Skip to contentSkip to conversation

part: add partition number to disk_partition_t

PATCH New Stephen Warren 4 participants 9 messages Oct 10, 2012 – Oct 17, 2012 VFAT/FAT/msdos filesystem

lore
Share & tools
lore
#0 lore raw

PATCH v3 1/4 part: add partition number to disk_partition_t

From: Stephen Warren <[email protected]>

The FAT filesystem code knows which partition ID it is operating on. Currently, this is passed to fat_register_device() as a parameter. In order to convert FAT to the more standardized fat_set_blk_dev(), the information needs to come from somewhere else, and the partition definition structure is the logical place.

---
v3: Set info->part in get_device_and_partition() when returning "whole
disk" partition.
v2: No change.

---
 disk/part.c    |    2 ++
 include/part.h |    1 +
 2 files changed, 3 insertions(+), 0 deletions(-)
disk/part.c +2 -0
index 8ba3cde..6d8135d 100644
@@ -391,6 +391,7 @@ int get_partition_info(block_dev_desc_t *dev_desc, int part
 	defined(CONFIG_MMC) || \
 	defined(CONFIG_SYSTEMACE)
 
+	info->part = part;
 #ifdef CONFIG_PARTITION_UUIDS
 	/* The common case is no UUID support */
 	info->uuid[0] = 0;
@@ -557,6 +558,7 @@ int get_device_and_partition(const char *ifname, const char *dev_part_str,
 			goto cleanup;
 		}
 
+		info->part = 0;
 		info->start = 0;
 		info->size = (*dev_desc)->lba;
 		info->blksz = (*dev_desc)->blksz;
include/part.h +1 -0
index 27ea283..ebdebd8 100644
@@ -88,6 +88,7 @@ typedef struct block_dev_desc {
 #define DEV_TYPE_OPDISK		0x07	/* optical disk */
 
 typedef struct disk_partition {
+	int	part;		/* Partition number			*/
 	ulong	start;		/* # of first block in partition	*/
 	ulong	size;		/* number of blocks in partition	*/
 	ulong	blksz;		/* block size in bytes			*/
Signature · 1 line
1.7.0.4
#1 lore raw

PATCH v3 2/4 FAT: make use of disk_partition_t.part

From: Stephen Warren <[email protected]>

This removes the standalone cur_part_nr variable, opening the way to replacing fat_register_device() with fat_set_blk_dev().

Note that when get_partition_info() fails and we use the entire disk, the correct partition number is 0 (whole disk) not 1 (first partition), so that change is also made here.

An alternative to this (and the previous) patch might be to simply remove the partition number from the printf(). I don't know how attached people are to it.

---
v3: No change.
v2: No change.

---
 fs/fat/fat.c |    6 ++----
 1 files changed, 2 insertions(+), 4 deletions(-)
fs/fat/fat.c +2 -4
index 80156c8..1e0d2a3 100644
@@ -46,7 +46,6 @@ static void downcase(char *str)
 }
 
 static block_dev_desc_t *cur_dev;
-static unsigned int cur_part_nr;
 static disk_partition_t cur_part_info;
 
 #define DOS_BOOT_MAGIC_OFFSET	0x1fe
@@ -79,7 +78,6 @@ int fat_register_device(block_dev_desc_t * dev_desc, int part_no)
 	/* Read the partition table, if present */
 	if (!get_partition_info(dev_desc, part_no, &cur_part_info)) {
 		cur_dev = dev_desc;
-		cur_part_nr = part_no;
 	}
 #endif
 
@@ -92,7 +90,7 @@ int fat_register_device(block_dev_desc_t * dev_desc, int part_no)
 		}
 
 		cur_dev = dev_desc;
-		cur_part_nr = 1;
+		cur_part_info.part = 0;
 		cur_part_info.start = 0;
 		cur_part_info.size = dev_desc->lba;
 		cur_part_info.blksz = dev_desc->blksz;
@@ -1235,7 +1233,7 @@ int file_fat_detectfs(void)
 	vol_label[11] = '\0';
 	volinfo.fs_type[5] = '\0';
 
-	printf("Partition %d: Filesystem: %s \"%s\"\n", cur_part_nr,
+	printf("Partition %d: Filesystem: %s \"%s\"\n", cur_part_info.part,
 		volinfo.fs_type, vol_label);
 
 	return 0;
Signature · 1 line
1.7.0.4
#2 lore raw

PATCH v3 2/4 FAT: make use of disk_partition_t.part

Hi

Quoted 9 lines · On Wednesday 10 October 2012 12:14:00 Stephen Warren wrote:

From: Stephen Warren <[email protected]>

This removes the standalone cur_part_nr variable, opening the way to replacing fat_register_device() with fat_set_blk_dev().

Note that when get_partition_info() fails and we use the entire disk, the correct partition number is 0 (whole disk) not 1 (first partition), so that change is also made here.

Quoted 5 lines

An alternative to this (and the previous) patch might be to simply remove the partition number from the printf(). I don't know how attached people are to it.

Just as a heads up, in the DM any difference between a partition and a whole block device (also between different interfaces for disks) is hidden from any user code (code other than the one keeping track of partitions/disks, that only uses such information to determine whether to scan for partitions), you only get some object that can read/write blocks if you ask it nicely, and you have to make do with that (if you need more then you're probably doing something wrong).
As a result, there is no notion of partition number, and the string you are patching up here (along with many others, due to unification of disk interfaces) is changed.

Best Regards
Pavel Herrmann

#3 lore raw

PATCH v3 2/4 FAT: make use of disk_partition_t.part

Quoted 18 lines · Pavel Herrmann

Hi

Quoted 9 lines · On Wednesday 10 October 2012 12:14:00 Stephen Warren wrote:

From: Stephen Warren <[email protected]>

This removes the standalone cur_part_nr variable, opening the way to replacing fat_register_device() with fat_set_blk_dev().

Note that when get_partition_info() fails and we use the entire disk, the correct partition number is 0 (whole disk) not 1 (first partition), so that change is also made here.

Quoted 5 lines

An alternative to this (and the previous) patch might be to simply remove the partition number from the printf(). I don't know how attached people are to it.

Quoted 10 lines

Just as a heads up, in the DM any difference between a partition and a whole block device (also between different interfaces for disks) is hidden from any user code (code other than the one keeping track of partitions/disks, that only uses such information to determine whether to scan for partitions), you only get some object that can read/write blocks if you ask it nicely, and you have to make do with that (if you need more then you're probably doing something wrong).
As a result, there is no notion of partition number, and the string you are patching up here (along with many others, due to unification of disk interfaces) is changed.

OK, so do you think it'd be better right now to drop patches 1 and 2 in this series, and just remove the partition number from fat.c's printf() call? That'd certainly be simple to do.

#4 lore raw

PATCH v3 2/4 FAT: make use of disk_partition_t.part

Quoted 31 lines · On Monday 15 of October 2012 10:40:25 Stephen Warren wrote:
Quoted 18 lines · Pavel Herrmann

Hi

Quoted 9 lines · On Wednesday 10 October 2012 12:14:00 Stephen Warren wrote:

From: Stephen Warren <[email protected]>

This removes the standalone cur_part_nr variable, opening the way to replacing fat_register_device() with fat_set_blk_dev().

Note that when get_partition_info() fails and we use the entire disk, the correct partition number is 0 (whole disk) not 1 (first partition), so that change is also made here.

Quoted 5 lines

An alternative to this (and the previous) patch might be to simply remove the partition number from the printf(). I don't know how attached people are to it.

Quoted 11 lines

Just as a heads up, in the DM any difference between a partition and a whole block device (also between different interfaces for disks) is hidden from any user code (code other than the one keeping track of partitions/disks, that only uses such information to determine whether to scan for partitions), you only get some object that can read/write blocks if you ask it nicely, and you have to make do with that (if you need more then you're probably doing something wrong).
As a result, there is no notion of partition number, and the string you are
patching up here (along with many others, due to unification of disk interfaces) is changed.

Quoted 3 lines

OK, so do you think it'd be better right now to drop patches 1 and 2 in this series, and just remove the partition number from fat.c's printf() call? That'd certainly be simple to do.

Well, in my case I have done a broader abstraction, that could be used for non-continuous partitions as well (think LVM) with minimal effort (think extend identifier structure used for searching to more than interface:number:partnumber, no changes in FS code), and partition number loses any meaning in that context.
Whether dropping the number now is an acceptable change would be up to Tom Rini, I would vote for it though, if that meant anything around here.

Pavel Herrmann

#5 lore raw

PATCH v3 2/4 FAT: make use of disk_partition_t.part

Quoted 36 lines · Pavel Herrmann
Quoted 31 lines · On Monday 15 of October 2012 10:40:25 Stephen Warren wrote:
Quoted 18 lines · Pavel Herrmann

Hi

Quoted 9 lines · On Wednesday 10 October 2012 12:14:00 Stephen Warren wrote:

From: Stephen Warren <[email protected]>

This removes the standalone cur_part_nr variable, opening the way to replacing fat_register_device() with fat_set_blk_dev().

Note that when get_partition_info() fails and we use the entire disk, the correct partition number is 0 (whole disk) not 1 (first partition), so that change is also made here.

Quoted 5 lines

An alternative to this (and the previous) patch might be to simply remove the partition number from the printf(). I don't know how attached people are to it.

Quoted 11 lines

Just as a heads up, in the DM any difference between a partition and a whole block device (also between different interfaces for disks) is hidden from any user code (code other than the one keeping track of partitions/disks, that only uses such information to determine whether to scan for partitions), you only get some object that can read/write blocks if you ask it nicely, and you have to make do with that (if you need more then you're probably doing something wrong).
As a result, there is no notion of partition number, and the string you are
patching up here (along with many others, due to unification of disk interfaces) is changed.

Quoted 3 lines

OK, so do you think it'd be better right now to drop patches 1 and 2 in this series, and just remove the partition number from fat.c's printf() call? That'd certainly be simple to do.

Quoted 7 lines

Well, in my case I have done a broader abstraction, that could be used for non-continuous partitions as well (think LVM) with minimal effort (think extend identifier structure used for searching to more than interface:number:partnumber, no changes in FS code), and partition number loses any meaning in that context.
Whether dropping the number now is an acceptable change would be up to Tom Rini, I would vote for it though, if that meant anything around here.

OK, lets rework this series as suggested here to make the DM work a little easier. Drop the print instead.

Signature · 8 lines
Tom
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 836 bytes
Desc: Digital signature
URL: <http://lists.denx.de/pipermail/u-boot/attachments/20121017/712e7b43/attachment.pgp>
#6 lore raw

PATCH v3 3/4 FAT: initialize all fields in cur_part_info, simplify init

In reply to Stephen Warren

From: Stephen Warren <[email protected]>

cur_part_info.{name,type} are strings. So, we don't need to memset() the entire thing, just put the NULL-termination in the first byte.

Add missing initialization of the bootable and uuid fields.

None of these fields are actually used by fat.c. However, since it stores the entire disk_partition_t, we should make sure that all fields are valid.

---
v3: New patch.

---
 fs/fat/fat.c |    8 ++++++--
 1 files changed, 6 insertions(+), 2 deletions(-)
fs/fat/fat.c +6 -2
index 1e0d2a3..2863c4b 100644
@@ -94,8 +94,12 @@ int fat_register_device(block_dev_desc_t * dev_desc, int part_no)
 		cur_part_info.start = 0;
 		cur_part_info.size = dev_desc->lba;
 		cur_part_info.blksz = dev_desc->blksz;
-		memset(cur_part_info.name, 0, sizeof(cur_part_info.name));
-		memset(cur_part_info.type, 0, sizeof(cur_part_info.type));
+		cur_part_info.name[0] = 0;
+		cur_part_info.type[0] = 0;
+		cur_part_info.bootable = 0;
+#ifdef CONFIG_PARTITION_UUIDS
+		cur_part_info.uuid[0] = 0;
+#endif
 	}
 
 	/* Make sure it has a valid FAT header */
Signature · 1 line
1.7.0.4
#7 lore raw

PATCH v3 4/4 FAT: implement fat_set_blk_dev(), convert cmd_fat.c

In reply to Stephen Warren

From: Stephen Warren <[email protected]>

This makes the FAT filesystem API more consistent with other block-based filesystems. If in the future standard multi-filesystem commands such as "ls" or "load" are implemented, having FAT work the same way as other filesystems will be necessary.

Convert cmd_fat.c to the new API, so the code looks more like other files implementing the same commands for other filesystems.

---
v3: Rebase on new fat.c cleanup patch.
v2: Fix inverted test of get_partition_info() result in fat_register_device().

---
 common/cmd_fat.c |    8 +++---
 fs/fat/fat.c     |   69 +++++++++++++++++++++++++-----------------------------
 include/fat.h    |    1 +
 3 files changed, 37 insertions(+), 41 deletions(-)
common/cmd_fat.c +4 -4
index 5a5698b..c38302d 100644
@@ -55,7 +55,7 @@ int do_fat_fsload (cmd_tbl_t *cmdtp, int flag, int argc, char * const argv[])
 		return 1;
 
 	dev = dev_desc->dev;
-	if (fat_register_device(dev_desc,part)!=0) {
+	if (fat_set_blk_dev(dev_desc, &info) != 0) {
 		printf("\n** Unable to use %s %d:%d for fatload **\n",
 			argv[1], dev, part);
 		return 1;
@@ -111,7 +111,7 @@ int do_fat_ls (cmd_tbl_t *cmdtp, int flag, int argc, char * const argv[])
 		return 1;
 
 	dev = dev_desc->dev;
-	if (fat_register_device(dev_desc,part)!=0) {
+	if (fat_set_blk_dev(dev_desc, &info) != 0) {
 		printf("\n** Unable to use %s %d:%d for fatls **\n",
 			argv[1], dev, part);
 		return 1;
@@ -149,7 +149,7 @@ int do_fat_fsinfo (cmd_tbl_t *cmdtp, int flag, int argc, char * const argv[])
 		return 1;
 
 	dev = dev_desc->dev;
-	if (fat_register_device(dev_desc,part)!=0) {
+	if (fat_set_blk_dev(dev_desc, &info) != 0) {
 		printf("\n** Unable to use %s %d:%d for fatinfo **\n",
 			argv[1], dev, part);
 		return 1;
@@ -185,7 +185,7 @@ static int do_fat_fswrite(cmd_tbl_t *cmdtp, int flag,
 
 	dev = dev_desc->dev;
 
-	if (fat_register_device(dev_desc, part) != 0) {
+	if (fat_set_blk_dev(dev_desc, &info) != 0) {
 		printf("\n** Unable to use %s %d:%d for fatwrite **\n",
 			argv[1], dev, part);
 		return 1;
fs/fat/fat.c +32 -37

84 lines not on this page · raw message

include/fat.h +1 -0
index cc85b06..706cd7a 100644
@@ -212,6 +212,7 @@ long file_fat_read_at(const char *filename, unsigned long pos, void *buffer,
 		      unsigned long maxsize);
 long file_fat_read(const char *filename, void *buffer, unsigned long maxsize);
 const char *file_getfsname(int idx);
+int fat_set_blk_dev(block_dev_desc_t *rbdd, disk_partition_t *info);
 int fat_register_device(block_dev_desc_t *dev_desc, int part_no);
 
 int file_fat_write(const char *filename, void *buffer, unsigned long maxsize);
Signature · 1 line
1.7.0.4
#8 lore raw

PATCH v3 4/4 FAT: implement fat_set_blk_dev(), convert cmd_fat.c

Quoted 175 lines · Stephen Warren

This makes the FAT filesystem API more consistent with other block-based
filesystems. If in the future standard multi-filesystem commands such as
"ls" or "load" are implemented, having FAT work the same way as other filesystems will be necessary.

Convert cmd_fat.c to the new API, so the code looks more like other files
implementing the same commands for other filesystems.

---
v3: Rebase on new fat.c cleanup patch.
v2: Fix inverted test of get_partition_info() result in fat_register_device().

---
 common/cmd_fat.c |    8 +++---
 fs/fat/fat.c     |   69
 +++++++++++++++++++++++++-----------------------------
 include/fat.h    |    1 +
 3 files changed, 37 insertions(+), 41 deletions(-)
common/cmd_fat.c +1 -1
index 5a5698b..c38302d 100644
@@ -55,7 +55,7 @@ int do_fat_fsload (cmd_tbl_t *cmdtp, int flag, int
 argc, char * const argv[])
 		return 1;
 
 	dev = dev_desc->dev;
-	if (fat_register_device(dev_desc,part)!=0) {
+	if (fat_set_blk_dev(dev_desc, &info) != 0) {
 		printf("\n** Unable to use %s %d:%d for fatload **\n",
 			argv[1], dev, part);
common/cmd_fat.c +1 -1
 		return 1;
@@ -111,7 +111,7 @@ int do_fat_ls (cmd_tbl_t *cmdtp, int flag, int
 argc, char * const argv[])
 		return 1;
 
 	dev = dev_desc->dev;
-	if (fat_register_device(dev_desc,part)!=0) {
+	if (fat_set_blk_dev(dev_desc, &info) != 0) {
 		printf("\n** Unable to use %s %d:%d for fatls **\n",
 			argv[1], dev, part);
common/cmd_fat.c +1 -1
 		return 1;
@@ -149,7 +149,7 @@ int do_fat_fsinfo (cmd_tbl_t *cmdtp, int flag,
 int argc, char * const argv[])
 		return 1;
 
 	dev = dev_desc->dev;
-	if (fat_register_device(dev_desc,part)!=0) {
+	if (fat_set_blk_dev(dev_desc, &info) != 0) {
 		printf("\n** Unable to use %s %d:%d for fatinfo **\n",
 			argv[1], dev, part);
common/cmd_fat.c +1 -1
 		return 1;
@@ -185,7 +185,7 @@ static int do_fat_fswrite(cmd_tbl_t *cmdtp, int
 flag,
 
 	dev = dev_desc->dev;
 
-	if (fat_register_device(dev_desc, part) != 0) {
+	if (fat_set_blk_dev(dev_desc, &info) != 0) {
 		printf("\n** Unable to use %s %d:%d for fatwrite **\n",
 			argv[1], dev, part);
common/cmd_fat.c +0 -0
 		return 1;
fs/fat/fat.c +3 -37
index 2863c4b..a6fc07c 100644
@@ -61,46 +61,12 @@ static int disk_read(__u32 block, __u32
 nr_blocks, void *buf)
 			cur_part_info.start + block, nr_blocks, buf);
 }
 
-int fat_register_device(block_dev_desc_t * dev_desc, int part_no)
+int fat_set_blk_dev(block_dev_desc_t *dev_desc, disk_partition_t
 *info)
 {
 	ALLOC_CACHE_ALIGN_BUFFER(unsigned char, buffer, dev_desc->blksz);
 
-	/* First close any currently found FAT filesystem */
-	cur_dev = NULL;
-
-#if (defined(CONFIG_CMD_IDE) || \
-     defined(CONFIG_CMD_SATA) || \
-     defined(CONFIG_CMD_SCSI) || \
-     defined(CONFIG_CMD_USB) || \
-     defined(CONFIG_MMC) || \
-     defined(CONFIG_SYSTEMACE) )
-
-	/* Read the partition table, if present */
-	if (!get_partition_info(dev_desc, part_no, &cur_part_info)) {
-		cur_dev = dev_desc;
-	}
-#endif
-
-	/* Otherwise it might be a superfloppy (whole-disk FAT filesystem)
 */
-	if (!cur_dev) {
-		if (part_no != 0) {
-			printf("** Partition %d not valid on device %d **\n",
-					part_no, dev_desc->dev);
-			return -1;
-		}
-
-		cur_dev = dev_desc;
-		cur_part_info.part = 0;
-		cur_part_info.start = 0;
-		cur_part_info.size = dev_desc->lba;
-		cur_part_info.blksz = dev_desc->blksz;
-		cur_part_info.name[0] = 0;
-		cur_part_info.type[0] = 0;
-		cur_part_info.bootable = 0;
-#ifdef CONFIG_PARTITION_UUIDS
-		cur_part_info.uuid[0] = 0;
-#endif
-	}
+	cur_dev = dev_desc;
+	cur_part_info = *info;
fs/fat/fat.c +29 -0
 	/* Make sure it has a valid FAT header */
 	if (disk_read(0, 1, buffer) != 1) {
@@ -124,6 +90,35 @@ int fat_register_device(block_dev_desc_t *
 dev_desc, int part_no)
 	return -1;
 }
 
+int fat_register_device(block_dev_desc_t *dev_desc, int part_no)
+{
+	disk_partition_t info;
+
+	/* First close any currently found FAT filesystem */
+	cur_dev = NULL;
+
+	/* Read the partition table, if present */
+	if (get_partition_info(dev_desc, part_no, &info)) {
+		if (part_no != 0) {
+			printf("** Partition %d not valid on device %d **\n",
+					part_no, dev_desc->dev);
+			return -1;
+		}
+
+		info.part = 0;
+		info.start = 0;
+		info.size = dev_desc->lba;
+		info.blksz = dev_desc->blksz;
+		info.name[0] = 0;
+		info.type[0] = 0;
+		info.bootable = 0;
+#ifdef CONFIG_PARTITION_UUIDS
+		info.uuid[0] = 0;
+#endif
+	}
+
+	return fat_set_blk_dev(dev_desc, &info);
+}
 
 /*

* Get the first occurence of a directory delimiter ('/' or '\') in a string.

include/fat.h +1 -0
index cc85b06..706cd7a 100644
@@ -212,6 +212,7 @@ long file_fat_read_at(const char *filename,
 unsigned long pos, void *buffer,
 		      unsigned long maxsize);
 long file_fat_read(const char *filename, void *buffer, unsigned long
 maxsize);
 const char *file_getfsname(int idx);
+int fat_set_blk_dev(block_dev_desc_t *rbdd, disk_partition_t *info);
 int fat_register_device(block_dev_desc_t *dev_desc, int part_no);
Quoted 2 lines

int file_fat_write(const char *filename, void *buffer, unsigned long maxsize);

For the V3 of this series:

Best regards,
Beno?t