| lists.openwall.net | lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC | |
|
Open Source and information security mailing list archives
| ||
|
Message-ID: <Z79XXQjp9Dz7OYYQ@LQ3V64L9R2>
Date: Wed, 26 Feb 2025 13:03:09 -0500
From: Joe Damato <jdamato@...tly.com>
To: Jason Wang <jasowang@...hat.com>
Cc: netdev@...r.kernel.org, mkarsten@...terloo.ca,
gerhard@...leder-embedded.com, xuanzhuo@...ux.alibaba.com,
kuba@...nel.org, "Michael S. Tsirkin" <mst@...hat.com>,
Eugenio Pérez <eperezma@...hat.com>,
Andrew Lunn <andrew+netdev@...n.ch>,
"David S. Miller" <davem@...emloft.net>,
Eric Dumazet <edumazet@...gle.com>, Paolo Abeni <pabeni@...hat.com>,
"open list:VIRTIO CORE AND NET DRIVERS" <virtualization@...ts.linux.dev>,
open list <linux-kernel@...r.kernel.org>
Subject: Re: [PATCH net-next v4 3/4] virtio-net: Map NAPIs to queues
On Wed, Feb 26, 2025 at 01:48:50PM +0800, Jason Wang wrote:
> On Tue, Feb 25, 2025 at 10:05 AM Joe Damato <jdamato@...tly.com> wrote:
> >
> > Use netif_queue_set_napi to map NAPIs to queue IDs so that the mapping
> > can be accessed by user apps, taking care to hold RTNL as needed.
>
> I may miss something but I wonder whether letting the caller hold the
> lock is better.
Hmm...
Double checking all the paths over again, here's what I see:
- refill_work, delayed work that needs RTNL so this change seems
right?
- virtnet_disable_queue_pair, called from virtnet_open and
virtnet_close. When called via NDO these are safe and hold RTNL,
but they can be called from power management and need RTNL.
- virtnet_enable_queue_pair called from virtnet_open, safe when
used via NDO but needs RTNL when used via power management.
- virtnet_rx_pause called in both paths as you mentioned, one
which needs RTNL and one which doesn't.
I think there are a couple ways to fix this:
1. Edit this patch to remove the virtnet_queue_set_napi helper,
and call netif_queue_set_napi from the napi_enable and
napi_disable helpers directly. Modify code calling into these
paths to hold rtnl (or not) as described above.
2. Modify virtnet_enable_queue_pair, virtnet_disable_queue_pair,
and virtnet_rx_pause to take a "bool need_rtnl" as an a
function argument and pass that through.
I'm not sure which is cleaner and I do not have a preference.
Can you let me know which you prefer? I am happy to implement either
one for the next revision.
[...]
> > ---
> > drivers/net/virtio_net.c | 73 ++++++++++++++++++++++++++++------------
> > 1 file changed, 52 insertions(+), 21 deletions(-)
> >
> > diff --git a/drivers/net/virtio_net.c b/drivers/net/virtio_net.c
> > index e578885c1093..13bb4a563073 100644
> > --- a/drivers/net/virtio_net.c
> > +++ b/drivers/net/virtio_net.c
> > @@ -2807,6 +2807,20 @@ static void skb_recv_done(struct virtqueue *rvq)
> > virtqueue_napi_schedule(&rq->napi, rvq);
> > }
> >
> > +static void virtnet_queue_set_napi(struct net_device *dev,
> > + struct napi_struct *napi,
> > + enum netdev_queue_type q_type, int qidx,
> > + bool need_rtnl)
> > +{
> > + if (need_rtnl)
> > + rtnl_lock();
> > +
> > + netif_queue_set_napi(dev, qidx, q_type, napi);
> > +
> > + if (need_rtnl)
> > + rtnl_unlock();
> > +}
> > +
> > static void virtnet_napi_do_enable(struct virtqueue *vq,
> > struct napi_struct *napi)
> > {
> > @@ -2821,15 +2835,21 @@ static void virtnet_napi_do_enable(struct virtqueue *vq,
> > local_bh_enable();
> > }
> >
> > -static void virtnet_napi_enable(struct receive_queue *rq)
> > +static void virtnet_napi_enable(struct receive_queue *rq, bool need_rtnl)
> > {
> > + struct virtnet_info *vi = rq->vq->vdev->priv;
> > + int qidx = vq2rxq(rq->vq);
> > +
> > virtnet_napi_do_enable(rq->vq, &rq->napi);
> > + virtnet_queue_set_napi(vi->dev, &rq->napi,
> > + NETDEV_QUEUE_TYPE_RX, qidx, need_rtnl);
> > }
> >
> > -static void virtnet_napi_tx_enable(struct send_queue *sq)
> > +static void virtnet_napi_tx_enable(struct send_queue *sq, bool need_rtnl)
> > {
> > struct virtnet_info *vi = sq->vq->vdev->priv;
> > struct napi_struct *napi = &sq->napi;
> > + int qidx = vq2txq(sq->vq);
> >
> > if (!napi->weight)
> > return;
> > @@ -2843,20 +2863,31 @@ static void virtnet_napi_tx_enable(struct send_queue *sq)
> > }
> >
> > virtnet_napi_do_enable(sq->vq, napi);
> > + virtnet_queue_set_napi(vi->dev, napi, NETDEV_QUEUE_TYPE_TX, qidx,
> > + need_rtnl);
> > }
> >
> > -static void virtnet_napi_tx_disable(struct send_queue *sq)
> > +static void virtnet_napi_tx_disable(struct send_queue *sq, bool need_rtnl)
> > {
> > + struct virtnet_info *vi = sq->vq->vdev->priv;
> > struct napi_struct *napi = &sq->napi;
> > + int qidx = vq2txq(sq->vq);
> >
> > - if (napi->weight)
> > + if (napi->weight) {
> > + virtnet_queue_set_napi(vi->dev, NULL, NETDEV_QUEUE_TYPE_TX,
> > + qidx, need_rtnl);
> > napi_disable(napi);
> > + }
> > }
> >
> > -static void virtnet_napi_disable(struct receive_queue *rq)
> > +static void virtnet_napi_disable(struct receive_queue *rq, bool need_rtnl)
> > {
> > + struct virtnet_info *vi = rq->vq->vdev->priv;
> > struct napi_struct *napi = &rq->napi;
> > + int qidx = vq2rxq(rq->vq);
> >
> > + virtnet_queue_set_napi(vi->dev, NULL, NETDEV_QUEUE_TYPE_TX, qidx,
> > + need_rtnl);
> > napi_disable(napi);
> > }
> >
> > @@ -2870,9 +2901,9 @@ static void refill_work(struct work_struct *work)
> > for (i = 0; i < vi->curr_queue_pairs; i++) {
> > struct receive_queue *rq = &vi->rq[i];
> >
> > - virtnet_napi_disable(rq);
> > + virtnet_napi_disable(rq, true);
> > still_empty = !try_fill_recv(vi, rq, GFP_KERNEL);
> > - virtnet_napi_enable(rq);
> > + virtnet_napi_enable(rq, true);
> >
> > /* In theory, this can happen: if we don't get any buffers in
> > * we will *never* try to fill again.
> > @@ -3069,8 +3100,8 @@ static int virtnet_poll(struct napi_struct *napi, int budget)
> >
> > static void virtnet_disable_queue_pair(struct virtnet_info *vi, int qp_index)
> > {
> > - virtnet_napi_tx_disable(&vi->sq[qp_index]);
> > - virtnet_napi_disable(&vi->rq[qp_index]);
> > + virtnet_napi_tx_disable(&vi->sq[qp_index], false);
> > + virtnet_napi_disable(&vi->rq[qp_index], false);
> > xdp_rxq_info_unreg(&vi->rq[qp_index].xdp_rxq);
> > }
> >
> > @@ -3089,8 +3120,8 @@ static int virtnet_enable_queue_pair(struct virtnet_info *vi, int qp_index)
> > if (err < 0)
> > goto err_xdp_reg_mem_model;
> >
> > - virtnet_napi_enable(&vi->rq[qp_index]);
> > - virtnet_napi_tx_enable(&vi->sq[qp_index]);
> > + virtnet_napi_enable(&vi->rq[qp_index], false);
> > + virtnet_napi_tx_enable(&vi->sq[qp_index], false);
> >
> > return 0;
> >
> > @@ -3342,7 +3373,7 @@ static void virtnet_rx_pause(struct virtnet_info *vi, struct receive_queue *rq)
> > bool running = netif_running(vi->dev);
> >
> > if (running) {
> > - virtnet_napi_disable(rq);
> > + virtnet_napi_disable(rq, true);
>
> During the resize, the rtnl lock has been held on the ethtool path
>
> rtnl_lock();
> rc = __dev_ethtool(net, ifr, useraddr, ethcmd, state);
> rtnl_unlock();
>
> virtnet_rx_resize()
> virtnet_rx_pause()
>
> and in the case of XSK binding, I see ASSERT_RTNL in xp_assign_dev() at least.
Thanks for catching this. I re-read all the paths and I think I've
outlined a few other issues above.
Please let me know which of the proposed methods above you'd like me
to implement to get this merged.
Thanks.
---
pw-bot: cr
Powered by blists - more mailing lists