Skip to contentSkip to conversation

Add virtio disk identification support

PATCH In review john cooper 2 participants 5 messages Mar 25, 2010 – Jun 3, 2010

Subsystems (3)
lore
Share & tools
lore
#0 lore raw

cover Add virtio disk identification support

This series adds the minimal support to qemu and virtio_blk to support passing of a virtio_blk serial id string from qemu through the guest driver and to the guest userland.

This is derived in part from a patch set posted by Rusty some time ago, but has been minimized to remove support for prior versions which attempted to provide the same functionality via pci config/io space. This version rather uses a virtio request as proposed in Rusty's example.

Also removed is the packaging of the serial/id string within the glorious bag of bits returned by the ATA_IDENTIFY command. Here we transfer only the 20 bytes of serial/id string from qemu to the guest userland. In the proposed interface, this is made available by an ioctl() into the virtio_blk driver however other interfaces (eg: /sys) have also been proposed. A code snippet is attached below as an example of ioctl usage.

The resulting code is quite minimal and I believe it addresses all concerns raised in prior versions.

-john

#include <stdio.h>
#include <strings.h>
#include <sys/types.h>
#include <fcntl.h>
#include <linux/hdreg.h>

#define IOCTL_CMD 'VBID'

main()

{
	int fd, rv;
	char buf[512];

	bzero(buf, sizeof (buf));
	if ((fd = open("/dev/vda", O_RDONLY)) < 0)
		perror("open");
	else if (ioctl(fd, IOCTL_CMD, buf) < 0)
		perror("ioctl");
	else
		printf("[%s]\n", buf);
}
Signature · 1 line
[email protected]
#1 lore raw
Quoted 34 lines · john cooper

This series adds the minimal support to qemu and virtio_blk to support passing of a virtio_blk serial id string from qemu through the guest driver and to the guest userland.

This is derived in part from a patch set posted by Rusty some time ago, but has been minimized to remove support for prior versions which attempted to provide the same functionality via pci config/io space. This version rather uses a virtio request as proposed in Rusty's example.

Also removed is the packaging of the serial/id string within the glorious bag of bits returned by the ATA_IDENTIFY command. Here we transfer only the 20 bytes of serial/id string from qemu to the guest userland. In the proposed interface, this is made available by an ioctl() into the virtio_blk driver however other interfaces (eg: /sys) have also been proposed. A code snippet is attached below as an example of ioctl usage.

The resulting code is quite minimal and I believe it addresses all concerns raised in prior versions.

-john

#include <stdio.h>
#include <strings.h>
#include <sys/types.h>
#include <fcntl.h>
#include <linux/hdreg.h>

#define IOCTL_CMD 'VBID'

main()

Quoted 12 lines
{
	int fd, rv;
	char buf[512];

	bzero(buf, sizeof (buf));
	if ((fd = open("/dev/vda", O_RDONLY)) < 0)
		perror("open");
	else if (ioctl(fd, IOCTL_CMD, buf) < 0)
		perror("ioctl");
	else
		printf("[%s]\n", buf);
}

Would we want to patch up blkid command to use this so distro stacks can extract the serial and build the typical /dev/disk/by-id/ links ?

Signature · 4 lines
Ryan Harper
Software Engineer; Linux Technology Center
IBM Corp., Austin, Tx
[email protected]
Acked-by on the v1 series
#2 lore raw
In reply to john cooper
Quoted 18 lines · john cooper

This series adds the minimal support to qemu and virtio_blk to support passing of a virtio_blk serial id string from qemu through the guest driver and to the guest userland.

This is derived in part from a patch set posted by Rusty some time ago, but has been minimized to remove support for prior versions which attempted to provide the same functionality via pci config/io space. This version rather uses a virtio request as proposed in Rusty's example.

Also removed is the packaging of the serial/id string within the glorious bag of bits returned by the ATA_IDENTIFY command. Here we transfer only the 20 bytes of serial/id string from qemu to the guest userland. In the proposed interface, this is made available by an ioctl() into the virtio_blk driver however other interfaces (eg: /sys) have also been proposed. A code snippet is attached below as an example of ioctl usage.

Quoted 4 lines

The resulting code is quite minimal and I believe it addresses all concerns raised in prior versions.

-john

I've applied the qemu and kernel side of this patch series and tested this out using the sample code below. I've also reworked this example into a virtioblk_id tool to work with udev to generate /dev/disk/by-id links; I'll be submitting a patch set to linux-hotplug with these changes, (and update for path_id) and some udev rules to persistent-storage script to autogenerate by-id and by-path symlinks for virtio-blk devices.

I've also got a patch to apply ontop of the qemu patches to generate a default serial number if one isn't specified (like we do for ide devices).

Anthony, is this series in your queue yet?

Quoted 28 lines

#include <stdio.h>
#include <strings.h>
#include <sys/types.h>
#include <fcntl.h>
#include <linux/hdreg.h>

#define IOCTL_CMD 'VBID'

main()

{
	int fd, rv;
	char buf[512];

	bzero(buf, sizeof (buf));
	if ((fd = open("/dev/vda", O_RDONLY)) < 0)
		perror("open");
	else if (ioctl(fd, IOCTL_CMD, buf) < 0)
		perror("ioctl");
	else
		printf("[%s]\n", buf);
}
Signature · 1 line
[email protected]
Signature · 4 lines
Ryan Harper
Software Engineer; Linux Technology Center
IBM Corp., Austin, Tx
[email protected]
#3 lore raw
Quoted 7 lines · Ryan Harper

I've applied the qemu and kernel side of this patch series and tested this out using the sample code below. I've also reworked this example into a virtioblk_id tool to work with udev to generate /dev/disk/by-id links; I'll be submitting a patch set to linux-hotplug with these changes, (and update for path_id) and some udev rules to persistent-storage script to autogenerate by-id and by-path symlinks for virtio-blk devices.

Sorry I didn't respond to you earlier. Actually the guest visible interface below was only intended as an example while waiting for (what I misinterpreted as Marc's intention to create) a /sys interface.

I'm all for putting this issue to rest, but if we're going to live with an ioctl interface retrieving the id string, let's make it a little more friendly from the user's perspective. I have a slightly modified version which basically implements the same interface but expects the sizeof the entire array to be preset in the first element. Otherwise the user has no way of informing the driver of the destination's size, nor a way for the driver to indicate when the data won't fit in the area specified by the user.

I should still have the patch on a test machine which ATM is unaccessible, but will have at first thing tomorrow. Let's hold off until then so we can address this nit and avoid yet another hiccup in nailing down this interface.

Thanks,

-john

Quoted 36 lines

I've also got a patch to apply ontop of the qemu patches to generate a default serial number if one isn't specified (like we do for ide devices).

Anthony, is this series in your queue yet?

Quoted 27 lines

#include <stdio.h>
#include <strings.h>
#include <sys/types.h>
#include <fcntl.h>
#include <linux/hdreg.h>

#define IOCTL_CMD 'VBID'

main()

{
	int fd, rv;
	char buf[512];

	bzero(buf, sizeof (buf));
	if ((fd = open("/dev/vda", O_RDONLY)) < 0)
		perror("open");
	else if (ioctl(fd, IOCTL_CMD, buf) < 0)
		perror("ioctl");
	else
		printf("[%s]\n", buf);
}
Signature · 1 line
[email protected]
Signature · 1 line
[email protected]
#4 lore raw
Quoted 4 lines · john cooper

I'm all for putting this issue to rest, but if we're going to live with an ioctl interface retrieving the id string, let's make it a little more friendly from the user's perspective.

The qemu side of the patch is ok as-is. The guest-user interface issue is contained in the driver. While I see the example ioctl patch has been incorporated into the virtio_blk driver, there can be no data retrieved through this interface as virtblk_get_id() will fail without the qemu counterpart. So we can clean up the details without concern of existing usage.

The only difference (as above) is allowing the caller to pass a buffer size to the driver and the driver informing the caller of the total number of bytes available:

#include <stdio.h>
#include <strings.h>
#include <fcntl.h>

#define IOCTL_CMD 'VBID'
#define BUFSZ 10

main()

{
        int fd, rv;
        char buf[255];

        bzero(buf, sizeof (buf));
        buf[0] = BUFSZ;
        if ((fd = open("/dev/vda", O_RDONLY)) < 0)
                perror("open");
        else if ((rv = ioctl(fd, IOCTL_CMD, buf)) < 0)
                perror("ioctl");
        else
                printf("[%s] %d of %d bytes\n", buf,
                       BUFSZ < rv ? BUFSZ : rv, rv);
}

---

drivers/block/virtio_blk.c +20 -6
index 83fa09a..6237732 100644
@@ -225,15 +225,29 @@ static int virtblk_ioctl(struct block_device *bdev, fmode_t mode,
 	struct gendisk *disk = bdev->bd_disk;
 	struct virtio_blk *vblk = disk->private_data;
 
+	/* user passes the address of a char[] for return of the id string
+	 * and has set char[0] to the array size.  copy id string to this
+	 * char[] and return the number of non-nul characters in the internal
+	 * id string.  The caller can then determine if all were received.
+	 */
 	if (cmd == 0x56424944) { /* 'VBID' */
 		void __user *usr_data = (void __user *)data;
 		char id_str[VIRTIO_BLK_ID_BYTES];
-		int err;
-
-		err = virtblk_get_id(disk, id_str);
-		if (!err && copy_to_user(usr_data, id_str, VIRTIO_BLK_ID_BYTES))
-			err = -EFAULT;
-		return err;
+		unsigned char idlen;
+		int rv;
+
+		if (copy_from_user(&idlen, usr_data, sizeof (idlen)))
+			return -EFAULT;
+		if (VIRTIO_BLK_ID_BYTES < idlen)
+			idlen = VIRTIO_BLK_ID_BYTES;
+		if ((rv = virtblk_get_id(disk, id_str)))
+			return rv;
+		if (copy_to_user(usr_data, id_str, idlen))
+			return -EFAULT;
+		for (rv = 0; rv < VIRTIO_BLK_ID_BYTES; ++rv)
+			if (!id_str[rv])
+				break;
+		return rv;
 	}
 	/*
 	 * Only allow the generic SCSI ioctls if the host can support it.
Signature · 1 line
[email protected]