block/fvd.c

   1 /*
   2  * Copyright (c) 2010-2011 IBM
   3  *
   4  * Authors:
   5  *         Chunqiang Tang <ctang@us.ibm.com>
   6  *
   7  * This work is licensed under the terms of the GNU GPL, version 2.
   8  * See the COPYING file in the top-level directory.
   9  */
  10
  11 /*=============================================================================
  12  *  A short description: this module implements the QEMU block device driver
  13  *  for the Fast Virtual Disk (FVD) format.  See the following companion
  14  *  papers for a detailed description of FVD:
  15  *  1. The so-called "FVD-cow paper":
  16  *          "FVD: a High-Performance Virtual Machine Image Format for Cloud",
  17  *      by Chunqiang Tang, 2010.
  18  *  2. The so-called "FVD-compact paper":
  19  *          "FVD: a High-Performance Virtual Machine Image Format for Cloud
  20  *           with Sparse Image Capability", by Chunqiang Tang, 2010.
  21  *============================================================================*/
  22
  23 #include "qemu/osdep.h"
  24 #include "block/fvd.h"
  25
  26 //#define ENABLE_TRACE_IO
  27 //#define DEBUG_MEMORY_LEAK
  28 //#define SIMULATED_TEST_WITH_QEMU_IO
  29
  30 #ifndef FVD_DEBUG
  31 #undef DEBUG_MEMORY_LEAK
  32 #undef ENABLE_TRACE_IO
  33 #undef SIMULATED_TEST_WITH_QEMU_IO
  34 #endif
  35
  36 /* Use include to avoid exposing too many FVD symbols, and to allow inline
  37  * function optimization. */
  38 #include "block/fvd-utils.c"
  39 #include "block/fvd-debug.c"
  40 #include "block/fvd-misc.c"
  41 #include "block/fvd-create.c"
  42 #include "block/fvd-open.c"
  43 #include "block/fvd-read.c"
  44 #include "block/fvd-write.c"
  45 #include "block/fvd-load.c"
  46 #include "block/fvd-store.c"
  47 #include "block/fvd-journal.c"
  48 #include "block/fvd-prefetch.c"
  49
  50 bool in_qemu_tool;
  51
  52 static AIOCBInfo fvd_aio_pool = {
  53     .aiocb_size = sizeof (FvdAIOCB),
  54 };
  55
  56 static BlockDriver bdrv_fvd = {
  57     .format_name = "fvd",
  58     .instance_size = sizeof (BDRVFvdState),
  59     .bdrv_create = fvd_create,
  60     .bdrv_probe = fvd_probe,
  61     .bdrv_file_open = fvd_open,
  62     .bdrv_close = fvd_close,
  63     .bdrv_co_get_block_status = fvd_get_block_status,
  64     .bdrv_co_flush_to_disk = fvd_flush,
  65     .bdrv_aio_readv = fvd_aio_readv,
  66     .bdrv_aio_writev = fvd_aio_writev,
  67     .bdrv_aio_flush = fvd_aio_flush,
  68     .create_opts = &fvd_create_opts,
  69     .bdrv_get_info = fvd_get_info,
  70     .bdrv_update = fvd_update,
  71     .bdrv_has_zero_init = fvd_has_zero_init
  72 };
  73
  74 static void bdrv_fvd_init (void)
  75 {
  76     bdrv_register (&bdrv_fvd);
  77 }
  78
  79 block_init (bdrv_fvd_init);
  80
  81 /*
  82  * Since bdrv_close may not be properly invoked on a VM shutdown, we
  83  * use a destructor to flush metadata to disk. This only affects
  84  * performance and does not affect correctness.
  85  * See Section 3.3.4 of the FVD-cow paper for the rationale.
  86  */
  87 extern QTAILQ_HEAD (, BlockDriverState) bdrv_states;
  88 static void __attribute__ ((destructor)) flush_fvd_bitmap_to_disk (void)
  89 {
  90     BlockDriverState *bs = NULL;
  91     while ((bs = bdrv_next(bs))) {
  92         if (bs->drv == &bdrv_fvd) {
  93             flush_metadata_to_disk_on_exit (bs);
  94
  95 #ifdef FVD_DEBUG
  96             dump_resource_summary (bs->opaque);
  97 #endif
  98         }
  99     }
 100 }
 101
 102 /*
 103  * TODOs: Below are some potential enhancements for future development:
 104  * 1. Handle storage leak on failure.
 105  *
 106  * 2. Profile-directed prefetch. See Section 3.4.1 of the FVD-cow paper.
 107  * Related metadata are FvdHeader.prefetch_profile_offset and
 108  * FvdHeader.prefetch_profile_entries,
 109  * FvdHeader.profile_directed_prefetch_start_delay,
 110  * FvdHeader.generate_prefetch_profile.
 111  *
 112  * 3.  Cap the prefetch throughput at the upper limit. See Section 3.4.2 of
 113  * the FVD-cow paper.  Related metadata are
 114  * FvdHeader.prefetch_max_read_throughput and
 115  * FvdHeader.prefetch_max_write_throughput.
 116  *
 117  * 4. Support write through to the base image. When a VM issues a write
 118  * request, in addition to saving the data in the FVD data file, also save the
 119  * data in the base image if the address of write request is not beyond the
 120  * size of the base image (this of course requires the base image NOT to be
 121  * 'read_only'. This feature changes the semantics of copy-on-write, but it
 122  * suits a different use case, where the base image is stored on a remote
 123  * storage server, and the FVD image is stored on a local disk and acts as a
 124  * write-through cache of the base image. This can be used to cache and
 125  * improve the performance of persistent storage on network-attached storage,
 126  * e.g., Amazon EBS.  This feature is not described in the FVD-cow paper as it
 127  * would complicate the discussion.  Related metadata are
 128  * FvdHeader.write_updates_base_img.
 129  */