Add option --uuid with same semantics that is provided by command
'mkswap'. By default a random UUID is generated, to not set any use
'btrfs filesystem mkswapfile -U clear swapfile'.
Issue: #581
Signed-off-by: David Sterba <dsterba@suse.com>
The checksum conversion is still experimental and still does not convert
all filesystems correctly. Do not use on valuable data.
Previous implementation copied the UUID conversion which was not a good
base for the checksum conversion so it left out basically all trees
except extent and checksum.
This update adds the base for the required safety features:
- let the old csum tree intact until the full conversion is done (ie.
all data are still verifiable)
- add on-disk status tracking item, this should keep the from/to
checksum conversion, last generation to catch potential updates of the
underlying filesystem if conversion is interrupted and the filesystem
mounted
- convert most of the fundamental trees, the subvolumes, tree log and
relocation trees are not converted
- trees are converted in-place to avoid potentially running out of space
but this might be better done by transaction protection with a
temporary tree
Known issues:
- not all trees are converted
- not all checksums are correctly inserted into the new tree and reading
the files leads to EIO
Issue: #438
Signed-off-by: David Sterba <dsterba@suse.com>
The main script ci/images/test-build should be most up to date and
copied to all docker templates, sync them. Fix descriptions that were
copy&pasted from musl. Add missing test-build script from musl image.
Signed-off-by: David Sterba <dsterba@suse.com>
A fresh build of Centos 8 fails due to changes in upstream repositoris:
"Error: Failed to download metadata for repo ‘appstream’: Cannot prepare
internal mirrorlist: No URLs in mirrorlist"
There are repositories no longer updated (and missing some packages like
reiserfs-utils) but for our CI purposes it's sufficient so make it work.
Signed-off-by: David Sterba <dsterba@suse.com>
The crc32c selection has been already using the pointer-based approach
so drop the cpuid detection and use our common code for that.
Signed-off-by: David Sterba <dsterba@suse.com>
The crc32c.c is going to be updated with CPU feature detection from
common/ which is not suitable for libbtrfs, so make own copy of the
file.
Signed-off-by: David Sterba <dsterba@suse.com>
Change how sha256 implementation is selected. Instead of an if-else
check in the block processing function select the best version and assign
the function pointer. This is slightly faster.
At this point the selection is not implemented properly in
hash-speedtest so all results are from the fastest version. This will
be fixed once all algorithms are converted.
Signed-off-by: David Sterba <dsterba@suse.com>
Change how blake2 implementation is selected. Instead of an if-else check
inside blake2b_compress each time, select the best one and assign the
function pointer. This is slightly faster.
At this point the selection is not implemented properly in
hash-speedtest so all results are from the fastest version. This will
be fixed once all algorithms are converted.
Signed-off-by: David Sterba <dsterba@suse.com>
Prepare a single location that will detect or set accelerated versions
of hash algorithms. Right now it's the crc32c, blake2 and sha256 do
an if-else switch while crc32c sets a function pointer.
Signed-off-by: David Sterba <dsterba@suse.com>
Add CPU_FLAG_NONE that's first in the list and represents a CPU without
any acceleration features. Can be used for reference/fallback
implementation.
Signed-off-by: David Sterba <dsterba@suse.com>
The conversion have been copy&pasted from one code but not all messages
reflect that and mistakenly say fsid instead of csum, etc. Also rename
functions converting the trees to more descriptive names.
Signed-off-by: David Sterba <dsterba@suse.com>
Update xxhash implementation from https://github.com/Cyan4973/xxHash.
This has moved a lot of code so the diff is huge, plus the code we don't
need now for btrfs has been removed (XXH3, XXH32).
There's no significant change in performance.
Signed-off-by: David Sterba <dsterba@suse.com>
The CI build on musl warns:
[CC] cmds/balance.o
In file included from cmds/inspect.c:19:
/usr/include/sys/fcntl.h:1:2: warning: #warning redirecting incorrect #include <sys/fcntl.h> to <fcntl.h> [-Wcpp]
1 | #warning redirecting incorrect #include <sys/fcntl.h> to <fcntl.h>
| ^~~~~~~
On glibc the header directly includes fcntl.h.
Signed-off-by: David Sterba <dsterba@suse.com>
Commit 0d5767e664 ("btrfs-progs: build: use -std=gnu11") was supposed
to set C level to gnu11 to match kernel but it was done only in the C=1
(checker) branch. The std used was still gnu90 that can be verified by
running with V=1.
Set the std option to gnu11 unconditionally and delete the m4 macros
that were detecting support on older gcc (4.5). The reference build on
lowest supported distro (Centos 7) compiles fine.
Signed-off-by: David Sterba <dsterba@suse.com>
The devel package for reiserfs is no longer available in openSUSE
Tumbleweed. Build auto-detects the support for convert.
Signed-off-by: David Sterba <dsterba@suse.com>
The messages of ioctl-test related build is not aligned with other
commands:
[LD] hash-speedtest
[LD] hash-vectest
[CC] ioctl-test.o
[LD] ioctl-test
Signed-off-by: David Sterba <dsterba@suse.com>
The test programs are not build by default but we want at least the
build coverage, so add them. Also make the build parallel.
Signed-off-by: David Sterba <dsterba@suse.com>
There are some stale headers that we don't need and the int types are
using the kernel types and pull kerncompat.h. As this is a basic header
that should minimize dependencies use the standard int types.
Signed-off-by: David Sterba <dsterba@suse.com>
Now that there are more implementations for the hashes test them all on
the vectors if the CPU supports that.
Signed-off-by: David Sterba <dsterba@suse.com>
Add test vectors that are longer that the internal block length.
Accelerated implementations may not be used on the short or unpaded
blocks but we need to test them as well.
Signed-off-by: David Sterba <dsterba@suse.com>
Copy sha256-x86.c from https://github.com/noloader/SHA-Intrinsics, that
uses the compiler intrinsics to implement the update step with the
native x86_64 instructions.
To avoid dependencies of the reference code and the x86 version, check
runtime support only if the compiler also supports -msha.
Signed-off-by: David Sterba <dsterba@suse.com>
Benchmark all accelerated implementations if the CPU supports them. Set
the level before each test, expecting that the implementation switches
the implementation dynamically.
Signed-off-by: David Sterba <dsterba@suse.com>
Copy AVX2 implementation from https://github.com/sneves/blake2-avx2 .
Though this is marked experimental, libsodium uses this version.
Signed-off-by: David Sterba <dsterba@suse.com>
Copy implementation from https://github.com/BLAKE2/BLAKE2, add runtime
detection of SSE2 and add the switch function.
Signed-off-by: David Sterba <dsterba@suse.com>
Add support for run-time detection of CPU features on x86_64 to allow
selection of accelerated implementations of hash algorithms.
When possible use the compiler builtin (works on gcc and clang).
The SHA extensions can't be detected by __builtin_cpu_supports and the
__cpuid/__cpuidex macros are not consistently provided in all supported
gcc and clang versions. Copy the __cpuidex and call it manually for the
SHA extensions. Complete list https://en.wikipedia.org/wiki/CPUID .
Signed-off-by: David Sterba <dsterba@suse.com>
[WARNING]
Clang 15.0.7 warns about several unused variables:
kernel-shared/zoned.c:829:6: warning: variable 'num_sequential' set but not used [-Wunused-but-set-variable]
u32 num_sequential = 0, num_conventional = 0;
^
cmds/scrub.c:1174:6: warning: variable 'n_skip' set but not used [-Wunused-but-set-variable]
int n_skip = 0;
^
mkfs/main.c:493:6: warning: variable 'total_block_count' set but not used [-Wunused-but-set-variable]
u64 total_block_count = 0;
^
image/main.c:2246:6: warning: variable 'bytenr' set but not used [-Wunused-but-set-variable]
u64 bytenr = 0;
^
[CAUSE]
Most of them are just straightforward set but not used variables.
The only exception is total_block_count, which has commented out code
relying on it.
[FIX]
Just remove those variables, and for @total_block_count, also remove the
comments.
Signed-off-by: Qu Wenruo <wqu@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
[WARNING]
Clang 15.0.7 gives the following warning:
image/main.c:95:2: warning: field '' with variable sized type 'union metadump_struct::(anonymous at image/main.c:95:2)' not at the end of a struct or class is a GNU extension [-Wgnu-variable-sized-type-not-at-end]
union {
^
[CAUSE]
The union contains meta_cluster, which variable sized:
struct meta_cluster {
struct meta_cluster_header header;
struct meta_cluster_item items[];
} __attribute__ ((__packed__));
Thus clang gives above warning since it's a GNU extension.
[FIX]
Just move the union to the end of the structure.
Signed-off-by: Qu Wenruo <wqu@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
[FALSE ALERT]
Unlike gcc, clang doesn't really understand the comments, thus it's
reportings tons of fall through related errors:
cmds/reflink.c:124:3: warning: unannotated fall-through between switch labels [-Wimplicit-fallthrough]
case 'r':
^
cmds/reflink.c:124:3: note: insert '__attribute__((fallthrough));' to silence this warning
case 'r':
^
__attribute__((fallthrough));
cmds/reflink.c:124:3: note: insert 'break;' to avoid fall-through
case 'r':
^
break;
[CAUSE]
Although gcc is fine with /* fallthrough */ comments, clang is not.
[FIX]
So just introduce a fallthrough macro to handle the situation properly,
and use that macro instead.
Signed-off-by: Qu Wenruo <wqu@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
[FALSE ALERT]
Clang 15.0.7 gives the following false alert in get_dev_extent_len():
kernel-shared/extent-tree.c:3328:2: warning: variable 'div' is used uninitialized whenever switch default is taken [-Wsometimes-uninitialized]
default:
^~~~~~~
kernel-shared/extent-tree.c:3332:24: note: uninitialized use occurs here
return map->ce.size / div;
^~~
kernel-shared/extent-tree.c:3311:9: note: initialize the variable 'div' to silence this warning
int div;
^
= 0
And one in btrfs_stripe_length() too:
kernel-shared/volumes.c:2781:2: warning: variable 'stripe_len' is used uninitialized whenever switch default is taken [-Wsometimes-uninitialized]
default:
^~~~~~~
kernel-shared/volumes.c:2785:9: note: uninitialized use occurs here
return stripe_len;
^~~~~~~~~~
kernel-shared/volumes.c:2754:16: note: initialize the variable 'stripe_len' to silence this warning
u64 stripe_len;
^
= 0
[CAUSE]
Clang doesn't really understand what BUG_ON() means, thus in that
default case, we won't get uninitialized value but crash directly.
[FIX]
Silent the errors by assigning the default value properly using the
value of SINGLE profile.
Signed-off-by: Qu Wenruo <wqu@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
[FALSE ALERT]
With clang 15.0.7, there is a false alert on uninitialized value in
ctree.c:
kernel-shared/ctree.c:3418:13: warning: variable 'offset' is used uninitialized whenever 'if' condition is false [-Wsometimes-uninitialized]
} else if (ret < 0) {
^~~~~~~
kernel-shared/ctree.c:3428:41: note: uninitialized use occurs here
write_extent_buffer(eb, &subvol_id_le, offset, sizeof(subvol_id_le));
^~~~~~
kernel-shared/ctree.c:3418:9: note: remove the 'if' if its condition is always true
} else if (ret < 0) {
^~~~~~~~~~~~~
kernel-shared/ctree.c:3380:22: note: initialize the variable 'offset' to silence this warning
unsigned long offset;
^
= 0
kernel-shared/ctree.c:3418:13: warning: variable 'eb' is used uninitialized whenever 'if' condition is false [-Wsometimes-uninitialized]
} else if (ret < 0) {
^~~~~~~
kernel-shared/ctree.c:3429:26: note: uninitialized use occurs here
btrfs_mark_buffer_dirty(eb);
^~
kernel-shared/ctree.c:3418:9: note: remove the 'if' if its condition is always true
} else if (ret < 0) {
^~~~~~~~~~~~~
kernel-shared/ctree.c:3378:26: note: initialize the variable 'eb' to silence this warning
struct extent_buffer *eb;
^
= NULL
[CAUSE]
The original code is handling the return value from
btrfs_insert_empty_item() like this:
ret = btrfs_insert_empty_item();
if (ret >= 0) {
/* Do something for it. */
} else if (ret == -EEXIST) {
/* Do something else. */
} else if (ret < 0) {
/* Error handling. */
}
But the problem is, the last one check is always true if we can reach
there.
Thus clang is providing the hint to remove the if () check.
[FIX]
Normally we prefer to do error handling first, so move the error
handling first so we don't need the if () else if () chain.
Signed-off-by: Qu Wenruo <wqu@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Before decompressing, we zero out the content of the entire output buffer,
so that we don't get any garbage after the last byte of data. We do this
for all compression algorithms. However zstd, at least with libzstd 1.5.2
on Debian (version 1.5.2+dfsg-1), the decompression routine can end up
touching the content of the output buffer beyond the last valid byte of
decompressed data, resulting in a corruption.
Example reproducer:
$ cat test.sh
#!/bin/bash
DEV=/dev/sdj
MNT=/mnt/sdj
rm -f /tmp/send.stream
umount $DEV &> /dev/null
mkfs.btrfs -f $DEV &> /dev/null || echo "MKFS failed!"
mount -o compress=zstd $DEV $MNT
# File foo is not sector size aligned, 127K.
xfs_io -f -c "pwrite -S 0xab 0 3" \
-c "pwrite -S 0xcd 3 130042" \
-c "pwrite -S 0xef 130045 3" $MNT/foo
# Now do an fallocate that increases the size of foo from 127K to 128K.
xfs_io -c "falloc 0 128K " $MNT/foo
btrfs subvolume snapshot -r $MNT $MNT/snap
btrfs send --compressed-data -f /tmp/send.stream $MNT/snap
echo -e "\nFile data in the original filesystem:\n"
od -A d -t x1 $MNT/snap/foo
umount $MNT
mkfs.btrfs -f $DEV &> /dev/null || echo "MKFS failed!"
mount $DEV $MNT
btrfs receive --force-decompress -f /tmp/send.stream $MNT
echo -e "\nFile data in the new filesystem:\n"
od -A d -t x1 $MNT/snap/foo
umount $MNT
Running the reproducer gives:
$ ./test.sh
(...)
File data in the original filesystem:
0000000 ab ab ab cd cd cd cd cd cd cd cd cd cd cd cd cd
0000016 cd cd cd cd cd cd cd cd cd cd cd cd cd cd cd cd
*
0130032 cd cd cd cd cd cd cd cd cd cd cd cd cd ef ef ef
0130048 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
*
0131072
At subvol snap
File data in the new filesystem:
0000000 ab ab ab cd cd cd cd cd cd cd cd cd cd cd cd cd
0000016 cd cd cd cd cd cd cd cd cd cd cd cd cd cd cd cd
*
0130032 cd cd cd cd cd cd cd cd cd cd cd cd cd ef ef ef
0130048 cd cd cd cd 00 00 00 00 00 00 00 00 00 00 00 00
0130064 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
*
0131072
The are 4 bytes with a value of 0xcd instead of 0x00, at file offset
127K (130048).
Fix this by explicitly zeroing out the part of the output buffer that was
not used after decompressing with zstd.
The decompression of compressed extents, sent when using the send v2
stream, happens in the following cases:
1) By explicitly passing --force-decompress to the receive command, as in
the reproducer above;
2) Calling the BTRFS_IOC_ENCODED_WRITE ioctl failed with -ENOTTY, meaning
the kernel on the receiving side is old and does not implement that
ioctl;
3) Calling the BTRFS_IOC_ENCODED_WRITE ioctl failed with -ENOSPC;
4) Calling the BTRFS_IOC_ENCODED_WRITE ioctl failed with -EINVAL.
Reviewed-by: Qu Wenruo <wqu@suse.com>
Signed-off-by: Filipe Manana <fdmanana@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Move the source file to own directory so it can be further split and
refactored. File needs to be renamed to main.c so the build magic works.
Signed-off-by: David Sterba <dsterba@suse.com>
By default print how many subvolumes are considered for checks, either
found or specified on the command line. Once a subvolume is removed from
the list print the progress from the total count.
$ btrfs subvolume sync /path
Waiting for 130 subvolumes
Subvolume id 256 is gone (1/130)
Subvolume id 257 is gone (2/130)
...
Subvolume id 384 is gone (129/130)
Subvolume id 385 is gone (130/130)
Signed-off-by: David Sterba <dsterba@suse.com>
Use the configured widths and print padding directly instead of the
embedded printf format and fixed width strings.
Signed-off-by: David Sterba <dsterba@suse.com>
To make option formatting a bit easier so the spacing is unified add
macros and formatting helpers.
Usage in the help text:
OPTLINE("-o value", "description")
Internally the option and description are delimiters by chars that are
not part of normal text, the formatter separates that and uses fixed
with for output. The description text can be of any length, multi-line
text should still end up as one token (i.e. newline without ',' between).
Signed-off-by: David Sterba <dsterba@suse.com>