async file I/O API

git-svn-id: svn://svn.savannah.nongnu.org/qemu/trunk@2075 c046a42c-6fe2-441c-8c8c-71466251a162

Changelog

+version 0.8.3:
+
+  - Support for relative paths in backing files for disk images
+  - Async file I/O API
+
 version 0.8.2:
 
   - ACPI support

Makefile

@@ -25,14 +25,22 @@ else
 DOCS=
 endif
 
+ifndef CONFIG_DARWIN
+ifndef CONFIG_WIN32
+ifndef CONFIG_SOLARIS
+LIBS+=-lrt
+endif
+endif
+endif
+
 all: $(TOOLS) $(DOCS) recurse-all
 
 subdir-%: dyngen$(EXESUF)
 	$(MAKE) -C $(subst subdir-,,$@) all
 
 recurse-all: $(patsubst %,subdir-%, $(TARGET_DIRS))
-        
-qemu-img$(EXESUF): qemu-img.c block.c block-cow.c block-qcow.c aes.c block-vmdk.c block-cloop.c block-dmg.c block-bochs.c block-vpc.c block-vvfat.c
+
+qemu-img$(EXESUF): qemu-img.c block.c block-raw.c block-cow.c block-qcow.c aes.c block-vmdk.c block-cloop.c block-dmg.c block-bochs.c block-vpc.c block-vvfat.c
 	$(CC) -DQEMU_TOOL $(CFLAGS) $(LDFLAGS) $(DEFINES) -o $@ $^ -lz $(LIBS)
 
 dyngen$(EXESUF): dyngen.c

Makefile.target

@@ -289,7 +289,8 @@ ifeq ($(ARCH),alpha)
 endif
 
 # must use static linking to avoid leaving stuff in virtual address space
-VL_OBJS=vl.o osdep.o block.o readline.o monitor.o pci.o console.o loader.o
+VL_OBJS=vl.o osdep.o readline.o monitor.o pci.o console.o loader.o
+VL_OBJS+=block.o block-raw.o
 VL_OBJS+=block-cow.o block-qcow.o aes.o block-vmdk.o block-cloop.o block-dmg.o block-bochs.o block-vpc.o block-vvfat.o
 ifdef CONFIG_WIN32
 VL_OBJS+=tap-win32.o

block-bochs.c

@@ -85,15 +85,15 @@ static int bochs_probe(const uint8_t *buf, int buf_size, const char *filename)
     return 0;
 }
 
-static int bochs_open(BlockDriverState *bs, const char *filename)
+static int bochs_open(BlockDriverState *bs, const char *filename, int flags)
 {
     BDRVBochsState *s = bs->opaque;
     int fd, i;
     struct bochs_header bochs;
 
-    fd = open(filename, O_RDWR | O_BINARY | O_LARGEFILE);
+    fd = open(filename, O_RDWR | O_BINARY);
     if (fd < 0) {
-        fd = open(filename, O_RDONLY | O_BINARY | O_LARGEFILE);
+        fd = open(filename, O_RDONLY | O_BINARY);
         if (fd < 0)
             return -1;
     }

block-cloop.c

@@ -50,14 +50,14 @@ static int cloop_probe(const uint8_t *buf, int buf_size, const char *filename)
     return 0;
 }
 
-static int cloop_open(BlockDriverState *bs, const char *filename)
+static int cloop_open(BlockDriverState *bs, const char *filename, int flags)
 {
     BDRVCloopState *s = bs->opaque;
     uint32_t offsets_size,max_compressed_block_size=1,i;
 
-    s->fd = open(filename, O_RDONLY | O_BINARY | O_LARGEFILE);
+    s->fd = open(filename, O_RDONLY | O_BINARY);
     if (s->fd < 0)
-        return -1;
+        return -errno;
     bs->read_only = 1;
 
     /* read header */

block-cow.c

@@ -62,7 +62,7 @@ static int cow_probe(const uint8_t *buf, int buf_size, const char *filename)
         return 0;
 }
 
-static int cow_open(BlockDriverState *bs, const char *filename)
+static int cow_open(BlockDriverState *bs, const char *filename, int flags)
 {
     BDRVCowState *s = bs->opaque;
     int fd;
@@ -93,22 +93,6 @@ static int cow_open(BlockDriverState *bs, const char *filename)
     pstrcpy(bs->backing_file, sizeof(bs->backing_file), 
             cow_header.backing_file);
     
-#if 0
-    if (cow_header.backing_file[0] != '\0') {
-        if (stat(cow_header.backing_file, &st) != 0) {
-            fprintf(stderr, "%s: could not find original disk image '%s'\n", filename, cow_header.backing_file);
-            goto fail;
-        }
-        if (st.st_mtime != be32_to_cpu(cow_header.mtime)) {
-            fprintf(stderr, "%s: original raw disk image '%s' does not match saved timestamp\n", filename, cow_header.backing_file);
-            goto fail;
-            }
-        fd = open(cow_header.backing_file, O_RDONLY | O_LARGEFILE);
-        if (fd < 0)
-            goto fail;
-        bs->fd = fd;
-    }
-#endif
     /* mmap the bitmap */
     s->cow_bitmap_size = ((bs->total_sectors + 7) >> 3) + sizeof(cow_header);
     s->cow_bitmap_addr = mmap(get_mmap_addr(s->cow_bitmap_size), 
@@ -179,8 +163,15 @@ static int cow_read(BlockDriverState *bs, int64_t sector_num,
             if (ret != n * 512) 
                 return -1;
         } else {
+            if (bs->backing_hd) {
+                /* read from the base image */
+                ret = bdrv_read(bs->backing_hd, sector_num, buf, n);
+                if (ret < 0)
+                    return -1;
+            } else {
             memset(buf, 0, n * 512);
         }
+        }
         nb_sectors -= n;
         sector_num += n;
         buf += n * 512;
@@ -220,7 +211,7 @@ static int cow_create(const char *filename, int64_t image_sectors,
     if (flags)
         return -ENOTSUP;
 
-    cow_fd = open(filename, O_WRONLY | O_CREAT | O_TRUNC | O_BINARY | O_LARGEFILE, 
+    cow_fd = open(filename, O_WRONLY | O_CREAT | O_TRUNC | O_BINARY, 
               0644);
     if (cow_fd < 0)
         return -1;
@@ -228,18 +219,23 @@ static int cow_create(const char *filename, int64_t image_sectors,
     cow_header.magic = cpu_to_be32(COW_MAGIC);
     cow_header.version = cpu_to_be32(COW_VERSION);
     if (image_filename) {
+        /* Note: if no file, we put a dummy mtime */
+        cow_header.mtime = cpu_to_be32(0);
+
         fd = open(image_filename, O_RDONLY | O_BINARY);
         if (fd < 0) {
             close(cow_fd);
-            return -1;
+            goto mtime_fail;
         }
         if (fstat(fd, &st) != 0) {
             close(fd);
-            return -1;
+            goto mtime_fail;
         }
         close(fd);
         cow_header.mtime = cpu_to_be32(st.st_mtime);
-        realpath(image_filename, cow_header.backing_file);
+    mtime_fail:
+        pstrcpy(cow_header.backing_file, sizeof(cow_header.backing_file),
+                image_filename);
     }
     cow_header.sectorsize = cpu_to_be32(512);
     cow_header.size = cpu_to_be64(image_sectors * 512);

block-dmg.c

@@ -73,16 +73,16 @@ static off_t read_uint32(int fd)
 	return be32_to_cpu(buffer);
 }
 
-static int dmg_open(BlockDriverState *bs, const char *filename)
+static int dmg_open(BlockDriverState *bs, const char *filename, int flags)
 {
     BDRVDMGState *s = bs->opaque;
     off_t info_begin,info_end,last_in_offset,last_out_offset;
     uint32_t count;
     uint32_t max_compressed_size=1,max_sectors_per_chunk=1,i;
 
-    s->fd = open(filename, O_RDONLY | O_BINARY | O_LARGEFILE);
+    s->fd = open(filename, O_RDONLY | O_BINARY);
     if (s->fd < 0)
-        return -1;
+        return -errno;
     bs->read_only = 1;
     s->n_chunks = 0;
     s->offsets = s->lengths = s->sectors = s->sectorcounts = 0;
@@ -93,7 +93,7 @@ dmg_close:
 	close(s->fd);
 	/* open raw instead */
 	bs->drv=&bdrv_raw;
-	return bs->drv->bdrv_open(bs,filename);
+	return bs->drv->bdrv_open(bs, filename, flags);
     }
     info_begin=read_off(s->fd);
     if(info_begin==0)

block-vmdk.c

@@ -89,7 +89,7 @@ static int vmdk_probe(const uint8_t *buf, int buf_size, const char *filename)
         return 0;
 }
 
-static int vmdk_open(BlockDriverState *bs, const char *filename)
+static int vmdk_open(BlockDriverState *bs, const char *filename, int flags)
 {
     BDRVVmdkState *s = bs->opaque;
     int fd, i;

block-vpc.c

@@ -86,19 +86,16 @@ static int vpc_probe(const uint8_t *buf, int buf_size, const char *filename)
     return 0;
 }
 
-static int vpc_open(BlockDriverState *bs, const char *filename)
+static int vpc_open(BlockDriverState *bs, const char *filename, int flags)
 {
     BDRVVPCState *s = bs->opaque;
     int fd, i;
     struct vpc_subheader header;
 
-    fd = open(filename, O_RDWR | O_BINARY | O_LARGEFILE);
-    if (fd < 0) {
-        fd = open(filename, O_RDONLY | O_BINARY | O_LARGEFILE);
-        if (fd < 0)
-            return -1;
-    }
-    
+    fd = open(filename, O_RDONLY | O_BINARY);
+    if (fd < 0)
+        return -1;
+
     bs->read_only = 1; // no write support yet
     
     s->fd = fd;

block-vvfat.c

@@ -351,13 +351,6 @@ typedef struct BDRVVVFATState {
 } BDRVVVFATState;
 
 
-static int vvfat_probe(const uint8_t *buf, int buf_size, const char *filename)
-{
-    if (strstart(filename, "fat:", NULL))
-	return 100;
-    return 0;
-}
-
 static void init_mbr(BDRVVVFATState* s)
 {
     /* TODO: if the files mbr.img and bootsect.img exist, use them */
@@ -954,18 +947,22 @@ static int init_directories(BDRVVVFATState* s,
     return 0;
 }
 
+#ifdef DEBUG
 static BDRVVVFATState *vvv = NULL;
+#endif
 
 static int enable_write_target(BDRVVVFATState *s);
 static int is_consistent(BDRVVVFATState *s);
 
-static int vvfat_open(BlockDriverState *bs, const char* dirname)
+static int vvfat_open(BlockDriverState *bs, const char* dirname, int flags)
 {
     BDRVVVFATState *s = bs->opaque;
     int floppy = 0;
     int i;
 
+#ifdef DEBUG
     vvv = s;
+#endif
 
 DLOG(if (stderr == NULL) {
     stderr = fopen("vvfat.log", "a");
@@ -1040,7 +1037,6 @@ DLOG(if (stderr == NULL) {
 	bs->heads = bs->cyls = bs->secs = 0;
 
     //    assert(is_consistent(s));
-
     return 0;
 }
 
@@ -2732,8 +2728,7 @@ static int enable_write_target(BDRVVVFATState *s)
     array_init(&(s->commits), sizeof(commit_t));
 
     s->qcow_filename = malloc(1024);
-    strcpy(s->qcow_filename, "/tmp/vl.XXXXXX");
-    get_tmp_filename(s->qcow_filename, strlen(s->qcow_filename) + 1);
+    get_tmp_filename(s->qcow_filename, 1024);
     if (bdrv_create(&bdrv_qcow,
 		s->qcow_filename, s->sector_count, "fat:", 0) < 0)
 	return -1;
@@ -2767,14 +2762,15 @@ static void vvfat_close(BlockDriverState *bs)
 BlockDriver bdrv_vvfat = {
     "vvfat",
     sizeof(BDRVVVFATState),
-    vvfat_probe,
+    NULL, /* no probe for protocols */
     vvfat_open,
     vvfat_read,
     vvfat_write,
     vvfat_close,
     NULL, /* ??? Not sure if we can do any meaningful flushing.  */
     NULL,
-    vvfat_is_allocated
+    vvfat_is_allocated,
+    .protocol_name = "fat",
 };
 
 #ifdef DEBUG

block_int.h

@@ -28,7 +28,7 @@ struct BlockDriver {
     const char *format_name;
     int instance_size;
     int (*bdrv_probe)(const uint8_t *buf, int buf_size, const char *filename);
-    int (*bdrv_open)(BlockDriverState *bs, const char *filename);
+    int (*bdrv_open)(BlockDriverState *bs, const char *filename, int flags);
     int (*bdrv_read)(BlockDriverState *bs, int64_t sector_num, 
                      uint8_t *buf, int nb_sectors);
     int (*bdrv_write)(BlockDriverState *bs, int64_t sector_num, 
@@ -41,11 +41,28 @@ struct BlockDriver {
                              int nb_sectors, int *pnum);
     int (*bdrv_set_key)(BlockDriverState *bs, const char *key);
     int (*bdrv_make_empty)(BlockDriverState *bs);
+    /* aio */
+    int (*bdrv_aio_new)(BlockDriverAIOCB *acb);
+    int (*bdrv_aio_read)(BlockDriverAIOCB *acb, int64_t sector_num,
+                              uint8_t *buf, int nb_sectors);
+    int (*bdrv_aio_write)(BlockDriverAIOCB *acb, int64_t sector_num,
+                          const uint8_t *buf, int nb_sectors);
+    void (*bdrv_aio_cancel)(BlockDriverAIOCB *acb);
+    void (*bdrv_aio_delete)(BlockDriverAIOCB *acb);
+
+    const char *protocol_name;
+    int (*bdrv_pread)(BlockDriverState *bs, int64_t offset, 
+                      uint8_t *buf, int count);
+    int (*bdrv_pwrite)(BlockDriverState *bs, int64_t offset, 
+                       const uint8_t *buf, int count);
+    int (*bdrv_truncate)(BlockDriverState *bs, int64_t offset);
+    int64_t (*bdrv_getlength)(BlockDriverState *bs);
+
     struct BlockDriver *next;
 };
 
 struct BlockDriverState {
-    int64_t total_sectors;
+    int64_t total_sectors; /* XXX: will be suppressed */
     int read_only; /* if true, the media is read only */
     int inserted; /* if true, the media is present */
     int removable; /* if true, the media can be removed */
@@ -67,6 +84,9 @@ struct BlockDriverState {
     int is_temporary;
     
     BlockDriverState *backing_hd;
+    /* sync read/write emulation */
+
+    BlockDriverAIOCB *sync_aiocb;
     
     /* NOTE: the following infos are only hints for real hardware
        drivers. They are not used by the block driver */
@@ -76,6 +96,14 @@ struct BlockDriverState {
     BlockDriverState *next;
 };
 
+struct BlockDriverAIOCB {
+    BlockDriverState *bs;
+    BlockDriverCompletionFunc *cb;
+    void *cb_opaque;
+    
+    void *opaque; /* driver opaque */
+};
+
 void get_tmp_filename(char *filename, int size);
 
 #endif /* BLOCK_INT_H */

vl.c

@@ -4771,6 +4771,77 @@ static int ram_load(QEMUFile *f, void *opaque, int version_id)
     return 0;
 }
 
+/***********************************************************/
+/* bottom halves (can be seen as timers which expire ASAP) */
+
+struct QEMUBH {
+    QEMUBHFunc *cb;
+    void *opaque;
+    int scheduled;
+    QEMUBH *next;
+};
+
+static QEMUBH *first_bh = NULL;
+
+QEMUBH *qemu_bh_new(QEMUBHFunc *cb, void *opaque)
+{
+    QEMUBH *bh;
+    bh = qemu_mallocz(sizeof(QEMUBH));
+    if (!bh)
+        return NULL;
+    bh->cb = cb;
+    bh->opaque = opaque;
+    return bh;
+}
+
+void qemu_bh_poll(void)
+{
+    QEMUBH *bh, **pbh;
+
+    for(;;) {
+        pbh = &first_bh;
+        bh = *pbh;
+        if (!bh)
+            break;
+        *pbh = bh->next;
+        bh->scheduled = 0;
+        bh->cb(bh->opaque);
+    }
+}
+
+void qemu_bh_schedule(QEMUBH *bh)
+{
+    CPUState *env = cpu_single_env;
+    if (bh->scheduled)
+        return;
+    bh->scheduled = 1;
+    bh->next = first_bh;
+    first_bh = bh;
+
+    /* stop the currently executing CPU to execute the BH ASAP */
+    if (env) {
+        cpu_interrupt(env, CPU_INTERRUPT_EXIT);
+    }
+}
+
+void qemu_bh_cancel(QEMUBH *bh)
+{
+    QEMUBH **pbh;
+    if (bh->scheduled) {
+        pbh = &first_bh;
+        while (*pbh != bh)
+            pbh = &(*pbh)->next;
+        *pbh = bh->next;
+        bh->scheduled = 0;
+    }
+}
+
+void qemu_bh_delete(QEMUBH *bh)
+{
+    qemu_bh_cancel(bh);
+    qemu_free(bh);
+}
+
 /***********************************************************/
 /* machine registration */
 
@@ -5030,6 +5101,8 @@ void main_loop_wait(int timeout)
 #ifdef _WIN32
     tap_win32_poll();
 #endif
+    qemu_aio_poll();
+    qemu_bh_poll();
 
     if (vm_running) {
         qemu_run_timers(&active_timers[QEMU_TIMER_VIRTUAL], 
@@ -6049,6 +6122,7 @@ int main(int argc, char **argv)
     
     init_timers();
     init_timer_alarm();
+    qemu_aio_init();
 
 #ifdef _WIN32
     socket_init();
@@ -6093,7 +6167,7 @@ int main(int argc, char **argv)
                 snprintf(buf, sizeof(buf), "hd%c", i + 'a');
                 bs_table[i] = bdrv_new(buf);
             }
-            if (bdrv_open(bs_table[i], hd_filename[i], snapshot) < 0) {
+            if (bdrv_open(bs_table[i], hd_filename[i], snapshot ? BDRV_O_SNAPSHOT : 0) < 0) {
                 fprintf(stderr, "qemu: could not open hard disk image '%s'\n",
                         hd_filename[i]);
                 exit(1);
@@ -6118,7 +6192,8 @@ int main(int argc, char **argv)
                 bdrv_set_type_hint(fd_table[i], BDRV_TYPE_FLOPPY);
             }
             if (fd_filename[i] != '\0') {
-                if (bdrv_open(fd_table[i], fd_filename[i], snapshot) < 0) {
+                if (bdrv_open(fd_table[i], fd_filename[i],
+                              snapshot ? BDRV_O_SNAPSHOT : 0) < 0) {
                     fprintf(stderr, "qemu: could not open floppy disk image '%s'\n",
                             fd_filename[i]);
                     exit(1);

vl.h

@@ -481,6 +481,16 @@ void qemu_put_timer(QEMUFile *f, QEMUTimer *ts);
 void cpu_save(QEMUFile *f, void *opaque);
 int cpu_load(QEMUFile *f, void *opaque, int version_id);
 
+/* bottom halves */
+typedef struct QEMUBH QEMUBH;
+typedef void QEMUBHFunc(void *opaque);
+
+QEMUBH *qemu_bh_new(QEMUBHFunc *cb, void *opaque);
+void qemu_bh_schedule(QEMUBH *bh);
+void qemu_bh_cancel(QEMUBH *bh);
+void qemu_bh_delete(QEMUBH *bh);
+void qemu_bh_poll(void);
+
 /* block.c */
 typedef struct BlockDriverState BlockDriverState;
 typedef struct BlockDriver BlockDriver;
@@ -495,6 +505,16 @@ extern BlockDriver bdrv_bochs;
 extern BlockDriver bdrv_vpc;
 extern BlockDriver bdrv_vvfat;
 
+#define BDRV_O_RDONLY      0x0000
+#define BDRV_O_RDWR        0x0002
+#define BDRV_O_ACCESS      0x0003
+#define BDRV_O_CREAT       0x0004 /* create an empty file */
+#define BDRV_O_SNAPSHOT    0x0008 /* open the file read only and save writes in a snapshot */
+#define BDRV_O_FILE        0x0010 /* open as a raw file (do not try to
+                                     use a disk image format on top of
+                                     it (default for
+                                     bdrv_file_open()) */
+
 void bdrv_init(void);
 BlockDriver *bdrv_find_format(const char *format_name);
 int bdrv_create(BlockDriver *drv, 
@@ -502,17 +522,44 @@ int bdrv_create(BlockDriver *drv,
                 const char *backing_file, int flags);
 BlockDriverState *bdrv_new(const char *device_name);
 void bdrv_delete(BlockDriverState *bs);
-int bdrv_open(BlockDriverState *bs, const char *filename, int snapshot);
-int bdrv_open2(BlockDriverState *bs, const char *filename, int snapshot,
+int bdrv_file_open(BlockDriverState **pbs, const char *filename, int flags);
+int bdrv_open(BlockDriverState *bs, const char *filename, int flags);
+int bdrv_open2(BlockDriverState *bs, const char *filename, int flags,
                BlockDriver *drv);
 void bdrv_close(BlockDriverState *bs);
 int bdrv_read(BlockDriverState *bs, int64_t sector_num, 
               uint8_t *buf, int nb_sectors);
 int bdrv_write(BlockDriverState *bs, int64_t sector_num, 
                const uint8_t *buf, int nb_sectors);
+int bdrv_pread(BlockDriverState *bs, int64_t offset, 
+               void *buf, int count);
+int bdrv_pwrite(BlockDriverState *bs, int64_t offset, 
+                const void *buf, int count);
+int bdrv_truncate(BlockDriverState *bs, int64_t offset);
+int64_t bdrv_getlength(BlockDriverState *bs);
 void bdrv_get_geometry(BlockDriverState *bs, int64_t *nb_sectors_ptr);
 int bdrv_commit(BlockDriverState *bs);
 void bdrv_set_boot_sector(BlockDriverState *bs, const uint8_t *data, int size);
+/* async block I/O */
+typedef struct BlockDriverAIOCB BlockDriverAIOCB;
+typedef void BlockDriverCompletionFunc(void *opaque, int ret);
+
+BlockDriverAIOCB *bdrv_aio_new(BlockDriverState *bs);
+int bdrv_aio_read(BlockDriverAIOCB *acb, int64_t sector_num,
+                  uint8_t *buf, int nb_sectors,
+                  BlockDriverCompletionFunc *cb, void *opaque);
+int bdrv_aio_write(BlockDriverAIOCB *acb, int64_t sector_num,
+                   const uint8_t *buf, int nb_sectors,
+                   BlockDriverCompletionFunc *cb, void *opaque);
+void bdrv_aio_cancel(BlockDriverAIOCB *acb);
+void bdrv_aio_delete(BlockDriverAIOCB *acb);
+
+void qemu_aio_init(void);
+void qemu_aio_poll(void);
+void qemu_aio_wait_start(void);
+void qemu_aio_wait(void);
+void qemu_aio_wait_end(void);
+
 /* Ensure contents are flushed to disk.  */
 void bdrv_flush(BlockDriverState *bs);
 
@@ -551,6 +598,13 @@ const char *bdrv_get_device_name(BlockDriverState *bs);
 int qcow_get_cluster_size(BlockDriverState *bs);
 int qcow_compress_cluster(BlockDriverState *bs, int64_t sector_num,
                           const uint8_t *buf);
+void bdrv_get_backing_filename(BlockDriverState *bs, 
+                               char *filename, int filename_size);
+
+int path_is_absolute(const char *path);
+void path_combine(char *dest, int dest_size,
+                  const char *base_path,
+                  const char *filename);
 
 #ifndef QEMU_TOOL