Topology ioctls

Topology ioctls

[Martin K. Petersen]

The original rationale for exporting the topology information via sysfs
was that we intended to support multiple heterogeneous regions within a
block device.  And that fit poorly with an ioctl approach.

However, with a single region per device it is trivial to provide the
topology.  And while mkfs.* will continue to use the libblkid interface,
there are users that would like to get access to this information
without having to traverse sysfs and stitch things together manually.

Example:

#include <stdio.h>
#include <stdlib.h>
#include <stdarg.h>
#include <errno.h>
#include <fcntl.h>
#include <sys/ioctl.h>

#define BLKSSZGET 	_IO(0x12,104)
#define BLKIOMIN	_IO(0x12,120)
#define BLKIOOPT	_IO(0x12,121)
#define BLKALIGNOFF	_IO(0x12,122)
#define BLKPBSZGET	_IO(0x12,123)

static int die(const char *fmt, ...)
{
        int err = errno;
        va_list val;

        va_start(val, fmt);
        vfprintf(stderr, fmt, val);
        if (err != 0)
                fprintf(stderr, ": %s", strerror(err));
        fprintf(stderr, "\n");
        va_end(val);

        exit(EXIT_FAILURE);
}

int main(int argc, char *argv[])
{
	int fd, lbs, pbs, min, opt, align;

	if (argc != 2)
		die("Usage: %s <dev>", argv[0]);

	fd = open(argv[1], O_RDONLY);
	if (fd < 0)
		die("Can't open %s", argv[1]);

	if (ioctl(fd, BLKSSZGET, &lbs) < 0)
		die("Can't get logical block size");

	if (ioctl(fd, BLKPBSZGET, &pbs) < 0)
		die("Can't get physical block size");

	if (ioctl(fd, BLKIOMIN, &min) < 0)
		die("Can't get preferred random I/O size");

	if (ioctl(fd, BLKIOOPT, &opt) < 0)
		die("Can't get preferred sustained I/O size");

	if (ioctl(fd, BLKALIGNOFF, &align) < 0)
		die("Can't get alignment offset");

	printf("%s:\n", argv[1]);
	printf("\tlogical block size: %u\n", lbs);
	printf("\tphysical block size: %u\n", pbs);
	printf("\trandom I/O size: %u\n", min);
	printf("\tsustained I/O size: %u\n", opt);

	if (align == -1)
		printf("\talignment offset: inconsistent\n");
	else
		printf("\talignment offset: %u\n", align);

	exit(EXIT_SUCCESS);
}


Patch:

block: Topology ioctls

Not all users of the topology information want to use libblkid.  Provide
the topology information through bdev ioctls.

Also clarify sector size comments for existing BLK ioctls.

Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

---

diff --git a/block/compat_ioctl.c b/block/compat_ioctl.c
index 7865a34..bcc8bec 100644
--- a/block/compat_ioctl.c
+++ b/block/compat_ioctl.c
@@ -734,6 +734,14 @@ long compat_blkdev_ioctl(struct file *file, unsigned cmd, unsigned long arg)
 	switch (cmd) {
 	case HDIO_GETGEO:
 		return compat_hdio_getgeo(disk, bdev, compat_ptr(arg));
+	case BLKPBSZGET:
+		return compat_put_int(arg, bdev_physical_block_size(bdev));
+	case BLKIOMIN:
+		return compat_put_int(arg, bdev_io_min(bdev));
+	case BLKIOOPT:
+		return compat_put_int(arg, bdev_io_opt(bdev));
+	case BLKALIGNOFF:
+		return compat_put_int(arg, bdev_alignment_offset(bdev));
 	case BLKFLSBUF:
 	case BLKROSET:
 	case BLKDISCARD:
diff --git a/block/ioctl.c b/block/ioctl.c
index d3e6b58..fea6f2c 100644
--- a/block/ioctl.c
+++ b/block/ioctl.c
@@ -263,10 +263,12 @@ int blkdev_ioctl(struct block_device *bdev, fmode_t mode, unsigned cmd,
 		return put_long(arg, (bdi->ra_pages * PAGE_CACHE_SIZE) / 512);
 	case BLKROGET:
 		return put_int(arg, bdev_read_only(bdev) != 0);
-	case BLKBSZGET: /* get the logical block size (cf. BLKSSZGET) */
+	case BLKBSZGET: /* get block device soft block size (cf. BLKSSZGET) */
 		return put_int(arg, block_size(bdev));
-	case BLKSSZGET: /* get block device hardware sector size */
+	case BLKSSZGET: /* get block device logical block size */
 		return put_int(arg, bdev_logical_block_size(bdev));
+	case BLKPBSZGET: /* get block device physical block size */
+		return put_int(arg, bdev_physical_block_size(bdev));
 	case BLKSECTGET:
 		return put_ushort(arg, queue_max_sectors(bdev_get_queue(bdev)));
 	case BLKRASET:
@@ -309,6 +311,12 @@ int blkdev_ioctl(struct block_device *bdev, fmode_t mode, unsigned cmd,
 		return put_ulong(arg, size >> 9);
 	case BLKGETSIZE64:
 		return put_u64(arg, bdev->bd_inode->i_size);
+	case BLKIOMIN:
+		return put_int(arg, bdev_io_min(bdev));
+	case BLKIOOPT:
+		return put_int(arg, bdev_io_opt(bdev));
+	case BLKALIGNOFF:
+		return put_int(arg, bdev_alignment_offset(bdev));
 	case BLKTRACESTART:
 	case BLKTRACESTOP:
 	case BLKTRACESETUP:
diff --git a/include/linux/blkdev.h b/include/linux/blkdev.h
index e23a86c..935bcb0 100644
--- a/include/linux/blkdev.h
+++ b/include/linux/blkdev.h
@@ -1080,16 +1080,31 @@ static inline unsigned int queue_physical_block_size(struct request_queue *q)
 	return q->limits.physical_block_size;
 }
 
+static inline unsigned int bdev_physical_block_size(struct block_device *bdev)
+{
+	return queue_physical_block_size(bdev_get_queue(bdev));
+}
+
 static inline unsigned int queue_io_min(struct request_queue *q)
 {
 	return q->limits.io_min;
 }
 
+static inline unsigned int bdev_io_min(struct block_device *bdev)
+{
+	return queue_io_min(bdev_get_queue(bdev));
+}
+
 static inline unsigned int queue_io_opt(struct request_queue *q)
 {
 	return q->limits.io_opt;
 }
 
+static inline unsigned int bdev_io_opt(struct block_device *bdev)
+{
+	return queue_io_opt(bdev_get_queue(bdev));
+}
+
 static inline int queue_alignment_offset(struct request_queue *q)
 {
 	if (q && q->limits.misaligned)
@@ -1108,6 +1123,19 @@ static inline int queue_sector_alignment_offset(struct request_queue *q,
 		& (q->limits.io_min - 1);
 }
 
+static inline int bdev_alignment_offset(struct block_device *bdev)
+{
+	struct request_queue *q = bdev_get_queue(bdev);
+
+	if (!q || q->limits.misaligned)
+		return -1;
+
+	if (bdev != bdev->bd_contains)
+		return bdev->bd_part->alignment_offset;
+
+	return q->limits.alignment_offset;
+}
+
 static inline int queue_dma_alignment(struct request_queue *q)
 {
 	return q ? q->dma_alignment : 511;
diff --git a/include/linux/fs.h b/include/linux/fs.h
index 90162fb..3f401fc 100644
--- a/include/linux/fs.h
+++ b/include/linux/fs.h
@@ -300,6 +300,10 @@ struct inodes_stat_t {
 #define BLKTRACESTOP _IO(0x12,117)
 #define BLKTRACETEARDOWN _IO(0x12,118)
 #define BLKDISCARD _IO(0x12,119)
+#define BLKIOMIN _IO(0x12,120)
+#define BLKIOOPT _IO(0x12,121)
+#define BLKALIGNOFF _IO(0x12,122)
+#define BLKPBSZGET _IO(0x12,123)
 
 #define BMAP_IOCTL 1		/* obsolete - kept for compatibility */
 #define FIBMAP	   _IO(0x00,1)	/* bmap access */

Re: Topology ioctls

[Jamie Lokier]

Martin K. Petersen wrote:
> 
> The original rationale for exporting the topology information via sysfs
> was that we intended to support multiple heterogeneous regions within a
> block device.  And that fit poorly with an ioctl approach.
> 
> However, with a single region per device it is trivial to provide the
> topology.  And while mkfs.* will continue to use the libblkid interface,
> there are users that would like to get access to this information
> without having to traverse sysfs and stitch things together manually.

Quite nice, I can see it coming in handy.

One more bit of information I'd like is the "write affected block
size".  When you write to a single disk, it's the sector size (512 or
soon to be 4096).  For a RAID, it's probably quite large, depending on
implementation details.  For some kind of flash - see threads from
Pavel and LWN about sector writes causing erase-block sizes to be
lost.

That size is useful to any program which does journalling/logging type
of write pattern, i.e. databases and filesystems-in-a-file, so they
can write new commit blocks sufficiently far apart.

-- Jamie

Re: Topology ioctls

[Martin K. Petersen]

>>>>> "Jamie" == Jamie Lokier <jamie@shareable.org> writes:

Jamie> One more bit of information I'd like is the "write affected block
Jamie> size".  When you write to a single disk, it's the sector size
Jamie> (512 or soon to be 4096).  For a RAID, it's probably quite large,
Jamie> depending on implementation details.

That's what BLKIOMIN is all about...

-- 
Martin K. Petersen	Oracle Linux Engineering