[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-ID: <CALDO+SZU6Nw3e_2OPhzvjr0d4jgm5x9s9gaDYrU92hq7htqdYw@mail.gmail.com>
Date: Thu, 30 May 2019 10:29:53 -0700
From: William Tu <u9012063@...il.com>
To: Gregory Rose <gvrose8192@...il.com>
Cc: Linux Kernel Network Developers <netdev@...r.kernel.org>
Subject: Re: [PATCHv2 net] net: ip6_gre: access skb data after skb_cow_head()
Please ignore this patch and allow me to investigate more.
On Thu, May 30, 2019 at 10:23 AM Gregory Rose <gvrose8192@...il.com> wrote:
>
>
> On 5/30/2019 9:59 AM, William Tu wrote:
> > When increases the headroom, skb's data pointer might get re-allocated.
> > As a result, the skb->data before the skb_cow_head becomes a dangling pointer,
> > and dereferences to daddr causes general protection fault at the following
> > line in __gre6_xmit():
> >
> > if (dev->header_ops && dev->type == ARPHRD_IP6GRE)
> > fl6->daddr = ((struct ipv6hdr *)skb->data)->daddr;
> >
Look again Dave's comment that
The fl6->daddr assignments are object copies, not pointer assignments.
So this shouldn't cause any problem after skb_cow_head.
I will work on the right fix.
Thanks,
William
> > general protection fault: 0000 [#1] SMP PTI
> > OE 4.15.0-43-generic #146-Ubuntu
> > Hardware name: VMware, Inc. VMware Virtual Platform 440BX Desktop Reference
> > Platform, BIOS 6.00 07/03/2018
> > RIP: 0010: __gre6_xmit+0x11f/0x2c0 [openvswitch]
> > RSP: 0018:ffffb8d5c44df6a8 EFLAGS: 00010286
> > RAX: 00000000ffffffea RBX: ffff8b1528a0000 RCX: 0000000000000036
> > RDX: ffff000000000000 RSI: 0000000000000000 RDI: ffff8db267829200
> > RBP: ffffb8d5c44df 700 R08: 0000000000005865 RĂ˜9: ffffb8d5c44df724
> > R10: 0000000000000002 R11: 0000000000000000 R12: ffff8db267829200
> > R13: 0000000000000000 R14: ffffb8d5c44df 728 R15: 00000000ffffffff
> > FS: 00007f8744df 2700(0000) GS:ffff8db27fc0000000000) knlGS:0000000000000000
> > CS: 0910 DS: 0000 ES: 9000 CRO: 0000000080050033
> > CR2: 00007f893ef92148 CR3: 0000000400462003 CR4: 00000000001626f8
> > Call Trace:
> > ip6gre_tunnel_xmit+0x1cc/0x530 [openvswitch]
> > ? skb_clone+0x58/0xc0
> > __ip6gre_tunnel_xmit+0x12/0x20 [openvswitch]
> > ovs_vport_send +0xd4/0x170 [openvswitch]
> > do_output+0x53/0x160 [openvswitch]
> > do_execute_actions+0x9a1/0x1880 [openvswitch]
> >
> > Fix it by moving skb_cow_head before accessing the skb->data pointer.
> >
> > Fixes: 01b8d064d58b4 ("net: ip6_gre: Request headroom in __gre6_xmit()")
> > Reported-by: Haichao Ma <haichaom@...are.com>
> > Signed-off-by: William Tu <u9012063@...il.com>
> > ---
> > v1-v2: add more details in commit message.
> > ---
> > net/ipv6/ip6_gre.c | 6 +++---
> > 1 file changed, 3 insertions(+), 3 deletions(-)
> >
> > diff --git a/net/ipv6/ip6_gre.c b/net/ipv6/ip6_gre.c
> > index 655e46b227f9..90b2b129b105 100644
> > --- a/net/ipv6/ip6_gre.c
> > +++ b/net/ipv6/ip6_gre.c
> > @@ -714,6 +714,9 @@ static netdev_tx_t __gre6_xmit(struct sk_buff *skb,
> > struct ip6_tnl *tunnel = netdev_priv(dev);
> > __be16 protocol;
> >
> > + if (skb_cow_head(skb, dev->needed_headroom ?: tunnel->hlen))
> > + return -ENOMEM;
> > +
> > if (dev->type == ARPHRD_ETHER)
> > IPCB(skb)->flags = 0;
> >
> > @@ -722,9 +725,6 @@ static netdev_tx_t __gre6_xmit(struct sk_buff *skb,
> > else
> > fl6->daddr = tunnel->parms.raddr;
> >
> > - if (skb_cow_head(skb, dev->needed_headroom ?: tunnel->hlen))
> > - return -ENOMEM;
> > -
> > /* Push GRE header. */
> > protocol = (dev->type == ARPHRD_ETHER) ? htons(ETH_P_TEB) : proto;
> >
>
> Tested-by: Greg Rose <gvrose8192@...il.com>
> Reviewed-by: Greg Rose <gvrose8192@...il.com>
>
Powered by blists - more mailing lists