Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]

Groups > linux.kernel > #1259146 > unrolled thread

[PATCH v4 1/6] virtio-net: Stop doing DMA from the stack

Back to article view | Back to linux.kernel

This discussion starts older than the indexed window; earlier articles aren't shown. The article labeled Started by below is the oldest one visible, not the original post.

Contents

  [PATCH v4 1/6] virtio-net: Stop doing DMA from the stack Andy Lutomirski <luto@kernel.org> - 2015-10-30 02:20 +0100
    Re: [PATCH v4 1/6] virtio-net: Stop doing DMA from the stack Christian Borntraeger <borntraeger@de.ibm.com> - 2015-10-30 15:00 +0100
      Re: [PATCH v4 1/6] virtio-net: Stop doing DMA from the stack Andy Lutomirski <luto@amacapital.net> - 2015-10-31 06:10 +0100

#1259146 — [PATCH v4 1/6] virtio-net: Stop doing DMA from the stack

From: "Michael S. Tsirkin" <mst@redhat.com>

Once virtio starts using the DMA API, we won't be able to safely DMA
from the stack.  virtio-net does a couple of config DMA requests
from small stack buffers -- switch to using dynamically-allocated
memory.

This should have no effect on any performance-critical code paths.

[I wrote the subject and commit message.  mst wrote the code. --luto]

Signed-off-by: Andy Lutomirski <luto@kernel.org>
signed-off-by: Michael S. Tsirkin <mst@redhat.com>
---
 drivers/net/virtio_net.c | 34 +++++++++++++++++++---------------
 1 file changed, 19 insertions(+), 15 deletions(-)

diff --git a/drivers/net/virtio_net.c b/drivers/net/virtio_net.c
index d8838dedb7a4..f94ab786088f 100644
--- a/drivers/net/virtio_net.c
+++ b/drivers/net/virtio_net.c
@@ -140,6 +140,12 @@ struct virtnet_info {
 
 	/* CPU hot plug notifier */
 	struct notifier_block nb;
+
+	/* Control VQ buffers: protected by the rtnl lock */
+	struct virtio_net_ctrl_hdr ctrl_hdr;
+	virtio_net_ctrl_ack ctrl_status;
+	u8 ctrl_promisc;
+	u8 ctrl_allmulti;
 };
 
 struct padded_vnet_hdr {
@@ -976,31 +982,30 @@ static bool virtnet_send_command(struct virtnet_info *vi, u8 class, u8 cmd,
 				 struct scatterlist *out)
 {
 	struct scatterlist *sgs[4], hdr, stat;
-	struct virtio_net_ctrl_hdr ctrl;
-	virtio_net_ctrl_ack status = ~0;
 	unsigned out_num = 0, tmp;
 
 	/* Caller should know better */
 	BUG_ON(!virtio_has_feature(vi->vdev, VIRTIO_NET_F_CTRL_VQ));
 
-	ctrl.class = class;
-	ctrl.cmd = cmd;
+	vi->ctrl_status = ~0;
+	vi->ctrl_hdr.class = class;
+	vi->ctrl_hdr.cmd = cmd;
 	/* Add header */
-	sg_init_one(&hdr, &ctrl, sizeof(ctrl));
+	sg_init_one(&hdr, &vi->ctrl_hdr, sizeof(vi->ctrl_hdr));
 	sgs[out_num++] = &hdr;
 
 	if (out)
 		sgs[out_num++] = out;
 
 	/* Add return status. */
-	sg_init_one(&stat, &status, sizeof(status));
+	sg_init_one(&stat, &vi->ctrl_status, sizeof(vi->ctrl_status));
 	sgs[out_num] = &stat;
 
 	BUG_ON(out_num + 1 > ARRAY_SIZE(sgs));
 	virtqueue_add_sgs(vi->cvq, sgs, out_num, 1, vi, GFP_ATOMIC);
 
 	if (unlikely(!virtqueue_kick(vi->cvq)))
-		return status == VIRTIO_NET_OK;
+		return vi->ctrl_status == VIRTIO_NET_OK;
 
 	/* Spin for a response, the kick causes an ioport write, trapping
 	 * into the hypervisor, so the request should be handled immediately.
@@ -1009,7 +1014,7 @@ static bool virtnet_send_command(struct virtnet_info *vi, u8 class, u8 cmd,
 	       !virtqueue_is_broken(vi->cvq))
 		cpu_relax();
 
-	return status == VIRTIO_NET_OK;
+	return vi->ctrl_status == VIRTIO_NET_OK;
 }
 
 static int virtnet_set_mac_address(struct net_device *dev, void *p)
@@ -1151,7 +1156,6 @@ static void virtnet_set_rx_mode(struct net_device *dev)
 {
 	struct virtnet_info *vi = netdev_priv(dev);
 	struct scatterlist sg[2];
-	u8 promisc, allmulti;
 	struct virtio_net_ctrl_mac *mac_data;
 	struct netdev_hw_addr *ha;
 	int uc_count;
@@ -1163,22 +1167,22 @@ static void virtnet_set_rx_mode(struct net_device *dev)
 	if (!virtio_has_feature(vi->vdev, VIRTIO_NET_F_CTRL_RX))
 		return;
 
-	promisc = ((dev->flags & IFF_PROMISC) != 0);
-	allmulti = ((dev->flags & IFF_ALLMULTI) != 0);
+	vi->ctrl_promisc = ((dev->flags & IFF_PROMISC) != 0);
+	vi->ctrl_allmulti = ((dev->flags & IFF_ALLMULTI) != 0);
 
-	sg_init_one(sg, &promisc, sizeof(promisc));
+	sg_init_one(sg, &vi->ctrl_promisc, sizeof(vi->ctrl_promisc));
 
 	if (!virtnet_send_command(vi, VIRTIO_NET_CTRL_RX,
 				  VIRTIO_NET_CTRL_RX_PROMISC, sg))
 		dev_warn(&dev->dev, "Failed to %sable promisc mode.\n",
-			 promisc ? "en" : "dis");
+			 vi->ctrl_promisc ? "en" : "dis");
 
-	sg_init_one(sg, &allmulti, sizeof(allmulti));
+	sg_init_one(sg, &vi->ctrl_allmulti, sizeof(vi->ctrl_allmulti));
 
 	if (!virtnet_send_command(vi, VIRTIO_NET_CTRL_RX,
 				  VIRTIO_NET_CTRL_RX_ALLMULTI, sg))
 		dev_warn(&dev->dev, "Failed to %sable allmulti mode.\n",
-			 allmulti ? "en" : "dis");
+			 vi->ctrl_allmulti ? "en" : "dis");
 
 	uc_count = netdev_uc_count(dev);
 	mc_count = netdev_mc_count(dev);
-- 
2.4.3

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/

[toc] | [next] | [standalone]

#1259510

Am 30.10.2015 um 02:09 schrieb Andy Lutomirski:
> From: "Michael S. Tsirkin" <mst@redhat.com>
> 
> Once virtio starts using the DMA API, we won't be able to safely DMA
> from the stack.  virtio-net does a couple of config DMA requests
> from small stack buffers -- switch to using dynamically-allocated
> memory.
> 
> This should have no effect on any performance-critical code paths.
> 
> [I wrote the subject and commit message.  mst wrote the code. --luto]
> 
> Signed-off-by: Andy Lutomirski <luto@kernel.org>
> signed-off-by: Michael S. Tsirkin <mst@redhat.com>

I still get an error when using multiqueue:

#  ethtool -L eth0 combined 4
[   33.534686] virtio_ccw 0.0.000d: DMA-API: device driver maps memory from stack [addr=00000000629e7c06]
[   33.534704] ------------[ cut here ]------------
[   33.534705] WARNING: at lib/dma-debug.c:1169
[   33.534706] Modules linked in: dm_multipath
[   33.534709] CPU: 1 PID: 1087 Comm: ethtool Not tainted 4.3.0-rc3+ #269
[   33.534710] task: 00000000616f9978 ti: 00000000629e4000 task.ti: 00000000629e4000
[   33.534712] Krnl PSW : 0704d00180000000 00000000005869d2 (check_for_stack+0xb2/0x118)
[   33.534716]            R:0 T:1 IO:1 EX:1 Key:0 M:1 W:0 P:0 AS:3 CC:1 PM:0 EA:3
Krnl GPRS: 000000000000006a 0000000000d60f44 000000000000005a 0000000064ee0870
[   33.534718]            00000000005869ce 0000000000000000 0000000000000001 00000000629e7c06
[   33.534719]            0000000000000000 0000000000000c06 0000000000000002 000000006467f800
[   33.534720]            0000000064673428 00000000629e7c06 00000000005869ce 00000000629e7928
[   33.534726] Krnl Code: 00000000005869c2: c0200024ad4e	larl	%r2,a1c45e
           00000000005869c8: c0e5ffe6d6fc	brasl	%r14,2617c0
          #00000000005869ce: a7f40001		brc	15,5869d0
          >00000000005869d2: c010003465eb	larl	%r1,c135a8
           00000000005869d8: e31010000012	lt	%r1,0(%r1)
           00000000005869de: a784000a		brc	8,5869f2
           00000000005869e2: e340f0b00004	lg	%r4,176(%r15)
           00000000005869e8: ebcff0a00004	lmg	%r12,%r15,160(%r15)
[   33.534736] Call Trace:
[   33.534737] ([<00000000005869ce>] check_for_stack+0xae/0x118)
[   33.534738]  [<0000000000586e3c>] debug_dma_map_page+0x114/0x160
[   33.534740]  [<00000000005a31f8>] vring_map_one_sg.isra.7+0x98/0xc0
[   33.534742]  [<00000000005a3b72>] virtqueue_add_sgs+0x1e2/0x788
[   33.534744]  [<0000000000618afc>] virtnet_send_command+0xcc/0x140
[   33.534745]  [<0000000000618c0c>] virtnet_set_queues+0x9c/0x110
[   33.534747]  [<0000000000619928>] virtnet_set_channels+0x78/0xe0
[   33.534748]  [<00000000006f63ea>] ethtool_set_channels+0x62/0x88
[   33.534750]  [<00000000006f8900>] dev_ethtool+0x10d8/0x1a48
[   33.534752]  [<000000000070c540>] dev_ioctl+0x190/0x510
[   33.534754]  [<00000000006cf2da>] sock_do_ioctl+0x7a/0x90
[   33.534755]  [<00000000006cf840>] sock_ioctl+0x1e8/0x2d0
[   33.534758]  [<00000000002e6c78>] do_vfs_ioctl+0x3a8/0x508
[   33.534759]  [<00000000002e6e7c>] SyS_ioctl+0xa4/0xb8
[   33.534762]  [<00000000008231ec>] system_call+0x244/0x264
[   33.534763]  [<000003ff922026d2>] 0x3ff922026d2
[   33.534764] Last Breaking-Event-Address:
[   33.534765]  [<00000000005869ce>] check_for_stack+0xae/0x118
[   33.534766] ---[ end trace 2379df65f4decfc4 ]---


> ---
>  drivers/net/virtio_net.c | 34 +++++++++++++++++++---------------
>  1 file changed, 19 insertions(+), 15 deletions(-)
> 
> diff --git a/drivers/net/virtio_net.c b/drivers/net/virtio_net.c
> index d8838dedb7a4..f94ab786088f 100644
> --- a/drivers/net/virtio_net.c
> +++ b/drivers/net/virtio_net.c
> @@ -140,6 +140,12 @@ struct virtnet_info {
> 
>  	/* CPU hot plug notifier */
>  	struct notifier_block nb;
> +
> +	/* Control VQ buffers: protected by the rtnl lock */
> +	struct virtio_net_ctrl_hdr ctrl_hdr;
> +	virtio_net_ctrl_ack ctrl_status;
> +	u8 ctrl_promisc;
> +	u8 ctrl_allmulti;
>  };
> 
>  struct padded_vnet_hdr {
> @@ -976,31 +982,30 @@ static bool virtnet_send_command(struct virtnet_info *vi, u8 class, u8 cmd,
>  				 struct scatterlist *out)
>  {
>  	struct scatterlist *sgs[4], hdr, stat;
> -	struct virtio_net_ctrl_hdr ctrl;
> -	virtio_net_ctrl_ack status = ~0;
>  	unsigned out_num = 0, tmp;
> 
>  	/* Caller should know better */
>  	BUG_ON(!virtio_has_feature(vi->vdev, VIRTIO_NET_F_CTRL_VQ));
> 
> -	ctrl.class = class;
> -	ctrl.cmd = cmd;
> +	vi->ctrl_status = ~0;
> +	vi->ctrl_hdr.class = class;
> +	vi->ctrl_hdr.cmd = cmd;
>  	/* Add header */
> -	sg_init_one(&hdr, &ctrl, sizeof(ctrl));
> +	sg_init_one(&hdr, &vi->ctrl_hdr, sizeof(vi->ctrl_hdr));
>  	sgs[out_num++] = &hdr;
> 
>  	if (out)
>  		sgs[out_num++] = out;
> 
>  	/* Add return status. */
> -	sg_init_one(&stat, &status, sizeof(status));
> +	sg_init_one(&stat, &vi->ctrl_status, sizeof(vi->ctrl_status));
>  	sgs[out_num] = &stat;
> 
>  	BUG_ON(out_num + 1 > ARRAY_SIZE(sgs));
>  	virtqueue_add_sgs(vi->cvq, sgs, out_num, 1, vi, GFP_ATOMIC);
> 
>  	if (unlikely(!virtqueue_kick(vi->cvq)))
> -		return status == VIRTIO_NET_OK;
> +		return vi->ctrl_status == VIRTIO_NET_OK;
> 
>  	/* Spin for a response, the kick causes an ioport write, trapping
>  	 * into the hypervisor, so the request should be handled immediately.
> @@ -1009,7 +1014,7 @@ static bool virtnet_send_command(struct virtnet_info *vi, u8 class, u8 cmd,
>  	       !virtqueue_is_broken(vi->cvq))
>  		cpu_relax();
> 
> -	return status == VIRTIO_NET_OK;
> +	return vi->ctrl_status == VIRTIO_NET_OK;
>  }
> 
>  static int virtnet_set_mac_address(struct net_device *dev, void *p)
> @@ -1151,7 +1156,6 @@ static void virtnet_set_rx_mode(struct net_device *dev)
>  {
>  	struct virtnet_info *vi = netdev_priv(dev);
>  	struct scatterlist sg[2];
> -	u8 promisc, allmulti;
>  	struct virtio_net_ctrl_mac *mac_data;
>  	struct netdev_hw_addr *ha;
>  	int uc_count;
> @@ -1163,22 +1167,22 @@ static void virtnet_set_rx_mode(struct net_device *dev)
>  	if (!virtio_has_feature(vi->vdev, VIRTIO_NET_F_CTRL_RX))
>  		return;
> 
> -	promisc = ((dev->flags & IFF_PROMISC) != 0);
> -	allmulti = ((dev->flags & IFF_ALLMULTI) != 0);
> +	vi->ctrl_promisc = ((dev->flags & IFF_PROMISC) != 0);
> +	vi->ctrl_allmulti = ((dev->flags & IFF_ALLMULTI) != 0);
> 
> -	sg_init_one(sg, &promisc, sizeof(promisc));
> +	sg_init_one(sg, &vi->ctrl_promisc, sizeof(vi->ctrl_promisc));
> 
>  	if (!virtnet_send_command(vi, VIRTIO_NET_CTRL_RX,
>  				  VIRTIO_NET_CTRL_RX_PROMISC, sg))
>  		dev_warn(&dev->dev, "Failed to %sable promisc mode.\n",
> -			 promisc ? "en" : "dis");
> +			 vi->ctrl_promisc ? "en" : "dis");
> 
> -	sg_init_one(sg, &allmulti, sizeof(allmulti));
> +	sg_init_one(sg, &vi->ctrl_allmulti, sizeof(vi->ctrl_allmulti));
> 
>  	if (!virtnet_send_command(vi, VIRTIO_NET_CTRL_RX,
>  				  VIRTIO_NET_CTRL_RX_ALLMULTI, sg))
>  		dev_warn(&dev->dev, "Failed to %sable allmulti mode.\n",
> -			 allmulti ? "en" : "dis");
> +			 vi->ctrl_allmulti ? "en" : "dis");
> 
>  	uc_count = netdev_uc_count(dev);
>  	mc_count = netdev_mc_count(dev);
> 

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/

[toc] | [prev] | [next] | [standalone]

#1259934

On Fri, Oct 30, 2015 at 6:55 AM, Christian Borntraeger
<borntraeger@de.ibm.com> wrote:
> Am 30.10.2015 um 02:09 schrieb Andy Lutomirski:
>> From: "Michael S. Tsirkin" <mst@redhat.com>
>>
>> Once virtio starts using the DMA API, we won't be able to safely DMA
>> from the stack.  virtio-net does a couple of config DMA requests
>> from small stack buffers -- switch to using dynamically-allocated
>> memory.
>>
>> This should have no effect on any performance-critical code paths.
>>
>> [I wrote the subject and commit message.  mst wrote the code. --luto]
>>
>> Signed-off-by: Andy Lutomirski <luto@kernel.org>
>> signed-off-by: Michael S. Tsirkin <mst@redhat.com>
>
> I still get an error when using multiqueue:
>
> #  ethtool -L eth0 combined 4
> [   33.534686] virtio_ccw 0.0.000d: DMA-API: device driver maps memory from stack [addr=00000000629e7c06]

Fixed in my branch, I think.

--Andy
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/

[toc] | [prev] | [standalone]

Back to top | Article view | linux.kernel

csiph-web