Diff for /qemu/block.c between versions 1.1.1.12 and 1.1.1.19

version 1.1.1.12, 2018/04/24 17:16:37 version 1.1.1.19, 2018/04/24 18:34:00
Line 22 Line 22
  * THE SOFTWARE.   * THE SOFTWARE.
  */   */
 #include "config-host.h"  #include "config-host.h"
 #ifdef _BSD  
 /* include native header before sys-queue.h */  
 #include <sys/queue.h>  
 #endif  
   
 #include "qemu-common.h"  #include "qemu-common.h"
 #include "console.h"  #include "trace.h"
   #include "monitor.h"
 #include "block_int.h"  #include "block_int.h"
   #include "module.h"
   #include "qemu-objects.h"
   
 #ifdef _BSD  #ifdef CONFIG_BSD
 #include <sys/types.h>  #include <sys/types.h>
 #include <sys/stat.h>  #include <sys/stat.h>
 #include <sys/ioctl.h>  #include <sys/ioctl.h>
   #include <sys/queue.h>
   #ifndef __DragonFly__
 #include <sys/disk.h>  #include <sys/disk.h>
 #endif  #endif
   #endif
   
 #define SECTOR_BITS 9  #ifdef _WIN32
 #define SECTOR_SIZE (1 << SECTOR_BITS)  #include <windows.h>
   #endif
 static AIOPool vectored_aio_pool;  
   
 typedef struct BlockDriverAIOCBSync {  
     BlockDriverAIOCB common;  
     QEMUBH *bh;  
     int ret;  
 } BlockDriverAIOCBSync;  
   
 static BlockDriverAIOCB *bdrv_aio_read_em(BlockDriverState *bs,  static BlockDriverAIOCB *bdrv_aio_readv_em(BlockDriverState *bs,
         int64_t sector_num, uint8_t *buf, int nb_sectors,          int64_t sector_num, QEMUIOVector *qiov, int nb_sectors,
           BlockDriverCompletionFunc *cb, void *opaque);
   static BlockDriverAIOCB *bdrv_aio_writev_em(BlockDriverState *bs,
           int64_t sector_num, QEMUIOVector *qiov, int nb_sectors,
           BlockDriverCompletionFunc *cb, void *opaque);
   static BlockDriverAIOCB *bdrv_aio_flush_em(BlockDriverState *bs,
         BlockDriverCompletionFunc *cb, void *opaque);          BlockDriverCompletionFunc *cb, void *opaque);
 static BlockDriverAIOCB *bdrv_aio_write_em(BlockDriverState *bs,  static BlockDriverAIOCB *bdrv_aio_noop_em(BlockDriverState *bs,
         int64_t sector_num, const uint8_t *buf, int nb_sectors,  
         BlockDriverCompletionFunc *cb, void *opaque);          BlockDriverCompletionFunc *cb, void *opaque);
 static void bdrv_aio_cancel_em(BlockDriverAIOCB *acb);  
 static int bdrv_read_em(BlockDriverState *bs, int64_t sector_num,  static int bdrv_read_em(BlockDriverState *bs, int64_t sector_num,
                         uint8_t *buf, int nb_sectors);                          uint8_t *buf, int nb_sectors);
 static int bdrv_write_em(BlockDriverState *bs, int64_t sector_num,  static int bdrv_write_em(BlockDriverState *bs, int64_t sector_num,
                          const uint8_t *buf, int nb_sectors);                           const uint8_t *buf, int nb_sectors);
   
 BlockDriverState *bdrv_first;  static QTAILQ_HEAD(, BlockDriverState) bdrv_states =
       QTAILQ_HEAD_INITIALIZER(bdrv_states);
   
   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 */
   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
   
 static BlockDriver *first_drv;      return strchr(path, ':') != NULL;
   }
   
 int path_is_absolute(const char *path)  int path_is_absolute(const char *path)
 {  {
Line 129  void path_combine(char *dest, int dest_s Line 167  void path_combine(char *dest, int dest_s
     }      }
 }  }
   
   void bdrv_register(BlockDriver *bdrv)
 static void bdrv_register(BlockDriver *bdrv)  
 {  {
     if (!bdrv->bdrv_aio_read) {      if (!bdrv->bdrv_aio_readv) {
         /* add AIO emulation layer */          /* add AIO emulation layer */
         bdrv->bdrv_aio_read = bdrv_aio_read_em;          bdrv->bdrv_aio_readv = bdrv_aio_readv_em;
         bdrv->bdrv_aio_write = bdrv_aio_write_em;          bdrv->bdrv_aio_writev = bdrv_aio_writev_em;
         bdrv->bdrv_aio_cancel = bdrv_aio_cancel_em;      } else if (!bdrv->bdrv_read) {
         bdrv->aiocb_size = sizeof(BlockDriverAIOCBSync);  
     } else if (!bdrv->bdrv_read && !bdrv->bdrv_pread) {  
         /* add synchronous IO emulation layer */          /* add synchronous IO emulation layer */
         bdrv->bdrv_read = bdrv_read_em;          bdrv->bdrv_read = bdrv_read_em;
         bdrv->bdrv_write = bdrv_write_em;          bdrv->bdrv_write = bdrv_write_em;
     }      }
     aio_pool_init(&bdrv->aio_pool, bdrv->aiocb_size, bdrv->bdrv_aio_cancel);  
     bdrv->next = first_drv;      if (!bdrv->bdrv_aio_flush)
     first_drv = bdrv;          bdrv->bdrv_aio_flush = bdrv_aio_flush_em;
   
       QLIST_INSERT_HEAD(&bdrv_drivers, bdrv, list);
 }  }
   
 /* 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 = qemu_mallocz(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;  
     }      }
     return bs;      return bs;
 }  }
Line 168  BlockDriverState *bdrv_new(const char *d Line 201  BlockDriverState *bdrv_new(const char *d
 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;
 }  }
   
 int bdrv_create(BlockDriver *drv,  static int bdrv_is_whitelisted(BlockDriver *drv)
                 const char *filename, int64_t size_in_sectors,  {
                 const char *backing_file, int flags)      static const char *whitelist[] = {
           CONFIG_BDRV_WHITELIST
       };
       const char **p;
   
       if (!whitelist[0])
           return 1;               /* no whitelist, anything goes */
   
       for (p = whitelist; *p; p++) {
           if (!strcmp(drv->format_name, *p)) {
               return 1;
           }
       }
       return 0;
   }
   
   BlockDriver *bdrv_find_whitelisted_format(const char *format_name)
   {
       BlockDriver *drv = bdrv_find_format(format_name);
       return drv && bdrv_is_whitelisted(drv) ? drv : NULL;
   }
   
   int bdrv_create(BlockDriver *drv, const char* filename,
       QEMUOptionParameter *options)
 {  {
     if (!drv->bdrv_create)      if (!drv->bdrv_create)
         return -ENOTSUP;          return -ENOTSUP;
     return drv->bdrv_create(filename, size_in_sectors, backing_file, flags);  
       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
Line 207  void get_tmp_filename(char *filename, in Line 277  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] == ':');  
 }  
   
 static 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_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_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;      return NULL;
 }  }
   
 /* XXX: force raw format if block or character device ? It would  static int find_image_format(const char *filename, BlockDriver **pdrv)
    simplify the BSD case */  
 static BlockDriver *find_image_format(const char *filename)  
 {  {
     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;
   
     /* detect host devices. By convention, /dev/cdrom[N] is always      ret = bdrv_file_open(&bs, filename, 0);
        recognized as a host CDROM */      if (ret < 0) {
     if (strstart(filename, "/dev/cdrom", NULL))          *pdrv = NULL;
         return &bdrv_host_device;          return ret;
 #ifdef _WIN32  
     if (is_windows_drive(filename))  
         return &bdrv_host_device;  
 #else  
     {  
         struct stat st;  
         if (stat(filename, &st) >= 0 &&  
             (S_ISCHR(st.st_mode) || S_ISBLK(st.st_mode))) {  
             return &bdrv_host_device;  
         }  
     }      }
 #endif  
   
     drv = find_protocol(filename);      /* Return the raw BlockDriver * to scsi-generic devices or empty drives */
     /* no need to test disk image formats for vvfat */      if (bs->sg || !bdrv_is_inserted(bs)) {
     if (drv == &bdrv_vvfat)          bdrv_delete(bs);
         return drv;          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 305  static BlockDriver *find_image_format(co Line 381  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;
   }
   
   /*
    * 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);
   
       bs->file = NULL;
       bs->total_sectors = 0;
       bs->encrypted = 0;
       bs->valid_key = 0;
       bs->open_flags = flags;
       /* buffer_alignment defaulted to 512, drivers can change this value */
       bs->buffer_alignment = 512;
   
       pstrcpy(bs->filename, sizeof(bs->filename), filename);
   
       if (use_bdrv_whitelist && !bdrv_is_whitelisted(drv)) {
           return -ENOTSUP;
       }
   
       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;
   
       /*
        * 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 writeable.
        */
       if (bs->is_temporary) {
           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;
       }
   
       bs->keep_read_only = bs->read_only = !(open_flags & BDRV_O_RDWR);
   
       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;
       }
       qemu_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 324  int bdrv_file_open(BlockDriverState **pb Line 525  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;      int ret;
     char tmp_filename[PATH_MAX];  
     char backing_filename[PATH_MAX];  
   
     bs->read_only = 0;  
     bs->is_temporary = 0;  
     bs->encrypted = 0;  
     bs->valid_key = 0;  
   
     if (flags & BDRV_O_SNAPSHOT) {      if (flags & BDRV_O_SNAPSHOT) {
         BlockDriverState *bs1;          BlockDriverState *bs1;
         int64_t total_size;          int64_t total_size;
         int is_protocol = 0;          int is_protocol = 0;
           BlockDriver *bdrv_qcow2;
           QEMUOptionParameter *options;
           char tmp_filename[PATH_MAX];
           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_open(bs1, filename, 0);          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) >> 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 369  int bdrv_open2(BlockDriverState *bs, con Line 565  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");
           options = parse_option_parameters("", bdrv_qcow2->create_options, NULL);
   
           set_option_parameter_int(options, BLOCK_OPT_SIZE, total_size);
           set_option_parameter(options, BLOCK_OPT_BACKING_FILE, backing_filename);
           if (drv) {
               set_option_parameter(options, BLOCK_OPT_BACKING_FMT,
                   drv->format_name);
           }
   
         ret = bdrv_create(&bdrv_qcow2, tmp_filename,          ret = bdrv_create(bdrv_qcow2, tmp_filename, options);
                           total_size, backing_filename, 0);          free_option_parameters(options);
         if (ret < 0) {          if (ret < 0) {
             return ret;              return ret;
         }          }
   
         filename = tmp_filename;          filename = tmp_filename;
           drv = bdrv_qcow2;
         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_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);      /* Open the image */
     /* Note: for compatibility, we open disk image files as RDWR, and      ret = bdrv_open_common(bs, filename, flags, drv);
        RDONLY as fallback */  
     if (!(flags & BDRV_O_FILE))  
         open_flags = BDRV_O_RDWR | (flags & BDRV_O_CACHE_MASK);  
     else  
         open_flags = flags & ~(BDRV_O_FILE | BDRV_O_SNAPSHOT);  
     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) >> 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;
   
         bs->backing_hd = bdrv_new("");          bs->backing_hd = bdrv_new("");
         path_combine(backing_filename, sizeof(backing_filename),  
                      filename, bs->backing_file);          if (path_has_protocol(bs->backing_file)) {
         ret = bdrv_open(bs->backing_hd, backing_filename, open_flags);              pstrcpy(backing_filename, sizeof(backing_filename),
                       bs->backing_file);
           } 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);
           }
   
           /* 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;
           }
     }      }
   
     /* call the change callback */      if (!bdrv_key_required(bs)) {
     bs->media_changed = 1;          /* call the change callback */
     if (bs->change_cb)          bs->media_changed = 1;
         bs->change_cb(bs->change_opaque);          if (bs->change_cb)
               bs->change_cb(bs->change_opaque, CHANGE_MEDIA);
       }
   
     return 0;      return 0;
   
   unlink_and_fail:
       if (bs->is_temporary) {
           unlink(filename);
       }
       return ret;
 }  }
   
 void bdrv_close(BlockDriverState *bs)  void bdrv_close(BlockDriverState *bs)
 {  {
     if (bs->drv) {      if (bs->drv) {
         if (bs->backing_hd)          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);          qemu_free(bs->opaque);
 #ifdef _WIN32  #ifdef _WIN32
Line 456  void bdrv_close(BlockDriverState *bs) Line 677  void bdrv_close(BlockDriverState *bs)
         bs->opaque = NULL;          bs->opaque = NULL;
         bs->drv = NULL;          bs->drv = NULL;
   
           if (bs->file != NULL) {
               bdrv_close(bs->file);
           }
   
         /* call the change callback */          /* call the change callback */
         bs->media_changed = 1;          bs->media_changed = 1;
         if (bs->change_cb)          if (bs->change_cb)
             bs->change_cb(bs->change_opaque);              bs->change_cb(bs->change_opaque, CHANGE_MEDIA);
     }      }
 }  }
   
 void bdrv_delete(BlockDriverState *bs)  void bdrv_close_all(void)
 {  {
     BlockDriverState **pbs;      BlockDriverState *bs;
   
     pbs = &bdrv_first;  
     while (*pbs != bs && *pbs != NULL)  
         pbs = &(*pbs)->next;  
     if (*pbs == bs)  
         *pbs = bs->next;  
   
     bdrv_close(bs);      QTAILQ_FOREACH(bs, &bdrv_states, list) {
     qemu_free(bs);          bdrv_close(bs);
       }
 }  }
   
 /* commit COW file into the raw image */  void bdrv_delete(BlockDriverState *bs)
 int bdrv_commit(BlockDriverState *bs)  
 {  {
     BlockDriver *drv = bs->drv;      assert(!bs->peer);
     int64_t i, total_sectors;  
     int n, j;  
     unsigned char sector[512];  
   
     if (!drv)  
         return -ENOMEDIUM;  
   
     if (bs->read_only) {      /* remove from list, if necessary */
         return -EACCES;      if (bs->device_name[0] != '\0') {
           QTAILQ_REMOVE(&bdrv_states, bs, list);
     }      }
   
     if (!bs->backing_hd) {      bdrv_close(bs);
         return -ENOTSUP;      if (bs->file != NULL) {
           bdrv_delete(bs->file);
     }      }
   
     total_sectors = bdrv_getlength(bs) >> SECTOR_BITS;      assert(bs != bs_snapshots);
     for (i = 0; i < total_sectors;) {      qemu_free(bs);
         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) {  int bdrv_attach(BlockDriverState *bs, DeviceState *qdev)
                     return -EIO;  {
                 }      if (bs->peer) {
                 i++;          return -EBUSY;
             }      }
         } else {      bs->peer = qdev;
             i += n;      return 0;
   }
   
   void bdrv_detach(BlockDriverState *bs, DeviceState *qdev)
   {
       assert(bs->peer == qdev);
       bs->peer = NULL;
   }
   
   DeviceState *bdrv_get_attached(BlockDriverState *bs)
   {
       return bs->peer;
   }
   
   /*
    * Run consistency checks on an image
    *
    * 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 occured. The results of the
    * check are stored in res.
    */
   int bdrv_check(BlockDriverState *bs, BdrvCheckResult *res)
   {
       if (bs->drv->bdrv_check == NULL) {
           return -ENOTSUP;
       }
   
       memset(res, 0, sizeof(*res));
       return bs->drv->bdrv_check(bs, res);
   }
   
   #define COMMIT_BUF_SECTORS 2048
   
   /* commit COW file into the raw image */
   int bdrv_commit(BlockDriverState *bs)
   {
       BlockDriver *drv = bs->drv;
       BlockDriver *backing_drv;
       int64_t sector, total_sectors;
       int n, ro, open_flags;
       int ret = 0, rw_ret = 0;
       uint8_t *buf;
       char filename[1024];
       BlockDriverState *bs_rw, *bs_ro;
   
       if (!drv)
           return -ENOMEDIUM;
       
       if (!bs->backing_hd) {
           return -ENOTSUP;
       }
   
       if (bs->backing_hd->keep_read_only) {
           return -EACCES;
       }
   
       backing_drv = bs->backing_hd->drv;
       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;
     }      }
   
     if (drv->bdrv_make_empty)      total_sectors = bdrv_getlength(bs) >> BDRV_SECTOR_BITS;
         return drv->bdrv_make_empty(bs);      buf = qemu_malloc(COMMIT_BUF_SECTORS * BDRV_SECTOR_SIZE);
   
     return 0;      for (sector = 0; sector < total_sectors; sector += n) {
           if (drv->bdrv_is_allocated(bs, sector, COMMIT_BUF_SECTORS, &n)) {
   
               if (bdrv_read(bs, sector, buf, n) != 0) {
                   ret = -EIO;
                   goto ro_cleanup;
               }
   
               if (bdrv_write(bs->backing_hd, sector, buf, n) != 0) {
                   ret = -EIO;
                   goto ro_cleanup;
               }
           }
       }
   
       if (drv->bdrv_make_empty) {
           ret = drv->bdrv_make_empty(bs);
           bdrv_flush(bs);
       }
   
       /*
        * 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:
       qemu_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 545  static int bdrv_check_byte_request(Block Line 914  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)
 {  {
     int64_t offset;      return bdrv_check_byte_request(bs, sector_num * BDRV_SECTOR_SIZE,
                                      nb_sectors * BDRV_SECTOR_SIZE);
     /* Deal with byte accesses */  
     if (sector_num < 0)  
         offset = -sector_num;  
     else  
         offset = sector_num * 512;  
   
     return bdrv_check_byte_request(bs, offset, nb_sectors * 512);  
 }  }
   
 /* return < 0 if error. See bdrv_write() for the return codes */  /* return < 0 if error. See bdrv_write() for the return codes */
Line 567  int bdrv_read(BlockDriverState *bs, int6 Line 929  int bdrv_read(BlockDriverState *bs, int6
     if (bdrv_check_request(bs, sector_num, nb_sectors))      if (bdrv_check_request(bs, sector_num, nb_sectors))
         return -EIO;          return -EIO;
   
     if (drv->bdrv_pread) {      return drv->bdrv_read(bs, sector_num, buf, nb_sectors);
         int ret, len;  }
         len = nb_sectors * 512;  
         ret = drv->bdrv_pread(bs, sector_num * 512, buf, len);  static void set_dirty_bitmap(BlockDriverState *bs, int64_t sector_num,
         if (ret < 0)                               int nb_sectors, int dirty)
             return ret;  {
         else if (ret != len)      int64_t start, end;
             return -EINVAL;      unsigned long val, idx, bit;
         else {  
             bs->rd_bytes += (unsigned) len;      start = sector_num / BDRV_SECTORS_PER_DIRTY_CHUNK;
             bs->rd_ops ++;      end = (sector_num + nb_sectors - 1) / BDRV_SECTORS_PER_DIRTY_CHUNK;
             return 0;  
         }      for (; start <= end; start++) {
     } else {          idx = start / (sizeof(unsigned long) * 8);
         return drv->bdrv_read(bs, sector_num, buf, nb_sectors);          bit = start % (sizeof(unsigned long) * 8);
           val = bs->dirty_bitmap[idx];
           if (dirty) {
               if (!(val & (1UL << bit))) {
                   bs->dirty_count++;
                   val |= 1UL << bit;
               }
           } else {
               if (val & (1UL << bit)) {
                   bs->dirty_count--;
                   val &= ~(1UL << bit);
               }
           }
           bs->dirty_bitmap[idx] = val;
     }      }
 }  }
   
Line 602  int bdrv_write(BlockDriverState *bs, int Line 977  int bdrv_write(BlockDriverState *bs, int
     if (bdrv_check_request(bs, sector_num, nb_sectors))      if (bdrv_check_request(bs, sector_num, nb_sectors))
         return -EIO;          return -EIO;
   
     if (drv->bdrv_pwrite) {      if (bs->dirty_bitmap) {
         int ret, len, count = 0;          set_dirty_bitmap(bs, sector_num, nb_sectors, 1);
         len = nb_sectors * 512;      }
         do {  
             ret = drv->bdrv_pwrite(bs, sector_num * 512, buf, len - count);      if (bs->wr_highest_sector < sector_num + nb_sectors - 1) {
             if (ret < 0) {          bs->wr_highest_sector = sector_num + nb_sectors - 1;
                 printf("bdrv_write ret=%d\n", ret);  
                 return ret;  
             }  
             count += ret;  
             buf += ret;  
         } while (count != len);  
         bs->wr_bytes += (unsigned) len;  
         bs->wr_ops ++;  
         return 0;  
     }      }
   
     return drv->bdrv_write(bs, sector_num, buf, nb_sectors);      return drv->bdrv_write(bs, sector_num, buf, nb_sectors);
 }  }
   
 static int bdrv_pread_em(BlockDriverState *bs, int64_t offset,  int bdrv_pread(BlockDriverState *bs, int64_t offset,
                          uint8_t *buf, int count1)                 void *buf, int count1)
 {  {
     uint8_t tmp_buf[SECTOR_SIZE];      uint8_t tmp_buf[BDRV_SECTOR_SIZE];
     int len, nb_sectors, count;      int len, nb_sectors, count;
     int64_t sector_num;      int64_t sector_num;
       int ret;
   
     count = count1;      count = count1;
     /* first read to align to sector start */      /* first read to align to sector start */
     len = (SECTOR_SIZE - offset) & (SECTOR_SIZE - 1);      len = (BDRV_SECTOR_SIZE - offset) & (BDRV_SECTOR_SIZE - 1);
     if (len > count)      if (len > count)
         len = count;          len = count;
     sector_num = offset >> SECTOR_BITS;      sector_num = offset >> BDRV_SECTOR_BITS;
     if (len > 0) {      if (len > 0) {
         if (bdrv_read(bs, sector_num, tmp_buf, 1) < 0)          if ((ret = bdrv_read(bs, sector_num, tmp_buf, 1)) < 0)
             return -EIO;              return ret;
         memcpy(buf, tmp_buf + (offset & (SECTOR_SIZE - 1)), len);          memcpy(buf, tmp_buf + (offset & (BDRV_SECTOR_SIZE - 1)), len);
         count -= len;          count -= len;
         if (count == 0)          if (count == 0)
             return count1;              return count1;
Line 646  static int bdrv_pread_em(BlockDriverStat Line 1014  static int bdrv_pread_em(BlockDriverStat
     }      }
   
     /* read the sectors "in place" */      /* read the sectors "in place" */
     nb_sectors = count >> SECTOR_BITS;      nb_sectors = count >> BDRV_SECTOR_BITS;
     if (nb_sectors > 0) {      if (nb_sectors > 0) {
         if (bdrv_read(bs, sector_num, buf, nb_sectors) < 0)          if ((ret = bdrv_read(bs, sector_num, buf, nb_sectors)) < 0)
             return -EIO;              return ret;
         sector_num += nb_sectors;          sector_num += nb_sectors;
         len = nb_sectors << SECTOR_BITS;          len = nb_sectors << BDRV_SECTOR_BITS;
         buf += len;          buf += len;
         count -= len;          count -= len;
     }      }
   
     /* add data from the last sector */      /* add data from the last sector */
     if (count > 0) {      if (count > 0) {
         if (bdrv_read(bs, sector_num, tmp_buf, 1) < 0)          if ((ret = bdrv_read(bs, sector_num, tmp_buf, 1)) < 0)
             return -EIO;              return ret;
         memcpy(buf, tmp_buf, count);          memcpy(buf, tmp_buf, count);
     }      }
     return count1;      return count1;
 }  }
   
 static int bdrv_pwrite_em(BlockDriverState *bs, int64_t offset,  int bdrv_pwrite(BlockDriverState *bs, int64_t offset,
                           const uint8_t *buf, int count1)                  const void *buf, int count1)
 {  {
     uint8_t tmp_buf[SECTOR_SIZE];      uint8_t tmp_buf[BDRV_SECTOR_SIZE];
     int len, nb_sectors, count;      int len, nb_sectors, count;
     int64_t sector_num;      int64_t sector_num;
       int ret;
   
     count = count1;      count = count1;
     /* first write to align to sector start */      /* first write to align to sector start */
     len = (SECTOR_SIZE - offset) & (SECTOR_SIZE - 1);      len = (BDRV_SECTOR_SIZE - offset) & (BDRV_SECTOR_SIZE - 1);
     if (len > count)      if (len > count)
         len = count;          len = count;
     sector_num = offset >> SECTOR_BITS;      sector_num = offset >> BDRV_SECTOR_BITS;
     if (len > 0) {      if (len > 0) {
         if (bdrv_read(bs, sector_num, tmp_buf, 1) < 0)          if ((ret = bdrv_read(bs, sector_num, tmp_buf, 1)) < 0)
             return -EIO;              return ret;
         memcpy(tmp_buf + (offset & (SECTOR_SIZE - 1)), buf, len);          memcpy(tmp_buf + (offset & (BDRV_SECTOR_SIZE - 1)), buf, len);
         if (bdrv_write(bs, sector_num, tmp_buf, 1) < 0)          if ((ret = bdrv_write(bs, sector_num, tmp_buf, 1)) < 0)
             return -EIO;              return ret;
         count -= len;          count -= len;
         if (count == 0)          if (count == 0)
             return count1;              return count1;
Line 692  static int bdrv_pwrite_em(BlockDriverSta Line 1061  static int bdrv_pwrite_em(BlockDriverSta
     }      }
   
     /* write the sectors "in place" */      /* write the sectors "in place" */
     nb_sectors = count >> SECTOR_BITS;      nb_sectors = count >> BDRV_SECTOR_BITS;
     if (nb_sectors > 0) {      if (nb_sectors > 0) {
         if (bdrv_write(bs, sector_num, buf, nb_sectors) < 0)          if ((ret = bdrv_write(bs, sector_num, buf, nb_sectors)) < 0)
             return -EIO;              return ret;
         sector_num += nb_sectors;          sector_num += nb_sectors;
         len = nb_sectors << SECTOR_BITS;          len = nb_sectors << BDRV_SECTOR_BITS;
         buf += len;          buf += len;
         count -= len;          count -= len;
     }      }
   
     /* add data from the last sector */      /* add data from the last sector */
     if (count > 0) {      if (count > 0) {
         if (bdrv_read(bs, sector_num, tmp_buf, 1) < 0)          if ((ret = bdrv_read(bs, sector_num, tmp_buf, 1)) < 0)
             return -EIO;              return ret;
         memcpy(tmp_buf, buf, count);          memcpy(tmp_buf, buf, count);
         if (bdrv_write(bs, sector_num, tmp_buf, 1) < 0)          if ((ret = bdrv_write(bs, sector_num, tmp_buf, 1)) < 0)
             return -EIO;              return ret;
     }      }
     return count1;      return count1;
 }  }
   
 /**  /*
  * Read with byte offsets (needed only for file protocols)   * Writes to the file and ensures that no writes are reordered across this
    * request (acts as a barrier)
    *
    * Returns 0 on success, -errno in error cases.
  */   */
 int bdrv_pread(BlockDriverState *bs, int64_t offset,  int bdrv_pwrite_sync(BlockDriverState *bs, int64_t offset,
                void *buf1, int count1)      const void *buf, int count)
 {  {
     BlockDriver *drv = bs->drv;      int ret;
   
     if (!drv)      ret = bdrv_pwrite(bs, offset, buf, count);
         return -ENOMEDIUM;      if (ret < 0) {
     if (bdrv_check_byte_request(bs, offset, count1))          return ret;
         return -EIO;      }
   
     if (!drv->bdrv_pread)      /* No flush needed for cache=writethrough, it uses O_DSYNC */
         return bdrv_pread_em(bs, offset, buf1, count1);      if ((bs->open_flags & BDRV_O_CACHE_MASK) != 0) {
     return drv->bdrv_pread(bs, offset, buf1, count1);          bdrv_flush(bs);
       }
   
       return 0;
 }  }
   
 /**  /*
  * Write with byte offsets (needed only for file protocols)   * Writes to the file and ensures that no writes are reordered across this
    * request (acts as a barrier)
    *
    * Returns 0 on success, -errno in error cases.
  */   */
 int bdrv_pwrite(BlockDriverState *bs, int64_t offset,  int bdrv_write_sync(BlockDriverState *bs, int64_t sector_num,
                 const void *buf1, int count1)      const uint8_t *buf, int nb_sectors)
 {  {
     BlockDriver *drv = bs->drv;      return bdrv_pwrite_sync(bs, BDRV_SECTOR_SIZE * sector_num,
           buf, BDRV_SECTOR_SIZE * nb_sectors);
     if (!drv)  
         return -ENOMEDIUM;  
     if (bdrv_check_byte_request(bs, offset, count1))  
         return -EIO;  
   
     if (!drv->bdrv_pwrite)  
         return bdrv_pwrite_em(bs, offset, buf1, count1);  
     return drv->bdrv_pwrite(bs, offset, buf1, count1);  
 }  }
   
 /**  /**
Line 755  int bdrv_pwrite(BlockDriverState *bs, in Line 1125  int bdrv_pwrite(BlockDriverState *bs, in
 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;
     return drv->bdrv_truncate(bs, offset);      if (bs->read_only)
           return -EACCES;
       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);
           if (bs->change_cb) {
               bs->change_cb(bs->change_opaque, CHANGE_SIZE);
           }
       }
       return ret;
 }  }
   
 /**  /**
Line 770  int64_t bdrv_getlength(BlockDriverState  Line 1152  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 */      /* Fixed size devices use the total_sectors value for speed instead of
         return bs->total_sectors * SECTOR_SIZE;         issuing a length query (like lseek) on each call.  Also, legacy block
          drivers don't provide a bdrv_getlength function and must use
          total_sectors. */
       if (!bs->growable || !drv->bdrv_getlength) {
           return bs->total_sectors * BDRV_SECTOR_SIZE;
     }      }
     return drv->bdrv_getlength(bs);      return drv->bdrv_getlength(bs);
 }  }
Line 785  void bdrv_get_geometry(BlockDriverState  Line 1171  void bdrv_get_geometry(BlockDriverState 
     if (length < 0)      if (length < 0)
         length = 0;          length = 0;
     else      else
         length = length >> SECTOR_BITS;          length = length >> BDRV_SECTOR_BITS;
     *nb_sectors_ptr = length;      *nb_sectors_ptr = length;
 }  }
   
Line 806  struct partition { Line 1192  struct partition {
 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 943  int bdrv_get_translation_hint(BlockDrive Line 1329  int bdrv_get_translation_hint(BlockDrive
     return bs->translation;      return bs->translation;
 }  }
   
   void bdrv_set_on_error(BlockDriverState *bs, BlockErrorAction on_read_error,
                          BlockErrorAction on_write_error)
   {
       bs->on_read_error = on_read_error;
       bs->on_write_error = on_write_error;
   }
   
   BlockErrorAction bdrv_get_on_error(BlockDriverState *bs, int is_read)
   {
       return is_read ? bs->on_read_error : bs->on_write_error;
   }
   
   void bdrv_set_removable(BlockDriverState *bs, int removable)
   {
       bs->removable = removable;
       if (removable && bs == bs_snapshots) {
           bs_snapshots = NULL;
       }
   }
   
 int bdrv_is_removable(BlockDriverState *bs)  int bdrv_is_removable(BlockDriverState *bs)
 {  {
     return bs->removable;      return bs->removable;
Line 958  int bdrv_is_sg(BlockDriverState *bs) Line 1364  int bdrv_is_sg(BlockDriverState *bs)
     return bs->sg;      return bs->sg;
 }  }
   
   int bdrv_enable_write_cache(BlockDriverState *bs)
   {
       return bs->enable_write_cache;
   }
   
 /* XXX: no longer used */  /* XXX: no longer used */
 void bdrv_set_change_cb(BlockDriverState *bs,  void bdrv_set_change_cb(BlockDriverState *bs,
                         void (*change_cb)(void *opaque), void *opaque)                          void (*change_cb)(void *opaque, int reason),
                           void *opaque)
 {  {
     bs->change_cb = change_cb;      bs->change_cb = change_cb;
     bs->change_opaque = opaque;      bs->change_opaque = opaque;
Line 992  int bdrv_set_key(BlockDriverState *bs, c Line 1404  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);
     bs->valid_key = (ret == 0);      if (ret < 0) {
           bs->valid_key = 0;
       } else if (!bs->valid_key) {
           bs->valid_key = 1;
           /* call the change callback now, we skipped it on open */
           bs->media_changed = 1;
           if (bs->change_cb)
               bs->change_cb(bs->change_opaque, CHANGE_MEDIA);
       }
     return ret;      return ret;
 }  }
   
Line 1013  void bdrv_iterate_format(void (*it)(void Line 1436  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 1022  BlockDriverState *bdrv_find(const char * Line 1445  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 1043  const char *bdrv_get_device_name(BlockDr Line 1475  const char *bdrv_get_device_name(BlockDr
     return bs->device_name;      return bs->device_name;
 }  }
   
 void bdrv_flush(BlockDriverState *bs)  int bdrv_flush(BlockDriverState *bs)
 {  {
     if (bs->drv->bdrv_flush)      if (bs->open_flags & BDRV_O_NO_FLUSH) {
         bs->drv->bdrv_flush(bs);          return 0;
     if (bs->backing_hd)      }
         bdrv_flush(bs->backing_hd);  
       if (bs->drv && bs->drv->bdrv_flush) {
           return bs->drv->bdrv_flush(bs);
       }
   
       /*
        * 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_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 (bs->drv && !bdrv_is_read_only(bs) &&
             (!bdrv_is_removable(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;
   }
   
   int bdrv_discard(BlockDriverState *bs, int64_t sector_num, int nb_sectors)
   {
       if (!bs->drv) {
           return -ENOMEDIUM;
       }
       if (!bs->drv->bdrv_discard) {
           return 0;
       }
       return bs->drv->bdrv_discard(bs, sector_num, nb_sectors);
 }  }
   
 /*  /*
Line 1088  int bdrv_is_allocated(BlockDriverState * Line 1559  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);
 }  }
   
 void bdrv_info(void)  void bdrv_mon_event(const BlockDriverState *bdrv,
                       BlockMonEventAction action, int is_read)
   {
       QObject *data;
       const char *action_str;
   
       switch (action) {
       case BDRV_ACTION_REPORT:
           action_str = "report";
           break;
       case BDRV_ACTION_IGNORE:
           action_str = "ignore";
           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);
   
       qobject_decref(data);
   }
   
   static void bdrv_print_dict(QObject *obj, void *opaque)
   {
       QDict *bs_dict;
       Monitor *mon = opaque;
   
       bs_dict = qobject_to_qdict(obj);
   
       monitor_printf(mon, "%s: type=%s removable=%d",
                           qdict_get_str(bs_dict, "device"),
                           qdict_get_str(bs_dict, "type"),
                           qdict_get_bool(bs_dict, "removable"));
   
       if (qdict_get_bool(bs_dict, "removable")) {
           monitor_printf(mon, " locked=%d", qdict_get_bool(bs_dict, "locked"));
       }
   
       if (qdict_haskey(bs_dict, "inserted")) {
           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);
   }
   
   void bdrv_info(Monitor *mon, QObject **ret_data)
 {  {
       QList *bs_list;
     BlockDriverState *bs;      BlockDriverState *bs;
   
     for (bs = bdrv_first; bs != NULL; bs = bs->next) {      bs_list = qlist_new();
         term_printf("%s:", bs->device_name);  
         term_printf(" type=");      QTAILQ_FOREACH(bs, &bdrv_states, list) {
           QObject *bs_obj;
           const char *type = "unknown";
   
         switch(bs->type) {          switch(bs->type) {
         case BDRV_TYPE_HD:          case BDRV_TYPE_HD:
             term_printf("hd");              type = "hd";
             break;              break;
         case BDRV_TYPE_CDROM:          case BDRV_TYPE_CDROM:
             term_printf("cdrom");              type = "cdrom";
             break;              break;
         case BDRV_TYPE_FLOPPY:          case BDRV_TYPE_FLOPPY:
             term_printf("floppy");              type = "floppy";
             break;              break;
         }          }
         term_printf(" removable=%d", bs->removable);  
         if (bs->removable) {          bs_obj = qobject_from_jsonf("{ 'device': %s, 'type': %s, "
             term_printf(" locked=%d", bs->locked);                                      "'removable': %i, 'locked': %i }",
         }                                      bs->device_name, type, bs->removable,
                                       bs->locked);
   
         if (bs->drv) {          if (bs->drv) {
             term_printf(" file=");              QObject *obj;
             term_print_filename(bs->filename);              QDict *bs_dict = qobject_to_qdict(bs_obj);
   
               obj = qobject_from_jsonf("{ 'file': %s, 'ro': %i, 'drv': %s, "
                                        "'encrypted': %i }",
                                        bs->filename, bs->read_only,
                                        bs->drv->format_name,
                                        bdrv_is_encrypted(bs));
             if (bs->backing_file[0] != '\0') {              if (bs->backing_file[0] != '\0') {
                 term_printf(" backing_file=");                  QDict *qdict = qobject_to_qdict(obj);
                 term_print_filename(bs->backing_file);                  qdict_put(qdict, "backing_file",
             }                            qstring_from_str(bs->backing_file));
             term_printf(" ro=%d", bs->read_only);              }
             term_printf(" drv=%s", bs->drv->format_name);  
             term_printf(" encrypted=%d", bdrv_is_encrypted(bs));              qdict_put_obj(bs_dict, "inserted", obj);
         } else {  
             term_printf(" [not inserted]");  
         }          }
         term_printf("\n");          qlist_append_obj(bs_list, bs_obj);
       }
   
       *ret_data = QOBJECT(bs_list);
   }
   
   static void bdrv_stats_iter(QObject *data, void *opaque)
   {
       QDict *qdict;
       Monitor *mon = opaque;
   
       qdict = qobject_to_qdict(data);
       monitor_printf(mon, "%s:", qdict_get_str(qdict, "device"));
   
       qdict = qobject_to_qdict(qdict_get(qdict, "stats"));
       monitor_printf(mon, " rd_bytes=%" PRId64
                           " wr_bytes=%" PRId64
                           " 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)
   {
       qlist_iter(qobject_to_qlist(data), bdrv_stats_iter, mon);
   }
   
   static QObject* bdrv_info_stats_bs(BlockDriverState *bs)
   {
       QObject *res;
       QDict *dict;
   
       res = qobject_from_jsonf("{ 'stats': {"
                                "'rd_bytes': %" PRId64 ","
                                "'wr_bytes': %" PRId64 ","
                                "'rd_operations': %" PRId64 ","
                                "'wr_operations': %" PRId64 ","
                                "'wr_highest_offset': %" PRId64
                                "} }",
                                bs->rd_bytes, bs->wr_bytes,
                                bs->rd_ops, bs->wr_ops,
                                bs->wr_highest_sector *
                                (uint64_t)BDRV_SECTOR_SIZE);
       dict  = qobject_to_qdict(res);
   
       if (*bs->device_name) {
           qdict_put(dict, "device", qstring_from_str(bs->device_name));
     }      }
   
       if (bs->file) {
           QObject *parent = bdrv_info_stats_bs(bs->file);
           qdict_put_obj(dict, "parent", parent);
       }
   
       return res;
 }  }
   
 /* The "info blockstats" command. */  void bdrv_info_stats(Monitor *mon, QObject **ret_data)
 void bdrv_info_stats (void)  
 {  {
       QObject *obj;
       QList *devices;
     BlockDriverState *bs;      BlockDriverState *bs;
   
     for (bs = bdrv_first; bs != NULL; bs = bs->next) {      devices = qlist_new();
         term_printf ("%s:"  
                      " rd_bytes=%" PRIu64      QTAILQ_FOREACH(bs, &bdrv_states, list) {
                      " wr_bytes=%" PRIu64          obj = bdrv_info_stats_bs(bs);
                      " rd_operations=%" PRIu64          qlist_append_obj(devices, obj);
                      " wr_operations=%" PRIu64  
                      "\n",  
                      bs->device_name,  
                      bs->rd_bytes, bs->wr_bytes,  
                      bs->rd_ops, bs->wr_ops);  
     }      }
   
       *ret_data = QOBJECT(devices);
 }  }
   
 const char *bdrv_get_encrypted_filename(BlockDriverState *bs)  const char *bdrv_get_encrypted_filename(BlockDriverState *bs)
Line 1158  const char *bdrv_get_encrypted_filename( Line 1764  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) {      if (!bs->backing_file) {
         pstrcpy(filename, filename_size, "");          pstrcpy(filename, filename_size, "");
     } else {      } else {
         pstrcpy(filename, filename_size, bs->backing_file);          pstrcpy(filename, filename_size, bs->backing_file);
Line 1175  int bdrv_write_compressed(BlockDriverSta Line 1781  int bdrv_write_compressed(BlockDriverSta
         return -ENOTSUP;          return -ENOTSUP;
     if (bdrv_check_request(bs, sector_num, nb_sectors))      if (bdrv_check_request(bs, sector_num, nb_sectors))
         return -EIO;          return -EIO;
   
       if (bs->dirty_bitmap) {
           set_dirty_bitmap(bs, sector_num, nb_sectors, 1);
       }
   
     return drv->bdrv_write_compressed(bs, sector_num, buf, nb_sectors);      return drv->bdrv_write_compressed(bs, sector_num, buf, nb_sectors);
 }  }
   
Line 1189  int bdrv_get_info(BlockDriverState *bs,  Line 1800  int bdrv_get_info(BlockDriverState *bs, 
     return drv->bdrv_get_info(bs, bdi);      return drv->bdrv_get_info(bs, bdi);
 }  }
   
 int bdrv_put_buffer(BlockDriverState *bs, const uint8_t *buf, int64_t pos, int size)  int bdrv_save_vmstate(BlockDriverState *bs, const uint8_t *buf,
                         int64_t pos, int size)
 {  {
     BlockDriver *drv = bs->drv;      BlockDriver *drv = bs->drv;
     if (!drv)      if (!drv)
         return -ENOMEDIUM;          return -ENOMEDIUM;
     if (!drv->bdrv_put_buffer)      if (drv->bdrv_save_vmstate)
         return -ENOTSUP;          return drv->bdrv_save_vmstate(bs, buf, pos, size);
     return drv->bdrv_put_buffer(bs, buf, pos, size);      if (bs->file)
           return bdrv_save_vmstate(bs->file, buf, pos, size);
       return -ENOTSUP;
 }  }
   
 int bdrv_get_buffer(BlockDriverState *bs, uint8_t *buf, int64_t pos, int size)  int bdrv_load_vmstate(BlockDriverState *bs, uint8_t *buf,
                         int64_t pos, int size)
 {  {
     BlockDriver *drv = bs->drv;      BlockDriver *drv = bs->drv;
     if (!drv)      if (!drv)
         return -ENOMEDIUM;          return -ENOMEDIUM;
     if (!drv->bdrv_get_buffer)      if (drv->bdrv_load_vmstate)
         return -ENOTSUP;          return drv->bdrv_load_vmstate(bs, buf, pos, size);
     return drv->bdrv_get_buffer(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 */
   
   int bdrv_can_snapshot(BlockDriverState *bs)
   {
       BlockDriver *drv = bs->drv;
       if (!drv || bdrv_is_removable(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;
 }  }
   
 /**************************************************************/  
 /* handling of snapshots */  
   
 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 1239  int bdrv_snapshot_delete(BlockDriverStat Line 1925  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 1250  int bdrv_snapshot_list(BlockDriverState  Line 1938  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 1332  char *bdrv_snapshot_dump(char *buf, int  Line 2038  char *bdrv_snapshot_dump(char *buf, int 
 /**************************************************************/  /**************************************************************/
 /* async I/Os */  /* async I/Os */
   
 typedef struct VectorTranslationAIOCB {  BlockDriverAIOCB *bdrv_aio_readv(BlockDriverState *bs, int64_t sector_num,
     BlockDriverAIOCB common;                                   QEMUIOVector *qiov, int nb_sectors,
     QEMUIOVector *iov;                                   BlockDriverCompletionFunc *cb, void *opaque)
     uint8_t *bounce;  
     int is_write;  
     BlockDriverAIOCB *aiocb;  
 } VectorTranslationAIOCB;  
   
 static void bdrv_aio_cancel_vector(BlockDriverAIOCB *_acb)  
 {  {
     VectorTranslationAIOCB *acb      BlockDriver *drv = bs->drv;
         = container_of(_acb, VectorTranslationAIOCB, common);      BlockDriverAIOCB *ret;
   
     bdrv_aio_cancel(acb->aiocb);      trace_bdrv_aio_readv(bs, sector_num, nb_sectors, opaque);
 }  
   
 static void bdrv_aio_rw_vector_cb(void *opaque, int ret)      if (!drv)
 {          return NULL;
     VectorTranslationAIOCB *s = (VectorTranslationAIOCB *)opaque;      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 (!s->is_write) {      if (ret) {
         qemu_iovec_from_buffer(s->iov, s->bounce, s->iov->size);          /* Update stats even though technically transfer has not happened. */
           bs->rd_bytes += (unsigned) nb_sectors * BDRV_SECTOR_SIZE;
           bs->rd_ops ++;
     }      }
     qemu_vfree(s->bounce);  
     s->common.cb(s->common.opaque, ret);      return ret;
     qemu_aio_release(s);  
 }  }
   
 static BlockDriverAIOCB *bdrv_aio_rw_vector(BlockDriverState *bs,  typedef struct BlockCompleteData {
                                             int64_t sector_num,      BlockDriverCompletionFunc *cb;
                                             QEMUIOVector *iov,      void *opaque;
                                             int nb_sectors,      BlockDriverState *bs;
                                             BlockDriverCompletionFunc *cb,      int64_t sector_num;
                                             void *opaque,      int nb_sectors;
                                             int is_write)  } BlockCompleteData;
   
   static void block_complete_cb(void *opaque, int ret)
 {  {
     VectorTranslationAIOCB *s = qemu_aio_get_pool(&vectored_aio_pool, bs,      BlockCompleteData *b = opaque;
                                                   cb, opaque);  
   
     s->iov = iov;      if (b->bs->dirty_bitmap) {
     s->bounce = qemu_memalign(512, nb_sectors * 512);          set_dirty_bitmap(b->bs, b->sector_num, b->nb_sectors, 1);
     s->is_write = is_write;  
     if (is_write) {  
         qemu_iovec_to_buffer(s->iov, s->bounce);  
         s->aiocb = bdrv_aio_write(bs, sector_num, s->bounce, nb_sectors,  
                                   bdrv_aio_rw_vector_cb, s);  
     } else {  
         s->aiocb = bdrv_aio_read(bs, sector_num, s->bounce, nb_sectors,  
                                  bdrv_aio_rw_vector_cb, s);  
     }  
     if (!s->aiocb) {  
         qemu_vfree(s->bounce);  
         qemu_aio_release(s);  
         return NULL;  
     }      }
     return &s->common;      b->cb(b->opaque, ret);
       qemu_free(b);
 }  }
   
 BlockDriverAIOCB *bdrv_aio_readv(BlockDriverState *bs, int64_t sector_num,  static BlockCompleteData *blk_dirty_cb_alloc(BlockDriverState *bs,
                                  QEMUIOVector *iov, int nb_sectors,                                               int64_t sector_num,
                                  BlockDriverCompletionFunc *cb, void *opaque)                                               int nb_sectors,
                                                BlockDriverCompletionFunc *cb,
                                                void *opaque)
 {  {
     if (bdrv_check_request(bs, sector_num, nb_sectors))      BlockCompleteData *blkdata = qemu_mallocz(sizeof(BlockCompleteData));
         return NULL;  
       blkdata->bs = bs;
       blkdata->cb = cb;
       blkdata->opaque = opaque;
       blkdata->sector_num = sector_num;
       blkdata->nb_sectors = nb_sectors;
   
     return bdrv_aio_rw_vector(bs, sector_num, iov, nb_sectors,      return blkdata;
                               cb, opaque, 0);  
 }  }
   
 BlockDriverAIOCB *bdrv_aio_writev(BlockDriverState *bs, int64_t sector_num,  BlockDriverAIOCB *bdrv_aio_writev(BlockDriverState *bs, int64_t sector_num,
                                   QEMUIOVector *iov, int nb_sectors,                                    QEMUIOVector *qiov, int nb_sectors,
                                   BlockDriverCompletionFunc *cb, void *opaque)                                    BlockDriverCompletionFunc *cb, void *opaque)
 {  {
     if (bdrv_check_request(bs, sector_num, nb_sectors))  
         return NULL;  
   
     return bdrv_aio_rw_vector(bs, sector_num, iov, nb_sectors,  
                               cb, opaque, 1);  
 }  
   
 BlockDriverAIOCB *bdrv_aio_read(BlockDriverState *bs, int64_t sector_num,  
                                 uint8_t *buf, int nb_sectors,  
                                 BlockDriverCompletionFunc *cb, void *opaque)  
 {  
     BlockDriver *drv = bs->drv;      BlockDriver *drv = bs->drv;
     BlockDriverAIOCB *ret;      BlockDriverAIOCB *ret;
       BlockCompleteData *blk_cb_data;
   
       trace_bdrv_aio_writev(bs, sector_num, nb_sectors, opaque);
   
     if (!drv)      if (!drv)
         return NULL;          return NULL;
       if (bs->read_only)
           return NULL;
     if (bdrv_check_request(bs, sector_num, nb_sectors))      if (bdrv_check_request(bs, sector_num, nb_sectors))
         return NULL;          return NULL;
   
     ret = drv->bdrv_aio_read(bs, sector_num, buf, nb_sectors, cb, opaque);      if (bs->dirty_bitmap) {
           blk_cb_data = blk_dirty_cb_alloc(bs, sector_num, nb_sectors, cb,
                                            opaque);
           cb = &block_complete_cb;
           opaque = blk_cb_data;
       }
   
       ret = drv->bdrv_aio_writev(bs, sector_num, qiov, nb_sectors,
                                  cb, opaque);
   
     if (ret) {      if (ret) {
         /* Update stats even though technically transfer has not happened. */          /* Update stats even though technically transfer has not happened. */
         bs->rd_bytes += (unsigned) nb_sectors * SECTOR_SIZE;          bs->wr_bytes += (unsigned) nb_sectors * BDRV_SECTOR_SIZE;
         bs->rd_ops ++;          bs->wr_ops ++;
           if (bs->wr_highest_sector < sector_num + nb_sectors - 1) {
               bs->wr_highest_sector = sector_num + nb_sectors - 1;
           }
     }      }
   
     return ret;      return ret;
 }  }
   
 BlockDriverAIOCB *bdrv_aio_write(BlockDriverState *bs, int64_t sector_num,  
                                  const uint8_t *buf, int nb_sectors,  typedef struct MultiwriteCB {
                                  BlockDriverCompletionFunc *cb, void *opaque)      int error;
       int num_requests;
       int num_callbacks;
       struct {
           BlockDriverCompletionFunc *cb;
           void *opaque;
           QEMUIOVector *free_qiov;
           void *free_buf;
       } callbacks[];
   } MultiwriteCB;
   
   static void multiwrite_user_cb(MultiwriteCB *mcb)
 {  {
     BlockDriver *drv = bs->drv;      int i;
     BlockDriverAIOCB *ret;  
   
     if (!drv)      for (i = 0; i < mcb->num_callbacks; i++) {
         return NULL;          mcb->callbacks[i].cb(mcb->callbacks[i].opaque, mcb->error);
     if (bs->read_only)          if (mcb->callbacks[i].free_qiov) {
         return NULL;              qemu_iovec_destroy(mcb->callbacks[i].free_qiov);
     if (bdrv_check_request(bs, sector_num, nb_sectors))          }
         return NULL;          qemu_free(mcb->callbacks[i].free_qiov);
           qemu_vfree(mcb->callbacks[i].free_buf);
       }
   }
   
     ret = drv->bdrv_aio_write(bs, sector_num, buf, nb_sectors, cb, opaque);  static void multiwrite_cb(void *opaque, int ret)
   {
       MultiwriteCB *mcb = opaque;
   
     if (ret) {      trace_multiwrite_cb(mcb, ret);
         /* Update stats even though technically transfer has not happened. */  
         bs->wr_bytes += (unsigned) nb_sectors * SECTOR_SIZE;      if (ret < 0 && !mcb->error) {
         bs->wr_ops ++;          mcb->error = ret;
     }      }
   
     return ret;      mcb->num_requests--;
       if (mcb->num_requests == 0) {
           multiwrite_user_cb(mcb);
           qemu_free(mcb);
       }
   }
   
   static int multiwrite_req_compare(const void *a, const void *b)
   {
       const BlockRequest *req1 = a, *req2 = b;
   
       /*
        * Note that we can't simply subtract req2->sector from req1->sector
        * here as that could overflow the return value.
        */
       if (req1->sector > req2->sector) {
           return 1;
       } else if (req1->sector < req2->sector) {
           return -1;
       } else {
           return 0;
       }
   }
   
   /*
    * Takes a bunch of requests and tries to merge them. Returns the number of
    * requests that remain after merging.
    */
   static int multiwrite_merge(BlockDriverState *bs, BlockRequest *reqs,
       int num_reqs, MultiwriteCB *mcb)
   {
       int i, outidx;
   
       // Sort requests by start sector
       qsort(reqs, num_reqs, sizeof(*reqs), &multiwrite_req_compare);
   
       // Check if adjacent requests touch the same clusters. If so, combine them,
       // filling up gaps with zero sectors.
       outidx = 0;
       for (i = 1; i < num_reqs; i++) {
           int merge = 0;
           int64_t oldreq_last = reqs[outidx].sector + reqs[outidx].nb_sectors;
   
           // This handles the cases that are valid for all block drivers, namely
           // exactly sequential writes and overlapping writes.
           if (reqs[i].sector <= oldreq_last) {
               merge = 1;
           }
   
           // The block driver may decide that it makes sense to combine requests
           // even if there is a gap of some sectors between them. In this case,
           // the gap is filled with zeros (therefore only applicable for yet
           // unused space in format like qcow2).
           if (!merge && bs->drv->bdrv_merge_requests) {
               merge = bs->drv->bdrv_merge_requests(bs, &reqs[outidx], &reqs[i]);
           }
   
           if (reqs[outidx].qiov->niov + reqs[i].qiov->niov + 1 > IOV_MAX) {
               merge = 0;
           }
   
           if (merge) {
               size_t size;
               QEMUIOVector *qiov = qemu_mallocz(sizeof(*qiov));
               qemu_iovec_init(qiov,
                   reqs[outidx].qiov->niov + reqs[i].qiov->niov + 1);
   
               // Add the first request to the merged one. If the requests are
               // overlapping, drop the last sectors of the first request.
               size = (reqs[i].sector - reqs[outidx].sector) << 9;
               qemu_iovec_concat(qiov, reqs[outidx].qiov, size);
   
               // We might need to add some zeros between the two requests
               if (reqs[i].sector > oldreq_last) {
                   size_t zero_bytes = (reqs[i].sector - oldreq_last) << 9;
                   uint8_t *buf = qemu_blockalign(bs, zero_bytes);
                   memset(buf, 0, zero_bytes);
                   qemu_iovec_add(qiov, buf, zero_bytes);
                   mcb->callbacks[i].free_buf = buf;
               }
   
               // Add the second request
               qemu_iovec_concat(qiov, reqs[i].qiov, reqs[i].qiov->size);
   
               reqs[outidx].nb_sectors = qiov->size >> 9;
               reqs[outidx].qiov = qiov;
   
               mcb->callbacks[i].free_qiov = reqs[outidx].qiov;
           } else {
               outidx++;
               reqs[outidx].sector     = reqs[i].sector;
               reqs[outidx].nb_sectors = reqs[i].nb_sectors;
               reqs[outidx].qiov       = reqs[i].qiov;
           }
       }
   
       return outidx + 1;
   }
   
   /*
    * Submit multiple AIO write requests at once.
    *
    * On success, the function returns 0 and all requests in the reqs array have
    * been submitted. In error case this function returns -1, and any of the
    * requests may or may not be submitted yet. In particular, this means that the
    * callback will be called for some of the requests, for others it won't. The
    * caller must check the error field of the BlockRequest to wait for the right
    * callbacks (if error != 0, no callback will be called).
    *
    * The implementation may modify the contents of the reqs array, e.g. to merge
    * requests. However, the fields opaque and error are left unmodified as they
    * are used to signal failure for a single request to the caller.
    */
   int bdrv_aio_multiwrite(BlockDriverState *bs, BlockRequest *reqs, int num_reqs)
   {
       BlockDriverAIOCB *acb;
       MultiwriteCB *mcb;
       int i;
   
       if (num_reqs == 0) {
           return 0;
       }
   
       // Create MultiwriteCB structure
       mcb = qemu_mallocz(sizeof(*mcb) + num_reqs * sizeof(*mcb->callbacks));
       mcb->num_requests = 0;
       mcb->num_callbacks = num_reqs;
   
       for (i = 0; i < num_reqs; i++) {
           mcb->callbacks[i].cb = reqs[i].cb;
           mcb->callbacks[i].opaque = reqs[i].opaque;
       }
   
       // Check for mergable requests
       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
        * successfully, fail all requests that are not yet submitted (we must
        * return failure for all requests anyway)
        *
        * num_requests cannot be set to the right value immediately: If
        * bdrv_aio_writev fails for some request, num_requests would be too high
        * and therefore multiwrite_cb() would never recognize the multiwrite
        * request as completed. We also cannot use the loop variable i to set it
        * when the first request fails because the callback may already have been
        * called for previously submitted requests. Thus, num_requests must be
        * incremented for each request that is submitted.
        *
        * The problem that callbacks may be called early also means that we need
        * to take care that num_requests doesn't become 0 before all requests are
        * submitted - multiwrite_cb() would consider the multiwrite request
        * completed. A dummy request that is "completed" by a manual call to
        * multiwrite_cb() takes care of this.
        */
       mcb->num_requests = 1;
   
       // Run the aio requests
       for (i = 0; i < num_reqs; i++) {
           mcb->num_requests++;
           acb = bdrv_aio_writev(bs, reqs[i].sector, reqs[i].qiov,
               reqs[i].nb_sectors, multiwrite_cb, mcb);
   
           if (acb == NULL) {
               // We can only fail the whole thing if no request has been
               // submitted yet. Otherwise we'll wait for the submitted AIOs to
               // complete and report the error in the callback.
               if (i == 0) {
                   trace_bdrv_aio_multiwrite_earlyfail(mcb);
                   goto fail;
               } else {
                   trace_bdrv_aio_multiwrite_latefail(mcb, i);
                   multiwrite_cb(mcb, -EIO);
                   break;
               }
           }
       }
   
       /* Complete the dummy request */
       multiwrite_cb(mcb, 0);
   
       return 0;
   
   fail:
       for (i = 0; i < mcb->num_callbacks; i++) {
           reqs[i].error = -EIO;
       }
       qemu_free(mcb);
       return -1;
   }
   
   BlockDriverAIOCB *bdrv_aio_flush(BlockDriverState *bs,
           BlockDriverCompletionFunc *cb, void *opaque)
   {
       BlockDriver *drv = bs->drv;
   
       if (bs->open_flags & BDRV_O_NO_FLUSH) {
           return bdrv_aio_noop_em(bs, cb, opaque);
       }
   
       if (!drv)
           return NULL;
       return drv->bdrv_aio_flush(bs, cb, opaque);
 }  }
   
 void bdrv_aio_cancel(BlockDriverAIOCB *acb)  void bdrv_aio_cancel(BlockDriverAIOCB *acb)
Line 1470  void bdrv_aio_cancel(BlockDriverAIOCB *a Line 2392  void bdrv_aio_cancel(BlockDriverAIOCB *a
 /**************************************************************/  /**************************************************************/
 /* async block device emulation */  /* async block device emulation */
   
   typedef struct BlockDriverAIOCBSync {
       BlockDriverAIOCB common;
       QEMUBH *bh;
       int ret;
       /* vector translation state */
       QEMUIOVector *qiov;
       uint8_t *bounce;
       int is_write;
   } BlockDriverAIOCBSync;
   
   static void bdrv_aio_cancel_em(BlockDriverAIOCB *blockacb)
   {
       BlockDriverAIOCBSync *acb =
           container_of(blockacb, BlockDriverAIOCBSync, common);
       qemu_bh_delete(acb->bh);
       acb->bh = NULL;
       qemu_aio_release(acb);
   }
   
   static AIOPool bdrv_em_aio_pool = {
       .aiocb_size         = sizeof(BlockDriverAIOCBSync),
       .cancel             = bdrv_aio_cancel_em,
   };
   
 static void bdrv_aio_bh_cb(void *opaque)  static void bdrv_aio_bh_cb(void *opaque)
 {  {
     BlockDriverAIOCBSync *acb = opaque;      BlockDriverAIOCBSync *acb = opaque;
   
       if (!acb->is_write)
           qemu_iovec_from_buffer(acb->qiov, acb->bounce, acb->qiov->size);
       qemu_vfree(acb->bounce);
     acb->common.cb(acb->common.opaque, acb->ret);      acb->common.cb(acb->common.opaque, acb->ret);
       qemu_bh_delete(acb->bh);
       acb->bh = NULL;
     qemu_aio_release(acb);      qemu_aio_release(acb);
 }  }
   
 static BlockDriverAIOCB *bdrv_aio_read_em(BlockDriverState *bs,  static BlockDriverAIOCB *bdrv_aio_rw_vector(BlockDriverState *bs,
         int64_t sector_num, uint8_t *buf, int nb_sectors,                                              int64_t sector_num,
         BlockDriverCompletionFunc *cb, void *opaque)                                              QEMUIOVector *qiov,
                                               int nb_sectors,
                                               BlockDriverCompletionFunc *cb,
                                               void *opaque,
                                               int is_write)
   
 {  {
     BlockDriverAIOCBSync *acb;      BlockDriverAIOCBSync *acb;
     int ret;  
   
     acb = qemu_aio_get(bs, cb, opaque);      acb = qemu_aio_get(&bdrv_em_aio_pool, bs, cb, opaque);
       acb->is_write = is_write;
       acb->qiov = qiov;
       acb->bounce = qemu_blockalign(bs, qiov->size);
   
     if (!acb->bh)      if (!acb->bh)
         acb->bh = qemu_bh_new(bdrv_aio_bh_cb, acb);          acb->bh = qemu_bh_new(bdrv_aio_bh_cb, acb);
     ret = bdrv_read(bs, sector_num, buf, nb_sectors);  
     acb->ret = ret;      if (is_write) {
           qemu_iovec_to_buffer(acb->qiov, acb->bounce);
           acb->ret = bdrv_write(bs, sector_num, acb->bounce, nb_sectors);
       } else {
           acb->ret = bdrv_read(bs, sector_num, acb->bounce, nb_sectors);
       }
   
     qemu_bh_schedule(acb->bh);      qemu_bh_schedule(acb->bh);
   
     return &acb->common;      return &acb->common;
 }  }
   
 static BlockDriverAIOCB *bdrv_aio_write_em(BlockDriverState *bs,  static BlockDriverAIOCB *bdrv_aio_readv_em(BlockDriverState *bs,
         int64_t sector_num, const uint8_t *buf, int nb_sectors,          int64_t sector_num, QEMUIOVector *qiov, int nb_sectors,
           BlockDriverCompletionFunc *cb, void *opaque)
   {
       return bdrv_aio_rw_vector(bs, sector_num, qiov, nb_sectors, cb, opaque, 0);
   }
   
   static BlockDriverAIOCB *bdrv_aio_writev_em(BlockDriverState *bs,
           int64_t sector_num, QEMUIOVector *qiov, int nb_sectors,
           BlockDriverCompletionFunc *cb, void *opaque)
   {
       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)          BlockDriverCompletionFunc *cb, void *opaque)
 {  {
     BlockDriverAIOCBSync *acb;      BlockDriverAIOCBSync *acb;
     int ret;  
   
     acb = qemu_aio_get(bs, cb, opaque);      acb = qemu_aio_get(&bdrv_em_aio_pool, bs, cb, opaque);
       acb->is_write = 1; /* don't bounce in the completion hadler */
       acb->qiov = NULL;
       acb->bounce = NULL;
       acb->ret = 0;
   
     if (!acb->bh)      if (!acb->bh)
         acb->bh = qemu_bh_new(bdrv_aio_bh_cb, acb);          acb->bh = qemu_bh_new(bdrv_aio_bh_cb, acb);
     ret = bdrv_write(bs, sector_num, buf, nb_sectors);  
     acb->ret = ret;      bdrv_flush(bs);
     qemu_bh_schedule(acb->bh);      qemu_bh_schedule(acb->bh);
     return &acb->common;      return &acb->common;
 }  }
   
 static void bdrv_aio_cancel_em(BlockDriverAIOCB *blockacb)  static BlockDriverAIOCB *bdrv_aio_noop_em(BlockDriverState *bs,
           BlockDriverCompletionFunc *cb, void *opaque)
 {  {
     BlockDriverAIOCBSync *acb = (BlockDriverAIOCBSync *)blockacb;      BlockDriverAIOCBSync *acb;
     qemu_bh_cancel(acb->bh);  
     qemu_aio_release(acb);      acb = qemu_aio_get(&bdrv_em_aio_pool, bs, cb, opaque);
       acb->is_write = 1; /* don't bounce in the completion handler */
       acb->qiov = NULL;
       acb->bounce = NULL;
       acb->ret = 0;
   
       if (!acb->bh) {
           acb->bh = qemu_bh_new(bdrv_aio_bh_cb, acb);
       }
   
       qemu_bh_schedule(acb->bh);
       return &acb->common;
 }  }
   
 /**************************************************************/  /**************************************************************/
Line 1531  static int bdrv_read_em(BlockDriverState Line 2527  static int bdrv_read_em(BlockDriverState
 {  {
     int async_ret;      int async_ret;
     BlockDriverAIOCB *acb;      BlockDriverAIOCB *acb;
       struct iovec iov;
       QEMUIOVector qiov;
   
       async_context_push();
   
     async_ret = NOT_DONE;      async_ret = NOT_DONE;
     acb = bdrv_aio_read(bs, sector_num, buf, nb_sectors,      iov.iov_base = (void *)buf;
                         bdrv_rw_em_cb, &async_ret);      iov.iov_len = nb_sectors * BDRV_SECTOR_SIZE;
     if (acb == NULL)      qemu_iovec_init_external(&qiov, &iov, 1);
         return -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) {      while (async_ret == NOT_DONE) {
         qemu_aio_wait();          qemu_aio_wait();
     }      }
   
   
   fail:
       async_context_pop();
     return async_ret;      return async_ret;
 }  }
   
Line 1550  static int bdrv_write_em(BlockDriverStat Line 2558  static int bdrv_write_em(BlockDriverStat
 {  {
     int async_ret;      int async_ret;
     BlockDriverAIOCB *acb;      BlockDriverAIOCB *acb;
       struct iovec iov;
       QEMUIOVector qiov;
   
       async_context_push();
   
     async_ret = NOT_DONE;      async_ret = NOT_DONE;
     acb = bdrv_aio_write(bs, sector_num, buf, nb_sectors,      iov.iov_base = (void *)buf;
                          bdrv_rw_em_cb, &async_ret);      iov.iov_len = nb_sectors * BDRV_SECTOR_SIZE;
     if (acb == NULL)      qemu_iovec_init_external(&qiov, &iov, 1);
         return -1;      acb = bdrv_aio_writev(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) {      while (async_ret == NOT_DONE) {
         qemu_aio_wait();          qemu_aio_wait();
     }      }
   
   fail:
       async_context_pop();
     return async_ret;      return async_ret;
 }  }
   
 void bdrv_init(void)  void bdrv_init(void)
 {  {
     aio_pool_init(&vectored_aio_pool, sizeof(VectorTranslationAIOCB),      module_call_init(MODULE_INIT_BLOCK);
                   bdrv_aio_cancel_vector);  }
   
     bdrv_register(&bdrv_raw);  void bdrv_init_with_whitelist(void)
     bdrv_register(&bdrv_host_device);  {
 #ifndef _WIN32      use_bdrv_whitelist = 1;
     bdrv_register(&bdrv_cow);      bdrv_init();
 #endif  
     bdrv_register(&bdrv_qcow);  
     bdrv_register(&bdrv_vmdk);  
     bdrv_register(&bdrv_cloop);  
     bdrv_register(&bdrv_dmg);  
     bdrv_register(&bdrv_bochs);  
     bdrv_register(&bdrv_vpc);  
     bdrv_register(&bdrv_vvfat);  
     bdrv_register(&bdrv_qcow2);  
     bdrv_register(&bdrv_parallels);  
     bdrv_register(&bdrv_nbd);  
 }  
   
 void aio_pool_init(AIOPool *pool, int aiocb_size,  
                    void (*cancel)(BlockDriverAIOCB *acb))  
 {  
     pool->aiocb_size = aiocb_size;  
     pool->cancel = cancel;  
     pool->free_aiocb = NULL;  
 }  }
   
 void *qemu_aio_get_pool(AIOPool *pool, BlockDriverState *bs,  void *qemu_aio_get(AIOPool *pool, BlockDriverState *bs,
                         BlockDriverCompletionFunc *cb, void *opaque)                     BlockDriverCompletionFunc *cb, void *opaque)
 {  {
     BlockDriverAIOCB *acb;      BlockDriverAIOCB *acb;
   
Line 1610  void *qemu_aio_get_pool(AIOPool *pool, B Line 2611  void *qemu_aio_get_pool(AIOPool *pool, B
     return acb;      return acb;
 }  }
   
 void *qemu_aio_get(BlockDriverState *bs, BlockDriverCompletionFunc *cb,  
                    void *opaque)  
 {  
     return qemu_aio_get_pool(&bs->drv->aio_pool, bs, cb, opaque);  
 }  
   
 void qemu_aio_release(void *p)  void qemu_aio_release(void *p)
 {  {
     BlockDriverAIOCB *acb = (BlockDriverAIOCB *)p;      BlockDriverAIOCB *acb = (BlockDriverAIOCB *)p;
Line 1637  int bdrv_is_inserted(BlockDriverState *b Line 2632  int bdrv_is_inserted(BlockDriverState *b
     if (!drv)      if (!drv)
         return 0;          return 0;
     if (!drv->bdrv_is_inserted)      if (!drv->bdrv_is_inserted)
         return 1;          return !bs->tray_open;
     ret = drv->bdrv_is_inserted(bs);      ret = drv->bdrv_is_inserted(bs);
     return ret;      return ret;
 }  }
Line 1664  int bdrv_media_changed(BlockDriverState  Line 2659  int bdrv_media_changed(BlockDriverState 
 /**  /**
  * If eject_flag is TRUE, eject the media. Otherwise, close the tray   * If eject_flag is TRUE, eject the media. Otherwise, close the tray
  */   */
 void bdrv_eject(BlockDriverState *bs, int eject_flag)  int bdrv_eject(BlockDriverState *bs, int eject_flag)
 {  {
     BlockDriver *drv = bs->drv;      BlockDriver *drv = bs->drv;
     int ret;      int ret;
   
       if (bs->locked) {
           return -EBUSY;
       }
   
     if (!drv || !drv->bdrv_eject) {      if (!drv || !drv->bdrv_eject) {
         ret = -ENOTSUP;          ret = -ENOTSUP;
     } else {      } else {
         ret = drv->bdrv_eject(bs, eject_flag);          ret = drv->bdrv_eject(bs, eject_flag);
     }      }
     if (ret == -ENOTSUP) {      if (ret == -ENOTSUP) {
         if (eject_flag)          ret = 0;
             bdrv_close(bs);      }
       if (ret >= 0) {
           bs->tray_open = eject_flag;
     }      }
   
       return ret;
 }  }
   
 int bdrv_is_locked(BlockDriverState *bs)  int bdrv_is_locked(BlockDriverState *bs)
Line 1709  int bdrv_ioctl(BlockDriverState *bs, uns Line 2712  int bdrv_ioctl(BlockDriverState *bs, uns
         return drv->bdrv_ioctl(bs, req, buf);          return drv->bdrv_ioctl(bs, req, buf);
     return -ENOTSUP;      return -ENOTSUP;
 }  }
   
   BlockDriverAIOCB *bdrv_aio_ioctl(BlockDriverState *bs,
           unsigned long int req, void *buf,
           BlockDriverCompletionFunc *cb, void *opaque)
   {
       BlockDriver *drv = bs->drv;
   
       if (drv && drv->bdrv_aio_ioctl)
           return drv->bdrv_aio_ioctl(bs, req, buf, cb, opaque);
       return NULL;
   }
   
   
   
   void *qemu_blockalign(BlockDriverState *bs, size_t size)
   {
       return qemu_memalign((bs && bs->buffer_alignment) ? bs->buffer_alignment : 512, size);
   }
   
   void bdrv_set_dirty_tracking(BlockDriverState *bs, int enable)
   {
       int64_t bitmap_size;
   
       bs->dirty_count = 0;
       if (enable) {
           if (!bs->dirty_bitmap) {
               bitmap_size = (bdrv_getlength(bs) >> BDRV_SECTOR_BITS) +
                       BDRV_SECTORS_PER_DIRTY_CHUNK * 8 - 1;
               bitmap_size /= BDRV_SECTORS_PER_DIRTY_CHUNK * 8;
   
               bs->dirty_bitmap = qemu_mallocz(bitmap_size);
           }
       } else {
           if (bs->dirty_bitmap) {
               qemu_free(bs->dirty_bitmap);
               bs->dirty_bitmap = NULL;
           }
       }
   }
   
   int bdrv_get_dirty(BlockDriverState *bs, int64_t sector)
   {
       int64_t chunk = sector / (int64_t)BDRV_SECTORS_PER_DIRTY_CHUNK;
   
       if (bs->dirty_bitmap &&
           (sector << BDRV_SECTOR_BITS) < bdrv_getlength(bs)) {
           return !!(bs->dirty_bitmap[chunk / (sizeof(unsigned long) * 8)] &
               (1UL << (chunk % (sizeof(unsigned long) * 8))));
       } else {
           return 0;
       }
   }
   
   void bdrv_reset_dirty(BlockDriverState *bs, int64_t cur_sector,
                         int nr_sectors)
   {
       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;
   }
   
   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;
       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
       if (get_option_parameter(param, BLOCK_OPT_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.1.1.1.12  
changed lines
  Added in v.1.1.1.19


unix.superglobalmegacorp.com