Add zstd support to zfs

This PR adds two new compression types, based on ZStandard:

- zstd: A basic ZStandard compression algorithm Available compression.
  Levels for zstd are zstd-1 through zstd-19, where the compression
  increases with every level, but speed decreases.

- zstd-fast: A faster version of the ZStandard compression algorithm
  zstd-fast is basically a "negative" level of zstd. The compression
  decreases with every level, but speed increases.

  Available compression levels for zstd-fast:
   - zstd-fast-1 through zstd-fast-10
   - zstd-fast-20 through zstd-fast-100 (in increments of 10)
   - zstd-fast-500 and zstd-fast-1000

For more information check the man page.

Implementation details:

Rather than treat each level of zstd as a different algorithm (as was
done historically with gzip), the block pointer `enum zio_compress`
value is simply zstd for all levels, including zstd-fast, since they all
use the same decompression function.

The compress= property (a 64bit unsigned integer) uses the lower 7 bits
to store the compression algorithm (matching the number of bits used in
a block pointer, as the 8th bit was borrowed for embedded block
pointers).  The upper bits are used to store the compression level.

It is necessary to be able to determine what compression level was used
when later reading a block back, so the concept used in LZ4, where the
first 32bits of the on-disk value are the size of the compressed data
(since the allocation is rounded up to the nearest ashift), was
extended, and we store the version of ZSTD and the level as well as the
compressed size. This value is returned when decompressing a block, so
that if the block needs to be recompressed (L2ARC, nop-write, etc), that
the same parameters will be used to result in the matching checksum.

All of the internal ZFS code ( `arc_buf_hdr_t`, `objset_t`,
`zio_prop_t`, etc.) uses the separated _compress and _complevel
variables.  Only the properties ZAP contains the combined/bit-shifted
value. The combined value is split when the compression_changed_cb()
callback is called, and sets both objset members (os_compress and
os_complevel).

The userspace tools all use the combined/bit-shifted value.

Additional notes:

zdb can now also decode the ZSTD compression header (flag -Z) and
inspect the size, version and compression level saved in that header.
For each record, if it is ZSTD compressed, the parameters of the decoded
compression header get printed.

ZSTD is included with all current tests and new tests are added
as-needed.

Per-dataset feature flags now get activated when the property is set.
If a compression algorithm requires a feature flag, zfs activates the
feature when the property is set, rather than waiting for the first
block to be born.  This is currently only used by zstd but can be
extended as needed.

Portions-Sponsored-By: The FreeBSD Foundation
Co-authored-by: Allan Jude <[email protected]>
Co-authored-by: Brian Behlendorf <[email protected]>
Co-authored-by: Sebastian Gottschall <[email protected]>
Co-authored-by: Kjeld Schouten-Lebbing <[email protected]>
Co-authored-by: Michael Niewöhner <[email protected]>
Signed-off-by: Allan Jude <[email protected]>
Signed-off-by: Allan Jude <[email protected]>
Signed-off-by: Brian Behlendorf <[email protected]>
Signed-off-by: Sebastian Gottschall <[email protected]>
Signed-off-by: Kjeld Schouten-Lebbing <[email protected]>
Signed-off-by: Michael Niewöhner <[email protected]>
Closes #6247
Closes #9024
Closes #10277
Closes #10278

1 files changed, 54 insertions, 22 deletions

diff --git a/module/zfs/arc.c b/module/zfs/arc.c
index 0512497d5..ff2621194 100644
--- a/module/zfs/arc.c
+++ b/module/zfs/arc.c
@@ -26,6 +26,8 @@
  * Copyright (c) 2017, Nexenta Systems, Inc.  All rights reserved.
  * Copyright (c) 2019, loli10K <[email protected]>. All rights reserved.
  * Copyright (c) 2020, George Amanakis. All rights reserved.
+ * Copyright (c) 2019, Klara Inc.
+ * Copyright (c) 2019, Allan Jude
  * Copyright (c) 2020, The FreeBSD Foundation [1]
  *
  * [1] Portions of this software were developed by Allan Jude
@@ -1362,6 +1364,12 @@ arc_hdr_get_compress(arc_buf_hdr_t *hdr)
 	    HDR_GET_COMPRESS(hdr) : ZIO_COMPRESS_OFF);
 }
 
+uint8_t
+arc_get_complevel(arc_buf_t *buf)
+{
+	return (buf->b_hdr->b_complevel);
+}
+
 static inline boolean_t
 arc_buf_is_shared(arc_buf_t *buf)
 {
@@ -1707,7 +1715,8 @@ arc_buf_try_copy_decompressed_data(arc_buf_t *buf)
 static arc_buf_hdr_t *
 arc_buf_alloc_l2only(size_t size, arc_buf_contents_t type, l2arc_dev_t *dev,
     dva_t dva, uint64_t daddr, int32_t psize, uint64_t birth,
-    enum zio_compress compress, boolean_t protected, boolean_t prefetch)
+    enum zio_compress compress, uint8_t complevel, boolean_t protected,
+    boolean_t prefetch)
 {
 	arc_buf_hdr_t	*hdr;
 
@@ -1720,6 +1729,7 @@ arc_buf_alloc_l2only(size_t size, arc_buf_contents_t type, l2arc_dev_t *dev,
 	HDR_SET_LSIZE(hdr, size);
 	HDR_SET_PSIZE(hdr, psize);
 	arc_hdr_set_compress(hdr, compress);
+	hdr->b_complevel = complevel;
 	if (protected)
 		arc_hdr_set_flags(hdr, ARC_FLAG_PROTECTED);
 	if (prefetch)
@@ -1779,9 +1789,8 @@ arc_hdr_authenticate(arc_buf_hdr_t *hdr, spa_t *spa, uint64_t dsobj)
 		tmpbuf = zio_buf_alloc(lsize);
 		abd = abd_get_from_buf(tmpbuf, lsize);
 		abd_take_ownership_of_buf(abd, B_TRUE);
-
 		csize = zio_compress_data(HDR_GET_COMPRESS(hdr),
-		    hdr->b_l1hdr.b_pabd, tmpbuf, lsize);
+		    hdr->b_l1hdr.b_pabd, tmpbuf, lsize, hdr->b_complevel);
 		ASSERT3U(csize, <=, psize);
 		abd_zero_off(abd, csize, psize - csize);
 	}
@@ -1867,7 +1876,7 @@ arc_hdr_decrypt(arc_buf_hdr_t *hdr, spa_t *spa, const zbookmark_phys_t *zb)
 
 		ret = zio_decompress_data(HDR_GET_COMPRESS(hdr),
 		    hdr->b_l1hdr.b_pabd, tmp, HDR_GET_PSIZE(hdr),
-		    HDR_GET_LSIZE(hdr));
+		    HDR_GET_LSIZE(hdr), &hdr->b_complevel);
 		if (ret != 0) {
 			abd_return_buf(cabd, tmp, arc_hdr_size(hdr));
 			goto error;
@@ -2114,7 +2123,8 @@ arc_buf_fill(arc_buf_t *buf, spa_t *spa, const zbookmark_phys_t *zb,
 		} else {
 			error = zio_decompress_data(HDR_GET_COMPRESS(hdr),
 			    hdr->b_l1hdr.b_pabd, buf->b_data,
-			    HDR_GET_PSIZE(hdr), HDR_GET_LSIZE(hdr));
+			    HDR_GET_PSIZE(hdr), HDR_GET_LSIZE(hdr),
+			    &hdr->b_complevel);
 
 			/*
 			 * Absent hardware errors or software bugs, this should
@@ -2865,10 +2875,10 @@ arc_loan_buf(spa_t *spa, boolean_t is_metadata, int size)
 
 arc_buf_t *
 arc_loan_compressed_buf(spa_t *spa, uint64_t psize, uint64_t lsize,
-    enum zio_compress compression_type)
+    enum zio_compress compression_type, uint8_t complevel)
 {
 	arc_buf_t *buf = arc_alloc_compressed_buf(spa, arc_onloan_tag,
-	    psize, lsize, compression_type);
+	    psize, lsize, compression_type, complevel);
 
 	arc_loaned_bytes_update(arc_buf_size(buf));
 
@@ -2879,10 +2889,11 @@ arc_buf_t *
 arc_loan_raw_buf(spa_t *spa, uint64_t dsobj, boolean_t byteorder,
     const uint8_t *salt, const uint8_t *iv, const uint8_t *mac,
     dmu_object_type_t ot, uint64_t psize, uint64_t lsize,
-    enum zio_compress compression_type)
+    enum zio_compress compression_type, uint8_t complevel)
 {
 	arc_buf_t *buf = arc_alloc_raw_buf(spa, arc_onloan_tag, dsobj,
-	    byteorder, salt, iv, mac, ot, psize, lsize, compression_type);
+	    byteorder, salt, iv, mac, ot, psize, lsize, compression_type,
+	    complevel);
 
 	atomic_add_64(&arc_loaned_bytes, psize);
 	return (buf);
@@ -3249,7 +3260,7 @@ arc_hdr_free_abd(arc_buf_hdr_t *hdr, boolean_t free_rdata)
 
 static arc_buf_hdr_t *
 arc_hdr_alloc(uint64_t spa, int32_t psize, int32_t lsize,
-    boolean_t protected, enum zio_compress compression_type,
+    boolean_t protected, enum zio_compress compression_type, uint8_t complevel,
     arc_buf_contents_t type, boolean_t alloc_rdata)
 {
 	arc_buf_hdr_t *hdr;
@@ -3272,6 +3283,7 @@ arc_hdr_alloc(uint64_t spa, int32_t psize, int32_t lsize,
 	hdr->b_flags = 0;
 	arc_hdr_set_flags(hdr, arc_bufc_to_flags(type) | ARC_FLAG_HAS_L1HDR);
 	arc_hdr_set_compress(hdr, compression_type);
+	hdr->b_complevel = complevel;
 	if (protected)
 		arc_hdr_set_flags(hdr, ARC_FLAG_PROTECTED);
 
@@ -3574,7 +3586,7 @@ arc_buf_t *
 arc_alloc_buf(spa_t *spa, void *tag, arc_buf_contents_t type, int32_t size)
 {
 	arc_buf_hdr_t *hdr = arc_hdr_alloc(spa_load_guid(spa), size, size,
-	    B_FALSE, ZIO_COMPRESS_OFF, type, B_FALSE);
+	    B_FALSE, ZIO_COMPRESS_OFF, 0, type, B_FALSE);
 
 	arc_buf_t *buf = NULL;
 	VERIFY0(arc_buf_alloc_impl(hdr, spa, NULL, tag, B_FALSE, B_FALSE,
@@ -3590,7 +3602,7 @@ arc_alloc_buf(spa_t *spa, void *tag, arc_buf_contents_t type, int32_t size)
  */
 arc_buf_t *
 arc_alloc_compressed_buf(spa_t *spa, void *tag, uint64_t psize, uint64_t lsize,
-    enum zio_compress compression_type)
+    enum zio_compress compression_type, uint8_t complevel)
 {
 	ASSERT3U(lsize, >, 0);
 	ASSERT3U(lsize, >=, psize);
@@ -3598,7 +3610,7 @@ arc_alloc_compressed_buf(spa_t *spa, void *tag, uint64_t psize, uint64_t lsize,
 	ASSERT3U(compression_type, <, ZIO_COMPRESS_FUNCTIONS);
 
 	arc_buf_hdr_t *hdr = arc_hdr_alloc(spa_load_guid(spa), psize, lsize,
-	    B_FALSE, compression_type, ARC_BUFC_DATA, B_FALSE);
+	    B_FALSE, compression_type, complevel, ARC_BUFC_DATA, B_FALSE);
 
 	arc_buf_t *buf = NULL;
 	VERIFY0(arc_buf_alloc_impl(hdr, spa, NULL, tag, B_FALSE,
@@ -3624,7 +3636,7 @@ arc_buf_t *
 arc_alloc_raw_buf(spa_t *spa, void *tag, uint64_t dsobj, boolean_t byteorder,
     const uint8_t *salt, const uint8_t *iv, const uint8_t *mac,
     dmu_object_type_t ot, uint64_t psize, uint64_t lsize,
-    enum zio_compress compression_type)
+    enum zio_compress compression_type, uint8_t complevel)
 {
 	arc_buf_hdr_t *hdr;
 	arc_buf_t *buf;
@@ -3637,7 +3649,7 @@ arc_alloc_raw_buf(spa_t *spa, void *tag, uint64_t dsobj, boolean_t byteorder,
 	ASSERT3U(compression_type, <, ZIO_COMPRESS_FUNCTIONS);
 
 	hdr = arc_hdr_alloc(spa_load_guid(spa), psize, lsize, B_TRUE,
-	    compression_type, type, B_TRUE);
+	    compression_type, complevel, type, B_TRUE);
 
 	hdr->b_crypt_hdr.b_dsobj = dsobj;
 	hdr->b_crypt_hdr.b_ot = ot;
@@ -5579,6 +5591,9 @@ arc_read_done(zio_t *zio)
 		} else {
 			hdr->b_l1hdr.b_byteswap = DMU_BSWAP_NUMFUNCS;
 		}
+		if (!HDR_L2_READING(hdr)) {
+			hdr->b_complevel = zio->io_prop.zp_complevel;
+		}
 	}
 
 	arc_hdr_clear_flags(hdr, ARC_FLAG_L2_EVICTED);
@@ -5982,7 +5997,7 @@ top:
 			arc_buf_hdr_t *exists = NULL;
 			arc_buf_contents_t type = BP_GET_BUFC_TYPE(bp);
 			hdr = arc_hdr_alloc(spa_load_guid(spa), psize, lsize,
-			    BP_IS_PROTECTED(bp), BP_GET_COMPRESS(bp), type,
+			    BP_IS_PROTECTED(bp), BP_GET_COMPRESS(bp), 0, type,
 			    encrypted_read);
 
 			if (!embedded_bp) {
@@ -6549,7 +6564,7 @@ arc_release(arc_buf_t *buf, void *tag)
 		 * buffer which will be freed in arc_write().
 		 */
 		nhdr = arc_hdr_alloc(spa, psize, lsize, protected,
-		    compress, type, HDR_HAS_RABD(hdr));
+		    compress, hdr->b_complevel, type, HDR_HAS_RABD(hdr));
 		ASSERT3P(nhdr->b_l1hdr.b_buf, ==, NULL);
 		ASSERT0(nhdr->b_l1hdr.b_bufcnt);
 		ASSERT0(zfs_refcount_count(&nhdr->b_l1hdr.b_refcnt));
@@ -6713,6 +6728,7 @@ arc_write_ready(zio_t *zio)
 	}
 	HDR_SET_PSIZE(hdr, psize);
 	arc_hdr_set_compress(hdr, compress);
+	hdr->b_complevel = zio->io_prop.zp_complevel;
 
 	if (zio->io_error != 0 || psize == 0)
 		goto out;
@@ -6902,6 +6918,7 @@ arc_write(zio_t *pio, spa_t *spa, uint64_t txg,
 		ASSERT(ARC_BUF_COMPRESSED(buf));
 		localprop.zp_encrypt = B_TRUE;
 		localprop.zp_compress = HDR_GET_COMPRESS(hdr);
+		localprop.zp_complevel = hdr->b_complevel;
 		localprop.zp_byteorder =
 		    (hdr->b_l1hdr.b_byteswap == DMU_BSWAP_NUMFUNCS) ?
 		    ZFS_HOST_BYTEORDER : !ZFS_HOST_BYTEORDER;
@@ -6920,6 +6937,7 @@ arc_write(zio_t *pio, spa_t *spa, uint64_t txg,
 	} else if (ARC_BUF_COMPRESSED(buf)) {
 		ASSERT3U(HDR_GET_LSIZE(hdr), !=, arc_buf_size(buf));
 		localprop.zp_compress = HDR_GET_COMPRESS(hdr);
+		localprop.zp_complevel = hdr->b_complevel;
 		zio_flags |= ZIO_FLAG_RAW_COMPRESS;
 	}
 	callback = kmem_zalloc(sizeof (arc_write_callback_t), KM_SLEEP);
@@ -8252,7 +8270,7 @@ l2arc_untransform(zio_t *zio, l2arc_read_callback_t *cb)
 
 		ret = zio_decompress_data(HDR_GET_COMPRESS(hdr),
 		    hdr->b_l1hdr.b_pabd, tmp, HDR_GET_PSIZE(hdr),
-		    HDR_GET_LSIZE(hdr));
+		    HDR_GET_LSIZE(hdr), &hdr->b_complevel);
 		if (ret != 0) {
 			abd_return_buf_copy(cabd, tmp, arc_hdr_size(hdr));
 			arc_free_data_abd(hdr, cabd, arc_hdr_size(hdr), hdr);
@@ -8351,6 +8369,7 @@ l2arc_read_done(zio_t *zio)
 	    (HDR_HAS_RABD(hdr) && zio->io_abd == hdr->b_crypt_hdr.b_rabd));
 	zio->io_bp_copy = cb->l2rcb_bp;	/* XXX fix in L2ARC 2.0	*/
 	zio->io_bp = &zio->io_bp_copy;	/* XXX fix in L2ARC 2.0	*/
+	zio->io_prop.zp_complevel = hdr->b_complevel;
 
 	valid_cksum = arc_cksum_is_equal(hdr, zio);
 
@@ -8763,7 +8782,18 @@ l2arc_apply_transforms(spa_t *spa, arc_buf_hdr_t *hdr, uint64_t asize,
 		cabd = abd_alloc_for_io(asize, ismd);
 		tmp = abd_borrow_buf(cabd, asize);
 
-		psize = zio_compress_data(compress, to_write, tmp, size);
+		psize = zio_compress_data(compress, to_write, tmp, size,
+		    hdr->b_complevel);
+
+		if (psize >= size) {
+			abd_return_buf(cabd, tmp, asize);
+			HDR_SET_COMPRESS(hdr, ZIO_COMPRESS_OFF);
+			to_write = cabd;
+			abd_copy(to_write, hdr->b_l1hdr.b_pabd, size);
+			if (size != asize)
+				abd_zero_off(to_write, size, asize - size);
+			goto encrypt;
+		}
 		ASSERT3U(psize, <=, HDR_GET_PSIZE(hdr));
 		if (psize < asize)
 			bzero((char *)tmp + psize, asize - psize);
@@ -8772,6 +8802,7 @@ l2arc_apply_transforms(spa_t *spa, arc_buf_hdr_t *hdr, uint64_t asize,
 		to_write = cabd;
 	}
 
+encrypt:
 	if (HDR_ENCRYPTED(hdr)) {
 		eabd = abd_alloc_for_io(asize, ismd);
 
@@ -9922,7 +9953,7 @@ l2arc_log_blk_read(l2arc_dev_t *dev,
 		abd_copy_from_buf_off(abd, this_lb, 0, asize);
 		if ((err = zio_decompress_data(
 		    L2BLK_GET_COMPRESS((this_lbp)->lbp_prop),
-		    abd, this_lb, asize, sizeof (*this_lb))) != 0) {
+		    abd, this_lb, asize, sizeof (*this_lb), NULL)) != 0) {
 			err = SET_ERROR(EINVAL);
 			goto cleanup;
 		}
@@ -10021,7 +10052,7 @@ l2arc_hdr_restore(const l2arc_log_ent_phys_t *le, l2arc_dev_t *dev)
 	hdr = arc_buf_alloc_l2only(L2BLK_GET_LSIZE((le)->le_prop), type,
 	    dev, le->le_dva, le->le_daddr,
 	    L2BLK_GET_PSIZE((le)->le_prop), le->le_birth,
-	    L2BLK_GET_COMPRESS((le)->le_prop),
+	    L2BLK_GET_COMPRESS((le)->le_prop), le->le_complevel,
 	    L2BLK_GET_PROTECTED((le)->le_prop),
 	    L2BLK_GET_PREFETCH((le)->le_prop));
 	asize = vdev_psize_to_asize(dev->l2ad_vdev,
@@ -10197,7 +10228,7 @@ l2arc_log_blk_commit(l2arc_dev_t *dev, zio_t *pio, l2arc_write_callback_t *cb)
 
 	/* try to compress the buffer */
 	psize = zio_compress_data(ZIO_COMPRESS_LZ4,
-	    abd_buf->abd, tmpbuf, sizeof (*lb));
+	    abd_buf->abd, tmpbuf, sizeof (*lb), 0);
 
 	/* a log block is never entirely zero */
 	ASSERT(psize != 0);
@@ -10354,6 +10385,7 @@ l2arc_log_blk_insert(l2arc_dev_t *dev, const arc_buf_hdr_t *hdr)
 	L2BLK_SET_LSIZE((le)->le_prop, HDR_GET_LSIZE(hdr));
 	L2BLK_SET_PSIZE((le)->le_prop, HDR_GET_PSIZE(hdr));
 	L2BLK_SET_COMPRESS((le)->le_prop, HDR_GET_COMPRESS(hdr));
+	le->le_complevel = hdr->b_complevel;
 	L2BLK_SET_TYPE((le)->le_prop, hdr->b_type);
 	L2BLK_SET_PROTECTED((le)->le_prop, !!(HDR_PROTECTED(hdr)));
 	L2BLK_SET_PREFETCH((le)->le_prop, !!(HDR_PREFETCH(hdr)));