On 16.08.2022 05:30, Bobby Eshleman wrote:

CC'ing virtio-dev@lists.oasis-open.org

On Mon, Aug 15, 2022 at 10:56:05AM -0700, Bobby Eshleman wrote:

quoted

This commit allows vsock implementations to return errors
to the socket layer other than -ENOMEM. One immediate effect
of this is that upon the sk_sndbuf threshold being reached -EAGAIN
will be returned and userspace may throttle appropriately.

Resultingly, a known issue with uperf is resolved[1].

Additionally, to preserve legacy behavior for non-virtio
implementations, hyperv/vmci force errors to be -ENOMEM so that behavior
is unchanged.

[1]: https://gitlab.com/vsock/vsock/-/issues/1

Signed-off-by: Bobby Eshleman <redacted>
---
 include/linux/virtio_vsock.h            | 3 +++
 net/vmw_vsock/af_vsock.c                | 3 ++-
 net/vmw_vsock/hyperv_transport.c        | 2 +-
 net/vmw_vsock/virtio_transport_common.c | 3 ---
 net/vmw_vsock/vmci_transport.c          | 9 ++++++++-
 5 files changed, 14 insertions(+), 6 deletions(-)

diff --git a/include/linux/virtio_vsock.h b/include/linux/virtio_vsock.h
index 17ed01466875..9a37eddbb87a 100644
--- a/include/linux/virtio_vsock.h
+++ b/include/linux/virtio_vsock.h

@@ -8,6 +8,9 @@
 #include <net/sock.h>
 #include <net/af_vsock.h>
 
+/* Threshold for detecting small packets to copy */
+#define GOOD_COPY_LEN  128
+
 enum virtio_vsock_metadata_flags {
 	VIRTIO_VSOCK_METADATA_FLAGS_REPLY		= BIT(0),
 	VIRTIO_VSOCK_METADATA_FLAGS_TAP_DELIVERED	= BIT(1),

diff --git a/net/vmw_vsock/af_vsock.c b/net/vmw_vsock/af_vsock.c
index e348b2d09eac..1893f8aafa48 100644
--- a/net/vmw_vsock/af_vsock.c
+++ b/net/vmw_vsock/af_vsock.c

@@ -1844,8 +1844,9 @@ static int vsock_connectible_sendmsg(struct socket *sock, struct msghdr *msg,
 			written = transport->stream_enqueue(vsk,
 					msg, len - total_written);
 		}
+
 		if (written < 0) {
-			err = -ENOMEM;
+			err = written;
 			goto out_err;
 		}

IIUC, for stream, this thing will have effect, only one first transport access fails. In this
case 'total_written' will be 0, so 'err' == 'written' will be returned. But when 'total_written > 0',
'err' will be overwritten by 'total_written' below, preserving current behaviour. Is it what You
supposed?

Thanks

quoted

diff --git a/net/vmw_vsock/hyperv_transport.c b/net/vmw_vsock/hyperv_transport.c
index fd98229e3db3..e99aea571f6f 100644
--- a/net/vmw_vsock/hyperv_transport.c
+++ b/net/vmw_vsock/hyperv_transport.c

@@ -687,7 +687,7 @@ static ssize_t hvs_stream_enqueue(struct vsock_sock *vsk, struct msghdr *msg,
 	if (bytes_written)
 		ret = bytes_written;
 	kfree(send_buf);
-	return ret;
+	return ret < 0 ? -ENOMEM : ret;
 }
 
 static s64 hvs_stream_has_data(struct vsock_sock *vsk)

diff --git a/net/vmw_vsock/virtio_transport_common.c b/net/vmw_vsock/virtio_transport_common.c
index 920578597bb9..d5780599fe93 100644
--- a/net/vmw_vsock/virtio_transport_common.c
+++ b/net/vmw_vsock/virtio_transport_common.c

@@ -23,9 +23,6 @@
 /* How long to wait for graceful shutdown of a connection */
 #define VSOCK_CLOSE_TIMEOUT (8 * HZ)
 
-/* Threshold for detecting small packets to copy */
-#define GOOD_COPY_LEN  128
-
 static const struct virtio_transport *
 virtio_transport_get_ops(struct vsock_sock *vsk)
 {

diff --git a/net/vmw_vsock/vmci_transport.c b/net/vmw_vsock/vmci_transport.c
index b14f0ed7427b..c927a90dc859 100644
--- a/net/vmw_vsock/vmci_transport.c
+++ b/net/vmw_vsock/vmci_transport.c

@@ -1838,7 +1838,14 @@ static ssize_t vmci_transport_stream_enqueue(
 	struct msghdr *msg,
 	size_t len)
 {
-	return vmci_qpair_enquev(vmci_trans(vsk)->qpair, msg, len, 0);
+	int err;
+
+	err = vmci_qpair_enquev(vmci_trans(vsk)->qpair, msg, len, 0);
+
+	if (err < 0)
+		err = -ENOMEM;
+
+	return err;
 }
 
 static s64 vmci_transport_stream_has_data(struct vsock_sock *vsk)

-- 
2.35.1

---------------------------------------------------------------------
To unsubscribe, e-mail: virtio-dev-unsubscribe@lists.oasis-open.org
For additional commands, e-mail: virtio-dev-help@lists.oasis-open.org