Le vendredi 08 octobre 2021 à 06:02 -0400, Michael S. Tsirkin a écrit : > On Fri, Oct 08, 2021 at 10:06:57AM +0200, Greg KH wrote: > > On Fri, Oct 08, 2021 at 12:17:26AM +0800, Xuan Zhuo wrote: > > > On Thu, 7 Oct 2021 17:25:02 +0200, Greg KH < > > > gregkh@xxxxxxxxxxxxxxxxxxx> wrote: > > > > On Thu, Oct 07, 2021 at 11:06:12PM +0800, Xuan Zhuo wrote: > > > > > On Thu, 07 Oct 2021 14:04:22 +0200, Corentin Noël < > > > > > corentin.noel@xxxxxxxxxxxxx> wrote: > > > > > > I've been experiencing crashes with 5.14-rc1 and above that > > > > > > do not > > > > > > occur with 5.13, > > > > > > > > > > I should have fixed this problem before. I don't know why, I > > > > > just looked at the > > > > > latest net code, and this commit seems to be lost. > > > > > > > > > > 1a8024239dacf53fcf39c0f07fbf2712af22864f virtio-net: fix > > > > > for skb_over_panic inside big mode > > > > > > > > > > Can you test this patch again? > > > > > > > > That commit showed up in 5.13-rc5, so 5.14-rc1 and 5.13 should > > > > have had > > > > it in it, right? > > > > > > > > > > Yes, it may be lost due to conflicts during a certain merge. > > > > Really? I tried to apply that again to 5.14 and it did not > > work. So I > > do not understand what to do here, can you try to explain it > > better? > > > > thanks, > > > > greg k-h > > Hmm, something like the following perhaps then? > Corentin would you like to try this? > Warning: untested. > > > diff --git a/drivers/net/virtio_net.c b/drivers/net/virtio_net.c > index 096c2ac6b7a6..18dd9f6d107d 100644 > --- a/drivers/net/virtio_net.c > +++ b/drivers/net/virtio_net.c > @@ -406,12 +406,13 @@ static struct sk_buff *page_to_skb(struct > virtnet_info *vi, > * add_recvbuf_mergeable() + get_mergeable_buf_len() > */ > truesize = headroom ? PAGE_SIZE : truesize; > - tailroom = truesize - len - headroom; > + tailroom = truesize - headroom; > buf = p - headroom; > > len -= hdr_len; > offset += hdr_padded_len; > p += hdr_padded_len; > + tailroom -= hdr_padded_len + len; > > shinfo_size = SKB_DATA_ALIGN(sizeof(struct skb_shared_info)); > > Thank you for the patch, I started bisecting the issue but your patch actually makes it work again. Regards, Corentin
