Add the ability to 'resize' the loop device on the fly.

One practical application is a loop file with XFS filesystem, already
mounted: You can easily enlarge the file (append some bytes) and then call
ioctl(fd, LOOP_SET_CAPACITY, new); The loop driver will learn about the
new size and you can use xfs_growfs later on, which will allow you to use
full capacity of the loop file without the need to unmount.

Test app:

#include <linux/fs.h>
#include <linux/loop.h>
#include <sys/ioctl.h>
#include <sys/stat.h>
#include <sys/types.h>
#include <assert.h>
#include <errno.h>
#include <fcntl.h>
#include <stdio.h>
#include <stdlib.h>
#include <unistd.h>

#define _GNU_SOURCE
#include <getopt.h>

char *me;

void usage(FILE *f)
{
	fprintf(f, "%s [options] loop_dev [backend_file]\n"
		"-s, --set new_size_in_bytes\n"
		"\twhen backend_file is given, "
		"it will be expanded too while keeping the original contents\n",
		me);
}

struct option opts[] = {
	{
		.name		= "set",
		.has_arg	= 1,
		.flag		= NULL,
		.val		= 's'
	},
	{
		.name		= "help",
		.has_arg	= 0,
		.flag		= NULL,
		.val		= 'h'
	}
};

void err_size(char *name, __u64 old)
{
	fprintf(stderr, "size must be larger than current %s (%llu)\n",
		name, old);
}

int main(int argc, char *argv[])
{
	int fd, err, c, i, bfd;
	ssize_t ssz;
	size_t sz;
	__u64 old, new, append;
	char a[BUFSIZ];
	struct stat st;
	FILE *out;
	char *backend, *dev;

	err = EINVAL;
	out = stderr;
	me = argv[0];
	new = 0;
	while ((c = getopt_long(argc, argv, "s:h", opts, &i)) != -1) {
		switch (c) {
		case 's':
			errno = 0;
			new = strtoull(optarg, NULL, 0);
			if (errno) {
				err = errno;
				perror(argv[i]);
				goto out;
			}
			break;

		case 'h':
			err = 0;
			out = stdout;
			goto err;

		default:
			perror(argv[i]);
			goto err;
		}
	}

	if (optind < argc)
		dev = argv[optind++];
	else
		goto err;

	fd = open(dev, O_RDONLY);
	if (fd < 0) {
		err = errno;
		perror(dev);
		goto out;
	}

	err = ioctl(fd, BLKGETSIZE64, &old);
	if (err) {
		err = errno;
		perror("ioctl BLKGETSIZE64");
		goto out;
	}

	if (!new) {
		printf("%llu\n", old);
		goto out;
	}

	if (new < old) {
		err = EINVAL;
		err_size(dev, old);
		goto out;
	}

	if (optind < argc) {
		backend = argv[optind++];
		bfd = open(backend, O_WRONLY|O_APPEND);
		if (bfd < 0) {
			err = errno;
			perror(backend);
			goto out;
		}
		err = fstat(bfd, &st);
		if (err) {
			err = errno;
			perror(backend);
			goto out;
		}
		if (new < st.st_size) {
			err = EINVAL;
			err_size(backend, st.st_size);
			goto out;
		}
		append = new - st.st_size;
		sz = sizeof(a);
		while (append > 0) {
			if (append < sz)
				sz = append;
			ssz = write(bfd, a, sz);
			if (ssz != sz) {
				err = errno;
				perror(backend);
				goto out;
			}
			append -= sz;
		}
		err = fsync(bfd);
		if (err) {
			err = errno;
			perror(backend);
			goto out;
		}
	}

	err = ioctl(fd, LOOP_SET_CAPACITY, new);
	if (err) {
		err = errno;
		perror("ioctl LOOP_SET_CAPACITY");
	}
	goto out;

 err:
	usage(out);
 out:
	return err;
}
Signed-off-by: NJ. R. Okajima <hooanon05@yahoo.co.jp>
Signed-off-by: NTomas Matejicek <tomas@slax.org>
Cc: <util-linux-ng@vger.kernel.org>
Cc: Karel Zak <kzak@redhat.com>
Cc: Jens Axboe <jens.axboe@oracle.com>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Cc: Christoph Hellwig <hch@lst.de>
Cc: Akinobu Mita <akinobu.mita@gmail.com>
Cc: <linux-api@vger.kernel.org>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

Impact: fix 15 make headers_check warnings:

include of <linux/types.h> is preferred over <asm/types.h>
Signed-off-by: NJaswinder Singh Rajput <jaswinderrajput@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Sam Ravnborg <sam@ravnborg.org>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

This allows a flag to be set on loop devices so that when they are
closed for the last time, they'll self-destruct.

In general, so that we can automatically allocate loop devices (as with
losetup -f) and have them disappear when we're done with them.

In particular, right now, so that we can stop relying on the hackish
special-case in umount(8) which kills off loop devices which were set up by
'mount -oloop'.  That means we can stop putting crap in /etc/mtab which
doesn't belong there, which means it can be a symlink to /proc/mounts, which
means yet another writable file on the root filesystem is eliminated and the
'stateless' folks get happier...  and OLPC trac #356 can be closed.

The mount(8) side of that is at
http://marc.info/?l=util-linux-ng&m=119362955431694&w=2

[akpm@linux-foundation.org: coding-style fixes]
Signed-off-by: NDavid Woodhouse <dwmw2@infradead.org>
Cc: Bernardo Innocenti <bernie@codewiz.org>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

include/linux/loop.h:66: error: expected specifier-qualifier-list before 'request_queue_t'

Cc: Sebastian Siewior <sebastian@breakpoint.cc>
Cc: Jens Axboe <jens.axboe@oracle.com>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

Some of the code has been gradually transitioned to using the proper
struct request_queue, but there's lots left. So do a full sweet of
the kernel and get rid of this typedef and replace its uses with
the proper type.
Signed-off-by: NJens Axboe <jens.axboe@oracle.com>

Remove artificial maximum 256 loop device that can be created due to a
legacy device number limit.  Searching through lkml archive, there are
several instances where users complained about the artificial limit that
the loop driver impose.  There is no reason to have such limit.

This patch rid the limit entirely and make loop device and associated block
queue instantiation on demand.  With on-demand instantiation, it also gives
the benefit of not wasting memory if these devices are not in use (compare
to current implementation that always create 8 loop devices), a net
improvement in both areas.  This version is both tested with creation of
large number of loop devices and is compatible with existing losetup/mount
user land tools.

There are a number of people who worked on this and provided valuable
suggestions, in no particular order, by:

Jens Axboe
Jan Engelhardt
Christoph Hellwig
Thomas M
Signed-off-by: NKen Chen <kenchen@google.com>
Cc: Jan Engelhardt <jengelh@linux01.gwdg.de>
Cc: Christoph Hellwig <hch@lst.de>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

Convert loop.c from the deprecated kernel_thread to kthread.  This patch
simplifies the code quite a bit and passes similar testing to the previous
submission on both emulated x86 and s390.

Changes since last submission:
	switched to using a rather simple loop based on
	wait_event_interruptible.
Signed-off-by: NSerge E. Hallyn <serue@us.ibm.com>
Signed-off-by: NAndrew Morton <akpm@osdl.org>
Signed-off-by: NLinus Torvalds <torvalds@osdl.org>

This reverts commit c7b2eff0.

Hugh Dickins explains:

 "It seems too little tested: "losetup -d /dev/loop0" fails with
  EINVAL because nothing sets lo_thread; but even when you patch
  loop_thread() to set lo->lo_thread = current, it can't survive
  more than a few dozen iterations of the loop below (with a tmpfs
  mounted on /tst):

	j=0
	cp /dev/zero /tst
	while :
	do
	    let j=j+1
	    echo "Doing pass $j"
	    losetup /dev/loop0 /tst/zero
	    mkfs -t ext2 -b 1024 /dev/loop0 >/dev/null 2>&1
	    mount -t ext2 /dev/loop0 /mnt
	    umount /mnt
	    losetup -d /dev/loop0
	done

  it collapses with failed ioctl then BUG_ON(!bio).

  I think the original lo_done completion was more subtle and safe
  than the kthread conversion has allowed for."
Signed-off-by: NLinus Torvalds <torvalds@osdl.org>

Update loop.c to use a kthread instead of a deprecated kernel_thread for
loop devices.

[akpm@osdl.org: don't change the thread's name]
Signed-off-by: NSerge E. Hallyn <serue@us.ibm.com>
Signed-off-by: NAndrew Morton <akpm@osdl.org>
Signed-off-by: NLinus Torvalds <torvalds@osdl.org>

Semaphore to mutex conversion.

The conversion was generated via scripts, and the result was validated
automatically via a script as well.
Signed-off-by: NIngo Molnar <mingo@elte.hu>
Cc: Jens Axboe <axboe@suse.de>
Signed-off-by: NAndrew Morton <akpm@osdl.org>
Signed-off-by: NLinus Torvalds <torvalds@osdl.org>

convert the block loop device from semaphores to completions.
Signed-off-by: NIngo Molnar <mingo@elte.hu>

Signed-off-by: NAl Viro <viro@zeniv.linux.org.uk>
Signed-off-by: NLinus Torvalds <torvalds@osdl.org>

Looks like locking can be optimised quite a lot.  Increase lock widths
slightly so lo_lock is taken fewer times per request.  Also it was quite
trivial to cover lo_pending with that lock, and remove the atomic
requirement.  This also makes memory ordering explicitly correct, which is
nice (not that I particularly saw any mem ordering bugs).

Test was reading 4 250MB files in parallel on ext2-on-tmpfs filesystem (1K
block size, 4K page size).  System is 2 socket Xeon with HT (4 thread).

intel:/home/npiggin# umount /dev/loop0 ; mount /dev/loop0 /mnt/loop ; /usr/bin/time ./mtloop.sh

Before:
0.24user 5.51system 0:02.84elapsed 202%CPU (0avgtext+0avgdata 0maxresident)k
0.19user 5.52system 0:02.88elapsed 198%CPU (0avgtext+0avgdata 0maxresident)k
0.19user 5.57system 0:02.89elapsed 198%CPU (0avgtext+0avgdata 0maxresident)k
0.22user 5.51system 0:02.90elapsed 197%CPU (0avgtext+0avgdata 0maxresident)k
0.19user 5.44system 0:02.91elapsed 193%CPU (0avgtext+0avgdata 0maxresident)k

After:
0.07user 2.34system 0:01.68elapsed 143%CPU (0avgtext+0avgdata 0maxresident)k
0.06user 2.37system 0:01.68elapsed 144%CPU (0avgtext+0avgdata 0maxresident)k
0.06user 2.39system 0:01.68elapsed 145%CPU (0avgtext+0avgdata 0maxresident)k
0.06user 2.36system 0:01.68elapsed 144%CPU (0avgtext+0avgdata 0maxresident)k
0.06user 2.42system 0:01.68elapsed 147%CPU (0avgtext+0avgdata 0maxresident)k
Signed-off-by: NNick Piggin <nickpiggin@yahoo.com.au>
Signed-off-by: NAndrew Morton <akpm@osdl.org>
Signed-off-by: NLinus Torvalds <torvalds@osdl.org>

Initial git repository build. I'm not bothering with the full history,
even though we have it. We can create a separate "historical" git
archive of that later if we want to, and in the meantime it's about
3.2GB when imported into git - space that would just make the early
git days unnecessarily complicated, when we don't have a lot of good
infrastructure for it.

Let it rip!