* BUG: vnet0 selects TX queue 11, but real number of TX queues is 11
@ 2019-01-30 2:16 George Amanakis
2019-01-30 2:52 ` Stanislav Fomichev
0 siblings, 1 reply; 3+ messages in thread
From: George Amanakis @ 2019-01-30 2:16 UTC (permalink / raw)
To: linux-kernel; +Cc: sdf, George Amanakis
Since 4.20.4 when running a KVM with vhost_net I am seeing in dmesg:
vnet0 selects TX queue 11, but real number of TX queues is 11
The corresponding part in the xml definition of the virtual machine is:
-------8<-------
<interface type='bridge'>
<mac address='xx:xx:xx:xx:xx:xx'/>
<source bridge='br0'/>
<model type='virtio'/>
<driver name='vhost' queues='12'>
</driver>
<address type='pci' domain='0x0000' bus='0x01' slot='0x00' function='0x0'/>
</interface>
-------8<-------
Doing a git-bisect with 4.20.3 last known good, and 4.20.4 as bad, this
commit turned up:
-------8<-------
commit 9ff0436e2c3575ffe64d359fb3b67aee237dc519
Author: Stanislav Fomichev <sdf@google.com>
Date: Mon Jan 7 13:38:38 2019 -0800
tun: publish tfile after it's fully initialized
[ Upstream commit 0b7959b6257322f7693b08a459c505d4938646f2 ]
BUG: unable to handle kernel NULL pointer dereference at
00000000000000d1
-------8<-------
Applying the following patch corrects it in 4.20.5. Would this be the
correct thing to do?
---
drivers/net/tun.c | 5 +++--
1 file changed, 3 insertions(+), 2 deletions(-)
diff --git a/drivers/net/tun.c b/drivers/net/tun.c
index 6658658246d2..e0dc004c6483 100644
--- a/drivers/net/tun.c
+++ b/drivers/net/tun.c
@@ -862,8 +862,6 @@ static int tun_attach(struct tun_struct *tun, struct file *file,
if (rtnl_dereference(tun->xdp_prog))
sock_set_flag(&tfile->sk, SOCK_XDP);
- tun_set_real_num_queues(tun);
-
/* device is allowed to go away first, so no need to hold extra
* refcnt.
*/
@@ -875,6 +873,9 @@ static int tun_attach(struct tun_struct *tun, struct file *file,
rcu_assign_pointer(tfile->tun, tun);
rcu_assign_pointer(tun->tfiles[tun->numqueues], tfile);
tun->numqueues++;
+
+ tun_set_real_num_queues(tun);
+
out:
return err;
}
--
2.20.1
^ permalink raw reply related [flat|nested] 3+ messages in thread* Re: BUG: vnet0 selects TX queue 11, but real number of TX queues is 11 2019-01-30 2:16 BUG: vnet0 selects TX queue 11, but real number of TX queues is 11 George Amanakis @ 2019-01-30 2:52 ` Stanislav Fomichev 2019-01-30 3:53 ` [PATCH] tun: move the call to tun_set_real_num_queues George Amanakis 0 siblings, 1 reply; 3+ messages in thread From: Stanislav Fomichev @ 2019-01-30 2:52 UTC (permalink / raw) To: George Amanakis; +Cc: linux-kernel, Netdev On Tue, Jan 29, 2019 at 6:16 PM George Amanakis <gamanakis@gmail.com> wrote: > > Since 4.20.4 when running a KVM with vhost_net I am seeing in dmesg: > vnet0 selects TX queue 11, but real number of TX queues is 11 > > The corresponding part in the xml definition of the virtual machine is: > -------8<------- > <interface type='bridge'> > <mac address='xx:xx:xx:xx:xx:xx'/> > <source bridge='br0'/> > <model type='virtio'/> > <driver name='vhost' queues='12'> > </driver> > <address type='pci' domain='0x0000' bus='0x01' slot='0x00' function='0x0'/> > </interface> > -------8<------- > > Doing a git-bisect with 4.20.3 last known good, and 4.20.4 as bad, this > commit turned up: > -------8<------- > commit 9ff0436e2c3575ffe64d359fb3b67aee237dc519 > Author: Stanislav Fomichev <sdf@google.com> > Date: Mon Jan 7 13:38:38 2019 -0800 > > tun: publish tfile after it's fully initialized > > [ Upstream commit 0b7959b6257322f7693b08a459c505d4938646f2 ] > > BUG: unable to handle kernel NULL pointer dereference at > 00000000000000d1 > -------8<------- > > > Applying the following patch corrects it in 4.20.5. Would this be the > correct thing to do? Ouch, tun_set_real_num_queues uses tun->numqueues internally :-( Your patch looks good to me, care to do a proper submission (with a Fixes tag)? I wonder whether we should use it as an opportunity to also do something like the following (to make it more explicit): diff --git a/drivers/net/tun.c b/drivers/net/tun.c index 18656c4094b3..ea9928b3b930 100644 --- a/drivers/net/tun.c +++ b/drivers/net/tun.c @@ -632,10 +632,10 @@ static inline bool tun_not_capable(struct tun_struct *tun) !ns_capable(net->user_ns, CAP_NET_ADMIN); } -static void tun_set_real_num_queues(struct tun_struct *tun) +static void tun_set_real_num_queues(struct tun_struct *tun, unsigned int nr) { - netif_set_real_num_tx_queues(tun->dev, tun->numqueues); - netif_set_real_num_rx_queues(tun->dev, tun->numqueues); + netif_set_real_num_tx_queues(tun->dev, nr); + netif_set_real_num_rx_queues(tun->dev, nr); } static void tun_disable_queue(struct tun_struct *tun, struct tun_file *tfile) @@ -712,7 +712,7 @@ static void __tun_detach(struct tun_file *tfile, bool clean) tun_flow_delete_by_queue(tun, tun->numqueues + 1); /* Drop read queue */ tun_queue_purge(tfile); - tun_set_real_num_queues(tun); + tun_set_real_num_queues(tun, tun->numqueues); } else if (tfile->detached && clean) { tun = tun_enable_queue(tfile); sock_put(&tfile->sk); @@ -866,8 +866,6 @@ static int tun_attach(struct tun_struct *tun, struct file *file, if (rtnl_dereference(tun->xdp_prog)) sock_set_flag(&tfile->sk, SOCK_XDP); - tun_set_real_num_queues(tun); - /* device is allowed to go away first, so no need to hold extra * refcnt. */ @@ -879,6 +877,7 @@ static int tun_attach(struct tun_struct *tun, struct file *file, rcu_assign_pointer(tfile->tun, tun); rcu_assign_pointer(tun->tfiles[tun->numqueues], tfile); tun->numqueues++; + tun_set_real_num_queues(tun, tun->numqueues); out: return err; } > --- > drivers/net/tun.c | 5 +++-- > 1 file changed, 3 insertions(+), 2 deletions(-) > > diff --git a/drivers/net/tun.c b/drivers/net/tun.c > index 6658658246d2..e0dc004c6483 100644 > --- a/drivers/net/tun.c > +++ b/drivers/net/tun.c > @@ -862,8 +862,6 @@ static int tun_attach(struct tun_struct *tun, struct file *file, > if (rtnl_dereference(tun->xdp_prog)) > sock_set_flag(&tfile->sk, SOCK_XDP); > > - tun_set_real_num_queues(tun); > - > /* device is allowed to go away first, so no need to hold extra > * refcnt. > */ > @@ -875,6 +873,9 @@ static int tun_attach(struct tun_struct *tun, struct file *file, > rcu_assign_pointer(tfile->tun, tun); > rcu_assign_pointer(tun->tfiles[tun->numqueues], tfile); > tun->numqueues++; > + > + tun_set_real_num_queues(tun); > + > out: > return err; > } > -- > 2.20.1 > ^ permalink raw reply related [flat|nested] 3+ messages in thread
* [PATCH] tun: move the call to tun_set_real_num_queues 2019-01-30 2:52 ` Stanislav Fomichev @ 2019-01-30 3:53 ` George Amanakis 0 siblings, 0 replies; 3+ messages in thread From: George Amanakis @ 2019-01-30 3:53 UTC (permalink / raw) To: davem; +Cc: linux-kernel, sdf, netdev, George Amanakis Call tun_set_real_num_queues() after the increment of tun->numqueues since the former depends on it. Otherwise, the number of queues is not correctly accounted for, which results to warnings similar to: "vnet0 selects TX queue 11, but real number of TX queues is 11". Fixes: 0b7959b62573 ("tun: publish tfile after it's fully initialized") Reported-and-tested-by: George Amanakis <gamanakis@gmail.com> Signed-off-by: George Amanakis <gamanakis@gmail.com> Signed-off-by: Stanislav Fomichev <sdf@google.com> --- drivers/net/tun.c | 3 +-- 1 file changed, 1 insertion(+), 2 deletions(-) diff --git a/drivers/net/tun.c b/drivers/net/tun.c index 18656c4094b3..fed298c0cb39 100644 --- a/drivers/net/tun.c +++ b/drivers/net/tun.c @@ -866,8 +866,6 @@ static int tun_attach(struct tun_struct *tun, struct file *file, if (rtnl_dereference(tun->xdp_prog)) sock_set_flag(&tfile->sk, SOCK_XDP); - tun_set_real_num_queues(tun); - /* device is allowed to go away first, so no need to hold extra * refcnt. */ @@ -879,6 +877,7 @@ static int tun_attach(struct tun_struct *tun, struct file *file, rcu_assign_pointer(tfile->tun, tun); rcu_assign_pointer(tun->tfiles[tun->numqueues], tfile); tun->numqueues++; + tun_set_real_num_queues(tun); out: return err; } -- 2.20.1 ^ permalink raw reply related [flat|nested] 3+ messages in thread
end of thread, other threads:[~2019-01-30 3:54 UTC | newest] Thread overview: 3+ messages (download: mbox.gz / follow: Atom feed) -- links below jump to the message on this page -- 2019-01-30 2:16 BUG: vnet0 selects TX queue 11, but real number of TX queues is 11 George Amanakis 2019-01-30 2:52 ` Stanislav Fomichev 2019-01-30 3:53 ` [PATCH] tun: move the call to tun_set_real_num_queues George Amanakis
This is a public inbox, see mirroring instructions for how to clone and mirror all data and code used for this inbox; as well as URLs for NNTP newsgroup(s).