 	struct stat st;
 
-	for (i = 0; i < ARRAY_SIZE(dev_variants); i++) {
-		snprintf(path, path_len, dev_variants[i], dev);
+	for (i = 0; i < ARRAY_SIZE(dev_loop_variants); i++) {
+		snprintf(path, path_len, dev_loop_variants[i], dev);
 
 		if (stat(path, &st) == 0 && S_ISBLK(st.st_mode))
-			return 1;
+			return 0;
 	}
 
-	return 0;
+	return 1;
 }
 
 int tst_find_free_loopdev(char *path, size_t path_len)
@@ -88,8 +88,8 @@ int tst_find_free_loopdev(char *path, size_t path_len)
 		rc = ioctl(ctl_fd, LOOP_CTL_GET_FREE);
 		close(ctl_fd);
 		if (rc >= 0) {
-			if (path)
-				set_dev_path(rc, path, path_len);
+			if (!path || set_dev_loop_path(rc, path, path_len) != 0)
+				tst_brkm(TBROK, NULL, "Could not stat loop device %i", rc);
 			tst_resm(TINFO, "Found free device %d '%s'",
 				rc, path ?: "");
 			return rc;
@@ -116,7 +116,7 @@ int tst_find_free_loopdev(char *path, size_t path_len)
 	 */
 	for (i = 0; i < 256; i++) {
 
-		if (!set_dev_path(i, buf, sizeof(buf)))
+		if (set_dev_loop_path(i, buf, sizeof(buf)) != 0)
 			continue;
 
 		dev_fd = open(buf, O_RDONLY);
-- 
2.40.0.rc1.284.g88254d51c5-goog


-- 
Mailing list info: https://lists.linux.it/listinfo/ltp

Re: [LTP] [PATCH v2 2/3] set_dev_loop_path: Refactor set_dev_path and check return value

[Petr Vorel]

Hi Edward,

> tst_find_free_loopdev does not check the return value of set_dev_path
> and will return the last attempted path even if it does not pass a stat
> check.  set_dev_path also has a return value that is not consistent with
> the other functions in this file.

This change and change of return is a bit burden in loop rename changes.
I'm ok it's in single patch, but it'd be more readable if it were separate.

> Renames the function to set_dev_loop_path, the const array to
> dev_loop_variants and changes the return value to 0 on success and 1 on
> failure.  Check the return value when called in tst_find_free_loopdev
> for failure to acquire a loop device.

> Signed-off-by: Edward Liaw <edliaw@google.com>
> ---
>  lib/tst_device.c | 18 +++++++++---------
>  1 file changed, 9 insertions(+), 9 deletions(-)

> diff --git a/lib/tst_device.c b/lib/tst_device.c
> index a61c5a748..ae665f7b6 100644
> --- a/lib/tst_device.c
> +++ b/lib/tst_device.c
> @@ -54,25 +54,25 @@ static char dev_path[PATH_MAX];
>  static int device_acquired;
>  static unsigned long prev_dev_sec_write;

> -static const char *dev_variants[] = {
> +static const char *dev_loop_variants[] = {
>  	"/dev/loop%i",
>  	"/dev/loop/%i",
>  	"/dev/block/loop%i"
>  };

> -static int set_dev_path(int dev, char *path, size_t path_len)
> +static int set_dev_loop_path(int dev, char *path, size_t path_len)
>  {
>  	unsigned int i;
>  	struct stat st;

> -	for (i = 0; i < ARRAY_SIZE(dev_variants); i++) {
> -		snprintf(path, path_len, dev_variants[i], dev);
> +	for (i = 0; i < ARRAY_SIZE(dev_loop_variants); i++) {
> +		snprintf(path, path_len, dev_loop_variants[i], dev);

>  		if (stat(path, &st) == 0 && S_ISBLK(st.st_mode))
> -			return 1;
> +			return 0;
>  	}

> -	return 0;
> +	return 1;
>  }

>  int tst_find_free_loopdev(char *path, size_t path_len)
> @@ -88,8 +88,8 @@ int tst_find_free_loopdev(char *path, size_t path_len)
>  		rc = ioctl(ctl_fd, LOOP_CTL_GET_FREE);
>  		close(ctl_fd);
>  		if (rc >= 0) {
> -			if (path)
> -				set_dev_path(rc, path, path_len);
> +			if (!path || set_dev_loop_path(rc, path, path_len) != 0)
> +				tst_brkm(TBROK, NULL, "Could not stat loop device %i", rc);

set_dev_path() is going to be called only if non-NULL path
(see include/tst_device.h). I haven't found a test which uses it this way,
but shouldn't it be checking path, instead of !path?

	if (path && set_dev_loop_path(rc, path, path_len) != 0)

Kind regards,
Petr

>  			tst_resm(TINFO, "Found free device %d '%s'",
>  				rc, path ?: "");
>  			return rc;
> @@ -116,7 +116,7 @@ int tst_find_free_loopdev(char *path, size_t path_len)
>  	 */
>  	for (i = 0; i < 256; i++) {

> -		if (!set_dev_path(i, buf, sizeof(buf)))
> +		if (set_dev_loop_path(i, buf, sizeof(buf)) != 0)
>  			continue;

>  		dev_fd = open(buf, O_RDONLY);

-- 
Mailing list info: https://lists.linux.it/listinfo/ltp

Re: [LTP] [PATCH v2 2/3] set_dev_loop_path: Refactor set_dev_path and check return value

[Edward Liaw via ltp]

On Thu, Mar 23, 2023 at 1:52 AM Petr Vorel <pvorel@suse.cz> wrote:
>
> Hi Edward,
>
> > tst_find_free_loopdev does not check the return value of set_dev_path
> > and will return the last attempted path even if it does not pass a stat
> > check.  set_dev_path also has a return value that is not consistent with
> > the other functions in this file.
>
> This change and change of return is a bit burden in loop rename changes.
> I'm ok it's in single patch, but it'd be more readable if it were separate.

Not a problem, I will split it.

> set_dev_path() is going to be called only if non-NULL path
> (see include/tst_device.h). I haven't found a test which uses it this way,
> but shouldn't it be checking path, instead of !path?
>
>         if (path && set_dev_loop_path(rc, path, path_len) != 0)
>
> Kind regards,
> Petr

Oops, I missed that in the comment and thought a NULL path should be
checked as an error.  You are right, I will change it.  Also, I wasn't
sure if I should be explicit with the "!= 0".

-- 
Mailing list info: https://lists.linux.it/listinfo/ltp

Re: [LTP] [PATCH v2 2/3] set_dev_loop_path: Refactor set_dev_path and check return value

[Petr Vorel]

> On Thu, Mar 23, 2023 at 1:52 AM Petr Vorel <pvorel@suse.cz> wrote:

> > Hi Edward,

> > > tst_find_free_loopdev does not check the return value of set_dev_path
> > > and will return the last attempted path even if it does not pass a stat
> > > check.  set_dev_path also has a return value that is not consistent with
> > > the other functions in this file.

> > This change and change of return is a bit burden in loop rename changes.
> > I'm ok it's in single patch, but it'd be more readable if it were separate.

> Not a problem, I will split it.

+1

Also, please rebase (I pushed some changes) which touch files you also modify.

> > set_dev_path() is going to be called only if non-NULL path
> > (see include/tst_device.h). I haven't found a test which uses it this way,
> > but shouldn't it be checking path, instead of !path?

> >         if (path && set_dev_loop_path(rc, path, path_len) != 0)

> > Kind regards,
> > Petr

> Oops, I missed that in the comment and thought a NULL path should be
> checked as an error.  You are right, I will change it.  Also, I wasn't
> sure if I should be explicit with the "!= 0".

I guess we are quite ok with just "!". We try to be precise at syscalls testing
(to check if the return value on error is exactly -1, not just < 0, but with
normal non-testing code like this in tst_kernel.c it's not needed.

Kind regards,
Petr

-- 
Mailing list info: https://lists.linux.it/listinfo/ltp

[LTP] [PATCH v2 3/3] tst_find_backing_dev: Also check /dev/block/ for backing device

[Edward Liaw via ltp]

On Android, the backing devices are created in /dev/block/ and will not
be found using the method added in e1b1ae66b240 ("tst_find_backing_dev:
Get dev name from /sys/dev/block/*/uevent").  Adds a check for
/dev/block/%s as well as /dev/%s.

Modified the function signature of tst_find_backing_dev to add the
length of the dev path string.  Updated the documentation and code that
references it.

Signed-off-by: Edward Liaw <edliaw@google.com>
---
 doc/c-test-api.txt                            |  2 +-
 include/tst_device.h                          |  3 +-
 lib/newlib_tests/tst_device.c                 |  2 +-
 lib/tst_device.c                              | 37 +++++++++++++------
 .../kernel/syscalls/ioctl/ioctl_loop05.c      |  2 +-
 5 files changed, 31 insertions(+), 15 deletions(-)

diff --git a/doc/c-test-api.txt b/doc/c-test-api.txt
index a7dd59dac..eddb5c893 100644
--- a/doc/c-test-api.txt
+++ b/doc/c-test-api.txt
@@ -1087,7 +1087,7 @@ is created for that intention.
 -------------------------------------------------------------------------------
 #include "tst_test.h"
 
-void tst_find_backing_dev(const char *path, char *dev);
+void tst_find_backing_dev(const char *path, char *dev, size_t dev_len);
 -------------------------------------------------------------------------------
 
 This function finds the block dev that this path belongs to, using uevent in sysfs.
diff --git a/include/tst_device.h b/include/tst_device.h
index 977427f1c..f0aff4473 100644
--- a/include/tst_device.h
+++ b/include/tst_device.h
@@ -110,8 +110,9 @@ void tst_purge_dir(const char *path);
  * Find the file or path belongs to which block dev
  * @path  Path to find the backing dev
  * @dev   The block dev
+ * @dev_len   The length of the dev string
  */
-void tst_find_backing_dev(const char *path, char *dev);
+void tst_find_backing_dev(const char *path, char *dev, size_t dev_len);
 
 /*
  * Stat the device mounted on a given path.
diff --git a/lib/newlib_tests/tst_device.c b/lib/newlib_tests/tst_device.c
index 87cec3961..53099f9bc 100644
--- a/lib/newlib_tests/tst_device.c
+++ b/lib/newlib_tests/tst_device.c
@@ -71,7 +71,7 @@ static void test_tst_find_backing_dev(void)
 {
 	char block_dev[100];
 
-	tst_find_backing_dev(mntpoint, block_dev);
+	tst_find_backing_dev(mntpoint, block_dev, sizeof(block_dev));
 
 	if (!strcmp(tst_device->dev, block_dev))
 		tst_res(TPASS, "%s belongs to %s block dev", mntpoint,
diff --git a/lib/tst_device.c b/lib/tst_device.c
index ae665f7b6..dbeae4c03 100644
--- a/lib/tst_device.c
+++ b/lib/tst_device.c
@@ -60,6 +60,11 @@ static const char *dev_loop_variants[] = {
 	"/dev/block/loop%i"
 };
 
+static const char *dev_variants[] = {
+	"/dev/%s",
+	"/dev/block/%s"
+};
+
 static int set_dev_loop_path(int dev, char *path, size_t path_len)
 {
 	unsigned int i;
@@ -75,6 +80,21 @@ static int set_dev_loop_path(int dev, char *path, size_t path_len)
 	return 1;
 }
 
+static int set_dev_path(char *dev, char *path, size_t path_len)
+{
+	unsigned int i;
+	struct stat st;
+
+	for (i = 0; i < ARRAY_SIZE(dev_variants); i++) {
+		snprintf(path, path_len, dev_variants[i], dev);
+
+		if (stat(path, &st) == 0 && S_ISBLK(st.st_mode))
+			return 0;
+	}
+
+	return 1;
+}
+
 int tst_find_free_loopdev(char *path, size_t path_len)
 {
 	int ctl_fd, dev_fd, rc, i;
@@ -511,7 +531,7 @@ unsigned long tst_dev_bytes_written(const char *dev)
 }
 
 __attribute__((nonnull))
-void tst_find_backing_dev(const char *path, char *dev)
+void tst_find_backing_dev(const char *path, char *dev, size_t dev_len)
 {
 	struct stat buf;
 	struct btrfs_ioctl_fs_info_args args = {0};
@@ -574,7 +594,7 @@ void tst_find_backing_dev(const char *path, char *dev)
 			sprintf(uevent_path, "%s/%s/uevent",
 				bdev_path, d->d_name);
 		} else {
-			tst_brkm(TBROK | TERRNO, NULL, "No backining device found while looking in %s.", bdev_path);
+			tst_brkm(TBROK | TERRNO, NULL, "No backing device found while looking in %s.", bdev_path);
 		}
 
 		if (SAFE_READDIR(NULL, dir))
@@ -590,17 +610,12 @@ void tst_find_backing_dev(const char *path, char *dev)
 	if (!access(uevent_path, R_OK)) {
 		FILE_LINES_SCANF(NULL, uevent_path, "DEVNAME=%s", dev_name);
 
-		if (dev_name[0])
-			sprintf(dev, "/dev/%s", dev_name);
+		if (!dev_name[0] || set_dev_path(dev_name, dev, dev_len) != 0)
+			tst_brkm(TBROK, NULL, "Could not stat backing device %s", dev);
+
 	} else {
 		tst_brkm(TBROK, NULL, "uevent file (%s) access failed", uevent_path);
 	}
-
-	if (stat(dev, &buf) < 0)
-		tst_brkm(TWARN | TERRNO, NULL, "stat(%s) failed", dev);
-
-	if (S_ISBLK(buf.st_mode) != 1)
-		tst_brkm(TCONF, NULL, "dev(%s) isn't a block dev", dev);
 }
 
 void tst_stat_mount_dev(const char *const mnt_path, struct stat *const st)
@@ -643,7 +658,7 @@ int tst_dev_block_size(const char *path)
 	int size;
 	char dev_name[PATH_MAX];
 
-	tst_find_backing_dev(path, dev_name);
+	tst_find_backing_dev(path, dev_name, sizeof(dev_name));
 
 	fd = SAFE_OPEN(NULL, dev_name, O_RDONLY);
 	SAFE_IOCTL(NULL, fd, BLKSSZGET, &size);
diff --git a/testcases/kernel/syscalls/ioctl/ioctl_loop05.c b/testcases/kernel/syscalls/ioctl/ioctl_loop05.c
index b4427f331..3a5d5afef 100644
--- a/testcases/kernel/syscalls/ioctl/ioctl_loop05.c
+++ b/testcases/kernel/syscalls/ioctl/ioctl_loop05.c
@@ -125,7 +125,7 @@ static void setup(void)
 	 *   needn't transform transfer.
 	 */
 	sprintf(backing_file_path, "%s/test.img", tst_get_tmpdir());
-	tst_find_backing_dev(backing_file_path, bd_path);
+	tst_find_backing_dev(backing_file_path, bd_path, sizeof(bd_path));
 	block_devfd = SAFE_OPEN(bd_path, O_RDWR);
 	SAFE_IOCTL(block_devfd, BLKSSZGET, &logical_block_size);
 	tst_res(TINFO, "backing dev(%s) logical_block_size is %d", bd_path, logical_block_size);
-- 
2.40.0.rc1.284.g88254d51c5-goog


-- 
Mailing list info: https://lists.linux.it/listinfo/ltp

Re: [LTP] [PATCH v2 3/3] tst_find_backing_dev: Also check /dev/block/ for backing device

[Petr Vorel]

Hi Edward,

> On Android, the backing devices are created in /dev/block/ and will not
> be found using the method added in e1b1ae66b240 ("tst_find_backing_dev:
> Get dev name from /sys/dev/block/*/uevent").  Adds a check for
> /dev/block/%s as well as /dev/%s.

Fixes: e1b1ae66b240 ("tst_find_backing_dev: Get dev name from /sys/dev/block/*/uevent")

Kind regards,
Petr

-- 
Mailing list info: https://lists.linux.it/listinfo/ltp

Re: [LTP] [PATCH v2 0/3] tst_device.c: Handle Android path for backing device

[Petr Vorel]

> e1b1ae66b240 ("tst_find_backing_dev: Get dev name from
> /sys/dev/block/*/uevent") added a hardcoded path to check for the
> backing device in /dev.  On Android, it needs to check /dev/block.

Cc the original autor.

I guess both problems were unnoticed because they are kind of corner cases
(minimal Linux and Android). But sure we're happily accept patches to fix them.

Kind regards,
Petr

> The set_dev_path function was renamed to set_dev_loop_path and its
> return value was changed from 1 on success to 0 on success to be more
> consistent with other functions.  A check was added to
> tst_find_free_loopdev in the event that the loop device is not ready
> yet, which appears to happen occasionally on Android.

> v1->v2:
> Changed the function signature of tst_find_backing_dev to add the length
> of the path string.  Updated all references for this function to include
> the added parameter.

> Edward Liaw (3):
>   tst_device.c: Use PATH_MAX more consistently
>   set_dev_loop_path: Refactor set_dev_path and check return value
>   tst_find_backing_dev: Also check /dev/block/ for backing device

>  doc/c-test-api.txt                            |  2 +-
>  include/tst_device.h                          |  3 +-
>  lib/newlib_tests/tst_device.c                 |  2 +-
>  lib/tst_device.c                              | 59 ++++++++++++-------
>  .../kernel/syscalls/ioctl/ioctl_loop05.c      |  2 +-
>  5 files changed, 42 insertions(+), 26 deletions(-)

-- 
Mailing list info: https://lists.linux.it/listinfo/ltp