@@ -49,6 +49,7 @@ struct QIOChannelSocket {
socklen_t remoteAddrLen;
ssize_t zero_copy_queued;
ssize_t zero_copy_sent;
+ bool new_zero_copy_sent_success;
};
@@ -37,6 +37,11 @@
#define SOCKET_MAX_FDS 16
+#ifdef QEMU_MSG_ZEROCOPY
+static int qio_channel_socket_flush_internal(QIOChannel *ioc,
+ Error **errp);
+#endif
+
SocketAddress *
qio_channel_socket_get_local_address(QIOChannelSocket *ioc,
Error **errp)
@@ -65,6 +70,7 @@ qio_channel_socket_new(void)
sioc->fd = -1;
sioc->zero_copy_queued = 0;
sioc->zero_copy_sent = 0;
+ sioc->new_zero_copy_sent_success = FALSE;
ioc = QIO_CHANNEL(sioc);
qio_channel_set_feature(ioc, QIO_CHANNEL_FEATURE_SHUTDOWN);
@@ -566,6 +572,7 @@ static ssize_t qio_channel_socket_writev(QIOChannel *ioc,
size_t fdsize = sizeof(int) * nfds;
struct cmsghdr *cmsg;
int sflags = 0;
+ bool zero_copy_flush_pending = TRUE;
memset(control, 0, CMSG_SPACE(sizeof(int) * SOCKET_MAX_FDS));
@@ -612,9 +619,21 @@ static ssize_t qio_channel_socket_writev(QIOChannel *ioc,
goto retry;
case ENOBUFS:
if (flags & QIO_CHANNEL_WRITE_FLAG_ZERO_COPY) {
- error_setg_errno(errp, errno,
- "Process can't lock enough memory for using MSG_ZEROCOPY");
- return -1;
+ if (zero_copy_flush_pending) {
+ ret = qio_channel_socket_flush_internal(ioc, errp);
+ if (ret < 0) {
+ error_setg_errno(errp, errno,
+ "Zerocopy flush failed");
+ return -1;
+ }
+ zero_copy_flush_pending = FALSE;
+ goto retry;
+ } else {
+ error_setg_errno(errp, errno,
+ "Process can't lock enough memory for "
+ "using MSG_ZEROCOPY");
+ return -1;
+ }
}
break;
}
@@ -725,8 +744,8 @@ static ssize_t qio_channel_socket_writev(QIOChannel *ioc,
#ifdef QEMU_MSG_ZEROCOPY
-static int qio_channel_socket_flush(QIOChannel *ioc,
- Error **errp)
+static int qio_channel_socket_flush_internal(QIOChannel *ioc,
+ Error **errp)
{
QIOChannelSocket *sioc = QIO_CHANNEL_SOCKET(ioc);
struct msghdr msg = {};
@@ -791,15 +810,34 @@ static int qio_channel_socket_flush(QIOChannel *ioc,
/* No errors, count successfully finished sendmsg()*/
sioc->zero_copy_sent += serr->ee_data - serr->ee_info + 1;
- /* If any sendmsg() succeeded using zero copy, return 0 at the end */
+ /* If any sendmsg() succeeded using zero copy, mark zerocopy success */
if (serr->ee_code != SO_EE_CODE_ZEROCOPY_COPIED) {
- ret = 0;
+ sioc->new_zero_copy_sent_success = TRUE;
}
}
return ret;
}
+static int qio_channel_socket_flush(QIOChannel *ioc,
+ Error **errp)
+{
+ QIOChannelSocket *sioc = QIO_CHANNEL_SOCKET(ioc);
+ int ret;
+
+ ret = qio_channel_socket_flush_internal(ioc, errp);
+ if (ret < 0) {
+ return ret;
+ }
+
+ if (sioc->new_zero_copy_sent_success) {
+ sioc->new_zero_copy_sent_success = FALSE;
+ ret = 0;
+ }
+
+ return ret;
+}
+
#endif /* QEMU_MSG_ZEROCOPY */
static int
We allocate extra metadata SKBs in case of zerocopy send. This metadata memory is accounted for in the OPTMEM limit. If there is any error with sending zerocopy data or if zerocopy was skipped, these metadata SKBs are queued in the socket error queue. This error queue is freed when userspace reads it. Usually, if there are continuous failures, we merge the metadata into a single SKB and free another one. However, if there is any out-of-order processing or an intermittent zerocopy failures, this error chain can grow significantly, exhausting the OPTMEM limit. As a result, all new sendmsg requests fail to allocate any new SKB, leading to an ENOBUF error. To workaround this, if we encounter an ENOBUF error with a zerocopy sendmsg, we flush the error queue and retry once more. Signed-off-by: Manish Mishra <manish.mishra@nutanix.com> --- include/io/channel-socket.h | 1 + io/channel-socket.c | 52 ++++++++++++++++++++++++++++++++----- 2 files changed, 46 insertions(+), 7 deletions(-)