diff mbox

[v5,06/11] nbd: Minimal structured read for server

Message ID 20171019222637.17890-7-eblake@redhat.com (mailing list archive)
State New, archived
Headers show

Commit Message

Eric Blake Oct. 19, 2017, 10:26 p.m. UTC
From: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>

Minimal implementation of structured read: one structured reply chunk,
no segmentation.
Minimal structured error implementation: no text message.
Support DF flag, but just ignore it, as there is no segmentation any
way.

Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Signed-off-by: Eric Blake <eblake@redhat.com>

---
v5: correct DF flag spelling, include errname in trace, handle any bogus
payload from option
v4: better _DF flag handling, convert errno to wire format, add
comments and tracing, rework structured error for less churn when adding
text message later, don't kill connection on redundant client option
---
 nbd/server.c     | 106 +++++++++++++++++++++++++++++++++++++++++++++++++------
 nbd/trace-events |   2 ++
 2 files changed, 98 insertions(+), 10 deletions(-)

Comments

Vladimir Sementsov-Ogievskiy Oct. 20, 2017, 7:03 p.m. UTC | #1
20.10.2017 01:26, Eric Blake wrote:
> From: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
>
> Minimal implementation of structured read: one structured reply chunk,
> no segmentation.
> Minimal structured error implementation: no text message.
> Support DF flag, but just ignore it, as there is no segmentation any
> way.
>
> Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
> Signed-off-by: Eric Blake <eblake@redhat.com>
>
> ---
> v5: correct DF flag spelling, include errname in trace, handle any bogus
> payload from option
> v4: better _DF flag handling, convert errno to wire format, add
> comments and tracing, rework structured error for less churn when adding
> text message later, don't kill connection on redundant client option
> ---
>   nbd/server.c     | 106 +++++++++++++++++++++++++++++++++++++++++++++++++------
>   nbd/trace-events |   2 ++
>   2 files changed, 98 insertions(+), 10 deletions(-)
>
> diff --git a/nbd/server.c b/nbd/server.c
> index b3f7e0b18e..9be93c4a52 100644
> --- a/nbd/server.c
> +++ b/nbd/server.c
> @@ -100,6 +100,8 @@ struct NBDClient {
>       QTAILQ_ENTRY(NBDClient) next;
>       int nb_requests;
>       bool closing;
> +
> +    bool structured_reply;
>   };
>
>   /* That's all folks */
> @@ -754,6 +756,22 @@ static int nbd_negotiate_options(NBDClient *client, uint16_t myflags,
>                                                        "TLS not configured");
>                   }
>                   break;
> +
> +            case NBD_OPT_STRUCTURED_REPLY:
> +                if (length) {
> +                    ret = nbd_check_zero_length(client, length, option, errp);

here same thing like in previous patch, about success in 
nbd_check_zero_length

> +                } else if (client->structured_reply) {
> +                    ret = nbd_negotiate_send_rep_err(
> +                        client->ioc, NBD_REP_ERR_INVALID, option, errp,
> +                        "structured reply already negotiated");
> +                } else {
> +                    ret = nbd_negotiate_send_rep(client->ioc, NBD_REP_ACK,
> +                                                 option, errp);
> +                }

you've dropped "if (ret < 0) { return ret }", but the following two 
lines should not be
executed if ret < 0.. May be it doesn't matter (we will abort connection 
anyway after switch {}) but
it looks strange.

> +                client->structured_reply = true;
> +                myflags |= NBD_FLAG_SEND_DF;
> +                break;
> +
>               default:
>                   if (nbd_drop(client->ioc, length, errp) < 0) {
>                       return -EIO;
> @@ -1228,6 +1246,60 @@ static int nbd_co_send_simple_reply(NBDClient *client,
>       return nbd_co_send_iov(client, iov, len ? 2 : 1, errp);
>   }

[...]
Eric Blake Oct. 20, 2017, 7:11 p.m. UTC | #2
On 10/20/2017 02:03 PM, Vladimir Sementsov-Ogievskiy wrote:
> 20.10.2017 01:26, Eric Blake wrote:
>> From: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
>>
>> Minimal implementation of structured read: one structured reply chunk,
>> no segmentation.
>> Minimal structured error implementation: no text message.
>> Support DF flag, but just ignore it, as there is no segmentation any
>> way.
>>

>> +
>> +            case NBD_OPT_STRUCTURED_REPLY:
>> +                if (length) {
>> +                    ret = nbd_check_zero_length(client, length,
>> option, errp);
> 
> here same thing like in previous patch, about success in
> nbd_check_zero_length

Here, successfully answering the client with an error message should NOT
stop negotiating, so THIS return is correct.

> 
>> +                } else if (client->structured_reply) {
>> +                    ret = nbd_negotiate_send_rep_err(
>> +                        client->ioc, NBD_REP_ERR_INVALID, option, errp,
>> +                        "structured reply already negotiated");
>> +                } else {
>> +                    ret = nbd_negotiate_send_rep(client->ioc,
>> NBD_REP_ACK,
>> +                                                 option, errp);
>> +                }
> 
> you've dropped "if (ret < 0) { return ret }", but the following two
> lines should not be
> executed if ret < 0.. May be it doesn't matter (we will abort connection
> anyway after switch {}) but
> it looks strange.
> 
>> +                client->structured_reply = true;
>> +                myflags |= NBD_FLAG_SEND_DF;
>> +                break;

Indeed, these two lines are harmless due to the catch-all 'ret < 0'
after the switch at the tail end of the loop (which is why I dropped the
'if' here).
Vladimir Sementsov-Ogievskiy Oct. 20, 2017, 7:30 p.m. UTC | #3
20.10.2017 22:11, Eric Blake wrote:
> On 10/20/2017 02:03 PM, Vladimir Sementsov-Ogievskiy wrote:

>> 20.10.2017 01:26, Eric Blake wrote:

>>> From: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>

>>>

>>> Minimal implementation of structured read: one structured reply chunk,

>>> no segmentation.

>>> Minimal structured error implementation: no text message.

>>> Support DF flag, but just ignore it, as there is no segmentation any

>>> way.

>>>

>>> +

>>> +            case NBD_OPT_STRUCTURED_REPLY:

>>> +                if (length) {

>>> +                    ret = nbd_check_zero_length(client, length,

>>> option, errp);

>> here same thing like in previous patch, about success in

>> nbd_check_zero_length

> Here, successfully answering the client with an error message should NOT

> stop negotiating, so THIS return is correct.


ok, you are right

>

>>> +                } else if (client->structured_reply) {

>>> +                    ret = nbd_negotiate_send_rep_err(

>>> +                        client->ioc, NBD_REP_ERR_INVALID, option, errp,

>>> +                        "structured reply already negotiated");

>>> +                } else {

>>> +                    ret = nbd_negotiate_send_rep(client->ioc,

>>> NBD_REP_ACK,

>>> +                                                 option, errp);

>>> +                }

>> you've dropped "if (ret < 0) { return ret }", but the following two

>> lines should not be

>> executed if ret < 0.. May be it doesn't matter (we will abort connection

>> anyway after switch {}) but

>> it looks strange.

>>

>>> +                client->structured_reply = true;

>>> +                myflags |= NBD_FLAG_SEND_DF;

>>> +                break;

> Indeed, these two lines are harmless due to the catch-all 'ret < 0'

> after the switch at the tail end of the loop (which is why I dropped the

> 'if' here).

>

>


yes. but it looks strange) Anyway,
Reviewed-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>



-- 
Best regards,
Vladimir
Eric Blake Oct. 21, 2017, 4:02 p.m. UTC | #4
On 10/20/2017 02:30 PM, Vladimir Sementsov-Ogievskiy wrote:

>>>> +                } else if (client->structured_reply) {
>>>> +                    ret = nbd_negotiate_send_rep_err(
>>>> +                        client->ioc, NBD_REP_ERR_INVALID, option,
>>>> errp,
>>>> +                        "structured reply already negotiated");
>>>> +                } else {
>>>> +                    ret = nbd_negotiate_send_rep(client->ioc,
>>>> NBD_REP_ACK,
>>>> +                                                 option, errp);
>>>> +                }
>>> you've dropped "if (ret < 0) { return ret }", but the following two
>>> lines should not be
>>> executed if ret < 0.. May be it doesn't matter (we will abort connection
>>> anyway after switch {}) but
>>> it looks strange.
>>>
>>>> +                client->structured_reply = true;
>>>> +                myflags |= NBD_FLAG_SEND_DF;
>>>> +                break;
>> Indeed, these two lines are harmless due to the catch-all 'ret < 0'
>> after the switch at the tail end of the loop (which is why I dropped the
>> 'if' here).
>>
>>
> 
> yes. but it looks strange) Anyway,
> Reviewed-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
> 

On second thought, modifying client->structured_reply when we
successfully replied with an error is wrong.  So v6 will hoist the
modifications into the last else branch.
diff mbox

Patch

diff --git a/nbd/server.c b/nbd/server.c
index b3f7e0b18e..9be93c4a52 100644
--- a/nbd/server.c
+++ b/nbd/server.c
@@ -100,6 +100,8 @@  struct NBDClient {
     QTAILQ_ENTRY(NBDClient) next;
     int nb_requests;
     bool closing;
+
+    bool structured_reply;
 };

 /* That's all folks */
@@ -754,6 +756,22 @@  static int nbd_negotiate_options(NBDClient *client, uint16_t myflags,
                                                      "TLS not configured");
                 }
                 break;
+
+            case NBD_OPT_STRUCTURED_REPLY:
+                if (length) {
+                    ret = nbd_check_zero_length(client, length, option, errp);
+                } else if (client->structured_reply) {
+                    ret = nbd_negotiate_send_rep_err(
+                        client->ioc, NBD_REP_ERR_INVALID, option, errp,
+                        "structured reply already negotiated");
+                } else {
+                    ret = nbd_negotiate_send_rep(client->ioc, NBD_REP_ACK,
+                                                 option, errp);
+                }
+                client->structured_reply = true;
+                myflags |= NBD_FLAG_SEND_DF;
+                break;
+
             default:
                 if (nbd_drop(client->ioc, length, errp) < 0) {
                     return -EIO;
@@ -1228,6 +1246,60 @@  static int nbd_co_send_simple_reply(NBDClient *client,
     return nbd_co_send_iov(client, iov, len ? 2 : 1, errp);
 }

+static inline void set_be_chunk(NBDStructuredReplyChunk *chunk, uint16_t flags,
+                                uint16_t type, uint64_t handle, uint32_t length)
+{
+    stl_be_p(&chunk->magic, NBD_STRUCTURED_REPLY_MAGIC);
+    stw_be_p(&chunk->flags, flags);
+    stw_be_p(&chunk->type, type);
+    stq_be_p(&chunk->handle, handle);
+    stl_be_p(&chunk->length, length);
+}
+
+static int coroutine_fn nbd_co_send_structured_read(NBDClient *client,
+                                                    uint64_t handle,
+                                                    uint64_t offset,
+                                                    void *data,
+                                                    size_t size,
+                                                    Error **errp)
+{
+    NBDStructuredRead chunk;
+    struct iovec iov[] = {
+        {.iov_base = &chunk, .iov_len = sizeof(chunk)},
+        {.iov_base = data, .iov_len = size}
+    };
+
+    trace_nbd_co_send_structured_read(handle, offset, data, size);
+    set_be_chunk(&chunk.h, NBD_REPLY_FLAG_DONE, NBD_REPLY_TYPE_OFFSET_DATA,
+                 handle, sizeof(chunk) - sizeof(chunk.h) + size);
+    stq_be_p(&chunk.offset, offset);
+
+    return nbd_co_send_iov(client, iov, 2, errp);
+}
+
+static int coroutine_fn nbd_co_send_structured_error(NBDClient *client,
+                                                     uint64_t handle,
+                                                     uint32_t error,
+                                                     Error **errp)
+{
+    NBDStructuredError chunk;
+    int nbd_err = system_errno_to_nbd_errno(error);
+    struct iovec iov[] = {
+        {.iov_base = &chunk, .iov_len = sizeof(chunk)},
+        /* FIXME: Support human-readable error message */
+    };
+
+    assert(nbd_err);
+    trace_nbd_co_send_structured_error(handle, nbd_err,
+                                       nbd_err_lookup(nbd_err));
+    set_be_chunk(&chunk.h, NBD_REPLY_FLAG_DONE, NBD_REPLY_TYPE_ERROR, handle,
+                 sizeof(chunk) - sizeof(chunk.h));
+    stl_be_p(&chunk.error, nbd_err);
+    stw_be_p(&chunk.message_length, 0);
+
+    return nbd_co_send_iov(client, iov, 1, errp);
+}
+
 /* nbd_co_receive_request
  * Collect a client request. Return 0 if request looks valid, -EIO to drop
  * connection right away, and any other negative value to report an error to
@@ -1238,6 +1310,7 @@  static int nbd_co_receive_request(NBDRequestData *req, NBDRequest *request,
                                   Error **errp)
 {
     NBDClient *client = req->client;
+    int valid_flags;

     g_assert(qemu_in_coroutine());
     assert(client->recv_coroutine == qemu_coroutine_self());
@@ -1299,13 +1372,15 @@  static int nbd_co_receive_request(NBDRequestData *req, NBDRequest *request,
                    (uint64_t)client->exp->size);
         return request->type == NBD_CMD_WRITE ? -ENOSPC : -EINVAL;
     }
-    if (request->flags & ~(NBD_CMD_FLAG_FUA | NBD_CMD_FLAG_NO_HOLE)) {
-        error_setg(errp, "unsupported flags (got 0x%x)", request->flags);
-        return -EINVAL;
+    valid_flags = NBD_CMD_FLAG_FUA;
+    if (request->type == NBD_CMD_READ && client->structured_reply) {
+        valid_flags |= NBD_CMD_FLAG_DF;
+    } else if (request->type == NBD_CMD_WRITE_ZEROES) {
+        valid_flags |= NBD_CMD_FLAG_NO_HOLE;
     }
-    if (request->type != NBD_CMD_WRITE_ZEROES &&
-        (request->flags & NBD_CMD_FLAG_NO_HOLE)) {
-        error_setg(errp, "unexpected flags (got 0x%x)", request->flags);
+    if (request->flags & ~valid_flags) {
+        error_setg(errp, "unsupported flags for command %s (got 0x%x)",
+                   nbd_cmd_lookup(request->type), request->flags);
         return -EINVAL;
     }

@@ -1443,10 +1518,21 @@  reply:
         local_err = NULL;
     }

-    if (nbd_co_send_simple_reply(req->client, request.handle,
-                                 ret < 0 ? -ret : 0,
-                                 req->data, reply_data_len, &local_err) < 0)
-    {
+    if (client->structured_reply && request.type == NBD_CMD_READ) {
+        if (ret < 0) {
+            ret = nbd_co_send_structured_error(req->client, request.handle,
+                                               -ret, &local_err);
+        } else {
+            ret = nbd_co_send_structured_read(req->client, request.handle,
+                                              request.from, req->data,
+                                              reply_data_len, &local_err);
+        }
+    } else {
+        ret = nbd_co_send_simple_reply(req->client, request.handle,
+                                       ret < 0 ? -ret : 0,
+                                       req->data, reply_data_len, &local_err);
+    }
+    if (ret < 0) {
         error_prepend(&local_err, "Failed to send reply: ");
         goto disconnect;
     }
diff --git a/nbd/trace-events b/nbd/trace-events
index ab3d7dad4f..6894f8bbb4 100644
--- a/nbd/trace-events
+++ b/nbd/trace-events
@@ -56,6 +56,8 @@  nbd_receive_request(uint32_t magic, uint16_t flags, uint16_t type, uint64_t from
 nbd_blk_aio_attached(const char *name, void *ctx) "Export %s: Attaching clients to AIO context %p\n"
 nbd_blk_aio_detach(const char *name, void *ctx) "Export %s: Detaching clients from AIO context %p\n"
 nbd_co_send_simple_reply(uint64_t handle, uint32_t error, const char *errname, int len) "Send simple reply: handle = %" PRIu64 ", error = %" PRIu32 " (%s), len = %d"
+nbd_co_send_structured_read(uint64_t handle, uint64_t offset, void *data, size_t size) "Send structured read data reply: handle = %" PRIu64 ", offset = %" PRIu64 ", data = %p, len = %zu"
+nbd_co_send_structured_error(uint64_t handle, int err, const char *errname) "Send structured error reply: handle = %" PRIu64 ", error = %d (%s)"
 nbd_co_receive_request_decode_type(uint64_t handle, uint16_t type, const char *name) "Decoding type: handle = %" PRIu64 ", type = %" PRIu16 " (%s)"
 nbd_co_receive_request_payload_received(uint64_t handle, uint32_t len) "Payload received: handle = %" PRIu64 ", len = %" PRIu32
 nbd_co_receive_request_cmd_write(uint32_t len) "Reading %" PRIu32 " byte(s)"