[<prev] [next>] [day] [month] [year] [list]
Message-ID: <20200712102743-mutt-send-email-mst@kernel.org>
Date: Sun, 12 Jul 2020 10:55:03 -0400
From: "Michael S. Tsirkin" <mst@...hat.com>
To: netdev@...r.kernel.org, dwoods@...lanox.com, lsun@...lanox.com
Cc: Darren Hart <dvhart@...radead.org>,
Vadim Pasternak <vadimp@...lanox.com>,
platform-driver-x86@...r.kernel.org, linux-kernel@...r.kernel.org,
Jason Wang <jasowang@...hat.com>
Subject: [PATCH] mlxbf-tmfifo: endian-ness wrappers for config access
Hello!
While working out some endian-ness issues with virtio,
I noticed that the following driver:
drivers/platform/mellanox/mlxbf-tmfifo.c
seems to access virtio net config directly.
E.g.:
if (ntohs(hdr.len) > config->mtu +
MLXBF_TMFIFO_NET_L2_OVERHEAD)
return;
This is not incorrect in that the specific device is at the
moment always legacy (no virtio 1).
However this throws up sparse warnings as the structure
is shared with modern devices which need the tagging for
correct virtio 1 endian-ness.
Using correct conversions will also allow virtio 1 support
in this driver down the road.
I'd like to merge the following patch. It's on top of
a branch config-endian in my tree which includes
the endian-ness tagging.
Would appreciate acks from relevant maintainers.
I also note that the console config field seems to be unused.
Would appreciate a confirmation.
Thanks!
--
mlxbf-tmfifo: endian-ness wrappers for config access
Signed-off-by: Michael S. Tsirkin <mst@...hat.com>
-->
diff --git a/drivers/platform/mellanox/mlxbf-tmfifo.c b/drivers/platform/mellanox/mlxbf-tmfifo.c
index 5739a9669b29..b1484206429f 100644
--- a/drivers/platform/mellanox/mlxbf-tmfifo.c
+++ b/drivers/platform/mellanox/mlxbf-tmfifo.c
@@ -608,6 +608,7 @@ static void mlxbf_tmfifo_rxtx_header(struct mlxbf_tmfifo_vring *vring,
{
struct mlxbf_tmfifo *fifo = vring->fifo;
struct virtio_net_config *config;
+ struct virtio_device *vdev;
struct mlxbf_tmfifo_msg_hdr hdr;
int vdev_id, hdr_len;
@@ -625,7 +626,8 @@ static void mlxbf_tmfifo_rxtx_header(struct mlxbf_tmfifo_vring *vring,
vdev_id = VIRTIO_ID_NET;
hdr_len = sizeof(struct virtio_net_hdr);
config = &fifo->vdev[vdev_id]->config.net;
- if (ntohs(hdr.len) > config->mtu +
+ vdev = &fifo->vdev[vdev_id]->vdev;
+ if (ntohs(hdr.len) > virtio16_to_cpu(vdev, config->mtu) +
MLXBF_TMFIFO_NET_L2_OVERHEAD)
return;
} else {
@@ -1231,8 +1233,14 @@ static int mlxbf_tmfifo_probe(struct platform_device *pdev)
/* Create the network vdev. */
memset(&net_config, 0, sizeof(net_config));
- net_config.mtu = ETH_DATA_LEN;
- net_config.status = VIRTIO_NET_S_LINK_UP;
+
+#defined MLXBF_TMFIFO_LITTLE_ENDIAN (virtio_legacy_is_little_endian() ||
+ (MLXBF_TMFIFO_NET_FEATURES & (1ULL << VIRTIO_F_VERSION_1)))
+
+ net_config.mtu = __cpu_to_virtio16(MLXBF_TMFIFO_LITTLE_ENDIAN,
+ ETH_DATA_LEN);
+ net_config.status = __cpu_to_virtio16(MLXBF_TMFIFO_LITTLE_ENDIAN,
+ VIRTIO_NET_S_LINK_UP);
mlxbf_tmfifo_get_cfg_mac(net_config.mac);
rc = mlxbf_tmfifo_create_vdev(dev, fifo, VIRTIO_ID_NET,
MLXBF_TMFIFO_NET_FEATURES, &net_config,
--
MST
Powered by blists - more mailing lists