Skip to contentSkip to conversation

common: filetype: do not redetect MBR-type devices as a FAT-type

PATCH In review Peter Mamonov 4 participants 7 messages Nov 9, 2015 – Nov 11, 2015

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

PATCH 1/1 common: filetype: do not redetect MBR-type devices as a FAT-type

Deleted pieces of code detect MBR-containig device as a FAT-type device, if it's first partition contains a FAT filesystem. This behaviour enabled one to mount the FAT FS which is either directly on the device (disk0) or on the first partition (disk0.0) using the same command:
mount /dev/disk0 /fat
However, the desired behaviour can be reached with a:

    mount /dev/disk0 /fat || mount /dev/disk0.0 /fat || echo "Mounting failed"
---
 common/filetype.c | 30 ------------------------------
 1 file changed, 30 deletions(-)
common/filetype.c +0 -30
index c59441d..f9e03f7 100644
@@ -345,21 +345,6 @@ enum filetype file_name_detect_type(const char *filename)
 
 	type = file_detect_type(buf, ret);
 
-	if (type == filetype_mbr) {
-		/*
-		 * Get the first partition start sector
-		 * and check for FAT in it
-		 */
-		is_fat_or_mbr(buf, &bootsec);
-		ret = lseek(fd, (bootsec) * 512, SEEK_SET);
-		if (ret < 0)
-			goto err_out;
-		ret = read(fd, buf, 512);
-		if (ret < 0)
-			goto err_out;
-		type = is_fat_or_mbr((u8 *)buf, NULL);
-	}
-
 err_out:
 	close(fd);
 	free(buf);
@@ -385,21 +370,6 @@ enum filetype cdev_detect_type(const char *name)
 
 	type = file_detect_type(buf, ret);
 
-	if (type == filetype_mbr) {
-		unsigned long bootsec;
-		/*
-		 * Get the first partition start sector
-		 * and check for FAT in it
-		 */
-		is_fat_or_mbr(buf, &bootsec);
-
-		ret = cdev_read(cdev, buf, 512, bootsec * 512, 0);
-		if (ret < 0)
-			goto err_out;
-
-		type = is_fat_or_mbr((u8 *)buf, NULL);
-	}
-
 err_out:
 	free(buf);
 	cdev_close(cdev);
Signature · 7 lines
2.1.4


_______________________________________________
barebox mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/barebox
#1 lore raw
Quoted 10 lines · On Mon, 9 Nov 2015 13:32:51 +0300 Peter Mamonov <[email protected]> wrote:

Deleted pieces of code detect MBR-containig device as a FAT-type device, if it's first partition contains a FAT filesystem. This behaviour enabled one to mount the FAT FS which is either directly on the device (disk0) or on the first partition (disk0.0) using the same command:
mount /dev/disk0 /fat
However, the desired behaviour can be reached with a:

    mount /dev/disk0 /fat || mount /dev/disk0.0 /fat || echo "Mounting failed"

^^^^^^^^^^

Ad memorandum. Sascha's ack can be found here: http://lists.infradead.org/pipermail/barebox/2015-October/024952.html

Quoted 60 lines
---
 common/filetype.c | 30 ------------------------------
 1 file changed, 30 deletions(-)
common/filetype.c +0 -30
index c59441d..f9e03f7 100644
@@ -345,21 +345,6 @@ enum filetype file_name_detect_type(const char *filename)
 
 	type = file_detect_type(buf, ret);
 
-	if (type == filetype_mbr) {
-		/*
-		 * Get the first partition start sector
-		 * and check for FAT in it
-		 */
-		is_fat_or_mbr(buf, &bootsec);
-		ret = lseek(fd, (bootsec) * 512, SEEK_SET);
-		if (ret < 0)
-			goto err_out;
-		ret = read(fd, buf, 512);
-		if (ret < 0)
-			goto err_out;
-		type = is_fat_or_mbr((u8 *)buf, NULL);
-	}
-
 err_out:
 	close(fd);
 	free(buf);
@@ -385,21 +370,6 @@ enum filetype cdev_detect_type(const char *name)
 
 	type = file_detect_type(buf, ret);
 
-	if (type == filetype_mbr) {
-		unsigned long bootsec;
-		/*
-		 * Get the first partition start sector
-		 * and check for FAT in it
-		 */
-		is_fat_or_mbr(buf, &bootsec);
-
-		ret = cdev_read(cdev, buf, 512, bootsec * 512, 0);
-		if (ret < 0)
-			goto err_out;
-
-		type = is_fat_or_mbr((u8 *)buf, NULL);
-	}
-
 err_out:
 	free(buf);
 	cdev_close(cdev);
Signature · 7 lines
2.1.4


_______________________________________________
barebox mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/barebox
Signature · 8 lines
-- 
Best regards,
  Antony Pavlov

_______________________________________________
barebox mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/barebox
#2 lore raw
In reply to Peter Mamonov
Quoted 7 lines · Peter Mamonov

Deleted pieces of code detect MBR-containig device as a FAT-type device, if it's first partition contains a FAT filesystem. This behaviour enabled one to mount the FAT FS which is either directly on the device (disk0) or on the first partition (disk0.0) using the same command:
mount /dev/disk0 /fat
However, the desired behaviour can be reached with a:

    mount /dev/disk0 /fat || mount /dev/disk0.0 /fat || echo "Mounting failed"

The in flash env for omap is found by mounting a FAT filesystem and reading an env file from it. The code that does this uses the disk as the device, rather than the partition. Which I thought was odd and wondered why/if that worked correctly....

Anyway, this change would break that code. I suspect than instead of coding the disk0 || disk0.0 method it could just mount disk0.0. Because the code does this (edited):

partname = asprintf("/dev/%s.0", diskdev);
ret = stat(partname, &s);
if (ret) {
pr_err("Failed to load environment: no device '%s'\n", diskdev); return 0;
}
ret = mount(diskdev, "fat", "/boot", NULL);

If disk0.0 didn't exist then it wouldn't try to mount it in the 1st place, so the ability of trying both disk0 and disk0.0 wouldn't be used.
_______________________________________________
barebox mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/barebox

#3 lore raw
Quoted 18 lines · Trent Piepho
Quoted 7 lines · Peter Mamonov

Deleted pieces of code detect MBR-containig device as a FAT-type device, if it's first partition contains a FAT filesystem. This behaviour enabled one to mount the FAT FS which is either directly on the device (disk0) or on the first partition (disk0.0) using the same command:
mount /dev/disk0 /fat
However, the desired behaviour can be reached with a:

    mount /dev/disk0 /fat || mount /dev/disk0.0 /fat || echo "Mounting failed"

The in flash env for omap is found by mounting a FAT filesystem and reading an env file from it. The code that does this uses the disk as the device, rather than the partition. Which I thought was odd and wondered why/if that worked correctly....

Anyway, this change would break that code. I suspect than instead of coding the disk0 || disk0.0 method it could just mount disk0.0. Because the code does this (edited):

Quoted 7 lines

partname = asprintf("/dev/%s.0", diskdev);
ret = stat(partname, &s);
if (ret) {
pr_err("Failed to load environment: no device '%s'\n", diskdev); return 0;
}
ret = mount(diskdev, "fat", "/boot", NULL);

This is really odd. The code first tests if the partition exists and tries to mount the whole disk afterwards. git blame points to:

commit be322768f04603df371c9c1f08b9621690dd74c6

Author: Sascha Hauer <[email protected]>
Date:   Sat Aug 24 12:48:50 2013 +0200

ARM: omap: Allow to set mmc devname used for booting

I think what we should do here is in the following patch. With this Trents patch can be applied as-is. Thanks for noting, Peter.

Sascha

---------------------------8<---------------------------

From 98fb34f2aa271138eab28a36aec733fe6572339c Mon Sep 17 00:00:00 2001
From: Sascha Hauer <[email protected]>
Date: Tue, 10 Nov 2015 08:58:16 +0100
Subject: [PATCH] ARM: omap: Use correct device to mount on /boot

The current code tests if a partition (i.e. disk0.0) exists and instead of mounting boot from this partition it uses the whole device (disk0). This only works because the the FAT code accepts a MBR as input and automatically skips it.

Let the code use the partition to mount /boot instead as it was intended. We don't have to stat() the partition device, since this error will be caught by mount() anyway, so remove the unnecessary stat().

---
 arch/arm/mach-omap/omap_generic.c | 21 +++++++--------------
 1 file changed, 7 insertions(+), 14 deletions(-)
arch/arm/mach-omap/omap_generic.c +7 -14
index 071a1bf..4e26c6b 100644
@@ -111,7 +111,6 @@ const char *omap_get_bootmmc_devname(void)
 #define ENV_PATH "/boot/barebox.env"
 static int omap_env_init(void)
 {
-	struct stat s;
 	char *partname;
 	const char *diskdev;
 	int ret;
@@ -128,25 +127,19 @@ static int omap_env_init(void)
 
 	partname = asprintf("/dev/%s.0", diskdev);
 
-	ret = stat(partname, &s);
-
-	free(partname);
-
-	if (ret) {
-		pr_err("Failed to load environment: no device '%s'\n", diskdev);
-		return 0;
-	}
-
 	mkdir("/boot", 0666);
-	ret = mount(diskdev, "fat", "/boot", NULL);
+	ret = mount(partname, "fat", "/boot", NULL);
 	if (ret) {
-		pr_err("Failed to load environment: mount %s failed (%d)\n", diskdev, ret);
-		return 0;
+		pr_err("Failed to load environment: mount %s failed (%d)\n", partname, ret);
+		goto out;
 	}
 
-	pr_debug("Loading default env from %s on device %s\n", ENV_PATH, diskdev);
+	pr_debug("Loading default env from %s on device %s\n", ENV_PATH, partname);
 	default_environment_path_set(ENV_PATH);
 
+out:
+	free(partname);
+
 	return 0;
 }
 late_initcall(omap_env_init);
Signature · 12 lines
2.6.1

-- 
Pengutronix e.K.                           |                             |
Industrial Linux Solutions                 | http://www.pengutronix.de/  |
Peiner Str. 6-8, 31137 Hildesheim, Germany | Phone: +49-5121-206917-0    |
Amtsgericht Hildesheim, HRA 2686           | Fax:   +49-5121-206917-5555 |

_______________________________________________
barebox mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/barebox
#4 lore raw
Quoted 30 lines · Sascha Hauer
Quoted 18 lines · Trent Piepho
Quoted 7 lines · Peter Mamonov

Deleted pieces of code detect MBR-containig device as a FAT-type device, if it's first partition contains a FAT filesystem. This behaviour enabled one to mount the FAT FS which is either directly on the device (disk0) or on the first partition (disk0.0) using the same command:
mount /dev/disk0 /fat
However, the desired behaviour can be reached with a:

    mount /dev/disk0 /fat || mount /dev/disk0.0 /fat || echo "Mounting failed"

The in flash env for omap is found by mounting a FAT filesystem and reading an env file from it. The code that does this uses the disk as the device, rather than the partition. Which I thought was odd and wondered why/if that worked correctly....

Anyway, this change would break that code. I suspect than instead of coding the disk0 || disk0.0 method it could just mount disk0.0. Because the code does this (edited):

Quoted 7 lines

partname = asprintf("/dev/%s.0", diskdev);
ret = stat(partname, &s);
if (ret) {
pr_err("Failed to load environment: no device '%s'\n", diskdev); return 0;
}
ret = mount(diskdev, "fat", "/boot", NULL);

This is really odd. The code first tests if the partition exists and tries to mount the whole disk afterwards. git blame points to:

Quoted 8 lines

commit be322768f04603df371c9c1f08b9621690dd74c6

Author: Sascha Hauer <[email protected]>
Date:   Sat Aug 24 12:48:50 2013 +0200

ARM: omap: Allow to set mmc devname used for booting

I think what we should do here is in the following patch. With this Trents patch can be applied as-is. Thanks for noting, Peter.

Sorry, I mixed you two up. I mean: With this Peters patch can be applied as-is. Thanks for noting, Trent.

Sascha

Signature · 9 lines
Pengutronix e.K.                           |                             |
Industrial Linux Solutions                 | http://www.pengutronix.de/  |
Peiner Str. 6-8, 31137 Hildesheim, Germany | Phone: +49-5121-206917-0    |
Amtsgericht Hildesheim, HRA 2686           | Fax:   +49-5121-206917-5555 |

_______________________________________________
barebox mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/barebox
#5 lore raw
In reply to Sascha Hauer
Quoted 9 lines · Sascha Hauer

From 98fb34f2aa271138eab28a36aec733fe6572339c Mon Sep 17 00:00:00 2001
From: Sascha Hauer <[email protected]>
Date: Tue, 10 Nov 2015 08:58:16 +0100
Subject: [PATCH] ARM: omap: Use correct device to mount on /boot

The current code tests if a partition (i.e. disk0.0) exists and instead of mounting boot from this partition it uses the whole device (disk0). This only works because the the FAT code accepts a MBR as input and automatically skips it.

I agree, this should fix it.

Quoted 5 lines

Let the code use the partition to mount /boot instead as it was intended. We don't have to stat() the partition device, since this error will be caught by mount() anyway, so remove the unnecessary stat().

There are copies of this code in rpi.c and socfpga/generic.c. They don't suffer from the whole disk flaw, but they do stat() first as well.

_______________________________________________
barebox mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/barebox

#6 lore raw
In reply to Peter Mamonov
Quoted 10 lines · Peter Mamonov

Deleted pieces of code detect MBR-containig device as a FAT-type device, if it's first partition contains a FAT filesystem. This behaviour enabled one to mount the FAT FS which is either directly on the device (disk0) or on the first partition (disk0.0) using the same command:
mount /dev/disk0 /fat
However, the desired behaviour can be reached with a:

    mount /dev/disk0 /fat || mount /dev/disk0.0 /fat || echo "Mounting failed"

Please don't add Acked-by for patches the person has never seen. I agreed to the approach of removing the check here, but I can only ever ack a patch that I have seen.

Applied this with my ack removed.

Sascha

Signature · 9 lines
Pengutronix e.K.                           |                             |
Industrial Linux Solutions                 | http://www.pengutronix.de/  |
Peiner Str. 6-8, 31137 Hildesheim, Germany | Phone: +49-5121-206917-0    |
Amtsgericht Hildesheim, HRA 2686           | Fax:   +49-5121-206917-5555 |

_______________________________________________
barebox mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/barebox