module/zfs/blkptr.c

   1 /*
   2  * CDDL HEADER START
   3  *
   4  * This file and its contents are supplied under the terms of the
   5  * Common Development and Distribution License ("CDDL"), version 1.0.
   6  * You may only use this file in accordance with the terms of version
   7  * 1.0 of the CDDL.
   8  *
   9  * A full copy of the text of the CDDL should have accompanied this
  10  * source.  A copy of the CDDL is also available via the Internet at
  11  * http://www.illumos.org/license/CDDL.
  12  *
  13  * CDDL HEADER END
  14  */
  15
  16 /*
  17  * Copyright (c) 2013, 2016 by Delphix. All rights reserved.
  18  */
  19
  20 #include <sys/blkptr.h>
  21 #include <sys/zfs_context.h>
  22 #include <sys/zio.h>
  23 #include <sys/zio_compress.h>
  24
  25 /*
  26  * Embedded-data Block Pointers
  27  *
  28  * Normally, block pointers point (via their DVAs) to a block which holds data.
  29  * If the data that we need to store is very small, this is an inefficient
  30  * use of space, because a block must be at minimum 1 sector (typically 512
  31  * bytes or 4KB).  Additionally, reading these small blocks tends to generate
  32  * more random reads.
  33  *
  34  * Embedded-data Block Pointers allow small pieces of data (the "payload",
  35  * up to 112 bytes) to be stored in the block pointer itself, instead of
  36  * being pointed to.  The "Pointer" part of this name is a bit of a
  37  * misnomer, as nothing is pointed to.
  38  *
  39  * BP_EMBEDDED_TYPE_DATA block pointers allow highly-compressible data to
  40  * be embedded in the block pointer.  The logic for this is handled in
  41  * the SPA, by the zio pipeline.  Therefore most code outside the zio
  42  * pipeline doesn't need special-cases to handle these block pointers.
  43  *
  44  * See spa.h for details on the exact layout of embedded block pointers.
  45  */
  46
  47 void
  48 encode_embedded_bp_compressed(blkptr_t *bp, void *data,
  49     enum zio_compress comp, int uncompressed_size, int compressed_size)
  50 {
  51         uint64_t *bp64 = (uint64_t *)bp;
  52         uint64_t w = 0;
  53         uint8_t *data8 = data;
  54
  55         ASSERT3U(compressed_size, <=, BPE_PAYLOAD_SIZE);
  56         ASSERT(uncompressed_size == compressed_size ||
  57             comp != ZIO_COMPRESS_OFF);
  58         ASSERT3U(comp, >=, ZIO_COMPRESS_OFF);
  59         ASSERT3U(comp, <, ZIO_COMPRESS_FUNCTIONS);
  60
  61         memset(bp, 0, sizeof (*bp));
  62         BP_SET_EMBEDDED(bp, B_TRUE);
  63         BP_SET_COMPRESS(bp, comp);
  64         BP_SET_BYTEORDER(bp, ZFS_HOST_BYTEORDER);
  65         BPE_SET_LSIZE(bp, uncompressed_size);
  66         BPE_SET_PSIZE(bp, compressed_size);
  67
  68         /*
  69          * Encode the byte array into the words of the block pointer.
  70          * First byte goes into low bits of first word (little endian).
  71          */
  72         for (int i = 0; i < compressed_size; i++) {
  73                 BF64_SET(w, (i % sizeof (w)) * NBBY, NBBY, data8[i]);
  74                 if (i % sizeof (w) == sizeof (w) - 1) {
  75                         /* we've reached the end of a word */
  76                         ASSERT3P(bp64, <, bp + 1);
  77                         *bp64 = w;
  78                         bp64++;
  79                         if (!BPE_IS_PAYLOADWORD(bp, bp64))
  80                                 bp64++;
  81                         w = 0;
  82                 }
  83         }
  84         /* write last partial word */
  85         if (bp64 < (uint64_t *)(bp + 1))
  86                 *bp64 = w;
  87 }
  88
  89 /*
  90  * buf must be at least BPE_GET_PSIZE(bp) bytes long (which will never be
  91  * more than BPE_PAYLOAD_SIZE bytes).
  92  */
  93 void
  94 decode_embedded_bp_compressed(const blkptr_t *bp, void *buf)
  95 {
  96         int psize;
  97         uint8_t *buf8 = buf;
  98         uint64_t w = 0;
  99         const uint64_t *bp64 = (const uint64_t *)bp;
 100
 101         ASSERT(BP_IS_EMBEDDED(bp));
 102
 103         psize = BPE_GET_PSIZE(bp);
 104
 105         /*
 106          * Decode the words of the block pointer into the byte array.
 107          * Low bits of first word are the first byte (little endian).
 108          */
 109         for (int i = 0; i < psize; i++) {
 110                 if (i % sizeof (w) == 0) {
 111                         /* beginning of a word */
 112                         ASSERT3P(bp64, <, bp + 1);
 113                         w = *bp64;
 114                         bp64++;
 115                         if (!BPE_IS_PAYLOADWORD(bp, bp64))
 116                                 bp64++;
 117                 }
 118                 buf8[i] = BF64_GET(w, (i % sizeof (w)) * NBBY, NBBY);
 119         }
 120 }
 121
 122 /*
 123  * Fill in the buffer with the (decompressed) payload of the embedded
 124  * blkptr_t.  Takes into account compression and byteorder (the payload is
 125  * treated as a stream of bytes).
 126  * Return 0 on success, or ENOSPC if it won't fit in the buffer.
 127  */
 128 int
 129 decode_embedded_bp(const blkptr_t *bp, void *buf, int buflen)
 130 {
 131         int lsize, psize;
 132
 133         ASSERT(BP_IS_EMBEDDED(bp));
 134
 135         lsize = BPE_GET_LSIZE(bp);
 136         psize = BPE_GET_PSIZE(bp);
 137
 138         if (lsize > buflen)
 139                 return (SET_ERROR(ENOSPC));
 140         ASSERT3U(lsize, ==, buflen);
 141
 142         if (BP_GET_COMPRESS(bp) != ZIO_COMPRESS_OFF) {
 143                 uint8_t dstbuf[BPE_PAYLOAD_SIZE];
 144                 decode_embedded_bp_compressed(bp, dstbuf);
 145                 abd_t cabd, dabd;
 146                 abd_get_from_buf_struct(&cabd, dstbuf, psize);
 147                 abd_get_from_buf_struct(&dabd, buf, buflen);
 148                 VERIFY0(zio_decompress_data(BP_GET_COMPRESS(bp), &cabd,
 149                     &dabd, psize, buflen, NULL));
 150                 abd_free(&dabd);
 151                 abd_free(&cabd);
 152         } else {
 153                 ASSERT3U(lsize, ==, psize);
 154                 decode_embedded_bp_compressed(bp, buf);
 155         }
 156
 157         return (0);
 158 }