block/qed.h

   1 /*
   2  * QEMU Enhanced Disk Format
   3  *
   4  * Copyright IBM, Corp. 2010
   5  *
   6  * Authors:
   7  *  Stefan Hajnoczi   <stefanha@linux.vnet.ibm.com>
   8  *  Anthony Liguori   <aliguori@us.ibm.com>
   9  *
  10  * This work is licensed under the terms of the GNU LGPL, version 2 or later.
  11  * See the COPYING.LIB file in the top-level directory.
  12  *
  13  */
  14
  15 #ifndef BLOCK_QED_H
  16 #define BLOCK_QED_H
  17
  18 #include "block_int.h"
  19
  20 /* The layout of a QED file is as follows:
  21  *
  22  * +--------+----------+----------+----------+-----+
  23  * | header | L1 table | cluster0 | cluster1 | ... |
  24  * +--------+----------+----------+----------+-----+
  25  *
  26  * There is a 2-level pagetable for cluster allocation:
  27  *
  28  *                     +----------+
  29  *                     | L1 table |
  30  *                     +----------+
  31  *                ,------'  |  '------.
  32  *           +----------+   |    +----------+
  33  *           | L2 table |  ...   | L2 table |
  34  *           +----------+        +----------+
  35  *       ,------'  |  '------.
  36  *  +----------+   |    +----------+
  37  *  |   Data   |  ...   |   Data   |
  38  *  +----------+        +----------+
  39  *
  40  * The L1 table is fixed size and always present.  L2 tables are allocated on
  41  * demand.  The L1 table size determines the maximum possible image size; it
  42  * can be influenced using the cluster_size and table_size values.
  43  *
  44  * All fields are little-endian on disk.
  45  */
  46
  47 enum {
  48     QED_MAGIC = 'Q' | 'E' << 8 | 'D' << 16 | '\0' << 24,
  49
  50     /* The image supports a backing file */
  51     QED_F_BACKING_FILE = 0x01,
  52
  53     /* The backing file format must not be probed, treat as raw image */
  54     QED_F_BACKING_FORMAT_NO_PROBE = 0x04,
  55
  56     /* Feature bits must be used when the on-disk format changes */
  57     QED_FEATURE_MASK = QED_F_BACKING_FILE | /* supported feature bits */
  58                        QED_F_BACKING_FORMAT_NO_PROBE,
  59     QED_COMPAT_FEATURE_MASK = 0,            /* supported compat feature bits */
  60     QED_AUTOCLEAR_FEATURE_MASK = 0,         /* supported autoclear feature bits */
  61
  62     /* Data is stored in groups of sectors called clusters.  Cluster size must
  63      * be large to avoid keeping too much metadata.  I/O requests that have
  64      * sub-cluster size will require read-modify-write.
  65      */
  66     QED_MIN_CLUSTER_SIZE = 4 * 1024, /* in bytes */
  67     QED_MAX_CLUSTER_SIZE = 64 * 1024 * 1024,
  68     QED_DEFAULT_CLUSTER_SIZE = 64 * 1024,
  69
  70     /* Allocated clusters are tracked using a 2-level pagetable.  Table size is
  71      * a multiple of clusters so large maximum image sizes can be supported
  72      * without jacking up the cluster size too much.
  73      */
  74     QED_MIN_TABLE_SIZE = 1,        /* in clusters */
  75     QED_MAX_TABLE_SIZE = 16,
  76     QED_DEFAULT_TABLE_SIZE = 4,
  77 };
  78
  79 typedef struct {
  80     uint32_t magic;                 /* QED\0 */
  81
  82     uint32_t cluster_size;          /* in bytes */
  83     uint32_t table_size;            /* for L1 and L2 tables, in clusters */
  84     uint32_t header_size;           /* in clusters */
  85
  86     uint64_t features;              /* format feature bits */
  87     uint64_t compat_features;       /* compatible feature bits */
  88     uint64_t autoclear_features;    /* self-resetting feature bits */
  89
  90     uint64_t l1_table_offset;       /* in bytes */
  91     uint64_t image_size;            /* total logical image size, in bytes */
  92
  93     /* if (features & QED_F_BACKING_FILE) */
  94     uint32_t backing_filename_offset; /* in bytes from start of header */
  95     uint32_t backing_filename_size;   /* in bytes */
  96 } QEDHeader;
  97
  98 typedef struct {
  99     BlockDriverState *bs;           /* device */
 100     uint64_t file_size;             /* length of image file, in bytes */
 101
 102     QEDHeader header;               /* always cpu-endian */
 103     uint32_t table_nelems;
 104     uint32_t l1_shift;
 105     uint32_t l2_shift;
 106     uint32_t l2_mask;
 107 } BDRVQEDState;
 108
 109 /**
 110  * Round down to the start of a cluster
 111  */
 112 static inline uint64_t qed_start_of_cluster(BDRVQEDState *s, uint64_t offset)
 113 {
 114     return offset & ~(uint64_t)(s->header.cluster_size - 1);
 115 }
 116
 117 /**
 118  * Test if a cluster offset is valid
 119  */
 120 static inline bool qed_check_cluster_offset(BDRVQEDState *s, uint64_t offset)
 121 {
 122     uint64_t header_size = (uint64_t)s->header.header_size *
 123                            s->header.cluster_size;
 124
 125     if (offset & (s->header.cluster_size - 1)) {
 126         return false;
 127     }
 128     return offset >= header_size && offset < s->file_size;
 129 }
 130
 131 /**
 132  * Test if a table offset is valid
 133  */
 134 static inline bool qed_check_table_offset(BDRVQEDState *s, uint64_t offset)
 135 {
 136     uint64_t end_offset = offset + (s->header.table_size - 1) *
 137                           s->header.cluster_size;
 138
 139     /* Overflow check */
 140     if (end_offset <= offset) {
 141         return false;
 142     }
 143
 144     return qed_check_cluster_offset(s, offset) &&
 145            qed_check_cluster_offset(s, end_offset);
 146 }
 147
 148 #endif /* BLOCK_QED_H */