Diff for /qemu/block.c between versions 1.1.1.17 and 1.1.1.22

version 1.1.1.17, 2018/04/24 18:16:35 version 1.1.1.22, 2018/04/24 19:17:20
Line 23 Line 23
  */   */
 #include "config-host.h"  #include "config-host.h"
 #include "qemu-common.h"  #include "qemu-common.h"
   #include "trace.h"
 #include "monitor.h"  #include "monitor.h"
 #include "block_int.h"  #include "block_int.h"
 #include "module.h"  #include "module.h"
 #include "qemu-objects.h"  #include "qjson.h"
   #include "qemu-coroutine.h"
   #include "qmp-commands.h"
   
 #ifdef CONFIG_BSD  #ifdef CONFIG_BSD
 #include <sys/types.h>  #include <sys/types.h>
Line 42 Line 45
 #include <windows.h>  #include <windows.h>
 #endif  #endif
   
   #define NOT_DONE 0x7fffffff /* used while emulated sync operation in progress */
   
   static void bdrv_dev_change_media_cb(BlockDriverState *bs, bool load);
 static BlockDriverAIOCB *bdrv_aio_readv_em(BlockDriverState *bs,  static BlockDriverAIOCB *bdrv_aio_readv_em(BlockDriverState *bs,
         int64_t sector_num, QEMUIOVector *qiov, int nb_sectors,          int64_t sector_num, QEMUIOVector *qiov, int nb_sectors,
         BlockDriverCompletionFunc *cb, void *opaque);          BlockDriverCompletionFunc *cb, void *opaque);
 static BlockDriverAIOCB *bdrv_aio_writev_em(BlockDriverState *bs,  static BlockDriverAIOCB *bdrv_aio_writev_em(BlockDriverState *bs,
         int64_t sector_num, QEMUIOVector *qiov, int nb_sectors,          int64_t sector_num, QEMUIOVector *qiov, int nb_sectors,
         BlockDriverCompletionFunc *cb, void *opaque);          BlockDriverCompletionFunc *cb, void *opaque);
 static BlockDriverAIOCB *bdrv_aio_flush_em(BlockDriverState *bs,  static int coroutine_fn bdrv_co_readv_em(BlockDriverState *bs,
         BlockDriverCompletionFunc *cb, void *opaque);                                           int64_t sector_num, int nb_sectors,
 static int bdrv_read_em(BlockDriverState *bs, int64_t sector_num,                                           QEMUIOVector *iov);
                         uint8_t *buf, int nb_sectors);  static int coroutine_fn bdrv_co_writev_em(BlockDriverState *bs,
 static int bdrv_write_em(BlockDriverState *bs, int64_t sector_num,                                           int64_t sector_num, int nb_sectors,
                          const uint8_t *buf, int nb_sectors);                                           QEMUIOVector *iov);
   static int coroutine_fn bdrv_co_do_readv(BlockDriverState *bs,
       int64_t sector_num, int nb_sectors, QEMUIOVector *qiov);
   static int coroutine_fn bdrv_co_do_writev(BlockDriverState *bs,
       int64_t sector_num, int nb_sectors, QEMUIOVector *qiov);
   static BlockDriverAIOCB *bdrv_co_aio_rw_vector(BlockDriverState *bs,
                                                  int64_t sector_num,
                                                  QEMUIOVector *qiov,
                                                  int nb_sectors,
                                                  BlockDriverCompletionFunc *cb,
                                                  void *opaque,
                                                  bool is_write);
   static void coroutine_fn bdrv_co_do_rw(void *opaque);
   
 BlockDriverState *bdrv_first;  static QTAILQ_HEAD(, BlockDriverState) bdrv_states =
       QTAILQ_HEAD_INITIALIZER(bdrv_states);
   
 static BlockDriver *first_drv;  static QLIST_HEAD(, BlockDriver) bdrv_drivers =
       QLIST_HEAD_INITIALIZER(bdrv_drivers);
   
   /* The device to use for VM snapshots */
   static BlockDriverState *bs_snapshots;
   
 /* If non-zero, use only whitelisted block drivers */  /* If non-zero, use only whitelisted block drivers */
 static int use_bdrv_whitelist;  static int use_bdrv_whitelist;
   
   #ifdef _WIN32
   static int is_windows_drive_prefix(const char *filename)
   {
       return (((filename[0] >= 'a' && filename[0] <= 'z') ||
                (filename[0] >= 'A' && filename[0] <= 'Z')) &&
               filename[1] == ':');
   }
   
   int is_windows_drive(const char *filename)
   {
       if (is_windows_drive_prefix(filename) &&
           filename[2] == '\0')
           return 1;
       if (strstart(filename, "\\\\.\\", NULL) ||
           strstart(filename, "//./", NULL))
           return 1;
       return 0;
   }
   #endif
   
   /* check if the path starts with "<protocol>:" */
   static int path_has_protocol(const char *path)
   {
   #ifdef _WIN32
       if (is_windows_drive(path) ||
           is_windows_drive_prefix(path)) {
           return 0;
       }
   #endif
   
       return strchr(path, ':') != NULL;
   }
   
 int path_is_absolute(const char *path)  int path_is_absolute(const char *path)
 {  {
     const char *p;      const char *p;
Line 128  void path_combine(char *dest, int dest_s Line 184  void path_combine(char *dest, int dest_s
   
 void bdrv_register(BlockDriver *bdrv)  void bdrv_register(BlockDriver *bdrv)
 {  {
     if (!bdrv->bdrv_aio_readv) {      /* Block drivers without coroutine functions need emulation */
         /* add AIO emulation layer */      if (!bdrv->bdrv_co_readv) {
         bdrv->bdrv_aio_readv = bdrv_aio_readv_em;          bdrv->bdrv_co_readv = bdrv_co_readv_em;
         bdrv->bdrv_aio_writev = bdrv_aio_writev_em;          bdrv->bdrv_co_writev = bdrv_co_writev_em;
     } else if (!bdrv->bdrv_read) {  
         /* add synchronous IO emulation layer */          /* bdrv_co_readv_em()/brdv_co_writev_em() work in terms of aio, so if
         bdrv->bdrv_read = bdrv_read_em;           * the block driver lacks aio we need to emulate that too.
         bdrv->bdrv_write = bdrv_write_em;           */
           if (!bdrv->bdrv_aio_readv) {
               /* add AIO emulation layer */
               bdrv->bdrv_aio_readv = bdrv_aio_readv_em;
               bdrv->bdrv_aio_writev = bdrv_aio_writev_em;
           }
     }      }
   
     if (!bdrv->bdrv_aio_flush)      QLIST_INSERT_HEAD(&bdrv_drivers, bdrv, list);
         bdrv->bdrv_aio_flush = bdrv_aio_flush_em;  
   
     bdrv->next = first_drv;  
     first_drv = bdrv;  
 }  }
   
 /* create a new block device (by default it is empty) */  /* create a new block device (by default it is empty) */
 BlockDriverState *bdrv_new(const char *device_name)  BlockDriverState *bdrv_new(const char *device_name)
 {  {
     BlockDriverState **pbs, *bs;      BlockDriverState *bs;
   
     bs = qemu_mallocz(sizeof(BlockDriverState));      bs = g_malloc0(sizeof(BlockDriverState));
     pstrcpy(bs->device_name, sizeof(bs->device_name), device_name);      pstrcpy(bs->device_name, sizeof(bs->device_name), device_name);
     if (device_name[0] != '\0') {      if (device_name[0] != '\0') {
         /* insert at the end */          QTAILQ_INSERT_TAIL(&bdrv_states, bs, list);
         pbs = &bdrv_first;  
         while (*pbs != NULL)  
             pbs = &(*pbs)->next;  
         *pbs = bs;  
     }      }
       bdrv_iostatus_disable(bs);
     return bs;      return bs;
 }  }
   
 BlockDriver *bdrv_find_format(const char *format_name)  BlockDriver *bdrv_find_format(const char *format_name)
 {  {
     BlockDriver *drv1;      BlockDriver *drv1;
     for(drv1 = first_drv; drv1 != NULL; drv1 = drv1->next) {      QLIST_FOREACH(drv1, &bdrv_drivers, list) {
         if (!strcmp(drv1->format_name, format_name))          if (!strcmp(drv1->format_name, format_name)) {
             return drv1;              return drv1;
           }
     }      }
     return NULL;      return NULL;
 }  }
Line 205  int bdrv_create(BlockDriver *drv, const  Line 260  int bdrv_create(BlockDriver *drv, const 
     return drv->bdrv_create(filename, options);      return drv->bdrv_create(filename, options);
 }  }
   
   int bdrv_create_file(const char* filename, QEMUOptionParameter *options)
   {
       BlockDriver *drv;
   
       drv = bdrv_find_protocol(filename);
       if (drv == NULL) {
           return -ENOENT;
       }
   
       return bdrv_create(drv, filename, options);
   }
   
 #ifdef _WIN32  #ifdef _WIN32
 void get_tmp_filename(char *filename, int size)  void get_tmp_filename(char *filename, int size)
 {  {
Line 228  void get_tmp_filename(char *filename, in Line 295  void get_tmp_filename(char *filename, in
 }  }
 #endif  #endif
   
 #ifdef _WIN32  /*
 static int is_windows_drive_prefix(const char *filename)   * Detect host devices. By convention, /dev/cdrom[N] is always
    * recognized as a host CDROM.
    */
   static BlockDriver *find_hdev_driver(const char *filename)
 {  {
     return (((filename[0] >= 'a' && filename[0] <= 'z') ||      int score_max = 0, score;
              (filename[0] >= 'A' && filename[0] <= 'Z')) &&      BlockDriver *drv = NULL, *d;
             filename[1] == ':');  
 }  
   
 int is_windows_drive(const char *filename)      QLIST_FOREACH(d, &bdrv_drivers, list) {
 {          if (d->bdrv_probe_device) {
     if (is_windows_drive_prefix(filename) &&              score = d->bdrv_probe_device(filename);
         filename[2] == '\0')              if (score > score_max) {
         return 1;                  score_max = score;
     if (strstart(filename, "\\\\.\\", NULL) ||                  drv = d;
         strstart(filename, "//./", NULL))              }
         return 1;          }
     return 0;      }
   
       return drv;
 }  }
 #endif  
   
 static BlockDriver *find_protocol(const char *filename)  BlockDriver *bdrv_find_protocol(const char *filename)
 {  {
     BlockDriver *drv1;      BlockDriver *drv1;
     char protocol[128];      char protocol[128];
     int len;      int len;
     const char *p;      const char *p;
   
 #ifdef _WIN32      /* TODO Drivers without bdrv_file_open must be specified explicitly */
     if (is_windows_drive(filename) ||  
         is_windows_drive_prefix(filename))      /*
         return bdrv_find_format("raw");       * XXX(hch): we really should not let host device detection
 #endif       * override an explicit protocol specification, but moving this
        * later breaks access to device names with colons in them.
        * Thanks to the brain-dead persistent naming schemes on udev-
        * based Linux systems those actually are quite common.
        */
       drv1 = find_hdev_driver(filename);
       if (drv1) {
           return drv1;
       }
   
       if (!path_has_protocol(filename)) {
           return bdrv_find_format("file");
       }
     p = strchr(filename, ':');      p = strchr(filename, ':');
     if (!p)      assert(p != NULL);
         return bdrv_find_format("raw");  
     len = p - filename;      len = p - filename;
     if (len > sizeof(protocol) - 1)      if (len > sizeof(protocol) - 1)
         len = sizeof(protocol) - 1;          len = sizeof(protocol) - 1;
     memcpy(protocol, filename, len);      memcpy(protocol, filename, len);
     protocol[len] = '\0';      protocol[len] = '\0';
     for(drv1 = first_drv; drv1 != NULL; drv1 = drv1->next) {      QLIST_FOREACH(drv1, &bdrv_drivers, list) {
         if (drv1->protocol_name &&          if (drv1->protocol_name &&
             !strcmp(drv1->protocol_name, protocol))              !strcmp(drv1->protocol_name, protocol)) {
             return drv1;              return drv1;
     }  
     return NULL;  
 }  
   
 /*  
  * Detect host devices. By convention, /dev/cdrom[N] is always  
  * recognized as a host CDROM.  
  */  
 static BlockDriver *find_hdev_driver(const char *filename)  
 {  
     int score_max = 0, score;  
     BlockDriver *drv = NULL, *d;  
   
     for (d = first_drv; d; d = d->next) {  
         if (d->bdrv_probe_device) {  
             score = d->bdrv_probe_device(filename);  
             if (score > score_max) {  
                 score_max = score;  
                 drv = d;  
             }  
         }          }
     }      }
       return NULL;
     return drv;  
 }  }
   
 static BlockDriver *find_image_format(const char *filename)  static int find_image_format(const char *filename, BlockDriver **pdrv)
 {  {
     int ret, score, score_max;      int ret, score, score_max;
     BlockDriver *drv1, *drv;      BlockDriver *drv1, *drv;
     uint8_t buf[2048];      uint8_t buf[2048];
     BlockDriverState *bs;      BlockDriverState *bs;
   
     drv = find_protocol(filename);      ret = bdrv_file_open(&bs, filename, 0);
     /* no need to test disk image formats for vvfat */      if (ret < 0) {
     if (drv && strcmp(drv->format_name, "vvfat") == 0)          *pdrv = NULL;
         return drv;          return ret;
       }
   
       /* Return the raw BlockDriver * to scsi-generic devices or empty drives */
       if (bs->sg || !bdrv_is_inserted(bs)) {
           bdrv_delete(bs);
           drv = bdrv_find_format("raw");
           if (!drv) {
               ret = -ENOENT;
           }
           *pdrv = drv;
           return ret;
       }
   
     ret = bdrv_file_open(&bs, filename, BDRV_O_RDONLY);  
     if (ret < 0)  
         return NULL;  
     ret = bdrv_pread(bs, 0, buf, sizeof(buf));      ret = bdrv_pread(bs, 0, buf, sizeof(buf));
     bdrv_delete(bs);      bdrv_delete(bs);
     if (ret < 0) {      if (ret < 0) {
         return NULL;          *pdrv = NULL;
           return ret;
     }      }
   
     score_max = 0;      score_max = 0;
     for(drv1 = first_drv; drv1 != NULL; drv1 = drv1->next) {      drv = NULL;
       QLIST_FOREACH(drv1, &bdrv_drivers, list) {
         if (drv1->bdrv_probe) {          if (drv1->bdrv_probe) {
             score = drv1->bdrv_probe(buf, ret, filename);              score = drv1->bdrv_probe(buf, ret, filename);
             if (score > score_max) {              if (score > score_max) {
Line 329  static BlockDriver *find_image_format(co Line 399  static BlockDriver *find_image_format(co
             }              }
         }          }
     }      }
     return drv;      if (!drv) {
           ret = -ENOENT;
       }
       *pdrv = drv;
       return ret;
   }
   
   /**
    * Set the current 'total_sectors' value
    */
   static int refresh_total_sectors(BlockDriverState *bs, int64_t hint)
   {
       BlockDriver *drv = bs->drv;
   
       /* Do not attempt drv->bdrv_getlength() on scsi-generic devices */
       if (bs->sg)
           return 0;
   
       /* query actual device if possible, otherwise just trust the hint */
       if (drv->bdrv_getlength) {
           int64_t length = drv->bdrv_getlength(bs);
           if (length < 0) {
               return length;
           }
           hint = length >> BDRV_SECTOR_BITS;
       }
   
       bs->total_sectors = hint;
       return 0;
   }
   
   /**
    * Set open flags for a given cache mode
    *
    * Return 0 on success, -1 if the cache mode was invalid.
    */
   int bdrv_parse_cache_flags(const char *mode, int *flags)
   {
       *flags &= ~BDRV_O_CACHE_MASK;
   
       if (!strcmp(mode, "off") || !strcmp(mode, "none")) {
           *flags |= BDRV_O_NOCACHE | BDRV_O_CACHE_WB;
       } else if (!strcmp(mode, "directsync")) {
           *flags |= BDRV_O_NOCACHE;
       } else if (!strcmp(mode, "writeback")) {
           *flags |= BDRV_O_CACHE_WB;
       } else if (!strcmp(mode, "unsafe")) {
           *flags |= BDRV_O_CACHE_WB;
           *flags |= BDRV_O_NO_FLUSH;
       } else if (!strcmp(mode, "writethrough")) {
           /* this is the default */
       } else {
           return -1;
       }
   
       return 0;
   }
   
   /*
    * Common part for opening disk images and files
    */
   static int bdrv_open_common(BlockDriverState *bs, const char *filename,
       int flags, BlockDriver *drv)
   {
       int ret, open_flags;
   
       assert(drv != NULL);
   
       trace_bdrv_open_common(bs, filename, flags, drv->format_name);
   
       bs->file = NULL;
       bs->total_sectors = 0;
       bs->encrypted = 0;
       bs->valid_key = 0;
       bs->sg = 0;
       bs->open_flags = flags;
       bs->growable = 0;
       bs->buffer_alignment = 512;
   
       pstrcpy(bs->filename, sizeof(bs->filename), filename);
       bs->backing_file[0] = '\0';
   
       if (use_bdrv_whitelist && !bdrv_is_whitelisted(drv)) {
           return -ENOTSUP;
       }
   
       bs->drv = drv;
       bs->opaque = g_malloc0(drv->instance_size);
   
       bs->enable_write_cache = !!(flags & BDRV_O_CACHE_WB);
   
       /*
        * Clear flags that are internal to the block layer before opening the
        * image.
        */
       open_flags = flags & ~(BDRV_O_SNAPSHOT | BDRV_O_NO_BACKING);
   
       /*
        * Snapshots should be writable.
        */
       if (bs->is_temporary) {
           open_flags |= BDRV_O_RDWR;
       }
   
       bs->keep_read_only = bs->read_only = !(open_flags & BDRV_O_RDWR);
   
       /* Open the image, either directly or using a protocol */
       if (drv->bdrv_file_open) {
           ret = drv->bdrv_file_open(bs, filename, open_flags);
       } else {
           ret = bdrv_file_open(&bs->file, filename, open_flags);
           if (ret >= 0) {
               ret = drv->bdrv_open(bs, open_flags);
           }
       }
   
       if (ret < 0) {
           goto free_and_fail;
       }
   
       ret = refresh_total_sectors(bs, bs->total_sectors);
       if (ret < 0) {
           goto free_and_fail;
       }
   
   #ifndef _WIN32
       if (bs->is_temporary) {
           unlink(filename);
       }
   #endif
       return 0;
   
   free_and_fail:
       if (bs->file) {
           bdrv_delete(bs->file);
           bs->file = NULL;
       }
       g_free(bs->opaque);
       bs->opaque = NULL;
       bs->drv = NULL;
       return ret;
 }  }
   
   /*
    * Opens a file using a protocol (file, host_device, nbd, ...)
    */
 int bdrv_file_open(BlockDriverState **pbs, const char *filename, int flags)  int bdrv_file_open(BlockDriverState **pbs, const char *filename, int flags)
 {  {
     BlockDriverState *bs;      BlockDriverState *bs;
       BlockDriver *drv;
     int ret;      int ret;
   
       drv = bdrv_find_protocol(filename);
       if (!drv) {
           return -ENOENT;
       }
   
     bs = bdrv_new("");      bs = bdrv_new("");
     ret = bdrv_open2(bs, filename, flags | BDRV_O_FILE, NULL);      ret = bdrv_open_common(bs, filename, flags, drv);
     if (ret < 0) {      if (ret < 0) {
         bdrv_delete(bs);          bdrv_delete(bs);
         return ret;          return ret;
Line 348  int bdrv_file_open(BlockDriverState **pb Line 567  int bdrv_file_open(BlockDriverState **pb
     return 0;      return 0;
 }  }
   
 int bdrv_open(BlockDriverState *bs, const char *filename, int flags)  /*
 {   * Opens a disk image (raw, qcow2, vmdk, ...)
     return bdrv_open2(bs, filename, flags, NULL);   */
 }  int bdrv_open(BlockDriverState *bs, const char *filename, int flags,
                 BlockDriver *drv)
 int bdrv_open2(BlockDriverState *bs, const char *filename, int flags,  
                BlockDriver *drv)  
 {  {
     int ret, open_flags, try_rw;      int ret;
     char tmp_filename[PATH_MAX];      char tmp_filename[PATH_MAX];
     char backing_filename[PATH_MAX];  
   
     bs->is_temporary = 0;  
     bs->encrypted = 0;  
     bs->valid_key = 0;  
     /* buffer_alignment defaulted to 512, drivers can change this value */  
     bs->buffer_alignment = 512;  
   
     if (flags & BDRV_O_SNAPSHOT) {      if (flags & BDRV_O_SNAPSHOT) {
         BlockDriverState *bs1;          BlockDriverState *bs1;
Line 372  int bdrv_open2(BlockDriverState *bs, con Line 582  int bdrv_open2(BlockDriverState *bs, con
         int is_protocol = 0;          int is_protocol = 0;
         BlockDriver *bdrv_qcow2;          BlockDriver *bdrv_qcow2;
         QEMUOptionParameter *options;          QEMUOptionParameter *options;
           char backing_filename[PATH_MAX];
   
         /* if snapshot, we create a temporary backing file and open it          /* if snapshot, we create a temporary backing file and open it
            instead of opening 'filename' directly */             instead of opening 'filename' directly */
   
         /* if there is a backing file, use it */          /* if there is a backing file, use it */
         bs1 = bdrv_new("");          bs1 = bdrv_new("");
         ret = bdrv_open2(bs1, filename, 0, drv);          ret = bdrv_open(bs1, filename, 0, drv);
         if (ret < 0) {          if (ret < 0) {
             bdrv_delete(bs1);              bdrv_delete(bs1);
             return ret;              return ret;
         }          }
         total_size = bdrv_getlength(bs1) >> BDRV_SECTOR_BITS;          total_size = bdrv_getlength(bs1) & BDRV_SECTOR_MASK;
   
         if (bs1->drv && bs1->drv->protocol_name)          if (bs1->drv && bs1->drv->protocol_name)
             is_protocol = 1;              is_protocol = 1;
Line 396  int bdrv_open2(BlockDriverState *bs, con Line 607  int bdrv_open2(BlockDriverState *bs, con
         if (is_protocol)          if (is_protocol)
             snprintf(backing_filename, sizeof(backing_filename),              snprintf(backing_filename, sizeof(backing_filename),
                      "%s", filename);                       "%s", filename);
         else          else if (!realpath(filename, backing_filename))
             realpath(filename, backing_filename);              return -errno;
   
         bdrv_qcow2 = bdrv_find_format("qcow2");          bdrv_qcow2 = bdrv_find_format("qcow2");
         options = parse_option_parameters("", bdrv_qcow2->create_options, NULL);          options = parse_option_parameters("", bdrv_qcow2->create_options, NULL);
   
         set_option_parameter_int(options, BLOCK_OPT_SIZE, total_size * 512);          set_option_parameter_int(options, BLOCK_OPT_SIZE, total_size);
         set_option_parameter(options, BLOCK_OPT_BACKING_FILE, backing_filename);          set_option_parameter(options, BLOCK_OPT_BACKING_FILE, backing_filename);
         if (drv) {          if (drv) {
             set_option_parameter(options, BLOCK_OPT_BACKING_FMT,              set_option_parameter(options, BLOCK_OPT_BACKING_FMT,
Line 410  int bdrv_open2(BlockDriverState *bs, con Line 621  int bdrv_open2(BlockDriverState *bs, con
         }          }
   
         ret = bdrv_create(bdrv_qcow2, tmp_filename, options);          ret = bdrv_create(bdrv_qcow2, tmp_filename, options);
           free_option_parameters(options);
         if (ret < 0) {          if (ret < 0) {
             return ret;              return ret;
         }          }
Line 419  int bdrv_open2(BlockDriverState *bs, con Line 631  int bdrv_open2(BlockDriverState *bs, con
         bs->is_temporary = 1;          bs->is_temporary = 1;
     }      }
   
     pstrcpy(bs->filename, sizeof(bs->filename), filename);      /* Find the right image format driver */
     if (flags & BDRV_O_FILE) {      if (!drv) {
         drv = find_protocol(filename);          ret = find_image_format(filename, &drv);
     } else if (!drv) {  
         drv = find_hdev_driver(filename);  
         if (!drv) {  
             drv = find_image_format(filename);  
         }  
     }      }
   
     if (!drv) {      if (!drv) {
         ret = -ENOENT;  
         goto unlink_and_fail;          goto unlink_and_fail;
     }      }
     bs->drv = drv;  
     bs->opaque = qemu_mallocz(drv->instance_size);  
   
     /*  
      * Yes, BDRV_O_NOCACHE aka O_DIRECT means we have to present a  
      * write cache to the guest.  We do need the fdatasync to flush  
      * out transactions for block allocations, and we maybe have a  
      * volatile write cache in our backing device to deal with.  
      */  
     if (flags & (BDRV_O_CACHE_WB|BDRV_O_NOCACHE))  
         bs->enable_write_cache = 1;  
   
     /* Note: for compatibility, we open disk image files as RDWR, and  
        RDONLY as fallback */  
     try_rw = !bs->read_only || bs->is_temporary;  
     if (!(flags & BDRV_O_FILE))  
         open_flags = (try_rw ? BDRV_O_RDWR : 0) |  
             (flags & (BDRV_O_CACHE_MASK|BDRV_O_NATIVE_AIO));  
     else  
         open_flags = flags & ~(BDRV_O_FILE | BDRV_O_SNAPSHOT);  
   
     bs->open_flags = open_flags;      /* Open the image */
     if (use_bdrv_whitelist && !bdrv_is_whitelisted(drv))      ret = bdrv_open_common(bs, filename, flags, drv);
         ret = -ENOTSUP;  
     else  
         ret = drv->bdrv_open(bs, filename, open_flags);  
     if ((ret == -EACCES || ret == -EPERM) && !(flags & BDRV_O_FILE)) {  
         ret = drv->bdrv_open(bs, filename, open_flags & ~BDRV_O_RDWR);  
         bs->read_only = 1;  
     }  
     if (ret < 0) {      if (ret < 0) {
         qemu_free(bs->opaque);          goto unlink_and_fail;
         bs->opaque = NULL;  
         bs->drv = NULL;  
     unlink_and_fail:  
         if (bs->is_temporary)  
             unlink(filename);  
         return ret;  
     }  
     if (drv->bdrv_getlength) {  
         bs->total_sectors = bdrv_getlength(bs) >> BDRV_SECTOR_BITS;  
     }  
 #ifndef _WIN32  
     if (bs->is_temporary) {  
         unlink(filename);  
     }      }
 #endif  
     if (bs->backing_file[0] != '\0') {      /* If there is a backing file, use it */
         /* if there is a backing file, use it */      if ((flags & BDRV_O_NO_BACKING) == 0 && bs->backing_file[0] != '\0') {
           char backing_filename[PATH_MAX];
           int back_flags;
         BlockDriver *back_drv = NULL;          BlockDriver *back_drv = NULL;
   
         bs->backing_hd = bdrv_new("");          bs->backing_hd = bdrv_new("");
         /* pass on read_only property to the backing_hd */  
         bs->backing_hd->read_only = bs->read_only;          if (path_has_protocol(bs->backing_file)) {
         path_combine(backing_filename, sizeof(backing_filename),              pstrcpy(backing_filename, sizeof(backing_filename),
                      filename, bs->backing_file);                      bs->backing_file);
         if (bs->backing_format[0] != '\0')          } else {
               path_combine(backing_filename, sizeof(backing_filename),
                            filename, bs->backing_file);
           }
   
           if (bs->backing_format[0] != '\0') {
             back_drv = bdrv_find_format(bs->backing_format);              back_drv = bdrv_find_format(bs->backing_format);
         ret = bdrv_open2(bs->backing_hd, backing_filename, open_flags,          }
                          back_drv);  
           /* backing files always opened read-only */
           back_flags =
               flags & ~(BDRV_O_RDWR | BDRV_O_SNAPSHOT | BDRV_O_NO_BACKING);
   
           ret = bdrv_open(bs->backing_hd, backing_filename, back_flags, back_drv);
         if (ret < 0) {          if (ret < 0) {
             bdrv_close(bs);              bdrv_close(bs);
             return ret;              return ret;
         }          }
           if (bs->is_temporary) {
               bs->backing_hd->keep_read_only = !(flags & BDRV_O_RDWR);
           } else {
               /* base image inherits from "parent" */
               bs->backing_hd->keep_read_only = bs->keep_read_only;
           }
     }      }
   
     if (!bdrv_key_required(bs)) {      if (!bdrv_key_required(bs)) {
         /* call the change callback */          bdrv_dev_change_media_cb(bs, true);
         bs->media_changed = 1;  
         if (bs->change_cb)  
             bs->change_cb(bs->change_opaque);  
     }      }
   
     return 0;      return 0;
 }  
   
 void bdrv_close(BlockDriverState *bs)  unlink_and_fail:
 {      if (bs->is_temporary) {
     if (bs->drv) {          unlink(filename);
         if (bs->backing_hd)      }
       return ret;
   }
   
   void bdrv_close(BlockDriverState *bs)
   {
       if (bs->drv) {
           if (bs == bs_snapshots) {
               bs_snapshots = NULL;
           }
           if (bs->backing_hd) {
             bdrv_delete(bs->backing_hd);              bdrv_delete(bs->backing_hd);
               bs->backing_hd = NULL;
           }
         bs->drv->bdrv_close(bs);          bs->drv->bdrv_close(bs);
         qemu_free(bs->opaque);          g_free(bs->opaque);
 #ifdef _WIN32  #ifdef _WIN32
         if (bs->is_temporary) {          if (bs->is_temporary) {
             unlink(bs->filename);              unlink(bs->filename);
Line 521  void bdrv_close(BlockDriverState *bs) Line 716  void bdrv_close(BlockDriverState *bs)
         bs->opaque = NULL;          bs->opaque = NULL;
         bs->drv = NULL;          bs->drv = NULL;
   
         /* call the change callback */          if (bs->file != NULL) {
         bs->media_changed = 1;              bdrv_close(bs->file);
         if (bs->change_cb)          }
             bs->change_cb(bs->change_opaque);  
           bdrv_dev_change_media_cb(bs, false);
       }
   }
   
   void bdrv_close_all(void)
   {
       BlockDriverState *bs;
   
       QTAILQ_FOREACH(bs, &bdrv_states, list) {
           bdrv_close(bs);
       }
   }
   
   /* make a BlockDriverState anonymous by removing from bdrv_state list.
      Also, NULL terminate the device_name to prevent double remove */
   void bdrv_make_anon(BlockDriverState *bs)
   {
       if (bs->device_name[0] != '\0') {
           QTAILQ_REMOVE(&bdrv_states, bs, list);
     }      }
       bs->device_name[0] = '\0';
 }  }
   
 void bdrv_delete(BlockDriverState *bs)  void bdrv_delete(BlockDriverState *bs)
 {  {
     BlockDriverState **pbs;      assert(!bs->dev);
   
     pbs = &bdrv_first;      /* remove from list, if necessary */
     while (*pbs != bs && *pbs != NULL)      bdrv_make_anon(bs);
         pbs = &(*pbs)->next;  
     if (*pbs == bs)  
         *pbs = bs->next;  
   
     bdrv_close(bs);      bdrv_close(bs);
     qemu_free(bs);      if (bs->file != NULL) {
           bdrv_delete(bs->file);
       }
   
       assert(bs != bs_snapshots);
       g_free(bs);
   }
   
   int bdrv_attach_dev(BlockDriverState *bs, void *dev)
   /* TODO change to DeviceState *dev when all users are qdevified */
   {
       if (bs->dev) {
           return -EBUSY;
       }
       bs->dev = dev;
       bdrv_iostatus_reset(bs);
       return 0;
   }
   
   /* TODO qdevified devices don't use this, remove when devices are qdevified */
   void bdrv_attach_dev_nofail(BlockDriverState *bs, void *dev)
   {
       if (bdrv_attach_dev(bs, dev) < 0) {
           abort();
       }
   }
   
   void bdrv_detach_dev(BlockDriverState *bs, void *dev)
   /* TODO change to DeviceState *dev when all users are qdevified */
   {
       assert(bs->dev == dev);
       bs->dev = NULL;
       bs->dev_ops = NULL;
       bs->dev_opaque = NULL;
       bs->buffer_alignment = 512;
   }
   
   /* TODO change to return DeviceState * when all users are qdevified */
   void *bdrv_get_attached_dev(BlockDriverState *bs)
   {
       return bs->dev;
   }
   
   void bdrv_set_dev_ops(BlockDriverState *bs, const BlockDevOps *ops,
                         void *opaque)
   {
       bs->dev_ops = ops;
       bs->dev_opaque = opaque;
       if (bdrv_dev_has_removable_media(bs) && bs == bs_snapshots) {
           bs_snapshots = NULL;
       }
   }
   
   static void bdrv_dev_change_media_cb(BlockDriverState *bs, bool load)
   {
       if (bs->dev_ops && bs->dev_ops->change_media_cb) {
           bs->dev_ops->change_media_cb(bs->dev_opaque, load);
       }
   }
   
   bool bdrv_dev_has_removable_media(BlockDriverState *bs)
   {
       return !bs->dev || (bs->dev_ops && bs->dev_ops->change_media_cb);
   }
   
   void bdrv_dev_eject_request(BlockDriverState *bs, bool force)
   {
       if (bs->dev_ops && bs->dev_ops->eject_request_cb) {
           bs->dev_ops->eject_request_cb(bs->dev_opaque, force);
       }
   }
   
   bool bdrv_dev_is_tray_open(BlockDriverState *bs)
   {
       if (bs->dev_ops && bs->dev_ops->is_tray_open) {
           return bs->dev_ops->is_tray_open(bs->dev_opaque);
       }
       return false;
   }
   
   static void bdrv_dev_resize_cb(BlockDriverState *bs)
   {
       if (bs->dev_ops && bs->dev_ops->resize_cb) {
           bs->dev_ops->resize_cb(bs->dev_opaque);
       }
   }
   
   bool bdrv_dev_is_medium_locked(BlockDriverState *bs)
   {
       if (bs->dev_ops && bs->dev_ops->is_medium_locked) {
           return bs->dev_ops->is_medium_locked(bs->dev_opaque);
       }
       return false;
 }  }
   
 /*  /*
  * Run consistency checks on an image   * Run consistency checks on an image
  *   *
  * Returns the number of errors or -errno when an internal error occurs   * Returns 0 if the check could be completed (it doesn't mean that the image is
    * free of errors) or -errno when an internal error occurred. The results of the
    * check are stored in res.
  */   */
 int bdrv_check(BlockDriverState *bs)  int bdrv_check(BlockDriverState *bs, BdrvCheckResult *res)
 {  {
     if (bs->drv->bdrv_check == NULL) {      if (bs->drv->bdrv_check == NULL) {
         return -ENOTSUP;          return -ENOTSUP;
     }      }
   
     return bs->drv->bdrv_check(bs);      memset(res, 0, sizeof(*res));
       return bs->drv->bdrv_check(bs, res);
 }  }
   
   #define COMMIT_BUF_SECTORS 2048
   
 /* commit COW file into the raw image */  /* commit COW file into the raw image */
 int bdrv_commit(BlockDriverState *bs)  int bdrv_commit(BlockDriverState *bs)
 {  {
     BlockDriver *drv = bs->drv;      BlockDriver *drv = bs->drv;
     int64_t i, total_sectors;      BlockDriver *backing_drv;
     int n, j;      int64_t sector, total_sectors;
     unsigned char sector[512];      int n, ro, open_flags;
       int ret = 0, rw_ret = 0;
       uint8_t *buf;
       char filename[1024];
       BlockDriverState *bs_rw, *bs_ro;
   
     if (!drv)      if (!drv)
         return -ENOMEDIUM;          return -ENOMEDIUM;
       
       if (!bs->backing_hd) {
           return -ENOTSUP;
       }
   
     if (bs->read_only) {      if (bs->backing_hd->keep_read_only) {
         return -EACCES;          return -EACCES;
     }      }
   
     if (!bs->backing_hd) {      backing_drv = bs->backing_hd->drv;
         return -ENOTSUP;      ro = bs->backing_hd->read_only;
       strncpy(filename, bs->backing_hd->filename, sizeof(filename));
       open_flags =  bs->backing_hd->open_flags;
   
       if (ro) {
           /* re-open as RW */
           bdrv_delete(bs->backing_hd);
           bs->backing_hd = NULL;
           bs_rw = bdrv_new("");
           rw_ret = bdrv_open(bs_rw, filename, open_flags | BDRV_O_RDWR,
               backing_drv);
           if (rw_ret < 0) {
               bdrv_delete(bs_rw);
               /* try to re-open read-only */
               bs_ro = bdrv_new("");
               ret = bdrv_open(bs_ro, filename, open_flags & ~BDRV_O_RDWR,
                   backing_drv);
               if (ret < 0) {
                   bdrv_delete(bs_ro);
                   /* drive not functional anymore */
                   bs->drv = NULL;
                   return ret;
               }
               bs->backing_hd = bs_ro;
               return rw_ret;
           }
           bs->backing_hd = bs_rw;
     }      }
   
     total_sectors = bdrv_getlength(bs) >> BDRV_SECTOR_BITS;      total_sectors = bdrv_getlength(bs) >> BDRV_SECTOR_BITS;
     for (i = 0; i < total_sectors;) {      buf = g_malloc(COMMIT_BUF_SECTORS * BDRV_SECTOR_SIZE);
         if (drv->bdrv_is_allocated(bs, i, 65536, &n)) {  
             for(j = 0; j < n; j++) {  
                 if (bdrv_read(bs, i, sector, 1) != 0) {  
                     return -EIO;  
                 }  
   
                 if (bdrv_write(bs->backing_hd, i, sector, 1) != 0) {      for (sector = 0; sector < total_sectors; sector += n) {
                     return -EIO;          if (drv->bdrv_is_allocated(bs, sector, COMMIT_BUF_SECTORS, &n)) {
                 }  
                 i++;              if (bdrv_read(bs, sector, buf, n) != 0) {
             }                  ret = -EIO;
         } else {                  goto ro_cleanup;
             i += n;              }
   
               if (bdrv_write(bs->backing_hd, sector, buf, n) != 0) {
                   ret = -EIO;
                   goto ro_cleanup;
               }
         }          }
     }      }
   
     if (drv->bdrv_make_empty)      if (drv->bdrv_make_empty) {
         return drv->bdrv_make_empty(bs);          ret = drv->bdrv_make_empty(bs);
           bdrv_flush(bs);
       }
   
     return 0;      /*
        * Make sure all data we wrote to the backing device is actually
        * stable on disk.
        */
       if (bs->backing_hd)
           bdrv_flush(bs->backing_hd);
   
   ro_cleanup:
       g_free(buf);
   
       if (ro) {
           /* re-open as RO */
           bdrv_delete(bs->backing_hd);
           bs->backing_hd = NULL;
           bs_ro = bdrv_new("");
           ret = bdrv_open(bs_ro, filename, open_flags & ~BDRV_O_RDWR,
               backing_drv);
           if (ret < 0) {
               bdrv_delete(bs_ro);
               /* drive not functional anymore */
               bs->drv = NULL;
               return ret;
           }
           bs->backing_hd = bs_ro;
           bs->backing_hd->keep_read_only = 0;
       }
   
       return ret;
   }
   
   void bdrv_commit_all(void)
   {
       BlockDriverState *bs;
   
       QTAILQ_FOREACH(bs, &bdrv_states, list) {
           bdrv_commit(bs);
       }
   }
   
   /*
    * Return values:
    * 0        - success
    * -EINVAL  - backing format specified, but no file
    * -ENOSPC  - can't update the backing file because no space is left in the
    *            image file header
    * -ENOTSUP - format driver doesn't support changing the backing file
    */
   int bdrv_change_backing_file(BlockDriverState *bs,
       const char *backing_file, const char *backing_fmt)
   {
       BlockDriver *drv = bs->drv;
   
       if (drv->bdrv_change_backing_file != NULL) {
           return drv->bdrv_change_backing_file(bs, backing_file, backing_fmt);
       } else {
           return -ENOTSUP;
       }
 }  }
   
 static int bdrv_check_byte_request(BlockDriverState *bs, int64_t offset,  static int bdrv_check_byte_request(BlockDriverState *bs, int64_t offset,
Line 624  static int bdrv_check_byte_request(Block Line 1025  static int bdrv_check_byte_request(Block
 static int bdrv_check_request(BlockDriverState *bs, int64_t sector_num,  static int bdrv_check_request(BlockDriverState *bs, int64_t sector_num,
                               int nb_sectors)                                int nb_sectors)
 {  {
     return bdrv_check_byte_request(bs, sector_num * 512, nb_sectors * 512);      return bdrv_check_byte_request(bs, sector_num * BDRV_SECTOR_SIZE,
                                      nb_sectors * BDRV_SECTOR_SIZE);
   }
   
   typedef struct RwCo {
       BlockDriverState *bs;
       int64_t sector_num;
       int nb_sectors;
       QEMUIOVector *qiov;
       bool is_write;
       int ret;
   } RwCo;
   
   static void coroutine_fn bdrv_rw_co_entry(void *opaque)
   {
       RwCo *rwco = opaque;
   
       if (!rwco->is_write) {
           rwco->ret = bdrv_co_do_readv(rwco->bs, rwco->sector_num,
                                        rwco->nb_sectors, rwco->qiov);
       } else {
           rwco->ret = bdrv_co_do_writev(rwco->bs, rwco->sector_num,
                                         rwco->nb_sectors, rwco->qiov);
       }
   }
   
   /*
    * Process a synchronous request using coroutines
    */
   static int bdrv_rw_co(BlockDriverState *bs, int64_t sector_num, uint8_t *buf,
                         int nb_sectors, bool is_write)
   {
       QEMUIOVector qiov;
       struct iovec iov = {
           .iov_base = (void *)buf,
           .iov_len = nb_sectors * BDRV_SECTOR_SIZE,
       };
       Coroutine *co;
       RwCo rwco = {
           .bs = bs,
           .sector_num = sector_num,
           .nb_sectors = nb_sectors,
           .qiov = &qiov,
           .is_write = is_write,
           .ret = NOT_DONE,
       };
   
       qemu_iovec_init_external(&qiov, &iov, 1);
   
       if (qemu_in_coroutine()) {
           /* Fast-path if already in coroutine context */
           bdrv_rw_co_entry(&rwco);
       } else {
           co = qemu_coroutine_create(bdrv_rw_co_entry);
           qemu_coroutine_enter(co, &rwco);
           while (rwco.ret == NOT_DONE) {
               qemu_aio_wait();
           }
       }
       return rwco.ret;
 }  }
   
 /* return < 0 if error. See bdrv_write() for the return codes */  /* return < 0 if error. See bdrv_write() for the return codes */
 int bdrv_read(BlockDriverState *bs, int64_t sector_num,  int bdrv_read(BlockDriverState *bs, int64_t sector_num,
               uint8_t *buf, int nb_sectors)                uint8_t *buf, int nb_sectors)
 {  {
     BlockDriver *drv = bs->drv;      return bdrv_rw_co(bs, sector_num, buf, nb_sectors, false);
   
     if (!drv)  
         return -ENOMEDIUM;  
     if (bdrv_check_request(bs, sector_num, nb_sectors))  
         return -EIO;  
   
     return drv->bdrv_read(bs, sector_num, buf, nb_sectors);  
 }  }
   
 static void set_dirty_bitmap(BlockDriverState *bs, int64_t sector_num,  static void set_dirty_bitmap(BlockDriverState *bs, int64_t sector_num,
Line 655  static void set_dirty_bitmap(BlockDriver Line 1108  static void set_dirty_bitmap(BlockDriver
         bit = start % (sizeof(unsigned long) * 8);          bit = start % (sizeof(unsigned long) * 8);
         val = bs->dirty_bitmap[idx];          val = bs->dirty_bitmap[idx];
         if (dirty) {          if (dirty) {
             val |= 1 << bit;              if (!(val & (1UL << bit))) {
                   bs->dirty_count++;
                   val |= 1UL << bit;
               }
         } else {          } else {
             val &= ~(1 << bit);              if (val & (1UL << bit)) {
                   bs->dirty_count--;
                   val &= ~(1UL << bit);
               }
         }          }
         bs->dirty_bitmap[idx] = val;          bs->dirty_bitmap[idx] = val;
     }      }
Line 672  static void set_dirty_bitmap(BlockDriver Line 1131  static void set_dirty_bitmap(BlockDriver
 int bdrv_write(BlockDriverState *bs, int64_t sector_num,  int bdrv_write(BlockDriverState *bs, int64_t sector_num,
                const uint8_t *buf, int nb_sectors)                 const uint8_t *buf, int nb_sectors)
 {  {
     BlockDriver *drv = bs->drv;      return bdrv_rw_co(bs, sector_num, (uint8_t *)buf, nb_sectors, true);
     if (!bs->drv)  
         return -ENOMEDIUM;  
     if (bs->read_only)  
         return -EACCES;  
     if (bdrv_check_request(bs, sector_num, nb_sectors))  
         return -EIO;  
   
     if (bs->dirty_bitmap) {  
         set_dirty_bitmap(bs, sector_num, nb_sectors, 1);  
     }  
   
     return drv->bdrv_write(bs, sector_num, buf, nb_sectors);  
 }  }
   
 int bdrv_pread(BlockDriverState *bs, int64_t offset,  int bdrv_pread(BlockDriverState *bs, int64_t offset,
Line 797  int bdrv_pwrite_sync(BlockDriverState *b Line 1244  int bdrv_pwrite_sync(BlockDriverState *b
         return ret;          return ret;
     }      }
   
     /* No flush needed for cache=writethrough, it uses O_DSYNC */      /* No flush needed for cache modes that use O_DSYNC */
     if ((bs->open_flags & BDRV_O_CACHE_MASK) != 0) {      if ((bs->open_flags & BDRV_O_CACHE_WB) != 0) {
         bdrv_flush(bs);          bdrv_flush(bs);
     }      }
   
Line 806  int bdrv_pwrite_sync(BlockDriverState *b Line 1253  int bdrv_pwrite_sync(BlockDriverState *b
 }  }
   
 /*  /*
  * Writes to the file and ensures that no writes are reordered across this   * Handle a read request in coroutine context
  * request (acts as a barrier)   */
  *  static int coroutine_fn bdrv_co_do_readv(BlockDriverState *bs,
  * Returns 0 on success, -errno in error cases.      int64_t sector_num, int nb_sectors, QEMUIOVector *qiov)
   {
       BlockDriver *drv = bs->drv;
   
       if (!drv) {
           return -ENOMEDIUM;
       }
       if (bdrv_check_request(bs, sector_num, nb_sectors)) {
           return -EIO;
       }
   
       return drv->bdrv_co_readv(bs, sector_num, nb_sectors, qiov);
   }
   
   int coroutine_fn bdrv_co_readv(BlockDriverState *bs, int64_t sector_num,
       int nb_sectors, QEMUIOVector *qiov)
   {
       trace_bdrv_co_readv(bs, sector_num, nb_sectors);
   
       return bdrv_co_do_readv(bs, sector_num, nb_sectors, qiov);
   }
   
   /*
    * Handle a write request in coroutine context
  */   */
 int bdrv_write_sync(BlockDriverState *bs, int64_t sector_num,  static int coroutine_fn bdrv_co_do_writev(BlockDriverState *bs,
     const uint8_t *buf, int nb_sectors)      int64_t sector_num, int nb_sectors, QEMUIOVector *qiov)
   {
       BlockDriver *drv = bs->drv;
       int ret;
   
       if (!bs->drv) {
           return -ENOMEDIUM;
       }
       if (bs->read_only) {
           return -EACCES;
       }
       if (bdrv_check_request(bs, sector_num, nb_sectors)) {
           return -EIO;
       }
   
       ret = drv->bdrv_co_writev(bs, sector_num, nb_sectors, qiov);
   
       if (bs->dirty_bitmap) {
           set_dirty_bitmap(bs, sector_num, nb_sectors, 1);
       }
   
       if (bs->wr_highest_sector < sector_num + nb_sectors - 1) {
           bs->wr_highest_sector = sector_num + nb_sectors - 1;
       }
   
       return ret;
   }
   
   int coroutine_fn bdrv_co_writev(BlockDriverState *bs, int64_t sector_num,
       int nb_sectors, QEMUIOVector *qiov)
 {  {
     return bdrv_pwrite_sync(bs, BDRV_SECTOR_SIZE * sector_num,      trace_bdrv_co_writev(bs, sector_num, nb_sectors);
         buf, BDRV_SECTOR_SIZE * nb_sectors);  
       return bdrv_co_do_writev(bs, sector_num, nb_sectors, qiov);
 }  }
   
 /**  /**
Line 824  int bdrv_write_sync(BlockDriverState *bs Line 1324  int bdrv_write_sync(BlockDriverState *bs
 int bdrv_truncate(BlockDriverState *bs, int64_t offset)  int bdrv_truncate(BlockDriverState *bs, int64_t offset)
 {  {
     BlockDriver *drv = bs->drv;      BlockDriver *drv = bs->drv;
       int ret;
     if (!drv)      if (!drv)
         return -ENOMEDIUM;          return -ENOMEDIUM;
     if (!drv->bdrv_truncate)      if (!drv->bdrv_truncate)
         return -ENOTSUP;          return -ENOTSUP;
     if (bs->read_only)      if (bs->read_only)
         return -EACCES;          return -EACCES;
     return drv->bdrv_truncate(bs, offset);      if (bdrv_in_use(bs))
           return -EBUSY;
       ret = drv->bdrv_truncate(bs, offset);
       if (ret == 0) {
           ret = refresh_total_sectors(bs, offset >> BDRV_SECTOR_BITS);
           bdrv_dev_resize_cb(bs);
       }
       return ret;
   }
   
   /**
    * Length of a allocated file in bytes. Sparse files are counted by actual
    * allocated space. Return < 0 if error or unknown.
    */
   int64_t bdrv_get_allocated_file_size(BlockDriverState *bs)
   {
       BlockDriver *drv = bs->drv;
       if (!drv) {
           return -ENOMEDIUM;
       }
       if (drv->bdrv_get_allocated_file_size) {
           return drv->bdrv_get_allocated_file_size(bs);
       }
       if (bs->file) {
           return bdrv_get_allocated_file_size(bs->file);
       }
       return -ENOTSUP;
 }  }
   
 /**  /**
Line 841  int64_t bdrv_getlength(BlockDriverState  Line 1368  int64_t bdrv_getlength(BlockDriverState 
     BlockDriver *drv = bs->drv;      BlockDriver *drv = bs->drv;
     if (!drv)      if (!drv)
         return -ENOMEDIUM;          return -ENOMEDIUM;
     if (!drv->bdrv_getlength) {  
         /* legacy mode */      if (bs->growable || bdrv_dev_has_removable_media(bs)) {
         return bs->total_sectors * BDRV_SECTOR_SIZE;          if (drv->bdrv_getlength) {
               return drv->bdrv_getlength(bs);
           }
     }      }
     return drv->bdrv_getlength(bs);      return bs->total_sectors * BDRV_SECTOR_SIZE;
 }  }
   
 /* return 0 as number of sectors if no device present or error */  /* return 0 as number of sectors if no device present or error */
Line 871  struct partition { Line 1400  struct partition {
         uint8_t end_cyl;            /* end cylinder */          uint8_t end_cyl;            /* end cylinder */
         uint32_t start_sect;        /* starting sector counting from 0 */          uint32_t start_sect;        /* starting sector counting from 0 */
         uint32_t nr_sects;          /* nr of sectors in partition */          uint32_t nr_sects;          /* nr of sectors in partition */
 } __attribute__((packed));  } QEMU_PACKED;
   
 /* try to guess the disk logical geometry from the MSDOS partition table. Return 0 if OK, -1 if could not guess */  /* try to guess the disk logical geometry from the MSDOS partition table. Return 0 if OK, -1 if could not guess */
 static int guess_disk_lchs(BlockDriverState *bs,  static int guess_disk_lchs(BlockDriverState *bs,
                            int *pcylinders, int *pheads, int *psectors)                             int *pcylinders, int *pheads, int *psectors)
 {  {
     uint8_t buf[512];      uint8_t buf[BDRV_SECTOR_SIZE];
     int ret, i, heads, sectors, cylinders;      int ret, i, heads, sectors, cylinders;
     struct partition *p;      struct partition *p;
     uint32_t nr_sects;      uint32_t nr_sects;
Line 984  void bdrv_set_geometry_hint(BlockDriverS Line 1513  void bdrv_set_geometry_hint(BlockDriverS
     bs->secs = secs;      bs->secs = secs;
 }  }
   
 void bdrv_set_type_hint(BlockDriverState *bs, int type)  
 {  
     bs->type = type;  
     bs->removable = ((type == BDRV_TYPE_CDROM ||  
                       type == BDRV_TYPE_FLOPPY));  
 }  
   
 void bdrv_set_translation_hint(BlockDriverState *bs, int translation)  void bdrv_set_translation_hint(BlockDriverState *bs, int translation)
 {  {
     bs->translation = translation;      bs->translation = translation;
Line 1004  void bdrv_get_geometry_hint(BlockDriverS Line 1526  void bdrv_get_geometry_hint(BlockDriverS
     *psecs = bs->secs;      *psecs = bs->secs;
 }  }
   
 int bdrv_get_type_hint(BlockDriverState *bs)  /* Recognize floppy formats */
 {  typedef struct FDFormat {
     return bs->type;      FDriveType drive;
       uint8_t last_sect;
       uint8_t max_track;
       uint8_t max_head;
   } FDFormat;
   
   static const FDFormat fd_formats[] = {
       /* First entry is default format */
       /* 1.44 MB 3"1/2 floppy disks */
       { FDRIVE_DRV_144, 18, 80, 1, },
       { FDRIVE_DRV_144, 20, 80, 1, },
       { FDRIVE_DRV_144, 21, 80, 1, },
       { FDRIVE_DRV_144, 21, 82, 1, },
       { FDRIVE_DRV_144, 21, 83, 1, },
       { FDRIVE_DRV_144, 22, 80, 1, },
       { FDRIVE_DRV_144, 23, 80, 1, },
       { FDRIVE_DRV_144, 24, 80, 1, },
       /* 2.88 MB 3"1/2 floppy disks */
       { FDRIVE_DRV_288, 36, 80, 1, },
       { FDRIVE_DRV_288, 39, 80, 1, },
       { FDRIVE_DRV_288, 40, 80, 1, },
       { FDRIVE_DRV_288, 44, 80, 1, },
       { FDRIVE_DRV_288, 48, 80, 1, },
       /* 720 kB 3"1/2 floppy disks */
       { FDRIVE_DRV_144,  9, 80, 1, },
       { FDRIVE_DRV_144, 10, 80, 1, },
       { FDRIVE_DRV_144, 10, 82, 1, },
       { FDRIVE_DRV_144, 10, 83, 1, },
       { FDRIVE_DRV_144, 13, 80, 1, },
       { FDRIVE_DRV_144, 14, 80, 1, },
       /* 1.2 MB 5"1/4 floppy disks */
       { FDRIVE_DRV_120, 15, 80, 1, },
       { FDRIVE_DRV_120, 18, 80, 1, },
       { FDRIVE_DRV_120, 18, 82, 1, },
       { FDRIVE_DRV_120, 18, 83, 1, },
       { FDRIVE_DRV_120, 20, 80, 1, },
       /* 720 kB 5"1/4 floppy disks */
       { FDRIVE_DRV_120,  9, 80, 1, },
       { FDRIVE_DRV_120, 11, 80, 1, },
       /* 360 kB 5"1/4 floppy disks */
       { FDRIVE_DRV_120,  9, 40, 1, },
       { FDRIVE_DRV_120,  9, 40, 0, },
       { FDRIVE_DRV_120, 10, 41, 1, },
       { FDRIVE_DRV_120, 10, 42, 1, },
       /* 320 kB 5"1/4 floppy disks */
       { FDRIVE_DRV_120,  8, 40, 1, },
       { FDRIVE_DRV_120,  8, 40, 0, },
       /* 360 kB must match 5"1/4 better than 3"1/2... */
       { FDRIVE_DRV_144,  9, 80, 0, },
       /* end */
       { FDRIVE_DRV_NONE, -1, -1, 0, },
   };
   
   void bdrv_get_floppy_geometry_hint(BlockDriverState *bs, int *nb_heads,
                                      int *max_track, int *last_sect,
                                      FDriveType drive_in, FDriveType *drive)
   {
       const FDFormat *parse;
       uint64_t nb_sectors, size;
       int i, first_match, match;
   
       bdrv_get_geometry_hint(bs, nb_heads, max_track, last_sect);
       if (*nb_heads != 0 && *max_track != 0 && *last_sect != 0) {
           /* User defined disk */
       } else {
           bdrv_get_geometry(bs, &nb_sectors);
           match = -1;
           first_match = -1;
           for (i = 0; ; i++) {
               parse = &fd_formats[i];
               if (parse->drive == FDRIVE_DRV_NONE) {
                   break;
               }
               if (drive_in == parse->drive ||
                   drive_in == FDRIVE_DRV_NONE) {
                   size = (parse->max_head + 1) * parse->max_track *
                       parse->last_sect;
                   if (nb_sectors == size) {
                       match = i;
                       break;
                   }
                   if (first_match == -1) {
                       first_match = i;
                   }
               }
           }
           if (match == -1) {
               if (first_match == -1) {
                   match = 1;
               } else {
                   match = first_match;
               }
               parse = &fd_formats[match];
           }
           *nb_heads = parse->max_head + 1;
           *max_track = parse->max_track;
           *last_sect = parse->last_sect;
           *drive = parse->drive;
       }
 }  }
   
 int bdrv_get_translation_hint(BlockDriverState *bs)  int bdrv_get_translation_hint(BlockDriverState *bs)
Line 1014  int bdrv_get_translation_hint(BlockDrive Line 1634  int bdrv_get_translation_hint(BlockDrive
     return bs->translation;      return bs->translation;
 }  }
   
 int bdrv_is_removable(BlockDriverState *bs)  void bdrv_set_on_error(BlockDriverState *bs, BlockErrorAction on_read_error,
                          BlockErrorAction on_write_error)
 {  {
     return bs->removable;      bs->on_read_error = on_read_error;
       bs->on_write_error = on_write_error;
 }  }
   
 int bdrv_is_read_only(BlockDriverState *bs)  BlockErrorAction bdrv_get_on_error(BlockDriverState *bs, int is_read)
 {  {
     return bs->read_only;      return is_read ? bs->on_read_error : bs->on_write_error;
 }  }
   
 int bdrv_set_read_only(BlockDriverState *bs, int read_only)  int bdrv_is_read_only(BlockDriverState *bs)
 {  {
     int ret = bs->read_only;      return bs->read_only;
     bs->read_only = read_only;  
     return ret;  
 }  }
   
 int bdrv_is_sg(BlockDriverState *bs)  int bdrv_is_sg(BlockDriverState *bs)
Line 1041  int bdrv_enable_write_cache(BlockDriverS Line 1661  int bdrv_enable_write_cache(BlockDriverS
     return bs->enable_write_cache;      return bs->enable_write_cache;
 }  }
   
 /* XXX: no longer used */  
 void bdrv_set_change_cb(BlockDriverState *bs,  
                         void (*change_cb)(void *opaque), void *opaque)  
 {  
     bs->change_cb = change_cb;  
     bs->change_opaque = opaque;  
 }  
   
 int bdrv_is_encrypted(BlockDriverState *bs)  int bdrv_is_encrypted(BlockDriverState *bs)
 {  {
     if (bs->backing_hd && bs->backing_hd->encrypted)      if (bs->backing_hd && bs->backing_hd->encrypted)
Line 1075  int bdrv_set_key(BlockDriverState *bs, c Line 1687  int bdrv_set_key(BlockDriverState *bs, c
         if (!bs->encrypted)          if (!bs->encrypted)
             return 0;              return 0;
     }      }
     if (!bs->encrypted || !bs->drv || !bs->drv->bdrv_set_key)      if (!bs->encrypted) {
         return -1;          return -EINVAL;
       } else if (!bs->drv || !bs->drv->bdrv_set_key) {
           return -ENOMEDIUM;
       }
     ret = bs->drv->bdrv_set_key(bs, key);      ret = bs->drv->bdrv_set_key(bs, key);
     if (ret < 0) {      if (ret < 0) {
         bs->valid_key = 0;          bs->valid_key = 0;
     } else if (!bs->valid_key) {      } else if (!bs->valid_key) {
         bs->valid_key = 1;          bs->valid_key = 1;
         /* call the change callback now, we skipped it on open */          /* call the change callback now, we skipped it on open */
         bs->media_changed = 1;          bdrv_dev_change_media_cb(bs, true);
         if (bs->change_cb)  
             bs->change_cb(bs->change_opaque);  
     }      }
     return ret;      return ret;
 }  }
Line 1104  void bdrv_iterate_format(void (*it)(void Line 1717  void bdrv_iterate_format(void (*it)(void
 {  {
     BlockDriver *drv;      BlockDriver *drv;
   
     for (drv = first_drv; drv != NULL; drv = drv->next) {      QLIST_FOREACH(drv, &bdrv_drivers, list) {
         it(opaque, drv->format_name);          it(opaque, drv->format_name);
     }      }
 }  }
Line 1113  BlockDriverState *bdrv_find(const char * Line 1726  BlockDriverState *bdrv_find(const char *
 {  {
     BlockDriverState *bs;      BlockDriverState *bs;
   
     for (bs = bdrv_first; bs != NULL; bs = bs->next) {      QTAILQ_FOREACH(bs, &bdrv_states, list) {
         if (!strcmp(name, bs->device_name))          if (!strcmp(name, bs->device_name)) {
             return bs;              return bs;
           }
     }      }
     return NULL;      return NULL;
 }  }
   
   BlockDriverState *bdrv_next(BlockDriverState *bs)
   {
       if (!bs) {
           return QTAILQ_FIRST(&bdrv_states);
       }
       return QTAILQ_NEXT(bs, list);
   }
   
 void bdrv_iterate(void (*it)(void *opaque, BlockDriverState *bs), void *opaque)  void bdrv_iterate(void (*it)(void *opaque, BlockDriverState *bs), void *opaque)
 {  {
     BlockDriverState *bs;      BlockDriverState *bs;
   
     for (bs = bdrv_first; bs != NULL; bs = bs->next) {      QTAILQ_FOREACH(bs, &bdrv_states, list) {
         it(opaque, bs);          it(opaque, bs);
     }      }
 }  }
Line 1134  const char *bdrv_get_device_name(BlockDr Line 1756  const char *bdrv_get_device_name(BlockDr
     return bs->device_name;      return bs->device_name;
 }  }
   
 void bdrv_flush(BlockDriverState *bs)  
 {  
     if (!bs->drv)  
         return;  
     if (bs->drv->bdrv_flush)  
         bs->drv->bdrv_flush(bs);  
     if (bs->backing_hd)  
         bdrv_flush(bs->backing_hd);  
 }  
   
 void bdrv_flush_all(void)  void bdrv_flush_all(void)
 {  {
     BlockDriverState *bs;      BlockDriverState *bs;
   
     for (bs = bdrv_first; bs != NULL; bs = bs->next)      QTAILQ_FOREACH(bs, &bdrv_states, list) {
         if (bs->drv && !bdrv_is_read_only(bs) &&           if (!bdrv_is_read_only(bs) && bdrv_is_inserted(bs)) {
             (!bdrv_is_removable(bs) || bdrv_is_inserted(bs)))  
             bdrv_flush(bs);              bdrv_flush(bs);
           }
       }
   }
   
   int bdrv_has_zero_init(BlockDriverState *bs)
   {
       assert(bs->drv);
   
       if (bs->drv->bdrv_has_zero_init) {
           return bs->drv->bdrv_has_zero_init(bs);
       }
   
       return 1;
 }  }
   
 /*  /*
Line 1181  int bdrv_is_allocated(BlockDriverState * Line 1805  int bdrv_is_allocated(BlockDriverState *
     return bs->drv->bdrv_is_allocated(bs, sector_num, nb_sectors, pnum);      return bs->drv->bdrv_is_allocated(bs, sector_num, nb_sectors, pnum);
 }  }
   
 static void bdrv_print_dict(QObject *obj, void *opaque)  void bdrv_mon_event(const BlockDriverState *bdrv,
                       BlockMonEventAction action, int is_read)
 {  {
     QDict *bs_dict;      QObject *data;
     Monitor *mon = opaque;      const char *action_str;
   
     bs_dict = qobject_to_qdict(obj);  
   
     monitor_printf(mon, "%s: type=%s removable=%d",      switch (action) {
                         qdict_get_str(bs_dict, "device"),      case BDRV_ACTION_REPORT:
                         qdict_get_str(bs_dict, "type"),          action_str = "report";
                         qdict_get_bool(bs_dict, "removable"));          break;
       case BDRV_ACTION_IGNORE:
     if (qdict_get_bool(bs_dict, "removable")) {          action_str = "ignore";
         monitor_printf(mon, " locked=%d", qdict_get_bool(bs_dict, "locked"));          break;
     }      case BDRV_ACTION_STOP:
           action_str = "stop";
           break;
       default:
           abort();
       }
   
       data = qobject_from_jsonf("{ 'device': %s, 'action': %s, 'operation': %s }",
                                 bdrv->device_name,
                                 action_str,
                                 is_read ? "read" : "write");
       monitor_protocol_event(QEVENT_BLOCK_IO_ERROR, data);
   
     if (qdict_haskey(bs_dict, "inserted")) {      qobject_decref(data);
         QDict *qdict = qobject_to_qdict(qdict_get(bs_dict, "inserted"));  
   
         monitor_printf(mon, " file=");  
         monitor_print_filename(mon, qdict_get_str(qdict, "file"));  
         if (qdict_haskey(qdict, "backing_file")) {  
             monitor_printf(mon, " backing_file=");  
             monitor_print_filename(mon, qdict_get_str(qdict, "backing_file"));  
         }  
         monitor_printf(mon, " ro=%d drv=%s encrypted=%d",  
                             qdict_get_bool(qdict, "ro"),  
                             qdict_get_str(qdict, "drv"),  
                             qdict_get_bool(qdict, "encrypted"));  
     } else {  
         monitor_printf(mon, " [not inserted]");  
     }  
   
     monitor_printf(mon, "\n");  
 }  
   
 void bdrv_info_print(Monitor *mon, const QObject *data)  
 {  
     qlist_iter(qobject_to_qlist(data), bdrv_print_dict, mon);  
 }  }
   
 /**  BlockInfoList *qmp_query_block(Error **errp)
  * bdrv_info(): Block devices information  
  *  
  * Each block device information is stored in a QDict and the  
  * returned QObject is a QList of all devices.  
  *  
  * The QDict contains the following:  
  *  
  * - "device": device name  
  * - "type": device type  
  * - "removable": true if the device is removable, false otherwise  
  * - "locked": true if the device is locked, false otherwise  
  * - "inserted": only present if the device is inserted, it is a QDict  
  *    containing the following:  
  *          - "file": device file name  
  *          - "ro": true if read-only, false otherwise  
  *          - "drv": driver format name  
  *          - "backing_file": backing file name if one is used  
  *          - "encrypted": true if encrypted, false otherwise  
  *  
  * Example:  
  *  
  * [ { "device": "ide0-hd0", "type": "hd", "removable": false, "locked": false,  
  *     "inserted": { "file": "/tmp/foobar", "ro": false, "drv": "qcow2" } },  
  *   { "device": "floppy0", "type": "floppy", "removable": true,  
  *     "locked": false } ]  
  */  
 void bdrv_info(Monitor *mon, QObject **ret_data)  
 {  {
     QList *bs_list;      BlockInfoList *head = NULL, *cur_item = NULL;
     BlockDriverState *bs;      BlockDriverState *bs;
   
     bs_list = qlist_new();      QTAILQ_FOREACH(bs, &bdrv_states, list) {
           BlockInfoList *info = g_malloc0(sizeof(*info));
   
     for (bs = bdrv_first; bs != NULL; bs = bs->next) {          info->value = g_malloc0(sizeof(*info->value));
         QObject *bs_obj;          info->value->device = g_strdup(bs->device_name);
         const char *type = "unknown";          info->value->type = g_strdup("unknown");
           info->value->locked = bdrv_dev_is_medium_locked(bs);
         switch(bs->type) {          info->value->removable = bdrv_dev_has_removable_media(bs);
         case BDRV_TYPE_HD:  
             type = "hd";          if (bdrv_dev_has_removable_media(bs)) {
             break;              info->value->has_tray_open = true;
         case BDRV_TYPE_CDROM:              info->value->tray_open = bdrv_dev_is_tray_open(bs);
             type = "cdrom";          }
             break;  
         case BDRV_TYPE_FLOPPY:          if (bdrv_iostatus_is_enabled(bs)) {
             type = "floppy";              info->value->has_io_status = true;
             break;              info->value->io_status = bs->iostatus;
         }          }
   
         bs_obj = qobject_from_jsonf("{ 'device': %s, 'type': %s, "  
                                     "'removable': %i, 'locked': %i }",  
                                     bs->device_name, type, bs->removable,  
                                     bs->locked);  
         assert(bs_obj != NULL);  
   
         if (bs->drv) {          if (bs->drv) {
             QObject *obj;              info->value->has_inserted = true;
             QDict *bs_dict = qobject_to_qdict(bs_obj);              info->value->inserted = g_malloc0(sizeof(*info->value->inserted));
               info->value->inserted->file = g_strdup(bs->filename);
             obj = qobject_from_jsonf("{ 'file': %s, 'ro': %i, 'drv': %s, "              info->value->inserted->ro = bs->read_only;
                                      "'encrypted': %i }",              info->value->inserted->drv = g_strdup(bs->drv->format_name);
                                      bs->filename, bs->read_only,              info->value->inserted->encrypted = bs->encrypted;
                                      bs->drv->format_name,              if (bs->backing_file[0]) {
                                      bdrv_is_encrypted(bs));                  info->value->inserted->has_backing_file = true;
             assert(obj != NULL);                  info->value->inserted->backing_file = g_strdup(bs->backing_file);
             if (bs->backing_file[0] != '\0') {  
                 QDict *qdict = qobject_to_qdict(obj);  
                 qdict_put(qdict, "backing_file",  
                           qstring_from_str(bs->backing_file));  
             }              }
           }
   
             qdict_put_obj(bs_dict, "inserted", obj);          /* XXX: waiting for the qapi to support GSList */
           if (!cur_item) {
               head = cur_item = info;
           } else {
               cur_item->next = info;
               cur_item = info;
         }          }
         qlist_append_obj(bs_list, bs_obj);  
     }      }
   
     *ret_data = QOBJECT(bs_list);      return head;
 }  }
   
 static void bdrv_stats_iter(QObject *data, void *opaque)  /* Consider exposing this as a full fledged QMP command */
   static BlockStats *qmp_query_blockstat(const BlockDriverState *bs, Error **errp)
 {  {
     QDict *qdict;      BlockStats *s;
     Monitor *mon = opaque;  
   
     qdict = qobject_to_qdict(data);      s = g_malloc0(sizeof(*s));
     monitor_printf(mon, "%s:", qdict_get_str(qdict, "device"));  
   
     qdict = qobject_to_qdict(qdict_get(qdict, "stats"));      if (bs->device_name[0]) {
     monitor_printf(mon, " rd_bytes=%" PRId64          s->has_device = true;
                         " wr_bytes=%" PRId64          s->device = g_strdup(bs->device_name);
                         " rd_operations=%" PRId64      }
                         " wr_operations=%" PRId64  
                         "\n",  
                         qdict_get_int(qdict, "rd_bytes"),  
                         qdict_get_int(qdict, "wr_bytes"),  
                         qdict_get_int(qdict, "rd_operations"),  
                         qdict_get_int(qdict, "wr_operations"));  
 }  
   
 void bdrv_stats_print(Monitor *mon, const QObject *data)      s->stats = g_malloc0(sizeof(*s->stats));
 {      s->stats->rd_bytes = bs->nr_bytes[BDRV_ACCT_READ];
     qlist_iter(qobject_to_qlist(data), bdrv_stats_iter, mon);      s->stats->wr_bytes = bs->nr_bytes[BDRV_ACCT_WRITE];
       s->stats->rd_operations = bs->nr_ops[BDRV_ACCT_READ];
       s->stats->wr_operations = bs->nr_ops[BDRV_ACCT_WRITE];
       s->stats->wr_highest_offset = bs->wr_highest_sector * BDRV_SECTOR_SIZE;
       s->stats->flush_operations = bs->nr_ops[BDRV_ACCT_FLUSH];
       s->stats->wr_total_time_ns = bs->total_time_ns[BDRV_ACCT_WRITE];
       s->stats->rd_total_time_ns = bs->total_time_ns[BDRV_ACCT_READ];
       s->stats->flush_total_time_ns = bs->total_time_ns[BDRV_ACCT_FLUSH];
   
       if (bs->file) {
           s->has_parent = true;
           s->parent = qmp_query_blockstat(bs->file, NULL);
       }
   
       return s;
 }  }
   
 /**  BlockStatsList *qmp_query_blockstats(Error **errp)
  * bdrv_info_stats(): show block device statistics  
  *  
  * Each device statistic information is stored in a QDict and  
  * the returned QObject is a QList of all devices.  
  *  
  * The QDict contains the following:  
  *  
  * - "device": device name  
  * - "stats": A QDict with the statistics information, it contains:  
  *     - "rd_bytes": bytes read  
  *     - "wr_bytes": bytes written  
  *     - "rd_operations": read operations  
  *     - "wr_operations": write operations  
  *   
  * Example:  
  *  
  * [ { "device": "ide0-hd0",  
  *               "stats": { "rd_bytes": 512,  
  *                          "wr_bytes": 0,  
  *                          "rd_operations": 1,  
  *                          "wr_operations": 0 } },  
  *   { "device": "ide1-cd0",  
  *               "stats": { "rd_bytes": 0,  
  *                          "wr_bytes": 0,  
  *                          "rd_operations": 0,  
  *                          "wr_operations": 0 } } ]  
  */  
 void bdrv_info_stats(Monitor *mon, QObject **ret_data)  
 {  {
     QObject *obj;      BlockStatsList *head = NULL, *cur_item = NULL;
     QList *devices;  
     BlockDriverState *bs;      BlockDriverState *bs;
   
     devices = qlist_new();      QTAILQ_FOREACH(bs, &bdrv_states, list) {
           BlockStatsList *info = g_malloc0(sizeof(*info));
     for (bs = bdrv_first; bs != NULL; bs = bs->next) {          info->value = qmp_query_blockstat(bs, NULL);
         obj = qobject_from_jsonf("{ 'device': %s, 'stats': {"  
                                  "'rd_bytes': %" PRId64 ","          /* XXX: waiting for the qapi to support GSList */
                                  "'wr_bytes': %" PRId64 ","          if (!cur_item) {
                                  "'rd_operations': %" PRId64 ","              head = cur_item = info;
                                  "'wr_operations': %" PRId64          } else {
                                  "} }",              cur_item->next = info;
                                  bs->device_name,              cur_item = info;
                                  bs->rd_bytes, bs->wr_bytes,          }
                                  bs->rd_ops, bs->wr_ops);  
         assert(obj != NULL);  
         qlist_append_obj(devices, obj);  
     }      }
   
     *ret_data = QOBJECT(devices);      return head;
 }  }
   
 const char *bdrv_get_encrypted_filename(BlockDriverState *bs)  const char *bdrv_get_encrypted_filename(BlockDriverState *bs)
Line 1393  const char *bdrv_get_encrypted_filename( Line 1948  const char *bdrv_get_encrypted_filename(
 void bdrv_get_backing_filename(BlockDriverState *bs,  void bdrv_get_backing_filename(BlockDriverState *bs,
                                char *filename, int filename_size)                                 char *filename, int filename_size)
 {  {
     if (!bs->backing_hd) {      pstrcpy(filename, filename_size, bs->backing_file);
         pstrcpy(filename, filename_size, "");  
     } else {  
         pstrcpy(filename, filename_size, bs->backing_file);  
     }  
 }  }
   
 int bdrv_write_compressed(BlockDriverState *bs, int64_t sector_num,  int bdrv_write_compressed(BlockDriverState *bs, int64_t sector_num,
Line 1435  int bdrv_save_vmstate(BlockDriverState * Line 1986  int bdrv_save_vmstate(BlockDriverState *
     BlockDriver *drv = bs->drv;      BlockDriver *drv = bs->drv;
     if (!drv)      if (!drv)
         return -ENOMEDIUM;          return -ENOMEDIUM;
     if (!drv->bdrv_save_vmstate)      if (drv->bdrv_save_vmstate)
         return -ENOTSUP;          return drv->bdrv_save_vmstate(bs, buf, pos, size);
     return drv->bdrv_save_vmstate(bs, buf, pos, size);      if (bs->file)
           return bdrv_save_vmstate(bs->file, buf, pos, size);
       return -ENOTSUP;
 }  }
   
 int bdrv_load_vmstate(BlockDriverState *bs, uint8_t *buf,  int bdrv_load_vmstate(BlockDriverState *bs, uint8_t *buf,
Line 1446  int bdrv_load_vmstate(BlockDriverState * Line 1999  int bdrv_load_vmstate(BlockDriverState *
     BlockDriver *drv = bs->drv;      BlockDriver *drv = bs->drv;
     if (!drv)      if (!drv)
         return -ENOMEDIUM;          return -ENOMEDIUM;
     if (!drv->bdrv_load_vmstate)      if (drv->bdrv_load_vmstate)
         return -ENOTSUP;          return drv->bdrv_load_vmstate(bs, buf, pos, size);
     return drv->bdrv_load_vmstate(bs, buf, pos, size);      if (bs->file)
           return bdrv_load_vmstate(bs->file, buf, pos, size);
       return -ENOTSUP;
   }
   
   void bdrv_debug_event(BlockDriverState *bs, BlkDebugEvent event)
   {
       BlockDriver *drv = bs->drv;
   
       if (!drv || !drv->bdrv_debug_event) {
           return;
       }
   
       return drv->bdrv_debug_event(bs, event);
   
 }  }
   
 /**************************************************************/  /**************************************************************/
 /* handling of snapshots */  /* handling of snapshots */
   
   int bdrv_can_snapshot(BlockDriverState *bs)
   {
       BlockDriver *drv = bs->drv;
       if (!drv || !bdrv_is_inserted(bs) || bdrv_is_read_only(bs)) {
           return 0;
       }
   
       if (!drv->bdrv_snapshot_create) {
           if (bs->file != NULL) {
               return bdrv_can_snapshot(bs->file);
           }
           return 0;
       }
   
       return 1;
   }
   
   int bdrv_is_snapshot(BlockDriverState *bs)
   {
       return !!(bs->open_flags & BDRV_O_SNAPSHOT);
   }
   
   BlockDriverState *bdrv_snapshots(void)
   {
       BlockDriverState *bs;
   
       if (bs_snapshots) {
           return bs_snapshots;
       }
   
       bs = NULL;
       while ((bs = bdrv_next(bs))) {
           if (bdrv_can_snapshot(bs)) {
               bs_snapshots = bs;
               return bs;
           }
       }
       return NULL;
   }
   
 int bdrv_snapshot_create(BlockDriverState *bs,  int bdrv_snapshot_create(BlockDriverState *bs,
                          QEMUSnapshotInfo *sn_info)                           QEMUSnapshotInfo *sn_info)
 {  {
     BlockDriver *drv = bs->drv;      BlockDriver *drv = bs->drv;
     if (!drv)      if (!drv)
         return -ENOMEDIUM;          return -ENOMEDIUM;
     if (!drv->bdrv_snapshot_create)      if (drv->bdrv_snapshot_create)
         return -ENOTSUP;          return drv->bdrv_snapshot_create(bs, sn_info);
     return drv->bdrv_snapshot_create(bs, sn_info);      if (bs->file)
           return bdrv_snapshot_create(bs->file, sn_info);
       return -ENOTSUP;
 }  }
   
 int bdrv_snapshot_goto(BlockDriverState *bs,  int bdrv_snapshot_goto(BlockDriverState *bs,
                        const char *snapshot_id)                         const char *snapshot_id)
 {  {
     BlockDriver *drv = bs->drv;      BlockDriver *drv = bs->drv;
       int ret, open_ret;
   
     if (!drv)      if (!drv)
         return -ENOMEDIUM;          return -ENOMEDIUM;
     if (!drv->bdrv_snapshot_goto)      if (drv->bdrv_snapshot_goto)
         return -ENOTSUP;          return drv->bdrv_snapshot_goto(bs, snapshot_id);
     return drv->bdrv_snapshot_goto(bs, snapshot_id);  
       if (bs->file) {
           drv->bdrv_close(bs);
           ret = bdrv_snapshot_goto(bs->file, snapshot_id);
           open_ret = drv->bdrv_open(bs, bs->open_flags);
           if (open_ret < 0) {
               bdrv_delete(bs->file);
               bs->drv = NULL;
               return open_ret;
           }
           return ret;
       }
   
       return -ENOTSUP;
 }  }
   
 int bdrv_snapshot_delete(BlockDriverState *bs, const char *snapshot_id)  int bdrv_snapshot_delete(BlockDriverState *bs, const char *snapshot_id)
Line 1481  int bdrv_snapshot_delete(BlockDriverStat Line 2105  int bdrv_snapshot_delete(BlockDriverStat
     BlockDriver *drv = bs->drv;      BlockDriver *drv = bs->drv;
     if (!drv)      if (!drv)
         return -ENOMEDIUM;          return -ENOMEDIUM;
     if (!drv->bdrv_snapshot_delete)      if (drv->bdrv_snapshot_delete)
         return -ENOTSUP;          return drv->bdrv_snapshot_delete(bs, snapshot_id);
     return drv->bdrv_snapshot_delete(bs, snapshot_id);      if (bs->file)
           return bdrv_snapshot_delete(bs->file, snapshot_id);
       return -ENOTSUP;
 }  }
   
 int bdrv_snapshot_list(BlockDriverState *bs,  int bdrv_snapshot_list(BlockDriverState *bs,
Line 1492  int bdrv_snapshot_list(BlockDriverState  Line 2118  int bdrv_snapshot_list(BlockDriverState 
     BlockDriver *drv = bs->drv;      BlockDriver *drv = bs->drv;
     if (!drv)      if (!drv)
         return -ENOMEDIUM;          return -ENOMEDIUM;
     if (!drv->bdrv_snapshot_list)      if (drv->bdrv_snapshot_list)
         return -ENOTSUP;          return drv->bdrv_snapshot_list(bs, psn_info);
     return drv->bdrv_snapshot_list(bs, psn_info);      if (bs->file)
           return bdrv_snapshot_list(bs->file, psn_info);
       return -ENOTSUP;
   }
   
   int bdrv_snapshot_load_tmp(BlockDriverState *bs,
           const char *snapshot_name)
   {
       BlockDriver *drv = bs->drv;
       if (!drv) {
           return -ENOMEDIUM;
       }
       if (!bs->read_only) {
           return -EINVAL;
       }
       if (drv->bdrv_snapshot_load_tmp) {
           return drv->bdrv_snapshot_load_tmp(bs, snapshot_name);
       }
       return -ENOTSUP;
 }  }
   
 #define NB_SUFFIXES 4  #define NB_SUFFIXES 4
Line 1570  char *bdrv_snapshot_dump(char *buf, int  Line 2214  char *bdrv_snapshot_dump(char *buf, int 
     return buf;      return buf;
 }  }
   
   
 /**************************************************************/  /**************************************************************/
 /* async I/Os */  /* async I/Os */
   
Line 1578  BlockDriverAIOCB *bdrv_aio_readv(BlockDr Line 2221  BlockDriverAIOCB *bdrv_aio_readv(BlockDr
                                  QEMUIOVector *qiov, int nb_sectors,                                   QEMUIOVector *qiov, int nb_sectors,
                                  BlockDriverCompletionFunc *cb, void *opaque)                                   BlockDriverCompletionFunc *cb, void *opaque)
 {  {
     BlockDriver *drv = bs->drv;      trace_bdrv_aio_readv(bs, sector_num, nb_sectors, opaque);
     BlockDriverAIOCB *ret;  
   
     if (!drv)  
         return NULL;  
     if (bdrv_check_request(bs, sector_num, nb_sectors))  
         return NULL;  
   
     ret = drv->bdrv_aio_readv(bs, sector_num, qiov, nb_sectors,  
                               cb, opaque);  
   
     if (ret) {      return bdrv_co_aio_rw_vector(bs, sector_num, qiov, nb_sectors,
         /* Update stats even though technically transfer has not happened. */                                   cb, opaque, false);
         bs->rd_bytes += (unsigned) nb_sectors * BDRV_SECTOR_SIZE;  
         bs->rd_ops ++;  
     }  
   
     return ret;  
 }  }
   
 BlockDriverAIOCB *bdrv_aio_writev(BlockDriverState *bs, int64_t sector_num,  BlockDriverAIOCB *bdrv_aio_writev(BlockDriverState *bs, int64_t sector_num,
                                   QEMUIOVector *qiov, int nb_sectors,                                    QEMUIOVector *qiov, int nb_sectors,
                                   BlockDriverCompletionFunc *cb, void *opaque)                                    BlockDriverCompletionFunc *cb, void *opaque)
 {  {
     BlockDriver *drv = bs->drv;      trace_bdrv_aio_writev(bs, sector_num, nb_sectors, opaque);
     BlockDriverAIOCB *ret;  
   
     if (!drv)  
         return NULL;  
     if (bs->read_only)  
         return NULL;  
     if (bdrv_check_request(bs, sector_num, nb_sectors))  
         return NULL;  
   
     if (bs->dirty_bitmap) {      return bdrv_co_aio_rw_vector(bs, sector_num, qiov, nb_sectors,
         set_dirty_bitmap(bs, sector_num, nb_sectors, 1);                                   cb, opaque, true);
     }  
   
     ret = drv->bdrv_aio_writev(bs, sector_num, qiov, nb_sectors,  
                                cb, opaque);  
   
     if (ret) {  
         /* Update stats even though technically transfer has not happened. */  
         bs->wr_bytes += (unsigned) nb_sectors * BDRV_SECTOR_SIZE;  
         bs->wr_ops ++;  
     }  
   
     return ret;  
 }  }
   
   
Line 1650  static void multiwrite_user_cb(Multiwrit Line 2259  static void multiwrite_user_cb(Multiwrit
         if (mcb->callbacks[i].free_qiov) {          if (mcb->callbacks[i].free_qiov) {
             qemu_iovec_destroy(mcb->callbacks[i].free_qiov);              qemu_iovec_destroy(mcb->callbacks[i].free_qiov);
         }          }
         qemu_free(mcb->callbacks[i].free_qiov);          g_free(mcb->callbacks[i].free_qiov);
         qemu_vfree(mcb->callbacks[i].free_buf);          qemu_vfree(mcb->callbacks[i].free_buf);
     }      }
 }  }
Line 1659  static void multiwrite_cb(void *opaque,  Line 2268  static void multiwrite_cb(void *opaque, 
 {  {
     MultiwriteCB *mcb = opaque;      MultiwriteCB *mcb = opaque;
   
       trace_multiwrite_cb(mcb, ret);
   
     if (ret < 0 && !mcb->error) {      if (ret < 0 && !mcb->error) {
         mcb->error = ret;          mcb->error = ret;
     }      }
Line 1666  static void multiwrite_cb(void *opaque,  Line 2277  static void multiwrite_cb(void *opaque, 
     mcb->num_requests--;      mcb->num_requests--;
     if (mcb->num_requests == 0) {      if (mcb->num_requests == 0) {
         multiwrite_user_cb(mcb);          multiwrite_user_cb(mcb);
         qemu_free(mcb);          g_free(mcb);
     }      }
 }  }
   
Line 1726  static int multiwrite_merge(BlockDriverS Line 2337  static int multiwrite_merge(BlockDriverS
   
         if (merge) {          if (merge) {
             size_t size;              size_t size;
             QEMUIOVector *qiov = qemu_mallocz(sizeof(*qiov));              QEMUIOVector *qiov = g_malloc0(sizeof(*qiov));
             qemu_iovec_init(qiov,              qemu_iovec_init(qiov,
                 reqs[outidx].qiov->niov + reqs[i].qiov->niov + 1);                  reqs[outidx].qiov->niov + reqs[i].qiov->niov + 1);
   
Line 1782  int bdrv_aio_multiwrite(BlockDriverState Line 2393  int bdrv_aio_multiwrite(BlockDriverState
     MultiwriteCB *mcb;      MultiwriteCB *mcb;
     int i;      int i;
   
       /* don't submit writes if we don't have a medium */
       if (bs->drv == NULL) {
           for (i = 0; i < num_reqs; i++) {
               reqs[i].error = -ENOMEDIUM;
           }
           return -1;
       }
   
     if (num_reqs == 0) {      if (num_reqs == 0) {
         return 0;          return 0;
     }      }
   
     // Create MultiwriteCB structure      // Create MultiwriteCB structure
     mcb = qemu_mallocz(sizeof(*mcb) + num_reqs * sizeof(*mcb->callbacks));      mcb = g_malloc0(sizeof(*mcb) + num_reqs * sizeof(*mcb->callbacks));
     mcb->num_requests = 0;      mcb->num_requests = 0;
     mcb->num_callbacks = num_reqs;      mcb->num_callbacks = num_reqs;
   
Line 1799  int bdrv_aio_multiwrite(BlockDriverState Line 2418  int bdrv_aio_multiwrite(BlockDriverState
     // Check for mergable requests      // Check for mergable requests
     num_reqs = multiwrite_merge(bs, reqs, num_reqs, mcb);      num_reqs = multiwrite_merge(bs, reqs, num_reqs, mcb);
   
       trace_bdrv_aio_multiwrite(mcb, mcb->num_callbacks, num_reqs);
   
     /*      /*
      * Run the aio requests. As soon as one request can't be submitted       * Run the aio requests. As soon as one request can't be submitted
      * successfully, fail all requests that are not yet submitted (we must       * successfully, fail all requests that are not yet submitted (we must
Line 1820  int bdrv_aio_multiwrite(BlockDriverState Line 2441  int bdrv_aio_multiwrite(BlockDriverState
      */       */
     mcb->num_requests = 1;      mcb->num_requests = 1;
   
       // Run the aio requests
     for (i = 0; i < num_reqs; i++) {      for (i = 0; i < num_reqs; i++) {
         mcb->num_requests++;          mcb->num_requests++;
         acb = bdrv_aio_writev(bs, reqs[i].sector, reqs[i].qiov,          acb = bdrv_aio_writev(bs, reqs[i].sector, reqs[i].qiov,
Line 1830  int bdrv_aio_multiwrite(BlockDriverState Line 2452  int bdrv_aio_multiwrite(BlockDriverState
             // submitted yet. Otherwise we'll wait for the submitted AIOs to              // submitted yet. Otherwise we'll wait for the submitted AIOs to
             // complete and report the error in the callback.              // complete and report the error in the callback.
             if (i == 0) {              if (i == 0) {
                   trace_bdrv_aio_multiwrite_earlyfail(mcb);
                 goto fail;                  goto fail;
             } else {              } else {
                   trace_bdrv_aio_multiwrite_latefail(mcb, i);
                 multiwrite_cb(mcb, -EIO);                  multiwrite_cb(mcb, -EIO);
                 break;                  break;
             }              }
Line 1847  fail: Line 2471  fail:
     for (i = 0; i < mcb->num_callbacks; i++) {      for (i = 0; i < mcb->num_callbacks; i++) {
         reqs[i].error = -EIO;          reqs[i].error = -EIO;
     }      }
     qemu_free(mcb);      g_free(mcb);
     return -1;      return -1;
 }  }
   
 BlockDriverAIOCB *bdrv_aio_flush(BlockDriverState *bs,  
         BlockDriverCompletionFunc *cb, void *opaque)  
 {  
     BlockDriver *drv = bs->drv;  
   
     if (!drv)  
         return NULL;  
   
     /*  
      * Note that unlike bdrv_flush the driver is reponsible for flushing a  
      * backing image if it exists.  
      */  
     return drv->bdrv_aio_flush(bs, cb, opaque);  
 }  
   
 void bdrv_aio_cancel(BlockDriverAIOCB *acb)  void bdrv_aio_cancel(BlockDriverAIOCB *acb)
 {  {
     acb->pool->cancel(acb);      acb->pool->cancel(acb);
Line 1887  typedef struct BlockDriverAIOCBSync { Line 2496  typedef struct BlockDriverAIOCBSync {
   
 static void bdrv_aio_cancel_em(BlockDriverAIOCB *blockacb)  static void bdrv_aio_cancel_em(BlockDriverAIOCB *blockacb)
 {  {
     BlockDriverAIOCBSync *acb = (BlockDriverAIOCBSync *)blockacb;      BlockDriverAIOCBSync *acb =
           container_of(blockacb, BlockDriverAIOCBSync, common);
     qemu_bh_delete(acb->bh);      qemu_bh_delete(acb->bh);
     acb->bh = NULL;      acb->bh = NULL;
     qemu_aio_release(acb);      qemu_aio_release(acb);
Line 1932  static BlockDriverAIOCB *bdrv_aio_rw_vec Line 2542  static BlockDriverAIOCB *bdrv_aio_rw_vec
   
     if (is_write) {      if (is_write) {
         qemu_iovec_to_buffer(acb->qiov, acb->bounce);          qemu_iovec_to_buffer(acb->qiov, acb->bounce);
         acb->ret = bdrv_write(bs, sector_num, acb->bounce, nb_sectors);          acb->ret = bs->drv->bdrv_write(bs, sector_num, acb->bounce, nb_sectors);
     } else {      } else {
         acb->ret = bdrv_read(bs, sector_num, acb->bounce, nb_sectors);          acb->ret = bs->drv->bdrv_read(bs, sector_num, acb->bounce, nb_sectors);
     }      }
   
     qemu_bh_schedule(acb->bh);      qemu_bh_schedule(acb->bh);
Line 1956  static BlockDriverAIOCB *bdrv_aio_writev Line 2566  static BlockDriverAIOCB *bdrv_aio_writev
     return bdrv_aio_rw_vector(bs, sector_num, qiov, nb_sectors, cb, opaque, 1);      return bdrv_aio_rw_vector(bs, sector_num, qiov, nb_sectors, cb, opaque, 1);
 }  }
   
 static BlockDriverAIOCB *bdrv_aio_flush_em(BlockDriverState *bs,  
         BlockDriverCompletionFunc *cb, void *opaque)  typedef struct BlockDriverAIOCBCoroutine {
       BlockDriverAIOCB common;
       BlockRequest req;
       bool is_write;
       QEMUBH* bh;
   } BlockDriverAIOCBCoroutine;
   
   static void bdrv_aio_co_cancel_em(BlockDriverAIOCB *blockacb)
 {  {
     BlockDriverAIOCBSync *acb;      qemu_aio_flush();
   }
   
     acb = qemu_aio_get(&bdrv_em_aio_pool, bs, cb, opaque);  static AIOPool bdrv_em_co_aio_pool = {
     acb->is_write = 1; /* don't bounce in the completion hadler */      .aiocb_size         = sizeof(BlockDriverAIOCBCoroutine),
     acb->qiov = NULL;      .cancel             = bdrv_aio_co_cancel_em,
     acb->bounce = NULL;  };
     acb->ret = 0;  
   
     if (!acb->bh)  static void bdrv_co_em_bh(void *opaque)
         acb->bh = qemu_bh_new(bdrv_aio_bh_cb, acb);  {
       BlockDriverAIOCBCoroutine *acb = opaque;
   
       acb->common.cb(acb->common.opaque, acb->req.error);
       qemu_bh_delete(acb->bh);
       qemu_aio_release(acb);
   }
   
     bdrv_flush(bs);  /* Invoke bdrv_co_do_readv/bdrv_co_do_writev */
   static void coroutine_fn bdrv_co_do_rw(void *opaque)
   {
       BlockDriverAIOCBCoroutine *acb = opaque;
       BlockDriverState *bs = acb->common.bs;
   
       if (!acb->is_write) {
           acb->req.error = bdrv_co_do_readv(bs, acb->req.sector,
               acb->req.nb_sectors, acb->req.qiov);
       } else {
           acb->req.error = bdrv_co_do_writev(bs, acb->req.sector,
               acb->req.nb_sectors, acb->req.qiov);
       }
   
       acb->bh = qemu_bh_new(bdrv_co_em_bh, acb);
     qemu_bh_schedule(acb->bh);      qemu_bh_schedule(acb->bh);
     return &acb->common;  
 }  }
   
 /**************************************************************/  static BlockDriverAIOCB *bdrv_co_aio_rw_vector(BlockDriverState *bs,
 /* sync block device emulation */                                                 int64_t sector_num,
                                                  QEMUIOVector *qiov,
                                                  int nb_sectors,
                                                  BlockDriverCompletionFunc *cb,
                                                  void *opaque,
                                                  bool is_write)
   {
       Coroutine *co;
       BlockDriverAIOCBCoroutine *acb;
   
       acb = qemu_aio_get(&bdrv_em_co_aio_pool, bs, cb, opaque);
       acb->req.sector = sector_num;
       acb->req.nb_sectors = nb_sectors;
       acb->req.qiov = qiov;
       acb->is_write = is_write;
   
 static void bdrv_rw_em_cb(void *opaque, int ret)      co = qemu_coroutine_create(bdrv_co_do_rw);
 {      qemu_coroutine_enter(co, acb);
     *(int *)opaque = ret;  
       return &acb->common;
 }  }
   
 #define NOT_DONE 0x7fffffff  static void coroutine_fn bdrv_aio_flush_co_entry(void *opaque)
   {
       BlockDriverAIOCBCoroutine *acb = opaque;
       BlockDriverState *bs = acb->common.bs;
   
 static int bdrv_read_em(BlockDriverState *bs, int64_t sector_num,      acb->req.error = bdrv_co_flush(bs);
                         uint8_t *buf, int nb_sectors)      acb->bh = qemu_bh_new(bdrv_co_em_bh, acb);
       qemu_bh_schedule(acb->bh);
   }
   
   BlockDriverAIOCB *bdrv_aio_flush(BlockDriverState *bs,
           BlockDriverCompletionFunc *cb, void *opaque)
 {  {
     int async_ret;      trace_bdrv_aio_flush(bs, opaque);
     BlockDriverAIOCB *acb;  
     struct iovec iov;  
     QEMUIOVector qiov;  
   
     async_context_push();      Coroutine *co;
       BlockDriverAIOCBCoroutine *acb;
   
     async_ret = NOT_DONE;      acb = qemu_aio_get(&bdrv_em_co_aio_pool, bs, cb, opaque);
     iov.iov_base = (void *)buf;      co = qemu_coroutine_create(bdrv_aio_flush_co_entry);
     iov.iov_len = nb_sectors * 512;      qemu_coroutine_enter(co, acb);
     qemu_iovec_init_external(&qiov, &iov, 1);  
     acb = bdrv_aio_readv(bs, sector_num, &qiov, nb_sectors,  
         bdrv_rw_em_cb, &async_ret);  
     if (acb == NULL) {  
         async_ret = -1;  
         goto fail;  
     }  
   
     while (async_ret == NOT_DONE) {      return &acb->common;
         qemu_aio_wait();  }
     }  
   
   static void coroutine_fn bdrv_aio_discard_co_entry(void *opaque)
   {
       BlockDriverAIOCBCoroutine *acb = opaque;
       BlockDriverState *bs = acb->common.bs;
   
 fail:      acb->req.error = bdrv_co_discard(bs, acb->req.sector, acb->req.nb_sectors);
     async_context_pop();      acb->bh = qemu_bh_new(bdrv_co_em_bh, acb);
     return async_ret;      qemu_bh_schedule(acb->bh);
 }  }
   
 static int bdrv_write_em(BlockDriverState *bs, int64_t sector_num,  BlockDriverAIOCB *bdrv_aio_discard(BlockDriverState *bs,
                          const uint8_t *buf, int nb_sectors)          int64_t sector_num, int nb_sectors,
           BlockDriverCompletionFunc *cb, void *opaque)
 {  {
     int async_ret;      Coroutine *co;
     BlockDriverAIOCB *acb;      BlockDriverAIOCBCoroutine *acb;
     struct iovec iov;  
     QEMUIOVector qiov;  
   
     async_context_push();      trace_bdrv_aio_discard(bs, sector_num, nb_sectors, opaque);
   
     async_ret = NOT_DONE;      acb = qemu_aio_get(&bdrv_em_co_aio_pool, bs, cb, opaque);
     iov.iov_base = (void *)buf;      acb->req.sector = sector_num;
     iov.iov_len = nb_sectors * 512;      acb->req.nb_sectors = nb_sectors;
     qemu_iovec_init_external(&qiov, &iov, 1);      co = qemu_coroutine_create(bdrv_aio_discard_co_entry);
     acb = bdrv_aio_writev(bs, sector_num, &qiov, nb_sectors,      qemu_coroutine_enter(co, acb);
         bdrv_rw_em_cb, &async_ret);  
     if (acb == NULL) {  
         async_ret = -1;  
         goto fail;  
     }  
     while (async_ret == NOT_DONE) {  
         qemu_aio_wait();  
     }  
   
 fail:      return &acb->common;
     async_context_pop();  
     return async_ret;  
 }  }
   
 void bdrv_init(void)  void bdrv_init(void)
Line 2065  void *qemu_aio_get(AIOPool *pool, BlockD Line 2707  void *qemu_aio_get(AIOPool *pool, BlockD
         acb = pool->free_aiocb;          acb = pool->free_aiocb;
         pool->free_aiocb = acb->next;          pool->free_aiocb = acb->next;
     } else {      } else {
         acb = qemu_mallocz(pool->aiocb_size);          acb = g_malloc0(pool->aiocb_size);
         acb->pool = pool;          acb->pool = pool;
     }      }
     acb->bs = bs;      acb->bs = bs;
Line 2083  void qemu_aio_release(void *p) Line 2725  void qemu_aio_release(void *p)
 }  }
   
 /**************************************************************/  /**************************************************************/
   /* Coroutine block device emulation */
   
   typedef struct CoroutineIOCompletion {
       Coroutine *coroutine;
       int ret;
   } CoroutineIOCompletion;
   
   static void bdrv_co_io_em_complete(void *opaque, int ret)
   {
       CoroutineIOCompletion *co = opaque;
   
       co->ret = ret;
       qemu_coroutine_enter(co->coroutine, NULL);
   }
   
   static int coroutine_fn bdrv_co_io_em(BlockDriverState *bs, int64_t sector_num,
                                         int nb_sectors, QEMUIOVector *iov,
                                         bool is_write)
   {
       CoroutineIOCompletion co = {
           .coroutine = qemu_coroutine_self(),
       };
       BlockDriverAIOCB *acb;
   
       if (is_write) {
           acb = bs->drv->bdrv_aio_writev(bs, sector_num, iov, nb_sectors,
                                          bdrv_co_io_em_complete, &co);
       } else {
           acb = bs->drv->bdrv_aio_readv(bs, sector_num, iov, nb_sectors,
                                         bdrv_co_io_em_complete, &co);
       }
   
       trace_bdrv_co_io_em(bs, sector_num, nb_sectors, is_write, acb);
       if (!acb) {
           return -EIO;
       }
       qemu_coroutine_yield();
   
       return co.ret;
   }
   
   static int coroutine_fn bdrv_co_readv_em(BlockDriverState *bs,
                                            int64_t sector_num, int nb_sectors,
                                            QEMUIOVector *iov)
   {
       return bdrv_co_io_em(bs, sector_num, nb_sectors, iov, false);
   }
   
   static int coroutine_fn bdrv_co_writev_em(BlockDriverState *bs,
                                            int64_t sector_num, int nb_sectors,
                                            QEMUIOVector *iov)
   {
       return bdrv_co_io_em(bs, sector_num, nb_sectors, iov, true);
   }
   
   static void coroutine_fn bdrv_flush_co_entry(void *opaque)
   {
       RwCo *rwco = opaque;
   
       rwco->ret = bdrv_co_flush(rwco->bs);
   }
   
   int coroutine_fn bdrv_co_flush(BlockDriverState *bs)
   {
       int ret;
   
       if (!bs->drv) {
           return 0;
       }
   
       /* Write back cached data to the OS even with cache=unsafe */
       if (bs->drv->bdrv_co_flush_to_os) {
           ret = bs->drv->bdrv_co_flush_to_os(bs);
           if (ret < 0) {
               return ret;
           }
       }
   
       /* But don't actually force it to the disk with cache=unsafe */
       if (bs->open_flags & BDRV_O_NO_FLUSH) {
           return 0;
       }
   
       if (bs->drv->bdrv_co_flush_to_disk) {
           return bs->drv->bdrv_co_flush_to_disk(bs);
       } else if (bs->drv->bdrv_aio_flush) {
           BlockDriverAIOCB *acb;
           CoroutineIOCompletion co = {
               .coroutine = qemu_coroutine_self(),
           };
   
           acb = bs->drv->bdrv_aio_flush(bs, bdrv_co_io_em_complete, &co);
           if (acb == NULL) {
               return -EIO;
           } else {
               qemu_coroutine_yield();
               return co.ret;
           }
       } else {
           /*
            * Some block drivers always operate in either writethrough or unsafe
            * mode and don't support bdrv_flush therefore. Usually qemu doesn't
            * know how the server works (because the behaviour is hardcoded or
            * depends on server-side configuration), so we can't ensure that
            * everything is safe on disk. Returning an error doesn't work because
            * that would break guests even if the server operates in writethrough
            * mode.
            *
            * Let's hope the user knows what he's doing.
            */
           return 0;
       }
   }
   
   void bdrv_invalidate_cache(BlockDriverState *bs)
   {
       if (bs->drv && bs->drv->bdrv_invalidate_cache) {
           bs->drv->bdrv_invalidate_cache(bs);
       }
   }
   
   void bdrv_invalidate_cache_all(void)
   {
       BlockDriverState *bs;
   
       QTAILQ_FOREACH(bs, &bdrv_states, list) {
           bdrv_invalidate_cache(bs);
       }
   }
   
   int bdrv_flush(BlockDriverState *bs)
   {
       Coroutine *co;
       RwCo rwco = {
           .bs = bs,
           .ret = NOT_DONE,
       };
   
       if (qemu_in_coroutine()) {
           /* Fast-path if already in coroutine context */
           bdrv_flush_co_entry(&rwco);
       } else {
           co = qemu_coroutine_create(bdrv_flush_co_entry);
           qemu_coroutine_enter(co, &rwco);
           while (rwco.ret == NOT_DONE) {
               qemu_aio_wait();
           }
       }
   
       return rwco.ret;
   }
   
   static void coroutine_fn bdrv_discard_co_entry(void *opaque)
   {
       RwCo *rwco = opaque;
   
       rwco->ret = bdrv_co_discard(rwco->bs, rwco->sector_num, rwco->nb_sectors);
   }
   
   int coroutine_fn bdrv_co_discard(BlockDriverState *bs, int64_t sector_num,
                                    int nb_sectors)
   {
       if (!bs->drv) {
           return -ENOMEDIUM;
       } else if (bdrv_check_request(bs, sector_num, nb_sectors)) {
           return -EIO;
       } else if (bs->read_only) {
           return -EROFS;
       } else if (bs->drv->bdrv_co_discard) {
           return bs->drv->bdrv_co_discard(bs, sector_num, nb_sectors);
       } else if (bs->drv->bdrv_aio_discard) {
           BlockDriverAIOCB *acb;
           CoroutineIOCompletion co = {
               .coroutine = qemu_coroutine_self(),
           };
   
           acb = bs->drv->bdrv_aio_discard(bs, sector_num, nb_sectors,
                                           bdrv_co_io_em_complete, &co);
           if (acb == NULL) {
               return -EIO;
           } else {
               qemu_coroutine_yield();
               return co.ret;
           }
       } else {
           return 0;
       }
   }
   
   int bdrv_discard(BlockDriverState *bs, int64_t sector_num, int nb_sectors)
   {
       Coroutine *co;
       RwCo rwco = {
           .bs = bs,
           .sector_num = sector_num,
           .nb_sectors = nb_sectors,
           .ret = NOT_DONE,
       };
   
       if (qemu_in_coroutine()) {
           /* Fast-path if already in coroutine context */
           bdrv_discard_co_entry(&rwco);
       } else {
           co = qemu_coroutine_create(bdrv_discard_co_entry);
           qemu_coroutine_enter(co, &rwco);
           while (rwco.ret == NOT_DONE) {
               qemu_aio_wait();
           }
       }
   
       return rwco.ret;
   }
   
   /**************************************************************/
 /* removable device support */  /* removable device support */
   
 /**  /**
Line 2091  void qemu_aio_release(void *p) Line 2947  void qemu_aio_release(void *p)
 int bdrv_is_inserted(BlockDriverState *bs)  int bdrv_is_inserted(BlockDriverState *bs)
 {  {
     BlockDriver *drv = bs->drv;      BlockDriver *drv = bs->drv;
     int ret;  
     if (!drv)      if (!drv)
         return 0;          return 0;
     if (!drv->bdrv_is_inserted)      if (!drv->bdrv_is_inserted)
         return 1;          return 1;
     ret = drv->bdrv_is_inserted(bs);      return drv->bdrv_is_inserted(bs);
     return ret;  
 }  }
   
 /**  /**
  * Return TRUE if the media changed since the last call to this   * Return whether the media changed since the last call to this
  * function. It is currently only used for floppy disks   * function, or -ENOTSUP if we don't know.  Most drivers don't know.
  */   */
 int bdrv_media_changed(BlockDriverState *bs)  int bdrv_media_changed(BlockDriverState *bs)
 {  {
     BlockDriver *drv = bs->drv;      BlockDriver *drv = bs->drv;
     int ret;  
   
     if (!drv || !drv->bdrv_media_changed)      if (drv && drv->bdrv_media_changed) {
         ret = -ENOTSUP;          return drv->bdrv_media_changed(bs);
     else      }
         ret = drv->bdrv_media_changed(bs);      return -ENOTSUP;
     if (ret == -ENOTSUP)  
         ret = bs->media_changed;  
     bs->media_changed = 0;  
     return ret;  
 }  }
   
 /**  /**
  * If eject_flag is TRUE, eject the media. Otherwise, close the tray   * If eject_flag is TRUE, eject the media. Otherwise, close the tray
  */   */
 int bdrv_eject(BlockDriverState *bs, int eject_flag)  void bdrv_eject(BlockDriverState *bs, int eject_flag)
 {  {
     BlockDriver *drv = bs->drv;      BlockDriver *drv = bs->drv;
     int ret;  
   
     if (bs->locked) {  
         return -EBUSY;  
     }  
   
     if (!drv || !drv->bdrv_eject) {      if (drv && drv->bdrv_eject) {
         ret = -ENOTSUP;          drv->bdrv_eject(bs, eject_flag);
     } else {  
         ret = drv->bdrv_eject(bs, eject_flag);  
     }  
     if (ret == -ENOTSUP) {  
         if (eject_flag)  
             bdrv_close(bs);  
         ret = 0;  
     }      }
   
     return ret;  
 }  
   
 int bdrv_is_locked(BlockDriverState *bs)  
 {  
     return bs->locked;  
 }  }
   
 /**  /**
  * Lock or unlock the media (if it is locked, the user won't be able   * Lock or unlock the media (if it is locked, the user won't be able
  * to eject it manually).   * to eject it manually).
  */   */
 void bdrv_set_locked(BlockDriverState *bs, int locked)  void bdrv_lock_medium(BlockDriverState *bs, bool locked)
 {  {
     BlockDriver *drv = bs->drv;      BlockDriver *drv = bs->drv;
   
     bs->locked = locked;      trace_bdrv_lock_medium(bs, locked);
     if (drv && drv->bdrv_set_locked) {  
         drv->bdrv_set_locked(bs, locked);      if (drv && drv->bdrv_lock_medium) {
           drv->bdrv_lock_medium(bs, locked);
     }      }
 }  }
   
Line 2186  BlockDriverAIOCB *bdrv_aio_ioctl(BlockDr Line 3018  BlockDriverAIOCB *bdrv_aio_ioctl(BlockDr
     return NULL;      return NULL;
 }  }
   
   void bdrv_set_buffer_alignment(BlockDriverState *bs, int align)
   {
       bs->buffer_alignment = align;
   }
   
 void *qemu_blockalign(BlockDriverState *bs, size_t size)  void *qemu_blockalign(BlockDriverState *bs, size_t size)
 {  {
Line 2197  void bdrv_set_dirty_tracking(BlockDriver Line 3032  void bdrv_set_dirty_tracking(BlockDriver
 {  {
     int64_t bitmap_size;      int64_t bitmap_size;
   
       bs->dirty_count = 0;
     if (enable) {      if (enable) {
         if (!bs->dirty_bitmap) {          if (!bs->dirty_bitmap) {
             bitmap_size = (bdrv_getlength(bs) >> BDRV_SECTOR_BITS) +              bitmap_size = (bdrv_getlength(bs) >> BDRV_SECTOR_BITS) +
                     BDRV_SECTORS_PER_DIRTY_CHUNK * 8 - 1;                      BDRV_SECTORS_PER_DIRTY_CHUNK * 8 - 1;
             bitmap_size /= BDRV_SECTORS_PER_DIRTY_CHUNK * 8;              bitmap_size /= BDRV_SECTORS_PER_DIRTY_CHUNK * 8;
   
             bs->dirty_bitmap = qemu_mallocz(bitmap_size);              bs->dirty_bitmap = g_malloc0(bitmap_size);
         }          }
     } else {      } else {
         if (bs->dirty_bitmap) {          if (bs->dirty_bitmap) {
             qemu_free(bs->dirty_bitmap);              g_free(bs->dirty_bitmap);
             bs->dirty_bitmap = NULL;              bs->dirty_bitmap = NULL;
         }          }
     }      }
Line 2219  int bdrv_get_dirty(BlockDriverState *bs, Line 3055  int bdrv_get_dirty(BlockDriverState *bs,
   
     if (bs->dirty_bitmap &&      if (bs->dirty_bitmap &&
         (sector << BDRV_SECTOR_BITS) < bdrv_getlength(bs)) {          (sector << BDRV_SECTOR_BITS) < bdrv_getlength(bs)) {
         return bs->dirty_bitmap[chunk / (sizeof(unsigned long) * 8)] &          return !!(bs->dirty_bitmap[chunk / (sizeof(unsigned long) * 8)] &
             (1 << (chunk % (sizeof(unsigned long) * 8)));              (1UL << (chunk % (sizeof(unsigned long) * 8))));
     } else {      } else {
         return 0;          return 0;
     }      }
Line 2231  void bdrv_reset_dirty(BlockDriverState * Line 3067  void bdrv_reset_dirty(BlockDriverState *
 {  {
     set_dirty_bitmap(bs, cur_sector, nr_sectors, 0);      set_dirty_bitmap(bs, cur_sector, nr_sectors, 0);
 }  }
   
   int64_t bdrv_get_dirty_count(BlockDriverState *bs)
   {
       return bs->dirty_count;
   }
   
   void bdrv_set_in_use(BlockDriverState *bs, int in_use)
   {
       assert(bs->in_use != in_use);
       bs->in_use = in_use;
   }
   
   int bdrv_in_use(BlockDriverState *bs)
   {
       return bs->in_use;
   }
   
   void bdrv_iostatus_enable(BlockDriverState *bs)
   {
       bs->iostatus_enabled = true;
       bs->iostatus = BLOCK_DEVICE_IO_STATUS_OK;
   }
   
   /* The I/O status is only enabled if the drive explicitly
    * enables it _and_ the VM is configured to stop on errors */
   bool bdrv_iostatus_is_enabled(const BlockDriverState *bs)
   {
       return (bs->iostatus_enabled &&
              (bs->on_write_error == BLOCK_ERR_STOP_ENOSPC ||
               bs->on_write_error == BLOCK_ERR_STOP_ANY    ||
               bs->on_read_error == BLOCK_ERR_STOP_ANY));
   }
   
   void bdrv_iostatus_disable(BlockDriverState *bs)
   {
       bs->iostatus_enabled = false;
   }
   
   void bdrv_iostatus_reset(BlockDriverState *bs)
   {
       if (bdrv_iostatus_is_enabled(bs)) {
           bs->iostatus = BLOCK_DEVICE_IO_STATUS_OK;
       }
   }
   
   /* XXX: Today this is set by device models because it makes the implementation
      quite simple. However, the block layer knows about the error, so it's
      possible to implement this without device models being involved */
   void bdrv_iostatus_set_err(BlockDriverState *bs, int error)
   {
       if (bdrv_iostatus_is_enabled(bs) &&
           bs->iostatus == BLOCK_DEVICE_IO_STATUS_OK) {
           assert(error >= 0);
           bs->iostatus = error == ENOSPC ? BLOCK_DEVICE_IO_STATUS_NOSPACE :
                                            BLOCK_DEVICE_IO_STATUS_FAILED;
       }
   }
   
   void
   bdrv_acct_start(BlockDriverState *bs, BlockAcctCookie *cookie, int64_t bytes,
           enum BlockAcctType type)
   {
       assert(type < BDRV_MAX_IOTYPE);
   
       cookie->bytes = bytes;
       cookie->start_time_ns = get_clock();
       cookie->type = type;
   }
   
   void
   bdrv_acct_done(BlockDriverState *bs, BlockAcctCookie *cookie)
   {
       assert(cookie->type < BDRV_MAX_IOTYPE);
   
       bs->nr_bytes[cookie->type] += cookie->bytes;
       bs->nr_ops[cookie->type]++;
       bs->total_time_ns[cookie->type] += get_clock() - cookie->start_time_ns;
   }
   
   int bdrv_img_create(const char *filename, const char *fmt,
                       const char *base_filename, const char *base_fmt,
                       char *options, uint64_t img_size, int flags)
   {
       QEMUOptionParameter *param = NULL, *create_options = NULL;
       QEMUOptionParameter *backing_fmt, *backing_file, *size;
       BlockDriverState *bs = NULL;
       BlockDriver *drv, *proto_drv;
       BlockDriver *backing_drv = NULL;
       int ret = 0;
   
       /* Find driver and parse its options */
       drv = bdrv_find_format(fmt);
       if (!drv) {
           error_report("Unknown file format '%s'", fmt);
           ret = -EINVAL;
           goto out;
       }
   
       proto_drv = bdrv_find_protocol(filename);
       if (!proto_drv) {
           error_report("Unknown protocol '%s'", filename);
           ret = -EINVAL;
           goto out;
       }
   
       create_options = append_option_parameters(create_options,
                                                 drv->create_options);
       create_options = append_option_parameters(create_options,
                                                 proto_drv->create_options);
   
       /* Create parameter list with default values */
       param = parse_option_parameters("", create_options, param);
   
       set_option_parameter_int(param, BLOCK_OPT_SIZE, img_size);
   
       /* Parse -o options */
       if (options) {
           param = parse_option_parameters(options, create_options, param);
           if (param == NULL) {
               error_report("Invalid options for file format '%s'.", fmt);
               ret = -EINVAL;
               goto out;
           }
       }
   
       if (base_filename) {
           if (set_option_parameter(param, BLOCK_OPT_BACKING_FILE,
                                    base_filename)) {
               error_report("Backing file not supported for file format '%s'",
                            fmt);
               ret = -EINVAL;
               goto out;
           }
       }
   
       if (base_fmt) {
           if (set_option_parameter(param, BLOCK_OPT_BACKING_FMT, base_fmt)) {
               error_report("Backing file format not supported for file "
                            "format '%s'", fmt);
               ret = -EINVAL;
               goto out;
           }
       }
   
       backing_file = get_option_parameter(param, BLOCK_OPT_BACKING_FILE);
       if (backing_file && backing_file->value.s) {
           if (!strcmp(filename, backing_file->value.s)) {
               error_report("Error: Trying to create an image with the "
                            "same filename as the backing file");
               ret = -EINVAL;
               goto out;
           }
       }
   
       backing_fmt = get_option_parameter(param, BLOCK_OPT_BACKING_FMT);
       if (backing_fmt && backing_fmt->value.s) {
           backing_drv = bdrv_find_format(backing_fmt->value.s);
           if (!backing_drv) {
               error_report("Unknown backing file format '%s'",
                            backing_fmt->value.s);
               ret = -EINVAL;
               goto out;
           }
       }
   
       // The size for the image must always be specified, with one exception:
       // If we are using a backing file, we can obtain the size from there
       size = get_option_parameter(param, BLOCK_OPT_SIZE);
       if (size && size->value.n == -1) {
           if (backing_file && backing_file->value.s) {
               uint64_t size;
               char buf[32];
   
               bs = bdrv_new("");
   
               ret = bdrv_open(bs, backing_file->value.s, flags, backing_drv);
               if (ret < 0) {
                   error_report("Could not open '%s'", backing_file->value.s);
                   goto out;
               }
               bdrv_get_geometry(bs, &size);
               size *= 512;
   
               snprintf(buf, sizeof(buf), "%" PRId64, size);
               set_option_parameter(param, BLOCK_OPT_SIZE, buf);
           } else {
               error_report("Image creation needs a size parameter");
               ret = -EINVAL;
               goto out;
           }
       }
   
       printf("Formatting '%s', fmt=%s ", filename, fmt);
       print_option_parameters(param);
       puts("");
   
       ret = bdrv_create(drv, filename, param);
   
       if (ret < 0) {
           if (ret == -ENOTSUP) {
               error_report("Formatting or formatting option not supported for "
                            "file format '%s'", fmt);
           } else if (ret == -EFBIG) {
               error_report("The image size is too large for file format '%s'",
                            fmt);
           } else {
               error_report("%s: error while creating %s: %s", filename, fmt,
                            strerror(-ret));
           }
       }
   
   out:
       free_option_parameters(create_options);
       free_option_parameters(param);
   
       if (bs) {
           bdrv_delete(bs);
       }
   
       return ret;
   }

Removed from v.1.1.1.17  
changed lines
  Added in v.1.1.1.22


unix.superglobalmegacorp.com