Re: vhost changes (batched) in linux-next after 12/13 trigger random crashes in KVM guests after reboot

[eperezma@redhat.com]

Hi Christian.

Could you try this patch on top of ("38ced0208491 vhost: use batched version by default")?

It will not solve your first random crash but it should help with the lost of network connectivity.

Please let me know how does it goes.

Thanks!

From 99f0f543f3939dbe803988c9153a95616ccccacd Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Eugenio=20P=C3=A9rez?= <eperezma@redhat.com>
Date: Thu, 6 Feb 2020 15:13:42 +0100
Subject: [PATCH] vhost: filter valid vhost descriptors flags
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

Previous commit copy _NEXT flag, and it complains if a copied descriptor
contains it.

Signed-off-by: Eugenio Pérez <eperezma@redhat.com>
---
 drivers/vhost/vhost.c | 6 ++++--
 1 file changed, 4 insertions(+), 2 deletions(-)

diff --git a/drivers/vhost/vhost.c b/drivers/vhost/vhost.c
index 27ae5b4872a0..56c5253056ee 100644
--- a/drivers/vhost/vhost.c
+++ b/drivers/vhost/vhost.c
@@ -2125,6 +2125,8 @@ static void pop_split_desc(struct vhost_virtqueue *vq)
 	--vq->ndescs;
 }
 
+#define VHOST_DESC_FLAGS (VRING_DESC_F_INDIRECT | VRING_DESC_F_WRITE | \
+			  VRING_DESC_F_NEXT)
 static int push_split_desc(struct vhost_virtqueue *vq, struct vring_desc *desc, u16 id)
 {
 	struct vhost_desc *h;
@@ -2134,7 +2136,7 @@ static int push_split_desc(struct vhost_virtqueue *vq, struct vring_desc *desc,
 	h = &vq->descs[vq->ndescs++];
 	h->addr = vhost64_to_cpu(vq, desc->addr);
 	h->len = vhost32_to_cpu(vq, desc->len);
-	h->flags = vhost16_to_cpu(vq, desc->flags);
+	h->flags = vhost16_to_cpu(vq, desc->flags) & VHOST_DESC_FLAGS;
 	h->id = id;
 
 	return 0;
@@ -2343,7 +2345,7 @@ int vhost_get_vq_desc(struct vhost_virtqueue *vq,
 		struct vhost_desc *desc = &vq->descs[i];
 		int access;
 
-		if (desc->flags & ~(VRING_DESC_F_INDIRECT | VRING_DESC_F_WRITE)) {
+		if (desc->flags & ~VHOST_DESC_FLAGS) {
 			vq_err(vq, "Unexpected flags: 0x%x at descriptor id 0x%x\n",
 			       desc->flags, desc->id);
 			ret = -EINVAL;
-- 
2.18.1


On Wed, 2020-01-22 at 20:32 +0100, Christian Borntraeger wrote:
> 
> On 20.01.20 07:27, Michael S. Tsirkin wrote:
> > On Tue, Jan 07, 2020 at 01:16:50PM +0100, Christian Borntraeger wrote:
> > > On 07.01.20 12:55, Michael S. Tsirkin wrote:
> > > 
> > > > I pushed batched-v3 - same head but bisect should work now.
> > > > 
> > > 
> > > With 
> > > commit 38ced0208491103b50f1056f0d1c8f28e2e13d08 (HEAD)
> > > Author:     Michael S. Tsirkin <mst@redhat.com>
> > > AuthorDate: Wed Dec 11 12:19:26 2019 -0500
> > > Commit:     Michael S. Tsirkin <mst@redhat.com>
> > > CommitDate: Tue Jan 7 06:52:42 2020 -0500
> > > 
> > >     vhost: use batched version by default
> > > 
> > > 
> > > I have exactly one successful ping and then the network inside the guest is broken (no packet
> > > anymore).
> > 
> > Does anything appear in host's dmesg when this happens?
> 
> I think there was nothing, but I am not sure. I would need to redo the test if this is important to know.
> 
> > 
> > > So you could consider this commit broken (but in a different way and also without any
> > > guest reboot necessary).
> > > 
> > > 
> > > bisect log:
> > > git bisect start
> > > # bad: [d2f6175f52062ee51ee69754a6925608213475d2] vhost: use vhost_desc instead of vhost_log
> > > git bisect bad d2f6175f52062ee51ee69754a6925608213475d2
> > > # good: [d1281e3a562ec6a08f944a876481dd043ba739b9] virtio-blk: remove VIRTIO_BLK_F_SCSI support
> > > git bisect good d1281e3a562ec6a08f944a876481dd043ba739b9
> > > # good: [fac7c0f46996e32d996f5c46121df24a6b95ec3b] vhost: option to fetch descriptors through an independent
> > > struct
> > > git bisect good fac7c0f46996e32d996f5c46121df24a6b95ec3b
> > > # bad: [539eb9d738f048cd7be61f404e8f9c7d9d2ff3cc] vhost: batching fetches
> > > git bisect bad 539eb9d738f048cd7be61f404e8f9c7d9d2ff3cc