A recent bugfix in the Linux kernel made it possible for the
LOOP_SET_STATUS64 ioctl to fail when called with a non-zero offset,
with an EAGAIN errno:
5db470e229e2 loop: drop caches if offset or block_size are changed
This fix changes a silent failure (where mount could sometimes access
the backing loop image through the cache without the specified offset)
to an explicit failure, and it has also been backported on stable
branches.
On a 5.0 kernel, other changes to the loop driver make it hard to get
generate the EAGAIN error, but this bugfix has also been backported to
stables branches, without these changes. At least with the 4.14 stable
branch, the EAGAIN error can be quickly generated with the following loop:
while mount -o loop,offset=239 disk point && umount point; do :; done
Retry the ioctl when it fails with EAGAIN, which means that mount or
losetup will eventually succeed when encountering this case.
[kzak@redhat.com: - use our local portable xusleep()]
Signed-off-by: Romain Izard <romain.izard.pro@gmail.com>
Signed-off-by: Karel Zak <kzak@redhat.com>
The partition scanner in kernel depends on blocksize. We need to set
the blocksize before we call LOOP_SET_STATUS64 (this ioctl triggers
the scanner).
This patch extends the internal API to save blocksize into loopdev
context to be usable later for loopcxt_setup_device().
Signed-off-by: Karel Zak <kzak@redhat.com>
Now the internal API uses loopcxt_set_...() to set context variables
as well as to call ioctls. This patch introduces loopcxt_ioctl_...()
to makes things more obvious to readers.
Signed-off-by: Karel Zak <kzak@redhat.com>
Let's make it possible to use debug.h without environment variables.
Suggested-by: J William Piggott <elseifthen@gmx.com>
Signed-off-by: Karel Zak <kzak@redhat.com>
* introduce new flag __UL_DEBUG_FL_NOADDR to suppress pointer address printing
* use __UL_DEBUG_FL_NOADDR when SUID
* move ul_debugobj() to debugobj.h, and require UL_DEBUG_CURRENT_MASK
to provide access to the current mask from ul_debugobj(). It's better
than modify all ul_debugobj() calls and use the global mask as
argument.
* remove never used UL_DEBUG_DEFINE_FLAG
Reported-by: halfdog <me@halfdog.net>
Signed-off-by: Karel Zak <kzak@redhat.com>
Kernel since 4.14 supports setting of logical block size[1]. It allows to
create loop devices that report logical block size different from 512.
Add support for this feature to losetup.
References:
[1] https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/drivers/block/loop.c?id=89e4fdecb51cf5535867026274bc97de9480ade5
[kzak@redhat.com: - fix loopcxt_get_blocksize()
- remove lo_blocksize from loop_info64]
Signed-off-by: Stanislav Brabec <sbrabec@suse.cz>
Cc: Ming Lei <ming.lei@redhat.com>
Cc: Hannes Reinecke <hare@suse.com>
Cc: Omar Sandoval <osandov@fb.com>
Cc: Jens Axboe <axboe@kernel.dk>
Signed-off-by: Karel Zak <kzak@redhat.com>
../lib/loopdev.c: In function 'loopcxt_next_from_sysfs':
../lib/loopdev.c:545:32: warning: '/loop/backing_file' directive output may be truncated writing 18 bytes into a region of size between 1 and 256 [-Wformat-truncation=]
snprintf(name, sizeof(name), "%s/loop/backing_file", d->d_name);
^~~~~~~~~~~~~~~~~~~~~~
../lib/loopdev.c:545:3: note: 'snprintf' output between 19 and 274 bytes into a destination of size 256
snprintf(name, sizeof(name), "%s/loop/backing_file", d->d_name);
^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
../lib/sysfs.c: In function 'sysfs_is_partition_dirent':
../lib/sysfs.c:343:31: warning: '/start' directive output may be truncated writing 6 bytes into a region of size between 1 and 256 [-Wformat-truncation=]
snprintf(path, sizeof(path), "%s/start", d->d_name);
^~~~~~~~~~
../lib/sysfs.c:343:2: note: 'snprintf' output between 7 and 262 bytes into a destination of size 256
snprintf(path, sizeof(path), "%s/start", d->d_name);
^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
../lib/sysfs.c: In function 'sysfs_partno_to_devno':
../lib/sysfs.c:372:32: warning: '/partition' directive output may be truncated writing 10 bytes into a region of size between 1 and 256 [-Wformat-truncation=]
snprintf(path, sizeof(path), "%s/partition", d->d_name);
^~~~~~~~~~~~~~
../lib/sysfs.c:372:3: note: 'snprintf' output between 11 and 266 bytes into a destination of size 256
snprintf(path, sizeof(path), "%s/partition", d->d_name);
^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
../lib/sysfs.c:377:33: warning: '/dev' directive output may be truncated writing 4 bytes into a region of size between 1 and 256 [-Wformat-truncation=]
snprintf(path, sizeof(path), "%s/dev", d->d_name);
^~~~~~~~
../lib/sysfs.c:377:4: note: 'snprintf' output between 5 and 260 bytes into a destination of size 256
snprintf(path, sizeof(path), "%s/dev", d->d_name);
^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Signed-off-by: Ruediger Meier <ruediger.meier@ga-group.nl>
The function is_loopdev does not set errno if the supplied string does
not reference a valid loop device. Fix this to avoid an error message
like this one:
losetup: /: failed to use device: Success
I prefer this one:
losetup: /: failed to use device: No such device
Signed-off-by: Tobias Stoeckmann <tobias@stoeckmann.org>
Implement stand-alone loopcxt_set_status(). It allows manipulation with some
loop device parameters even if it is initialized.
Its function is limited by the kernel implementation, and only a small subset of
changes is allowed.
For more see linux/drivers/block/loop.c:loop_set_status()
Signed-off-by: Stanislav Brabec <sbrabec@suse.cz>
This patch introduces overlap detections and loop devices
re-use for losetup(8). We already support this feature for mount(8)
where it's enabled by default (because we mount filesystems and it's
always mistake to share the same filesystem between more loop
devices).
Stanislav has suggested to enable this feature also for losetup by
default. I'm not sure about it, IMHO it's better to keep losetup(8)
simple and stupid by default, and inform users about possible problems
and solutions in the man page.
The feature forces losetup to scan all loop devices always when new
one is requested. This maybe disadvantage (especially when we use
control-loop to avoid /sys or /dev scans) on system with huge number
of loop devices.
Co-Author: Stanislav Brabec <sbrabec@suse.cz>
Signed-off-by: Karel Zak <kzak@redhat.com>
The current code scans loopdevs to detect already used loop device and
another scan to detect overlap.
Let's use one scan only, for this purpose loopcxt_find_overlap() has
been modified to return info (rc==2) about full size and offset match.
Signed-off-by: Karel Zak <kzak@redhat.com>
Fully safe checks of loop device need to check sizelimit. To prevent need of two
nearly equal functions, introduce sizelimit parameter to several internal
functions:
loopdev_is_used()
loopdev_find_by_backing_file()
loopcxt_is_used()
loopcxt_find_by_backing_file()
If sizelimit is zero, fall back to the old behavior (ignoring of sizelimit).
Signed-off-by: Stanislav Brabec <sbrabec@suse.cz>
Fix various typos in error messages, warnings, debug strings,
comments and names of static functions.
Signed-off-by: Sebastian Rasmussen <sebras@gmail.com>
The file is no portable (#ifdef HAVE_SYS_SYSMACROS_H is necessary),
but needed on many places. It seems better to keep it in c.h.
Signed-off-by: Karel Zak <kzak@redhat.com>
I have validated that we are still compatible at least back to
- openSUSE 11.4
- SLE 11
- RHEL/CentOS 6
- OSX 10.10.x, (Xcode 6.3)
- FreeBSD 10.2
Confirmed incompatibility:
- OSX 10.9.x, (Xcode 6.2)
Signed-off-by: Ruediger Meier <ruediger.meier@ga-group.nl>
From v4.4, linux kernel starts to support direct I/O and
AIO to backing file for loop driver, so allow losetup to
enable the feature by using LOOP_SET_DIRECT_IO ioctl cmd.
Signed-off-by: Ming Lei <tom.leiming@gmail.com>
loopback lib clears the existing FD unconditionally at error in
loopcxt_setup_device(). This is done even after EBUSY, thus the second call
actually clears the previous setup wrongly.
Author: Takashi Iwai <tiwai@suse.de>
Signed-off-by: Stanislav Brabec <sbrabec@suse.cz>
On system with /dev/lop-control the udevd creates /dev/loopN nodes.
It seems better to wait a moment after unsuccessful open(/dev/loopN)
and try it to open again.
The problem is pretty visible on systems where udevd also modifies
permission for loopN devices, then open() fails with EACCES when
losetup executed by non-root user (but user who is in "disk" group).
Addresses: https://bugzilla.redhat.com/show_bug.cgi?id=1045432
Signed-off-by: Karel Zak <kzak@redhat.com>
Kernel aligns the device size, but the offset where the device starts
is not required to be aligned.
# losetup --offset 32 -f file.img
is just fine, the final size of the look device will be (in sectors)
(backing_file_size - offset) >> 9
so we have to do the same in userspace when we check for successful
set capacity ioctl.
Signed-off-by: Karel Zak <kzak@redhat.com>
Now we use LOOP_CTL_GET_FREE ioctl to ask for free device, for example
losetup -f foo.img
Unfortunately, losetup(8) allows to ask for specified device
losetup /dev/loop100 foo.img
and in this case we assume that the device already exists in the
system. This is incorrect, we should be able to use loop-control
LOOP_CTL_ADD ioctl to ask for the specified device.
Signed-off-by: Karel Zak <kzak@redhat.com>
The loopcxt_check_size() is workaround for kernels < v3.9, kernel has
been fixed by commit 541c742a7559eb65f0e36d3e2338c2ca532a3e61.
The function sets loopdev size according to backing file size. The
problem is that the backing file could be a block device where
stat.st_size is zero, so we have to use blkdev_get_size() for block
devices.
Addresses: https://bugs.archlinux.org/task/35193
Reported-by: Dave Reisner <d@falconindy.com>
Signed-off-by: Karel Zak <kzak@redhat.com>
I recently tried to mount an hfsplus file system from an image file with
a partition table by using the loop offset and sizelimit options to specify
the location of the file system.
hfsplus stores some metadata at a set offset from the end of the partition,
so it's sensitive to the device size reported by the kernel.
It worked with this:
But failed with this:
/dev/loop0: [0089]:2 (<imagefile>), offset 32768, sizelimit 102400000
/dev/loop1: [0089]:2 (<imagefile>), offset 32768, sizelimit 102400000
/proc/partitions shows the correct number of blocks to match the sizelimit.
But if I set a breakpoint in mount before the mount syscall, I could see:
102400000
102432768
The kernel loop driver will set the gendisk capacity of the device at
LOOP_SET_STATUS64 but won't sync it to the block device until one of two
conditions are met: All open file descriptors referring to the device are
closed (and it will sync when re-opened) or if the LOOP_SET_CAPACITY ioctl
is called to sync it. Since mount opens the device and passes it directly
to the mount syscall after LOOP_SET_STATUS64 without closing and reopening
it, the sizelimit argument is effectively ignroed. The capacity needs to
be synced immediately for it to work as expected.
This patch adds the LOOP_SET_CAPACITY call to loopctx_setup_device since
the device isn't yet released to the user, so it's safe to sync the capacity
immediately.
[kzak@redhat.com: - port to the current git HEAD,
- use uint64_t]
Signed-off-by: Jeff Mahoney <jeffm@suse.com>
Signed-off-by: Karel Zak <kzak@redhat.com>
We use the code from include/ and lib/ on many places, so use public
domain if possible or LGPL for code copied from libs.
Signed-off-by: Karel Zak <kzak@redhat.com>
