block: Allow BDRV_REQ_FUA through blk_pwrite()

We have several block drivers that understand BDRV_REQ_FUA,
and emulate it in the block layer for the rest by a full flush.
But without a way to actually request BDRV_REQ_FUA during a
pass-through blk_pwrite(), FUA-aware block drivers like NBD are
forced to repeat the emulation logic of a full flush regardless
of whether the backend they are writing to could do it more
efficiently.

This patch just wires up a flags argument; followup patches
will actually make use of it in the NBD driver and in qemu-io.

Signed-off-by: Eric Blake <[email protected]>
Acked-by: Denis V. Lunev <[email protected]>
Signed-off-by: Kevin Wolf <[email protected]>

block/block-backend.c

@@ -953,9 +953,11 @@ int blk_pread(BlockBackend *blk, int64_t offset, void *buf, int count)
     return count;
 }
 
-int blk_pwrite(BlockBackend *blk, int64_t offset, const void *buf, int count)
+int blk_pwrite(BlockBackend *blk, int64_t offset, const void *buf, int count,
+               BdrvRequestFlags flags)
 {
-    int ret = blk_prw(blk, offset, (void*) buf, count, blk_write_entry, 0);
+    int ret = blk_prw(blk, offset, (void *) buf, count, blk_write_entry,
+                      flags);
     if (ret < 0) {
         return ret;
     }

block/crypto.c

@@ -91,7 +91,7 @@ static ssize_t block_crypto_write_func(QCryptoBlock *block,
     struct BlockCryptoCreateData *data = opaque;
     ssize_t ret;
 
-    ret = blk_pwrite(data->blk, offset, buf, buflen);
+    ret = blk_pwrite(data->blk, offset, buf, buflen, 0);
     if (ret < 0) {
         error_setg_errno(errp, -ret, "Could not write encryption header");
         return ret;

block/parallels.c

@@ -512,7 +512,7 @@ static int parallels_create(const char *filename, QemuOpts *opts, Error **errp)
     memset(tmp, 0, sizeof(tmp));
     memcpy(tmp, &header, sizeof(header));
 
-    ret = blk_pwrite(file, 0, tmp, BDRV_SECTOR_SIZE);
+    ret = blk_pwrite(file, 0, tmp, BDRV_SECTOR_SIZE, 0);
     if (ret < 0) {
         goto exit;
     }

block/qcow.c

@@ -853,14 +853,14 @@ static int qcow_create(const char *filename, QemuOpts *opts, Error **errp)
     }
 
     /* write all the data */
-    ret = blk_pwrite(qcow_blk, 0, &header, sizeof(header));
+    ret = blk_pwrite(qcow_blk, 0, &header, sizeof(header), 0);
     if (ret != sizeof(header)) {
         goto exit;
     }
 
     if (backing_file) {
         ret = blk_pwrite(qcow_blk, sizeof(header),
-            backing_file, backing_filename_len);
+                         backing_file, backing_filename_len, 0);
         if (ret != backing_filename_len) {
             goto exit;
         }
@@ -869,8 +869,8 @@ static int qcow_create(const char *filename, QemuOpts *opts, Error **errp)
     tmp = g_malloc0(BDRV_SECTOR_SIZE);
     for (i = 0; i < ((sizeof(uint64_t)*l1_size + BDRV_SECTOR_SIZE - 1)/
         BDRV_SECTOR_SIZE); i++) {
-        ret = blk_pwrite(qcow_blk, header_size +
-            BDRV_SECTOR_SIZE*i, tmp, BDRV_SECTOR_SIZE);
+        ret = blk_pwrite(qcow_blk, header_size + BDRV_SECTOR_SIZE * i,
+                         tmp, BDRV_SECTOR_SIZE, 0);
         if (ret != BDRV_SECTOR_SIZE) {
             g_free(tmp);
             goto exit;

block/qcow2.c

@@ -2207,7 +2207,7 @@ static int qcow2_create2(const char *filename, int64_t total_size,
             cpu_to_be64(QCOW2_COMPAT_LAZY_REFCOUNTS);
     }
 
-    ret = blk_pwrite(blk, 0, header, cluster_size);
+    ret = blk_pwrite(blk, 0, header, cluster_size, 0);
     g_free(header);
     if (ret < 0) {
         error_setg_errno(errp, -ret, "Could not write qcow2 header");
@@ -2217,7 +2217,7 @@ static int qcow2_create2(const char *filename, int64_t total_size,
     /* Write a refcount table with one refcount block */
     refcount_table = g_malloc0(2 * cluster_size);
     refcount_table[0] = cpu_to_be64(2 * cluster_size);
-    ret = blk_pwrite(blk, cluster_size, refcount_table, 2 * cluster_size);
+    ret = blk_pwrite(blk, cluster_size, refcount_table, 2 * cluster_size, 0);
     g_free(refcount_table);
 
     if (ret < 0) {

block/qed.c

@@ -601,18 +601,18 @@ static int qed_create(const char *filename, uint32_t cluster_size,
     }
 
     qed_header_cpu_to_le(&header, &le_header);
-    ret = blk_pwrite(blk, 0, &le_header, sizeof(le_header));
+    ret = blk_pwrite(blk, 0, &le_header, sizeof(le_header), 0);
     if (ret < 0) {
         goto out;
     }
     ret = blk_pwrite(blk, sizeof(le_header), backing_file,
-                     header.backing_filename_size);
+                     header.backing_filename_size, 0);
     if (ret < 0) {
         goto out;
     }
 
     l1_table = g_malloc0(l1_size);
-    ret = blk_pwrite(blk, header.l1_table_offset, l1_table, l1_size);
+    ret = blk_pwrite(blk, header.l1_table_offset, l1_table, l1_size, 0);
     if (ret < 0) {
         goto out;
     }

block/sheepdog.c

@@ -1681,7 +1681,7 @@ static int sd_prealloc(const char *filename, Error **errp)
         if (ret < 0) {
             goto out;
         }
-        ret = blk_pwrite(blk, idx * buf_size, buf, buf_size);
+        ret = blk_pwrite(blk, idx * buf_size, buf, buf_size, 0);
         if (ret < 0) {
             goto out;
         }

block/vdi.c

@@ -827,7 +827,7 @@ static int vdi_create(const char *filename, QemuOpts *opts, Error **errp)
     vdi_header_print(&header);
 #endif
     vdi_header_to_le(&header);
-    ret = blk_pwrite(blk, offset, &header, sizeof(header));
+    ret = blk_pwrite(blk, offset, &header, sizeof(header), 0);
     if (ret < 0) {
         error_setg(errp, "Error writing header to %s", filename);
         goto exit;
@@ -848,7 +848,7 @@ static int vdi_create(const char *filename, QemuOpts *opts, Error **errp)
                 bmap[i] = VDI_UNALLOCATED;
             }
         }
-        ret = blk_pwrite(blk, offset, bmap, bmap_size);
+        ret = blk_pwrite(blk, offset, bmap, bmap_size, 0);
         if (ret < 0) {
             error_setg(errp, "Error writing bmap to %s", filename);
             goto exit;

block/vhdx.c

@@ -1856,13 +1856,14 @@ static int vhdx_create(const char *filename, QemuOpts *opts, Error **errp)
     creator = g_utf8_to_utf16("QEMU v" QEMU_VERSION, -1, NULL,
                               &creator_items, NULL);
     signature = cpu_to_le64(VHDX_FILE_SIGNATURE);
-    ret = blk_pwrite(blk, VHDX_FILE_ID_OFFSET, &signature, sizeof(signature));
+    ret = blk_pwrite(blk, VHDX_FILE_ID_OFFSET, &signature, sizeof(signature),
+                     0);
     if (ret < 0) {
         goto delete_and_exit;
     }
     if (creator) {
         ret = blk_pwrite(blk, VHDX_FILE_ID_OFFSET + sizeof(signature),
-                         creator, creator_items * sizeof(gunichar2));
+                         creator, creator_items * sizeof(gunichar2), 0);
         if (ret < 0) {
             goto delete_and_exit;
         }

block/vmdk.c

@@ -1808,12 +1808,12 @@ static int vmdk_create_extent(const char *filename, int64_t filesize,
     header.check_bytes[3] = 0xa;
 
     /* write all the data */
-    ret = blk_pwrite(blk, 0, &magic, sizeof(magic));
+    ret = blk_pwrite(blk, 0, &magic, sizeof(magic), 0);
     if (ret < 0) {
         error_setg(errp, QERR_IO_ERROR);
         goto exit;
     }
-    ret = blk_pwrite(blk, sizeof(magic), &header, sizeof(header));
+    ret = blk_pwrite(blk, sizeof(magic), &header, sizeof(header), 0);
     if (ret < 0) {
         error_setg(errp, QERR_IO_ERROR);
         goto exit;
@@ -1833,7 +1833,7 @@ static int vmdk_create_extent(const char *filename, int64_t filesize,
         gd_buf[i] = cpu_to_le32(tmp);
     }
     ret = blk_pwrite(blk, le64_to_cpu(header.rgd_offset) * BDRV_SECTOR_SIZE,
-                     gd_buf, gd_buf_size);
+                     gd_buf, gd_buf_size, 0);
     if (ret < 0) {
         error_setg(errp, QERR_IO_ERROR);
         goto exit;
@@ -1845,7 +1845,7 @@ static int vmdk_create_extent(const char *filename, int64_t filesize,
         gd_buf[i] = cpu_to_le32(tmp);
     }
     ret = blk_pwrite(blk, le64_to_cpu(header.gd_offset) * BDRV_SECTOR_SIZE,
-                     gd_buf, gd_buf_size);
+                     gd_buf, gd_buf_size, 0);
     if (ret < 0) {
         error_setg(errp, QERR_IO_ERROR);
         goto exit;
@@ -2120,7 +2120,7 @@ static int vmdk_create(const char *filename, QemuOpts *opts, Error **errp)
 
     blk_set_allow_write_beyond_eof(new_blk, true);
 
-    ret = blk_pwrite(new_blk, desc_offset, desc, desc_len);
+    ret = blk_pwrite(new_blk, desc_offset, desc, desc_len, 0);
     if (ret < 0) {
         error_setg_errno(errp, -ret, "Could not write description");
         goto exit;

block/vpc.c

@@ -788,13 +788,13 @@ static int create_dynamic_disk(BlockBackend *blk, uint8_t *buf,
     block_size = 0x200000;
     num_bat_entries = (total_sectors + block_size / 512) / (block_size / 512);
 
-    ret = blk_pwrite(blk, offset, buf, HEADER_SIZE);
+    ret = blk_pwrite(blk, offset, buf, HEADER_SIZE, 0);
     if (ret < 0) {
         goto fail;
     }
 
     offset = 1536 + ((num_bat_entries * 4 + 511) & ~511);
-    ret = blk_pwrite(blk, offset, buf, HEADER_SIZE);
+    ret = blk_pwrite(blk, offset, buf, HEADER_SIZE, 0);
     if (ret < 0) {
         goto fail;
     }
@@ -804,7 +804,7 @@ static int create_dynamic_disk(BlockBackend *blk, uint8_t *buf,
 
     memset(buf, 0xFF, 512);
     for (i = 0; i < (num_bat_entries * 4 + 511) / 512; i++) {
-        ret = blk_pwrite(blk, offset, buf, 512);
+        ret = blk_pwrite(blk, offset, buf, 512, 0);
         if (ret < 0) {
             goto fail;
         }
@@ -831,7 +831,7 @@ static int create_dynamic_disk(BlockBackend *blk, uint8_t *buf,
     /* Write the header */
     offset = 512;
 
-    ret = blk_pwrite(blk, offset, buf, 1024);
+    ret = blk_pwrite(blk, offset, buf, 1024, 0);
     if (ret < 0) {
         goto fail;
     }
@@ -853,7 +853,7 @@ static int create_fixed_disk(BlockBackend *blk, uint8_t *buf,
         return ret;
     }
 
-    ret = blk_pwrite(blk, total_size - HEADER_SIZE, buf, HEADER_SIZE);
+    ret = blk_pwrite(blk, total_size - HEADER_SIZE, buf, HEADER_SIZE, 0);
     if (ret < 0) {
         return ret;
     }

hw/nvram/spapr_nvram.c

@@ -124,7 +124,7 @@ static void rtas_nvram_store(PowerPCCPU *cpu, sPAPRMachineState *spapr,
 
     alen = len;
     if (nvram->blk) {
-        alen = blk_pwrite(nvram->blk, offset, membuf, len);
+        alen = blk_pwrite(nvram->blk, offset, membuf, len, 0);
     }
 
     assert(nvram->buf);
@@ -190,7 +190,7 @@ static int spapr_nvram_post_load(void *opaque, int version_id)
     sPAPRNVRAM *nvram = VIO_SPAPR_NVRAM(opaque);
 
     if (nvram->blk) {
-        int alen = blk_pwrite(nvram->blk, 0, nvram->buf, nvram->size);
+        int alen = blk_pwrite(nvram->blk, 0, nvram->buf, nvram->size, 0);
 
         if (alen < 0) {
             return alen;

include/sysemu/block-backend.h

@@ -102,7 +102,8 @@ BlockAIOCB *blk_aio_write_zeroes(BlockBackend *blk, int64_t sector_num,
                                  int nb_sectors, BdrvRequestFlags flags,
                                  BlockCompletionFunc *cb, void *opaque);
 int blk_pread(BlockBackend *blk, int64_t offset, void *buf, int count);
-int blk_pwrite(BlockBackend *blk, int64_t offset, const void *buf, int count);
+int blk_pwrite(BlockBackend *blk, int64_t offset, const void *buf, int count,
+               BdrvRequestFlags flags);
 int64_t blk_getlength(BlockBackend *blk);
 void blk_get_geometry(BlockBackend *blk, uint64_t *nb_sectors_ptr);
 int64_t blk_nb_sectors(BlockBackend *blk);

nbd/server.c

@@ -1115,7 +1115,7 @@ static void nbd_trip(void *opaque)
         TRACE("Writing to device");
 
         ret = blk_pwrite(exp->blk, request.from + exp->dev_offset,
-                        req->data, request.len);
+                         req->data, request.len, 0);
         if (ret < 0) {
             LOG("writing to file failed");
             reply.error = -ret;

qemu-io-cmds.c

@@ -474,7 +474,7 @@ static int do_pwrite(BlockBackend *blk, char *buf, int64_t offset,
         return -ERANGE;
     }
 
-    *total = blk_pwrite(blk, offset, (uint8_t *)buf, count);
+    *total = blk_pwrite(blk, offset, (uint8_t *)buf, count, 0);
     if (*total < 0) {
         return *total;
     }