[v3,3/4] btrfs: Add zstd support
diff mbox

Message ID 20170720212743.3351201-4-terrelln@fb.com
State New
Headers show

Commit Message

Nick Terrell July 20, 2017, 9:27 p.m. UTC
Add zstd compression and decompression support to BtrFS. zstd at its
fastest level compresses almost as well as zlib, while offering much
faster compression and decompression, approaching lzo speeds.

I benchmarked btrfs with zstd compression against no compression, lzo
compression, and zlib compression. I benchmarked two scenarios. Copying
a set of files to btrfs, and then reading the files. Copying a tarball
to btrfs, extracting it to btrfs, and then reading the extracted files.
After every operation, I call `sync` and include the sync time.
Between every pair of operations I unmount and remount the filesystem
to avoid caching. The benchmark files can be found in the upstream
zstd source repository under
`contrib/linux-kernel/{btrfs-benchmark.sh,btrfs-extract-benchmark.sh}`
[1] [2].

I ran the benchmarks on a Ubuntu 14.04 VM with 2 cores and 4 GiB of RAM.
The VM is running on a MacBook Pro with a 3.1 GHz Intel Core i7 processor,
16 GB of RAM, and a SSD.

The first compression benchmark is copying 10 copies of the unzipped
Silesia corpus [3] into a BtrFS filesystem mounted with
`-o compress-force=Method`. The decompression benchmark times how long
it takes to `tar` all 10 copies into `/dev/null`. The compression ratio is
measured by comparing the output of `df` and `du`. See the benchmark file
[1] for details. I benchmarked multiple zstd compression levels, although
the patch uses zstd level 1.

| Method  | Ratio | Compression MB/s | Decompression speed |
|---------|-------|------------------|---------------------|
| None    |  0.99 |              504 |                 686 |
| lzo     |  1.66 |              398 |                 442 |
| zlib    |  2.58 |               65 |                 241 |
| zstd 1  |  2.57 |              260 |                 383 |
| zstd 3  |  2.71 |              174 |                 408 |
| zstd 6  |  2.87 |               70 |                 398 |
| zstd 9  |  2.92 |               43 |                 406 |
| zstd 12 |  2.93 |               21 |                 408 |
| zstd 15 |  3.01 |               11 |                 354 |

The next benchmark first copies `linux-4.11.6.tar` [4] to btrfs. Then it
measures the compression ratio, extracts the tar, and deletes the tar.
Then it measures the compression ratio again, and `tar`s the extracted
files into `/dev/null`. See the benchmark file [2] for details.

| Method | Tar Ratio | Extract Ratio | Copy (s) | Extract (s)| Read (s) |
|--------|-----------|---------------|----------|------------|----------|
| None   |      0.97 |          0.78 |    0.981 |      5.501 |    8.807 |
| lzo    |      2.06 |          1.38 |    1.631 |      8.458 |    8.585 |
| zlib   |      3.40 |          1.86 |    7.750 |     21.544 |   11.744 |
| zstd 1 |      3.57 |          1.85 |    2.579 |     11.479 |    9.389 |

[1] https://github.com/facebook/zstd/blob/dev/contrib/linux-kernel/btrfs-benchmark.sh
[2] https://github.com/facebook/zstd/blob/dev/contrib/linux-kernel/btrfs-extract-benchmark.sh
[3] http://sun.aei.polsl.pl/~sdeor/index.php?page=silesia
[4] https://cdn.kernel.org/pub/linux/kernel/v4.x/linux-4.11.6.tar.xz

zstd source repository: https://github.com/facebook/zstd

Signed-off-by: Nick Terrell <terrelln@fb.com>
---
v2 -> v3:
- Port upstream BtrFS commits e1ddce71d6, 389a6cfc2a, and 6acafd1eff
- Change default compression level for BtrFS to 3

 fs/btrfs/Kconfig           |   2 +
 fs/btrfs/Makefile          |   2 +-
 fs/btrfs/compression.c     |   1 +
 fs/btrfs/compression.h     |   6 +-
 fs/btrfs/ctree.h           |   1 +
 fs/btrfs/disk-io.c         |   2 +
 fs/btrfs/ioctl.c           |   6 +-
 fs/btrfs/props.c           |   6 +
 fs/btrfs/super.c           |  12 +-
 fs/btrfs/sysfs.c           |   2 +
 fs/btrfs/zstd.c            | 435 +++++++++++++++++++++++++++++++++++++++++++++
 include/uapi/linux/btrfs.h |   8 +-
 12 files changed, 471 insertions(+), 12 deletions(-)
 create mode 100644 fs/btrfs/zstd.c

--
2.9.3
--
To unsubscribe from this list: send the line "unsubscribe linux-btrfs" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Comments

kernel test robot July 23, 2017, 7:28 p.m. UTC | #1
Hi Nick,

[auto build test WARNING on linus/master]
[also build test WARNING on v4.13-rc1 next-20170721]
[if your patch is applied to the wrong git tree, please drop us a note to help improve the system]

url:    https://github.com/0day-ci/linux/commits/Nick-Terrell/Add-xxhash-and-zstd-modules/20170723-092845
config: parisc-allyesconfig (attached as .config)
compiler: hppa-linux-gnu-gcc (Debian 6.1.1-9) 6.1.1 20160705
reproduce:
        wget https://raw.githubusercontent.com/01org/lkp-tests/master/sbin/make.cross -O ~/bin/make.cross
        chmod +x ~/bin/make.cross
        # save the attached .config to linux build tree
        make.cross ARCH=parisc 

All warnings (new ones prefixed by >>):

   lib/xxhash.c: In function 'xxh64':
>> lib/xxhash.c:236:1: warning: the frame size of 1688 bytes is larger than 1024 bytes [-Wframe-larger-than=]
    }
    ^
   lib/xxhash.c: In function 'xxh64_update':
   lib/xxhash.c:441:1: warning: the frame size of 1072 bytes is larger than 1024 bytes [-Wframe-larger-than=]
    }
    ^

vim +236 lib/xxhash.c

09c83807 Nick Terrell 2017-07-20  171  
09c83807 Nick Terrell 2017-07-20  172  uint64_t xxh64(const void *input, const size_t len, const uint64_t seed)
09c83807 Nick Terrell 2017-07-20  173  {
09c83807 Nick Terrell 2017-07-20  174  	const uint8_t *p = (const uint8_t *)input;
09c83807 Nick Terrell 2017-07-20  175  	const uint8_t *const b_end = p + len;
09c83807 Nick Terrell 2017-07-20  176  	uint64_t h64;
09c83807 Nick Terrell 2017-07-20  177  
09c83807 Nick Terrell 2017-07-20  178  	if (len >= 32) {
09c83807 Nick Terrell 2017-07-20  179  		const uint8_t *const limit = b_end - 32;
09c83807 Nick Terrell 2017-07-20  180  		uint64_t v1 = seed + PRIME64_1 + PRIME64_2;
09c83807 Nick Terrell 2017-07-20  181  		uint64_t v2 = seed + PRIME64_2;
09c83807 Nick Terrell 2017-07-20  182  		uint64_t v3 = seed + 0;
09c83807 Nick Terrell 2017-07-20  183  		uint64_t v4 = seed - PRIME64_1;
09c83807 Nick Terrell 2017-07-20  184  
09c83807 Nick Terrell 2017-07-20  185  		do {
09c83807 Nick Terrell 2017-07-20  186  			v1 = xxh64_round(v1, get_unaligned_le64(p));
09c83807 Nick Terrell 2017-07-20  187  			p += 8;
09c83807 Nick Terrell 2017-07-20  188  			v2 = xxh64_round(v2, get_unaligned_le64(p));
09c83807 Nick Terrell 2017-07-20  189  			p += 8;
09c83807 Nick Terrell 2017-07-20  190  			v3 = xxh64_round(v3, get_unaligned_le64(p));
09c83807 Nick Terrell 2017-07-20  191  			p += 8;
09c83807 Nick Terrell 2017-07-20  192  			v4 = xxh64_round(v4, get_unaligned_le64(p));
09c83807 Nick Terrell 2017-07-20  193  			p += 8;
09c83807 Nick Terrell 2017-07-20  194  		} while (p <= limit);
09c83807 Nick Terrell 2017-07-20  195  
09c83807 Nick Terrell 2017-07-20  196  		h64 = xxh_rotl64(v1, 1) + xxh_rotl64(v2, 7) +
09c83807 Nick Terrell 2017-07-20  197  			xxh_rotl64(v3, 12) + xxh_rotl64(v4, 18);
09c83807 Nick Terrell 2017-07-20  198  		h64 = xxh64_merge_round(h64, v1);
09c83807 Nick Terrell 2017-07-20  199  		h64 = xxh64_merge_round(h64, v2);
09c83807 Nick Terrell 2017-07-20  200  		h64 = xxh64_merge_round(h64, v3);
09c83807 Nick Terrell 2017-07-20  201  		h64 = xxh64_merge_round(h64, v4);
09c83807 Nick Terrell 2017-07-20  202  
09c83807 Nick Terrell 2017-07-20  203  	} else {
09c83807 Nick Terrell 2017-07-20  204  		h64  = seed + PRIME64_5;
09c83807 Nick Terrell 2017-07-20  205  	}
09c83807 Nick Terrell 2017-07-20  206  
09c83807 Nick Terrell 2017-07-20  207  	h64 += (uint64_t)len;
09c83807 Nick Terrell 2017-07-20  208  
09c83807 Nick Terrell 2017-07-20  209  	while (p + 8 <= b_end) {
09c83807 Nick Terrell 2017-07-20  210  		const uint64_t k1 = xxh64_round(0, get_unaligned_le64(p));
09c83807 Nick Terrell 2017-07-20  211  
09c83807 Nick Terrell 2017-07-20  212  		h64 ^= k1;
09c83807 Nick Terrell 2017-07-20  213  		h64 = xxh_rotl64(h64, 27) * PRIME64_1 + PRIME64_4;
09c83807 Nick Terrell 2017-07-20  214  		p += 8;
09c83807 Nick Terrell 2017-07-20  215  	}
09c83807 Nick Terrell 2017-07-20  216  
09c83807 Nick Terrell 2017-07-20  217  	if (p + 4 <= b_end) {
09c83807 Nick Terrell 2017-07-20  218  		h64 ^= (uint64_t)(get_unaligned_le32(p)) * PRIME64_1;
09c83807 Nick Terrell 2017-07-20  219  		h64 = xxh_rotl64(h64, 23) * PRIME64_2 + PRIME64_3;
09c83807 Nick Terrell 2017-07-20  220  		p += 4;
09c83807 Nick Terrell 2017-07-20  221  	}
09c83807 Nick Terrell 2017-07-20  222  
09c83807 Nick Terrell 2017-07-20  223  	while (p < b_end) {
09c83807 Nick Terrell 2017-07-20  224  		h64 ^= (*p) * PRIME64_5;
09c83807 Nick Terrell 2017-07-20  225  		h64 = xxh_rotl64(h64, 11) * PRIME64_1;
09c83807 Nick Terrell 2017-07-20  226  		p++;
09c83807 Nick Terrell 2017-07-20  227  	}
09c83807 Nick Terrell 2017-07-20  228  
09c83807 Nick Terrell 2017-07-20  229  	h64 ^= h64 >> 33;
09c83807 Nick Terrell 2017-07-20  230  	h64 *= PRIME64_2;
09c83807 Nick Terrell 2017-07-20  231  	h64 ^= h64 >> 29;
09c83807 Nick Terrell 2017-07-20  232  	h64 *= PRIME64_3;
09c83807 Nick Terrell 2017-07-20  233  	h64 ^= h64 >> 32;
09c83807 Nick Terrell 2017-07-20  234  
09c83807 Nick Terrell 2017-07-20  235  	return h64;
09c83807 Nick Terrell 2017-07-20 @236  }
09c83807 Nick Terrell 2017-07-20  237  EXPORT_SYMBOL(xxh64);
09c83807 Nick Terrell 2017-07-20  238  

:::::: The code at line 236 was first introduced by commit
:::::: 09c83807485b28f2b1813bcd3b7e2c7c3720234e lib: Add xxhash module

:::::: TO: Nick Terrell <terrelln@fb.com>
:::::: CC: 0day robot <fengguang.wu@intel.com>

---
0-DAY kernel test infrastructure                Open Source Technology Center
https://lists.01.org/pipermail/kbuild-all                   Intel Corporation
kernel test robot July 24, 2017, 5:25 p.m. UTC | #2
Hi Nick,

[auto build test WARNING on linus/master]
[also build test WARNING on v4.13-rc2 next-20170724]
[if your patch is applied to the wrong git tree, please drop us a note to help improve the system]

url:    https://github.com/0day-ci/linux/commits/Nick-Terrell/Add-xxhash-and-zstd-modules/20170723-092845
config: x86_64-randconfig-a0-07242221 (attached as .config)
compiler: gcc-4.4 (Debian 4.4.7-8) 4.4.7
reproduce:
        # save the attached .config to linux build tree
        make ARCH=x86_64 

All warnings (new ones prefixed by >>):

   lib/zstd/decompress.c: In function 'ZSTD_decodeSequence':
>> lib/zstd/decompress.c:1001: warning: 'seq.match' is used uninitialized in this function

vim +1001 lib/zstd/decompress.c

dc22844b Nick Terrell 2017-07-20   930  
dc22844b Nick Terrell 2017-07-20   931  static seq_t ZSTD_decodeSequence(seqState_t *seqState)
dc22844b Nick Terrell 2017-07-20   932  {
dc22844b Nick Terrell 2017-07-20   933  	seq_t seq;
dc22844b Nick Terrell 2017-07-20   934  
dc22844b Nick Terrell 2017-07-20   935  	U32 const llCode = FSE_peekSymbol(&seqState->stateLL);
dc22844b Nick Terrell 2017-07-20   936  	U32 const mlCode = FSE_peekSymbol(&seqState->stateML);
dc22844b Nick Terrell 2017-07-20   937  	U32 const ofCode = FSE_peekSymbol(&seqState->stateOffb); /* <= maxOff, by table construction */
dc22844b Nick Terrell 2017-07-20   938  
dc22844b Nick Terrell 2017-07-20   939  	U32 const llBits = LL_bits[llCode];
dc22844b Nick Terrell 2017-07-20   940  	U32 const mlBits = ML_bits[mlCode];
dc22844b Nick Terrell 2017-07-20   941  	U32 const ofBits = ofCode;
dc22844b Nick Terrell 2017-07-20   942  	U32 const totalBits = llBits + mlBits + ofBits;
dc22844b Nick Terrell 2017-07-20   943  
dc22844b Nick Terrell 2017-07-20   944  	static const U32 LL_base[MaxLL + 1] = {0,  1,  2,  3,  4,  5,  6,  7,  8,    9,     10,    11,    12,    13,     14,     15,     16,     18,
dc22844b Nick Terrell 2017-07-20   945  					       20, 22, 24, 28, 32, 40, 48, 64, 0x80, 0x100, 0x200, 0x400, 0x800, 0x1000, 0x2000, 0x4000, 0x8000, 0x10000};
dc22844b Nick Terrell 2017-07-20   946  
dc22844b Nick Terrell 2017-07-20   947  	static const U32 ML_base[MaxML + 1] = {3,  4,  5,  6,  7,  8,  9,  10,   11,    12,    13,    14,    15,     16,     17,     18,     19,     20,
dc22844b Nick Terrell 2017-07-20   948  					       21, 22, 23, 24, 25, 26, 27, 28,   29,    30,    31,    32,    33,     34,     35,     37,     39,     41,
dc22844b Nick Terrell 2017-07-20   949  					       43, 47, 51, 59, 67, 83, 99, 0x83, 0x103, 0x203, 0x403, 0x803, 0x1003, 0x2003, 0x4003, 0x8003, 0x10003};
dc22844b Nick Terrell 2017-07-20   950  
dc22844b Nick Terrell 2017-07-20   951  	static const U32 OF_base[MaxOff + 1] = {0,       1,	1,	5,	0xD,      0x1D,      0x3D,      0x7D,      0xFD,     0x1FD,
dc22844b Nick Terrell 2017-07-20   952  						0x3FD,   0x7FD,    0xFFD,    0x1FFD,   0x3FFD,   0x7FFD,    0xFFFD,    0x1FFFD,   0x3FFFD,  0x7FFFD,
dc22844b Nick Terrell 2017-07-20   953  						0xFFFFD, 0x1FFFFD, 0x3FFFFD, 0x7FFFFD, 0xFFFFFD, 0x1FFFFFD, 0x3FFFFFD, 0x7FFFFFD, 0xFFFFFFD};
dc22844b Nick Terrell 2017-07-20   954  
dc22844b Nick Terrell 2017-07-20   955  	/* sequence */
dc22844b Nick Terrell 2017-07-20   956  	{
dc22844b Nick Terrell 2017-07-20   957  		size_t offset;
dc22844b Nick Terrell 2017-07-20   958  		if (!ofCode)
dc22844b Nick Terrell 2017-07-20   959  			offset = 0;
dc22844b Nick Terrell 2017-07-20   960  		else {
dc22844b Nick Terrell 2017-07-20   961  			offset = OF_base[ofCode] + BIT_readBitsFast(&seqState->DStream, ofBits); /* <=  (ZSTD_WINDOWLOG_MAX-1) bits */
dc22844b Nick Terrell 2017-07-20   962  			if (ZSTD_32bits())
dc22844b Nick Terrell 2017-07-20   963  				BIT_reloadDStream(&seqState->DStream);
dc22844b Nick Terrell 2017-07-20   964  		}
dc22844b Nick Terrell 2017-07-20   965  
dc22844b Nick Terrell 2017-07-20   966  		if (ofCode <= 1) {
dc22844b Nick Terrell 2017-07-20   967  			offset += (llCode == 0);
dc22844b Nick Terrell 2017-07-20   968  			if (offset) {
dc22844b Nick Terrell 2017-07-20   969  				size_t temp = (offset == 3) ? seqState->prevOffset[0] - 1 : seqState->prevOffset[offset];
dc22844b Nick Terrell 2017-07-20   970  				temp += !temp; /* 0 is not valid; input is corrupted; force offset to 1 */
dc22844b Nick Terrell 2017-07-20   971  				if (offset != 1)
dc22844b Nick Terrell 2017-07-20   972  					seqState->prevOffset[2] = seqState->prevOffset[1];
dc22844b Nick Terrell 2017-07-20   973  				seqState->prevOffset[1] = seqState->prevOffset[0];
dc22844b Nick Terrell 2017-07-20   974  				seqState->prevOffset[0] = offset = temp;
dc22844b Nick Terrell 2017-07-20   975  			} else {
dc22844b Nick Terrell 2017-07-20   976  				offset = seqState->prevOffset[0];
dc22844b Nick Terrell 2017-07-20   977  			}
dc22844b Nick Terrell 2017-07-20   978  		} else {
dc22844b Nick Terrell 2017-07-20   979  			seqState->prevOffset[2] = seqState->prevOffset[1];
dc22844b Nick Terrell 2017-07-20   980  			seqState->prevOffset[1] = seqState->prevOffset[0];
dc22844b Nick Terrell 2017-07-20   981  			seqState->prevOffset[0] = offset;
dc22844b Nick Terrell 2017-07-20   982  		}
dc22844b Nick Terrell 2017-07-20   983  		seq.offset = offset;
dc22844b Nick Terrell 2017-07-20   984  	}
dc22844b Nick Terrell 2017-07-20   985  
dc22844b Nick Terrell 2017-07-20   986  	seq.matchLength = ML_base[mlCode] + ((mlCode > 31) ? BIT_readBitsFast(&seqState->DStream, mlBits) : 0); /* <=  16 bits */
dc22844b Nick Terrell 2017-07-20   987  	if (ZSTD_32bits() && (mlBits + llBits > 24))
dc22844b Nick Terrell 2017-07-20   988  		BIT_reloadDStream(&seqState->DStream);
dc22844b Nick Terrell 2017-07-20   989  
dc22844b Nick Terrell 2017-07-20   990  	seq.litLength = LL_base[llCode] + ((llCode > 15) ? BIT_readBitsFast(&seqState->DStream, llBits) : 0); /* <=  16 bits */
dc22844b Nick Terrell 2017-07-20   991  	if (ZSTD_32bits() || (totalBits > 64 - 7 - (LLFSELog + MLFSELog + OffFSELog)))
dc22844b Nick Terrell 2017-07-20   992  		BIT_reloadDStream(&seqState->DStream);
dc22844b Nick Terrell 2017-07-20   993  
dc22844b Nick Terrell 2017-07-20   994  	/* ANS state update */
dc22844b Nick Terrell 2017-07-20   995  	FSE_updateState(&seqState->stateLL, &seqState->DStream); /* <=  9 bits */
dc22844b Nick Terrell 2017-07-20   996  	FSE_updateState(&seqState->stateML, &seqState->DStream); /* <=  9 bits */
dc22844b Nick Terrell 2017-07-20   997  	if (ZSTD_32bits())
dc22844b Nick Terrell 2017-07-20   998  		BIT_reloadDStream(&seqState->DStream);		   /* <= 18 bits */
dc22844b Nick Terrell 2017-07-20   999  	FSE_updateState(&seqState->stateOffb, &seqState->DStream); /* <=  8 bits */
dc22844b Nick Terrell 2017-07-20  1000  
dc22844b Nick Terrell 2017-07-20 @1001  	return seq;
dc22844b Nick Terrell 2017-07-20  1002  }
dc22844b Nick Terrell 2017-07-20  1003  

:::::: The code at line 1001 was first introduced by commit
:::::: dc22844bd66de07e2ff858aaebfe678b94df529c lib: Add zstd modules

:::::: TO: Nick Terrell <terrelln@fb.com>
:::::: CC: 0day robot <fengguang.wu@intel.com>

---
0-DAY kernel test infrastructure                Open Source Technology Center
https://lists.01.org/pipermail/kbuild-all                   Intel Corporation
Giovanni Cabiddu July 25, 2017, 11:19 p.m. UTC | #3
Hi Nick,

On Thu, Jul 20, 2017 at 10:27:42PM +0100, Nick Terrell wrote:
> Add zstd compression and decompression support to BtrFS. zstd at its
> fastest level compresses almost as well as zlib, while offering much
> faster compression and decompression, approaching lzo speeds.
Can we look at integrating the zstd implementation below the acomp API
available in the crypto subsystem?
(https://github.com/torvalds/linux/blob/master/crypto/acompress.c)
Acomp was designed to provide a generic and uniform API for compression
in the kernel which hides algorithm specific details to frameworks.
In future it would be nice to see btrfs using exclusively acomp
for compression. This way when a new compression algorithm is exposed
through acomp, it will be available immediately in btrfs.
Furthermore, any framework in the kernel that will use acomp will be
automatically enabled to use zstd.
What do you think?

Here is a prototype that shows how btrfs can be integrated with
acomp: https://patchwork.kernel.org/patch/9201741/

Regards,
David Sterba Aug. 18, 2017, 3:37 p.m. UTC | #4
On Wed, Jul 26, 2017 at 12:19:29AM +0100, Giovanni Cabiddu wrote:
> Hi Nick,
> 
> On Thu, Jul 20, 2017 at 10:27:42PM +0100, Nick Terrell wrote:
> > Add zstd compression and decompression support to BtrFS. zstd at its
> > fastest level compresses almost as well as zlib, while offering much
> > faster compression and decompression, approaching lzo speeds.
> Can we look at integrating the zstd implementation below the acomp API
> available in the crypto subsystem?
> (https://github.com/torvalds/linux/blob/master/crypto/acompress.c)
> Acomp was designed to provide a generic and uniform API for compression
> in the kernel which hides algorithm specific details to frameworks.
> In future it would be nice to see btrfs using exclusively acomp
> for compression. This way when a new compression algorithm is exposed
> through acomp, it will be available immediately in btrfs.

The compression algorithm is considered part of the filesystem format
and must be covered by an incompatibility bit. Regarding the acomp API,
I don't see much point adding the extra abstraction layer for the two
currently supported algorithms. We only call 2 functions, for
compression and decompression, no need to allocate the acomp helper
structures and call the functions indirectly.
--
To unsubscribe from this list: send the line "unsubscribe linux-btrfs" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Patch
diff mbox

diff --git a/fs/btrfs/Kconfig b/fs/btrfs/Kconfig
index 80e9c18..a26c63b 100644
--- a/fs/btrfs/Kconfig
+++ b/fs/btrfs/Kconfig
@@ -6,6 +6,8 @@  config BTRFS_FS
 	select ZLIB_DEFLATE
 	select LZO_COMPRESS
 	select LZO_DECOMPRESS
+	select ZSTD_COMPRESS
+	select ZSTD_DECOMPRESS
 	select RAID6_PQ
 	select XOR_BLOCKS
 	select SRCU
diff --git a/fs/btrfs/Makefile b/fs/btrfs/Makefile
index 128ce17..962a95a 100644
--- a/fs/btrfs/Makefile
+++ b/fs/btrfs/Makefile
@@ -6,7 +6,7 @@  btrfs-y += super.o ctree.o extent-tree.o print-tree.o root-tree.o dir-item.o \
 	   transaction.o inode.o file.o tree-defrag.o \
 	   extent_map.o sysfs.o struct-funcs.o xattr.o ordered-data.o \
 	   extent_io.o volumes.o async-thread.o ioctl.o locking.o orphan.o \
-	   export.o tree-log.o free-space-cache.o zlib.o lzo.o \
+	   export.o tree-log.o free-space-cache.o zlib.o lzo.o zstd.o \
 	   compression.o delayed-ref.o relocation.o delayed-inode.o scrub.o \
 	   reada.o backref.o ulist.o qgroup.o send.o dev-replace.o raid56.o \
 	   uuid-tree.o props.o hash.o free-space-tree.o
diff --git a/fs/btrfs/compression.c b/fs/btrfs/compression.c
index d2ef9ac..4ff42d1 100644
--- a/fs/btrfs/compression.c
+++ b/fs/btrfs/compression.c
@@ -704,6 +704,7 @@  static struct {
 static const struct btrfs_compress_op * const btrfs_compress_op[] = {
 	&btrfs_zlib_compress,
 	&btrfs_lzo_compress,
+	&btrfs_zstd_compress,
 };

 void __init btrfs_init_compress(void)
diff --git a/fs/btrfs/compression.h b/fs/btrfs/compression.h
index 87f6d33..2269e00 100644
--- a/fs/btrfs/compression.h
+++ b/fs/btrfs/compression.h
@@ -99,8 +99,9 @@  enum btrfs_compression_type {
 	BTRFS_COMPRESS_NONE  = 0,
 	BTRFS_COMPRESS_ZLIB  = 1,
 	BTRFS_COMPRESS_LZO   = 2,
-	BTRFS_COMPRESS_TYPES = 2,
-	BTRFS_COMPRESS_LAST  = 3,
+	BTRFS_COMPRESS_ZSTD  = 3,
+	BTRFS_COMPRESS_TYPES = 3,
+	BTRFS_COMPRESS_LAST  = 4,
 };

 struct btrfs_compress_op {
@@ -128,5 +129,6 @@  struct btrfs_compress_op {

 extern const struct btrfs_compress_op btrfs_zlib_compress;
 extern const struct btrfs_compress_op btrfs_lzo_compress;
+extern const struct btrfs_compress_op btrfs_zstd_compress;

 #endif
diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h
index 3f3eb7b..845d77c 100644
--- a/fs/btrfs/ctree.h
+++ b/fs/btrfs/ctree.h
@@ -270,6 +270,7 @@  struct btrfs_super_block {
 	 BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS |		\
 	 BTRFS_FEATURE_INCOMPAT_BIG_METADATA |		\
 	 BTRFS_FEATURE_INCOMPAT_COMPRESS_LZO |		\
+	 BTRFS_FEATURE_INCOMPAT_COMPRESS_ZSTD |		\
 	 BTRFS_FEATURE_INCOMPAT_RAID56 |		\
 	 BTRFS_FEATURE_INCOMPAT_EXTENDED_IREF |		\
 	 BTRFS_FEATURE_INCOMPAT_SKINNY_METADATA |	\
diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c
index 080e2eb..04632f4 100644
--- a/fs/btrfs/disk-io.c
+++ b/fs/btrfs/disk-io.c
@@ -2828,6 +2828,8 @@  int open_ctree(struct super_block *sb,
 	features |= BTRFS_FEATURE_INCOMPAT_MIXED_BACKREF;
 	if (fs_info->compress_type == BTRFS_COMPRESS_LZO)
 		features |= BTRFS_FEATURE_INCOMPAT_COMPRESS_LZO;
+	else if (fs_info->compress_type == BTRFS_COMPRESS_ZSTD)
+		features |= BTRFS_FEATURE_INCOMPAT_COMPRESS_ZSTD;

 	if (features & BTRFS_FEATURE_INCOMPAT_SKINNY_METADATA)
 		btrfs_info(fs_info, "has skinny extents");
diff --git a/fs/btrfs/ioctl.c b/fs/btrfs/ioctl.c
index fa1b78c..b9963d9 100644
--- a/fs/btrfs/ioctl.c
+++ b/fs/btrfs/ioctl.c
@@ -327,8 +327,10 @@  static int btrfs_ioctl_setflags(struct file *file, void __user *arg)

 		if (fs_info->compress_type == BTRFS_COMPRESS_LZO)
 			comp = "lzo";
-		else
+		else if (fs_info->compress_type == BTRFS_COMPRESS_ZLIB)
 			comp = "zlib";
+		else
+			comp = "zstd";
 		ret = btrfs_set_prop(inode, "btrfs.compression",
 				     comp, strlen(comp), 0);
 		if (ret)
@@ -1466,6 +1468,8 @@  int btrfs_defrag_file(struct inode *inode, struct file *file,

 	if (range->compress_type == BTRFS_COMPRESS_LZO) {
 		btrfs_set_fs_incompat(fs_info, COMPRESS_LZO);
+	} else if (range->compress_type == BTRFS_COMPRESS_ZSTD) {
+		btrfs_set_fs_incompat(fs_info, COMPRESS_ZSTD);
 	}

 	ret = defrag_count;
diff --git a/fs/btrfs/props.c b/fs/btrfs/props.c
index 4b23ae5..20631e9 100644
--- a/fs/btrfs/props.c
+++ b/fs/btrfs/props.c
@@ -390,6 +390,8 @@  static int prop_compression_validate(const char *value, size_t len)
 		return 0;
 	else if (!strncmp("zlib", value, len))
 		return 0;
+	else if (!strncmp("zstd", value, len))
+		return 0;

 	return -EINVAL;
 }
@@ -412,6 +414,8 @@  static int prop_compression_apply(struct inode *inode,
 		type = BTRFS_COMPRESS_LZO;
 	else if (!strncmp("zlib", value, len))
 		type = BTRFS_COMPRESS_ZLIB;
+	else if (!strncmp("zstd", value, len))
+		type = BTRFS_COMPRESS_ZSTD;
 	else
 		return -EINVAL;

@@ -429,6 +433,8 @@  static const char *prop_compression_extract(struct inode *inode)
 		return "zlib";
 	case BTRFS_COMPRESS_LZO:
 		return "lzo";
+	case BTRFS_COMPRESS_ZSTD:
+		return "zstd";
 	}

 	return NULL;
diff --git a/fs/btrfs/super.c b/fs/btrfs/super.c
index 12540b6..c370dea 100644
--- a/fs/btrfs/super.c
+++ b/fs/btrfs/super.c
@@ -513,6 +513,14 @@  int btrfs_parse_options(struct btrfs_fs_info *info, char *options,
 				btrfs_clear_opt(info->mount_opt, NODATASUM);
 				btrfs_set_fs_incompat(info, COMPRESS_LZO);
 				no_compress = 0;
+			} else if (strcmp(args[0].from, "zstd") == 0) {
+				compress_type = "zstd";
+				info->compress_type = BTRFS_COMPRESS_ZSTD;
+				btrfs_set_opt(info->mount_opt, COMPRESS);
+				btrfs_clear_opt(info->mount_opt, NODATACOW);
+				btrfs_clear_opt(info->mount_opt, NODATASUM);
+				btrfs_set_fs_incompat(info, COMPRESS_ZSTD);
+				no_compress = 0;
 			} else if (strncmp(args[0].from, "no", 2) == 0) {
 				compress_type = "no";
 				btrfs_clear_opt(info->mount_opt, COMPRESS);
@@ -1227,8 +1235,10 @@  static int btrfs_show_options(struct seq_file *seq, struct dentry *dentry)
 	if (btrfs_test_opt(info, COMPRESS)) {
 		if (info->compress_type == BTRFS_COMPRESS_ZLIB)
 			compress_type = "zlib";
-		else
+		else if (info->compress_type == BTRFS_COMPRESS_LZO)
 			compress_type = "lzo";
+		else
+			compress_type = "zstd";
 		if (btrfs_test_opt(info, FORCE_COMPRESS))
 			seq_printf(seq, ",compress-force=%s", compress_type);
 		else
diff --git a/fs/btrfs/sysfs.c b/fs/btrfs/sysfs.c
index c2d5f35..2b6d37c 100644
--- a/fs/btrfs/sysfs.c
+++ b/fs/btrfs/sysfs.c
@@ -200,6 +200,7 @@  BTRFS_FEAT_ATTR_INCOMPAT(mixed_backref, MIXED_BACKREF);
 BTRFS_FEAT_ATTR_INCOMPAT(default_subvol, DEFAULT_SUBVOL);
 BTRFS_FEAT_ATTR_INCOMPAT(mixed_groups, MIXED_GROUPS);
 BTRFS_FEAT_ATTR_INCOMPAT(compress_lzo, COMPRESS_LZO);
+BTRFS_FEAT_ATTR_INCOMPAT(compress_zstd, COMPRESS_ZSTD);
 BTRFS_FEAT_ATTR_INCOMPAT(big_metadata, BIG_METADATA);
 BTRFS_FEAT_ATTR_INCOMPAT(extended_iref, EXTENDED_IREF);
 BTRFS_FEAT_ATTR_INCOMPAT(raid56, RAID56);
@@ -212,6 +213,7 @@  static struct attribute *btrfs_supported_feature_attrs[] = {
 	BTRFS_FEAT_ATTR_PTR(default_subvol),
 	BTRFS_FEAT_ATTR_PTR(mixed_groups),
 	BTRFS_FEAT_ATTR_PTR(compress_lzo),
+	BTRFS_FEAT_ATTR_PTR(compress_zstd),
 	BTRFS_FEAT_ATTR_PTR(big_metadata),
 	BTRFS_FEAT_ATTR_PTR(extended_iref),
 	BTRFS_FEAT_ATTR_PTR(raid56),
diff --git a/fs/btrfs/zstd.c b/fs/btrfs/zstd.c
new file mode 100644
index 0000000..1822068
--- /dev/null
+++ b/fs/btrfs/zstd.c
@@ -0,0 +1,435 @@ 
+/*
+ * Copyright (c) 2016-present, Facebook, Inc.
+ * All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License v2 as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public
+ * License along with this program; if not, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+#include <linux/kernel.h>
+#include <linux/slab.h>
+#include <linux/vmalloc.h>
+#include <linux/init.h>
+#include <linux/err.h>
+#include <linux/sched.h>
+#include <linux/pagemap.h>
+#include <linux/bio.h>
+#include <linux/zstd.h>
+#include "compression.h"
+
+#define ZSTD_BTRFS_MAX_WINDOWLOG 17
+#define ZSTD_BTRFS_MAX_INPUT (1 << ZSTD_BTRFS_MAX_WINDOWLOG)
+#define ZSTD_BTRFS_DEFAULT_LEVEL 3
+
+static ZSTD_parameters zstd_get_btrfs_parameters(size_t src_len)
+{
+	ZSTD_parameters params = ZSTD_getParams(ZSTD_BTRFS_DEFAULT_LEVEL, src_len, 0);
+
+	if (params.cParams.windowLog > ZSTD_BTRFS_MAX_WINDOWLOG)
+		params.cParams.windowLog = ZSTD_BTRFS_MAX_WINDOWLOG;
+	WARN_ON(src_len > ZSTD_BTRFS_MAX_INPUT);
+	return params;
+}
+
+struct workspace {
+	void *mem;
+	size_t size;
+	char *buf;
+	struct list_head list;
+};
+
+static void zstd_free_workspace(struct list_head *ws)
+{
+	struct workspace *workspace = list_entry(ws, struct workspace, list);
+
+	kvfree(workspace->mem);
+	kfree(workspace->buf);
+	kfree(workspace);
+}
+
+static struct list_head *zstd_alloc_workspace(void)
+{
+	ZSTD_parameters params =
+			zstd_get_btrfs_parameters(ZSTD_BTRFS_MAX_INPUT);
+	struct workspace *workspace;
+
+	workspace = kzalloc(sizeof(*workspace), GFP_KERNEL);
+	if (!workspace)
+		return ERR_PTR(-ENOMEM);
+
+	workspace->size = max_t(size_t,
+			ZSTD_CStreamWorkspaceBound(params.cParams),
+			ZSTD_DStreamWorkspaceBound(ZSTD_BTRFS_MAX_INPUT));
+	workspace->mem = kvmalloc(workspace->size, GFP_KERNEL);
+	workspace->buf = kmalloc(PAGE_SIZE, GFP_KERNEL);
+	if (!workspace->mem || !workspace->buf)
+		goto fail;
+
+	INIT_LIST_HEAD(&workspace->list);
+
+	return &workspace->list;
+fail:
+	zstd_free_workspace(&workspace->list);
+	return ERR_PTR(-ENOMEM);
+}
+
+static int zstd_compress_pages(struct list_head *ws,
+		struct address_space *mapping,
+		u64 start,
+		struct page **pages,
+		unsigned long *out_pages,
+		unsigned long *total_in,
+		unsigned long *total_out)
+{
+	struct workspace *workspace = list_entry(ws, struct workspace, list);
+	ZSTD_CStream *stream;
+	int ret = 0;
+	int nr_pages = 0;
+	struct page *in_page = NULL;  /* The current page to read */
+	struct page *out_page = NULL; /* The current page to write to */
+	ZSTD_inBuffer in_buf = { NULL, 0, 0 };
+	ZSTD_outBuffer out_buf = { NULL, 0, 0 };
+	unsigned long tot_in = 0;
+	unsigned long tot_out = 0;
+	unsigned long len = *total_out;
+	const unsigned long nr_dest_pages = *out_pages;
+	unsigned long max_out = nr_dest_pages * PAGE_SIZE;
+	ZSTD_parameters params = zstd_get_btrfs_parameters(len);
+
+	*out_pages = 0;
+	*total_out = 0;
+	*total_in = 0;
+
+	/* Initialize the stream */
+	stream = ZSTD_initCStream(params, len, workspace->mem,
+			workspace->size);
+	if (!stream) {
+		pr_warn("BTRFS: ZSTD_initCStream failed\n");
+		ret = -EIO;
+		goto out;
+	}
+
+	/* map in the first page of input data */
+	in_page = find_get_page(mapping, start >> PAGE_SHIFT);
+	in_buf.src = kmap(in_page);
+	in_buf.pos = 0;
+	in_buf.size = min_t(size_t, len, PAGE_SIZE);
+
+
+	/* Allocate and map in the output buffer */
+	out_page = alloc_page(GFP_NOFS | __GFP_HIGHMEM);
+	if (out_page == NULL) {
+		ret = -ENOMEM;
+		goto out;
+	}
+	pages[nr_pages++] = out_page;
+	out_buf.dst = kmap(out_page);
+	out_buf.pos = 0;
+	out_buf.size = min_t(size_t, max_out, PAGE_SIZE);
+
+	while (1) {
+		size_t ret2;
+
+		ret2 = ZSTD_compressStream(stream, &out_buf, &in_buf);
+		if (ZSTD_isError(ret2)) {
+			pr_debug("BTRFS: ZSTD_compressStream returned %d\n",
+					ZSTD_getErrorCode(ret2));
+			ret = -EIO;
+			goto out;
+		}
+
+		/* Check to see if we are making it bigger */
+		if (tot_in + in_buf.pos > 8192 &&
+				tot_in + in_buf.pos <
+				tot_out + out_buf.pos) {
+			ret = -E2BIG;
+			goto out;
+		}
+
+		/* We've reached the end of our output range */
+		if (out_buf.pos >= max_out) {
+			tot_out += out_buf.pos;
+			ret = -E2BIG;
+			goto out;
+		}
+
+		/* Check if we need more output space */
+		if (out_buf.pos == out_buf.size) {
+			tot_out += PAGE_SIZE;
+			max_out -= PAGE_SIZE;
+			kunmap(out_page);
+			if (nr_pages == nr_dest_pages) {
+				out_page = NULL;
+				ret = -E2BIG;
+				goto out;
+			}
+			out_page = alloc_page(GFP_NOFS | __GFP_HIGHMEM);
+			if (out_page == NULL) {
+				ret = -ENOMEM;
+				goto out;
+			}
+			pages[nr_pages++] = out_page;
+			out_buf.dst = kmap(out_page);
+			out_buf.pos = 0;
+			out_buf.size = min_t(size_t, max_out, PAGE_SIZE);
+		}
+
+		/* We've reached the end of the input */
+		if (in_buf.pos >= len) {
+			tot_in += in_buf.pos;
+			break;
+		}
+
+		/* Check if we need more input */
+		if (in_buf.pos == in_buf.size) {
+			tot_in += PAGE_SIZE;
+			kunmap(in_page);
+			put_page(in_page);
+
+			start += PAGE_SIZE;
+			len -= PAGE_SIZE;
+			in_page = find_get_page(mapping, start >> PAGE_SHIFT);
+			in_buf.src = kmap(in_page);
+			in_buf.pos = 0;
+			in_buf.size = min_t(size_t, len, PAGE_SIZE);
+		}
+	}
+	while (1) {
+		size_t ret2;
+
+		ret2 = ZSTD_endStream(stream, &out_buf);
+		if (ZSTD_isError(ret2)) {
+			pr_debug("BTRFS: ZSTD_endStream returned %d\n",
+					ZSTD_getErrorCode(ret2));
+			ret = -EIO;
+			goto out;
+		}
+		if (ret2 == 0) {
+			tot_out += out_buf.pos;
+			break;
+		}
+		if (out_buf.pos >= max_out) {
+			tot_out += out_buf.pos;
+			ret = -E2BIG;
+			goto out;
+		}
+
+		tot_out += PAGE_SIZE;
+		max_out -= PAGE_SIZE;
+		kunmap(out_page);
+		if (nr_pages == nr_dest_pages) {
+			out_page = NULL;
+			ret = -E2BIG;
+			goto out;
+		}
+		out_page = alloc_page(GFP_NOFS | __GFP_HIGHMEM);
+		if (out_page == NULL) {
+			ret = -ENOMEM;
+			goto out;
+		}
+		pages[nr_pages++] = out_page;
+		out_buf.dst = kmap(out_page);
+		out_buf.pos = 0;
+		out_buf.size = min_t(size_t, max_out, PAGE_SIZE);
+	}
+
+	if (tot_out >= tot_in) {
+		ret = -E2BIG;
+		goto out;
+	}
+
+	ret = 0;
+	*total_in = tot_in;
+	*total_out = tot_out;
+out:
+	*out_pages = nr_pages;
+	/* Cleanup */
+	if (in_page) {
+		kunmap(in_page);
+		put_page(in_page);
+	}
+	if (out_page)
+		kunmap(out_page);
+	return ret;
+}
+
+static int zstd_decompress_bio(struct list_head *ws, struct compressed_bio *cb)
+{
+	struct workspace *workspace = list_entry(ws, struct workspace, list);
+	struct page **pages_in = cb->compressed_pages;
+	u64 disk_start = cb->start;
+	struct bio *orig_bio = cb->orig_bio;
+	size_t srclen = cb->compressed_len;
+	ZSTD_DStream *stream;
+	int ret = 0;
+	unsigned long page_in_index = 0;
+	unsigned long total_pages_in = DIV_ROUND_UP(srclen, PAGE_SIZE);
+	unsigned long buf_start;
+	unsigned long total_out = 0;
+	ZSTD_inBuffer in_buf = { NULL, 0, 0 };
+	ZSTD_outBuffer out_buf = { NULL, 0, 0 };
+
+	stream = ZSTD_initDStream(
+			ZSTD_BTRFS_MAX_INPUT, workspace->mem, workspace->size);
+	if (!stream) {
+		pr_debug("BTRFS: ZSTD_initDStream failed\n");
+		ret = -EIO;
+		goto done;
+	}
+
+	in_buf.src = kmap(pages_in[page_in_index]);
+	in_buf.pos = 0;
+	in_buf.size = min_t(size_t, srclen, PAGE_SIZE);
+
+	out_buf.dst = workspace->buf;
+	out_buf.pos = 0;
+	out_buf.size = PAGE_SIZE;
+
+	while (1) {
+		size_t ret2;
+
+		ret2 = ZSTD_decompressStream(stream, &out_buf, &in_buf);
+		if (ZSTD_isError(ret2)) {
+			pr_debug("BTRFS: ZSTD_decompressStream returned %d\n",
+					ZSTD_getErrorCode(ret2));
+			ret = -EIO;
+			goto done;
+		}
+		buf_start = total_out;
+		total_out += out_buf.pos;
+		out_buf.pos = 0;
+
+		ret = btrfs_decompress_buf2page(out_buf.dst, buf_start,
+				total_out, disk_start, orig_bio);
+		if (ret == 0)
+			break;
+
+		if (in_buf.pos >= srclen)
+			break;
+
+		/* Check if we've hit the end of a frame */
+		if (ret2 == 0)
+			break;
+
+		if (in_buf.pos == in_buf.size) {
+			kunmap(pages_in[page_in_index++]);
+			if (page_in_index >= total_pages_in) {
+				in_buf.src = NULL;
+				ret = -EIO;
+				goto done;
+			}
+			srclen -= PAGE_SIZE;
+			in_buf.src = kmap(pages_in[page_in_index]);
+			in_buf.pos = 0;
+			in_buf.size = min_t(size_t, srclen, PAGE_SIZE);
+		}
+	}
+	ret = 0;
+	zero_fill_bio(orig_bio);
+done:
+	if (in_buf.src)
+		kunmap(pages_in[page_in_index]);
+	return ret;
+}
+
+static int zstd_decompress(struct list_head *ws, unsigned char *data_in,
+		struct page *dest_page,
+		unsigned long start_byte,
+		size_t srclen, size_t destlen)
+{
+	struct workspace *workspace = list_entry(ws, struct workspace, list);
+	ZSTD_DStream *stream;
+	int ret = 0;
+	size_t ret2;
+	ZSTD_inBuffer in_buf = { NULL, 0, 0 };
+	ZSTD_outBuffer out_buf = { NULL, 0, 0 };
+	unsigned long total_out = 0;
+	unsigned long pg_offset = 0;
+	char *kaddr;
+
+	stream = ZSTD_initDStream(
+			ZSTD_BTRFS_MAX_INPUT, workspace->mem, workspace->size);
+	if (!stream) {
+		pr_warn("BTRFS: ZSTD_initDStream failed\n");
+		ret = -EIO;
+		goto finish;
+	}
+
+	destlen = min_t(size_t, destlen, PAGE_SIZE);
+
+	in_buf.src = data_in;
+	in_buf.pos = 0;
+	in_buf.size = srclen;
+
+	out_buf.dst = workspace->buf;
+	out_buf.pos = 0;
+	out_buf.size = PAGE_SIZE;
+
+	ret2 = 1;
+	while (pg_offset < destlen && in_buf.pos < in_buf.size) {
+		unsigned long buf_start;
+		unsigned long buf_offset;
+		unsigned long bytes;
+
+		/* Check if the frame is over and we still need more input */
+		if (ret2 == 0) {
+			pr_debug("BTRFS: ZSTD_decompressStream ended early\n");
+			ret = -EIO;
+			goto finish;
+		}
+		ret2 = ZSTD_decompressStream(stream, &out_buf, &in_buf);
+		if (ZSTD_isError(ret2)) {
+			pr_debug("BTRFS: ZSTD_decompressStream returned %d\n",
+					ZSTD_getErrorCode(ret2));
+			ret = -EIO;
+			goto finish;
+		}
+
+		buf_start = total_out;
+		total_out += out_buf.pos;
+		out_buf.pos = 0;
+
+		if (total_out <= start_byte)
+			continue;
+
+		if (total_out > start_byte && buf_start < start_byte)
+			buf_offset = start_byte - buf_start;
+		else
+			buf_offset = 0;
+
+		bytes = min_t(unsigned long, destlen - pg_offset,
+				out_buf.size - buf_offset);
+
+		kaddr = kmap_atomic(dest_page);
+		memcpy(kaddr + pg_offset, out_buf.dst + buf_offset, bytes);
+		kunmap_atomic(kaddr);
+
+		pg_offset += bytes;
+	}
+	ret = 0;
+finish:
+	if (pg_offset < destlen) {
+		kaddr = kmap_atomic(dest_page);
+		memset(kaddr + pg_offset, 0, destlen - pg_offset);
+		kunmap_atomic(kaddr);
+	}
+	return ret;
+}
+
+const struct btrfs_compress_op btrfs_zstd_compress = {
+	.alloc_workspace = zstd_alloc_workspace,
+	.free_workspace = zstd_free_workspace,
+	.compress_pages = zstd_compress_pages,
+	.decompress_bio = zstd_decompress_bio,
+	.decompress = zstd_decompress,
+};
diff --git a/include/uapi/linux/btrfs.h b/include/uapi/linux/btrfs.h
index 9aa74f3..378230c 100644
--- a/include/uapi/linux/btrfs.h
+++ b/include/uapi/linux/btrfs.h
@@ -255,13 +255,7 @@  struct btrfs_ioctl_fs_info_args {
 #define BTRFS_FEATURE_INCOMPAT_DEFAULT_SUBVOL	(1ULL << 1)
 #define BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS	(1ULL << 2)
 #define BTRFS_FEATURE_INCOMPAT_COMPRESS_LZO	(1ULL << 3)
-/*
- * some patches floated around with a second compression method
- * lets save that incompat here for when they do get in
- * Note we don't actually support it, we're just reserving the
- * number
- */
-#define BTRFS_FEATURE_INCOMPAT_COMPRESS_LZOv2	(1ULL << 4)
+#define BTRFS_FEATURE_INCOMPAT_COMPRESS_ZSTD	(1ULL << 4)

 /*
  * older kernels tried to do bigger metadata blocks, but the