[PATCH 2/3] nbd: Add .bdrv_known_zeroes() client support

[Eric Blake]

Using the new NBD extension of NBD_INFO_INIT_STATE, we can pass on the
information when a server reports that an image initially reads as all
zeroes.  The server information is treated as stale the moment we
request a write operation, even across reconnections to the server,
which is fine since our intended usage of BDRV_ZERO_OPEN is to
optimize qemu-img at startup, and not something relied on during later
image use.

Update iotests to reflect improved output of 'qemu-nbd --list'.

As NBD still cannot create or resize images, we don't need to worry
about BDRV_ZERO_CREATE or BDRV_ZERO_TRUNCATE.

Signed-off-by: Eric Blake <address@hidden>
---
 block/nbd.c                | 15 +++++++++++++++
 include/block/nbd.h        |  4 ++++
 nbd/client.c               | 24 ++++++++++++++++++++----
 nbd/trace-events           |  1 +
 qemu-nbd.c                 | 13 +++++++++++++
 tests/qemu-iotests/223.out |  4 ++++
 tests/qemu-iotests/233.out |  1 +
 7 files changed, 58 insertions(+), 4 deletions(-)

diff --git a/block/nbd.c b/block/nbd.c
index d085554f21ea..2e1fbd6152f6 100644
--- a/block/nbd.c
+++ b/block/nbd.c
@@ -1204,6 +1204,7 @@ static int nbd_client_co_pwritev(BlockDriverState *bs, 
uint64_t offset,
     };

     assert(!(s->info.flags & NBD_FLAG_READ_ONLY));
+    s->info.modified = true;
     if (flags & BDRV_REQ_FUA) {
         assert(s->info.flags & NBD_FLAG_SEND_FUA);
         request.flags |= NBD_CMD_FLAG_FUA;
@@ -1276,6 +1277,7 @@ static int nbd_client_co_pdiscard(BlockDriverState *bs, 
int64_t offset,
     };

     assert(!(s->info.flags & NBD_FLAG_READ_ONLY));
+    s->info.modified = true;
     if (!(s->info.flags & NBD_FLAG_SEND_TRIM) || !bytes) {
         return 0;
     }
@@ -1909,6 +1911,16 @@ static int nbd_co_flush(BlockDriverState *bs)
     return nbd_client_co_flush(bs);
 }

+static int nbd_known_zeroes(BlockDriverState *bs)
+{
+    BDRVNBDState *s = bs->opaque;
+
+    if (!s->info.modified && s->info.init_state & NBD_INIT_ZERO) {
+        return BDRV_ZERO_OPEN;
+    }
+    return 0;
+}
+
 static void nbd_refresh_limits(BlockDriverState *bs, Error **errp)
 {
     BDRVNBDState *s = (BDRVNBDState *)bs->opaque;
@@ -2027,6 +2039,7 @@ static BlockDriver bdrv_nbd = {
     .bdrv_close                 = nbd_close,
     .bdrv_co_flush_to_os        = nbd_co_flush,
     .bdrv_co_pdiscard           = nbd_client_co_pdiscard,
+    .bdrv_known_zeroes          = nbd_known_zeroes,
     .bdrv_refresh_limits        = nbd_refresh_limits,
     .bdrv_getlength             = nbd_getlength,
     .bdrv_detach_aio_context    = nbd_client_detach_aio_context,
@@ -2052,6 +2065,7 @@ static BlockDriver bdrv_nbd_tcp = {
     .bdrv_close                 = nbd_close,
     .bdrv_co_flush_to_os        = nbd_co_flush,
     .bdrv_co_pdiscard           = nbd_client_co_pdiscard,
+    .bdrv_known_zeroes          = nbd_known_zeroes,
     .bdrv_refresh_limits        = nbd_refresh_limits,
     .bdrv_getlength             = nbd_getlength,
     .bdrv_detach_aio_context    = nbd_client_detach_aio_context,
@@ -2077,6 +2091,7 @@ static BlockDriver bdrv_nbd_unix = {
     .bdrv_close                 = nbd_close,
     .bdrv_co_flush_to_os        = nbd_co_flush,
     .bdrv_co_pdiscard           = nbd_client_co_pdiscard,
+    .bdrv_known_zeroes          = nbd_known_zeroes,
     .bdrv_refresh_limits        = nbd_refresh_limits,
     .bdrv_getlength             = nbd_getlength,
     .bdrv_detach_aio_context    = nbd_client_detach_aio_context,
diff --git a/include/block/nbd.h b/include/block/nbd.h
index 0de020904a37..5103053bed49 100644
--- a/include/block/nbd.h
+++ b/include/block/nbd.h
@@ -307,6 +307,7 @@ struct NBDExportInfo {
     uint32_t min_block;
     uint32_t opt_block;
     uint32_t max_block;
+    uint16_t init_state;

     uint32_t context_id;

@@ -314,6 +315,9 @@ struct NBDExportInfo {
     char *description;
     int n_contexts;
     char **contexts;
+
+    /* Set during runtime to track if init_state is still trustworthy. */
+    bool modified;
 };
 typedef struct NBDExportInfo NBDExportInfo;

diff --git a/nbd/client.c b/nbd/client.c
index ba173108baab..199a8a2bc49e 100644
--- a/nbd/client.c
+++ b/nbd/client.c
@@ -350,16 +350,17 @@ static int nbd_opt_info_or_go(QIOChannel *ioc, uint32_t 
opt,

     assert(opt == NBD_OPT_GO || opt == NBD_OPT_INFO);
     trace_nbd_opt_info_go_start(nbd_opt_lookup(opt), info->name);
-    buf = g_malloc(4 + len + 2 + 2 * info->request_sizes + 1);
+    buf = g_malloc(4 + len + 2 + 2 * (info->request_sizes + 1) + 1);
     stl_be_p(buf, len);
     memcpy(buf + 4, info->name, len);
-    /* At most one request, everything else up to server */
-    stw_be_p(buf + 4 + len, info->request_sizes);
+    /* One or two requests, everything else up to server */
+    stw_be_p(buf + 4 + len, info->request_sizes + 1);
     if (info->request_sizes) {
         stw_be_p(buf + 4 + len + 2, NBD_INFO_BLOCK_SIZE);
     }
+    stw_be_p(buf + 4 + len + 2 + 2 * info->request_sizes, NBD_INFO_INIT_STATE);
     error = nbd_send_option_request(ioc, opt,
-                                    4 + len + 2 + 2 * info->request_sizes,
+                                    4 + len + 2 + 2 * (info->request_sizes + 
1),
                                     buf, errp);
     g_free(buf);
     if (error < 0) {
@@ -484,6 +485,21 @@ static int nbd_opt_info_or_go(QIOChannel *ioc, uint32_t 
opt,
                                           info->max_block);
             break;

+        case NBD_INFO_INIT_STATE:
+            if (len != sizeof(info->init_state)) {
+                error_setg(errp, "remaining export info len %" PRIu32
+                           " is unexpected size", len);
+                nbd_send_opt_abort(ioc);
+                return -1;
+            }
+            if (nbd_read16(ioc, &info->init_state, "info init state",
+                           errp) < 0) {
+                nbd_send_opt_abort(ioc);
+                return -1;
+            }
+            trace_nbd_opt_info_init_state(info->init_state);
+            break;
+
         default:
             /*
              * Not worth the bother to check if NBD_INFO_NAME or
diff --git a/nbd/trace-events b/nbd/trace-events
index a955918e9707..12589b2afb84 100644
--- a/nbd/trace-events
+++ b/nbd/trace-events
@@ -10,6 +10,7 @@ nbd_opt_info_go_start(const char *opt, const char *name) 
"Attempting %s for expo
 nbd_opt_info_go_success(const char *opt) "Export is ready after %s request"
 nbd_opt_info_unknown(int info, const char *name) "Ignoring unknown info %d 
(%s)"
 nbd_opt_info_block_size(uint32_t minimum, uint32_t preferred, uint32_t 
maximum) "Block sizes are 0x%" PRIx32 ", 0x%" PRIx32 ", 0x%" PRIx32
+nbd_opt_info_init_state(unsigned int flags) "Initial state flags 0x%x"
 nbd_receive_query_exports_start(const char *wantname) "Querying export list 
for '%s'"
 nbd_receive_query_exports_success(const char *wantname) "Found desired export 
name '%s'"
 nbd_receive_starttls_new_client(void) "Setting up TLS"
diff --git a/qemu-nbd.c b/qemu-nbd.c
index 4aa005004ebd..856df85823bc 100644
--- a/qemu-nbd.c
+++ b/qemu-nbd.c
@@ -220,6 +220,19 @@ static int qemu_nbd_client_list(SocketAddress *saddr, 
QCryptoTLSCreds *tls,
             printf("  opt block: %u\n", list[i].opt_block);
             printf("  max block: %u\n", list[i].max_block);
         }
+        {
+            static const char *const init_names[] = {
+                [NBD_INIT_SPARSE_BIT]            = "sparse",
+                [NBD_INIT_ZERO_BIT]              = "zero",
+            };
+            printf("  init state: 0x%x (", list[i].init_state);
+            for (size_t bit = 0; bit < ARRAY_SIZE(init_names); bit++) {
+                if (init_names[bit] && (list[i].init_state & (1 << bit))) {
+                    printf(" %s", init_names[bit]);
+                }
+            }
+            printf(" )\n");
+        }
         if (list[i].n_contexts) {
             printf("  available meta contexts: %d\n", list[i].n_contexts);
             for (j = 0; j < list[i].n_contexts; j++) {
diff --git a/tests/qemu-iotests/223.out b/tests/qemu-iotests/223.out
index 80c0cf65095b..ce7945aa7cf6 100644
--- a/tests/qemu-iotests/223.out
+++ b/tests/qemu-iotests/223.out
@@ -59,6 +59,7 @@ exports available: 2
   min block: 1
   opt block: 4096
   max block: 33554432
+  init state: 0x0 ( )
   available meta contexts: 2
    base:allocation
    qemu:dirty-bitmap:b
@@ -69,6 +70,7 @@ exports available: 2
   min block: 1
   opt block: 4096
   max block: 33554432
+  init state: 0x0 ( )
   available meta contexts: 2
    base:allocation
    qemu:dirty-bitmap:b2
@@ -140,6 +142,7 @@ exports available: 2
   min block: 1
   opt block: 4096
   max block: 33554432
+  init state: 0x0 ( )
   available meta contexts: 2
    base:allocation
    qemu:dirty-bitmap:b
@@ -150,6 +153,7 @@ exports available: 2
   min block: 1
   opt block: 4096
   max block: 33554432
+  init state: 0x0 ( )
   available meta contexts: 2
    base:allocation
    qemu:dirty-bitmap:b2
diff --git a/tests/qemu-iotests/233.out b/tests/qemu-iotests/233.out
index c3c344811b2b..5be30d6b7c9c 100644
--- a/tests/qemu-iotests/233.out
+++ b/tests/qemu-iotests/233.out
@@ -43,6 +43,7 @@ exports available: 1
   min block: 1
   opt block: 4096
   max block: 33554432
+  init state: 0x0 ( )
   available meta contexts: 1
    base:allocation

-- 
2.24.1