From: Christoph Paasch <christoph.paasch@gmail.com>
To: Vasily Averin <vvs@virtuozzo.com>
Cc: Eric Dumazet <eric.dumazet@gmail.com>,
"David S. Miller" <davem@davemloft.net>,
Hideaki YOSHIFUJI <yoshfuji@linux-ipv6.org>,
David Ahern <dsahern@kernel.org>,
Jakub Kicinski <kuba@kernel.org>, netdev <netdev@vger.kernel.org>,
LKML <linux-kernel@vger.kernel.org>,
kernel@openvz.org, Julian Wiedmann <jwi@linux.ibm.com>
Subject: Re: [PATCH NET-NEXT] ipv6: skb_expand_head() adjust skb->truesize incorrectly
Date: Wed, 25 Aug 2021 10:49:15 -0700 [thread overview]
Message-ID: <CALMXkpZYGC5HNkJAi4wCuawC-9CVNjN1LqO073YJvUF5ONwupA@mail.gmail.com> (raw)
In-Reply-To: <61da35a1-e609-02d5-609d-5228e184e43f@virtuozzo.com>
On Tue, Aug 24, 2021 at 10:22 AM Vasily Averin <vvs@virtuozzo.com> wrote:
>
> On 8/24/21 11:50 AM, Vasily Averin wrote:
> > On 8/24/21 1:23 AM, Eric Dumazet wrote:
> >> On 8/23/21 2:51 PM, Eric Dumazet wrote:
> >>> On 8/23/21 2:45 PM, Eric Dumazet wrote:
> >>>> On 8/23/21 10:25 AM, Christoph Paasch wrote:
> >>>>> Hello,
> >>>>>
> >>>>> On Mon, Aug 23, 2021 at 12:56 AM Vasily Averin <vvs@virtuozzo.com> wrote:
> >>>>>>
> >>>>>> Christoph Paasch reports [1] about incorrect skb->truesize
> >>>>>> after skb_expand_head() call in ip6_xmit.
> >>>>>> This happen because skb_set_owner_w() for newly clone skb is called
> >>>>>> too early, before pskb_expand_head() where truesize is adjusted for
> >>>>>> (!skb-sk) case.
> >>>>>>
> >>>>>> [1] https://lkml.org/lkml/2021/8/20/1082
> >>>>>>
> >>>>>> Reported-by: Christoph Paasch <christoph.paasch@gmail.com>
> >>>>>> Signed-off-by: Vasily Averin <vvs@virtuozzo.com>
> >>>>>> ---
> >>>>>> net/core/skbuff.c | 24 +++++++++++++-----------
> >>>>>> 1 file changed, 13 insertions(+), 11 deletions(-)
> >>>>>>
> >>>>>> diff --git a/net/core/skbuff.c b/net/core/skbuff.c
> >>>>>> index f931176..508d5c4 100644
> >>>>>> --- a/net/core/skbuff.c
> >>>>>> +++ b/net/core/skbuff.c
> >>>>>> @@ -1803,6 +1803,8 @@ struct sk_buff *skb_realloc_headroom(struct sk_buff *skb, unsigned int headroom)
> >>>>>>
> >>>>>> struct sk_buff *skb_expand_head(struct sk_buff *skb, unsigned int headroom)
> >>>>>> {
> >>>>>> + struct sk_buff *oskb = skb;
> >>>>>> + struct sk_buff *nskb = NULL;
> >>>>>> int delta = headroom - skb_headroom(skb);
> >>>>>>
> >>>>>> if (WARN_ONCE(delta <= 0,
> >>>>>> @@ -1811,21 +1813,21 @@ struct sk_buff *skb_expand_head(struct sk_buff *skb, unsigned int headroom)
> >>>>>>
> >>>>>> /* pskb_expand_head() might crash, if skb is shared */
> >>>>>> if (skb_shared(skb)) {
> >>>>>> - struct sk_buff *nskb = skb_clone(skb, GFP_ATOMIC);
> >>>>>> -
> >>>>>> - if (likely(nskb)) {
> >>>>>> - if (skb->sk)
> >>>>>> - skb_set_owner_w(nskb, skb->sk);
> >>>>>> - consume_skb(skb);
> >>>>>> - } else {
> >>>>>> - kfree_skb(skb);
> >>>>>> - }
> >>>>>> + nskb = skb_clone(skb, GFP_ATOMIC);
> >>>>>> skb = nskb;
> >>>>>> }
> >>>>>> if (skb &&
> >>>>>> - pskb_expand_head(skb, SKB_DATA_ALIGN(delta), 0, GFP_ATOMIC)) {
> >>>>>> - kfree_skb(skb);
> >>>>>> + pskb_expand_head(skb, SKB_DATA_ALIGN(delta), 0, GFP_ATOMIC))
> >>>>>> skb = NULL;
> >>>>>> +
> >>>>>> + if (!skb) {
> >>>>>> + kfree_skb(oskb);
> >>>>>> + if (nskb)
> >>>>>> + kfree_skb(nskb);
> >>>>>> + } else if (nskb) {
> >>>>>> + if (oskb->sk)
> >>>>>> + skb_set_owner_w(nskb, oskb->sk);
> >>>>>> + consume_skb(oskb);
> >>>>>
> >>>>> sorry, this does not fix the problem. The syzkaller repro still
> >>>>> triggers the WARN.
> >>>>>
> >>>>> When it happens, the skb in ip6_xmit() is not shared as it comes from
> >>>>> __tcp_transmit_skb, where it is skb_clone()'d.
> >>>>>
> >>>>>
> >>>>
> >>>> Old code (in skb_realloc_headroom())
> >>>> was first calling skb2 = skb_clone(skb, GFP_ATOMIC);
> >>>>
> >>>> At this point, skb2->sk was NULL
> >>>> So pskb_expand_head(skb2, SKB_DATA_ALIGN(delta), 0, ...) was able to tweak skb2->truesize
> >>>>
> >>>> I would try :
> >>>>
> >>>> diff --git a/net/core/skbuff.c b/net/core/skbuff.c
> >>>> index f9311762cc475bd38d87c33e988d7c983b902e56..326749a8938637b044a616cc33b6a19ed191ac41 100644
> >>>> --- a/net/core/skbuff.c
> >>>> +++ b/net/core/skbuff.c
> >>>> @@ -1804,6 +1804,7 @@ EXPORT_SYMBOL(skb_realloc_headroom);
> >>>> struct sk_buff *skb_expand_head(struct sk_buff *skb, unsigned int headroom)
> >>>> {
> >>>> int delta = headroom - skb_headroom(skb);
> >>>> + struct sk_buff *oskb = NULL;
> >>>>
> >>>> if (WARN_ONCE(delta <= 0,
> >>>> "%s is expecting an increase in the headroom", __func__))
> >>>> @@ -1813,19 +1814,21 @@ struct sk_buff *skb_expand_head(struct sk_buff *skb, unsigned int headroom)
> >>>> if (skb_shared(skb)) {
> >>>> struct sk_buff *nskb = skb_clone(skb, GFP_ATOMIC);
> >>>>
> >>>> - if (likely(nskb)) {
> >>>> - if (skb->sk)
> >>>> - skb_set_owner_w(nskb, skb->sk);
> >>>> - consume_skb(skb);
> >>>> - } else {
> >>>> + if (unlikely(!nskb)) {
> >>>> kfree_skb(skb);
> >>>> + return NULL;
> >>>> }
> >>>> + oskb = skb;
> >>>> skb = nskb;
> >>>> }
> >>>> - if (skb &&
> >>>> - pskb_expand_head(skb, SKB_DATA_ALIGN(delta), 0, GFP_ATOMIC)) {
> >>>> + if (pskb_expand_head(skb, SKB_DATA_ALIGN(delta), 0, GFP_ATOMIC)) {
> >>>> kfree_skb(skb);
> >>>> - skb = NULL;
> >>>> + kfree_skb(oskb);
> >>>> + return NULL;
> >>>> + }
> >>>> + if (oskb) {
> >>>> + skb_set_owner_w(skb, oskb->sk);
> >>>> + consume_skb(oskb);
> >>>> }
> >>>> return skb;
> >>>> }
> >>>
> >>>
> >>> Oh well, probably not going to work.
> >>>
> >>> We have to find a way to properly increase skb->truesize, even if skb_clone() is _not_ called.
> >
> > Can we adjust truesize outside pskb_expand_head()?
> > Could you please explain why it can be not safe?
>
> Do you mean truesize change should not break balance of sk->sk_wmem_alloc?
AFAICS, that's the problem around adjusting truesize. So, maybe "just"
refcount_add the increase of the truesize.
The below does fix the syzkaller bug for me and seems to do the right
thing overall. But I honestly think that this is becoming too hacky
and not worth it... and who knows what other corner-cases this now
exposes...
Maybe a revert is a better course of action?
---
diff --git a/net/core/skbuff.c b/net/core/skbuff.c
index f9311762cc47..9cc18a0fdd1c 100644
--- a/net/core/skbuff.c
+++ b/net/core/skbuff.c
@@ -71,6 +71,7 @@
#include <net/mpls.h>
#include <net/mptcp.h>
#include <net/page_pool.h>
+#include <net/tcp.h>
#include <linux/uaccess.h>
#include <trace/events/skb.h>
@@ -1756,9 +1757,14 @@ int pskb_expand_head(struct sk_buff *skb, int
nhead, int ntail,
* For the moment, we really care of rx path, or
* when skb is orphaned (not attached to a socket).
*/
- if (!skb->sk || skb->destructor == sock_edemux)
+ if (!skb->sk || skb->destructor == sock_edemux || skb->destructor ==
tcp_wfree) {
skb->truesize += size - osize;
+ if (skb->sk && skb->destructor == tcp_wfree) {
+ refcount_add(size - osize, &skb->sk->sk_wmem_alloc);
+ }
+ }
+
return 0;
nofrags:
next prev parent reply other threads:[~2021-08-25 17:49 UTC|newest]
Thread overview: 106+ messages / expand[flat|nested] mbox.gz Atom feed top
[not found] <cover.1625665132.git.vvs@virtuozzo.com>
2021-07-07 14:04 ` [PATCH IPV6 1/1] ipv6: allocate enough headroom in ip6_finish_output2() Vasily Averin
2021-07-07 14:45 ` David Ahern
2021-07-07 16:42 ` Jakub Kicinski
2021-07-07 17:41 ` Eric Dumazet
2021-07-07 17:53 ` Vasily Averin
2021-07-07 18:30 ` Jakub Kicinski
2021-07-07 18:50 ` Eric Dumazet
2021-07-09 9:04 ` [PATCH IPV6 v2 0/4] " Vasily Averin
2021-07-12 6:44 ` [PATCH IPV6 v3 0/1] " Vasily Averin
[not found] ` <cover.1626069562.git.vvs@virtuozzo.com>
2021-07-12 6:45 ` [PATCH IPV6 v3 1/1] " Vasily Averin
2021-07-12 18:30 ` patchwork-bot+netdevbpf
2021-07-13 7:46 ` Vasily Averin
2021-07-13 12:01 ` [PATCH NET v4 0/1] " Vasily Averin
[not found] ` <cover.1626177047.git.vvs@virtuozzo.com>
2021-07-13 12:01 ` [PATCH NET v4 1/1] " Vasily Averin
2021-07-18 10:44 ` Vasily Averin
2021-07-18 15:22 ` David Ahern
2021-07-18 17:04 ` David Miller
2021-07-19 7:55 ` [PATCH NET] ipv6: ip6_finish_output2: set sk into newly allocated nskb Vasily Averin
2021-07-20 10:10 ` patchwork-bot+netdevbpf
2021-07-13 12:31 ` [PATCH IPV6 v3 1/1] ipv6: allocate enough headroom in ip6_finish_output2() Vasily Averin
2021-07-12 13:26 ` [PATCH NET 0/7] skbuff: introduce pskb_realloc_headroom() Vasily Averin
[not found] ` <cover.1626093470.git.vvs@virtuozzo.com>
2021-07-12 13:26 ` [PATCH NET 1/7] " Vasily Averin
2021-07-12 17:53 ` Jakub Kicinski
2021-07-12 18:45 ` Vasily Averin
2021-07-13 20:57 ` [PATCH NET v2 0/7] skbuff: introduce skb_expand_head() Vasily Averin
2021-08-02 8:52 ` [PATCH NET v3 " Vasily Averin
[not found] ` <cover.1627891754.git.vvs@virtuozzo.com>
2021-08-02 8:52 ` [PATCH NET v3 1/7] " Vasily Averin
2021-08-02 8:52 ` [PATCH NET v3 2/7] ipv6: use skb_expand_head in ip6_finish_output2 Vasily Averin
2021-08-02 8:52 ` [PATCH NET v3 3/7] ipv6: use skb_expand_head in ip6_xmit Vasily Averin
2021-08-02 8:52 ` [PATCH NET v3 4/7] ipv4: use skb_expand_head in ip_finish_output2 Vasily Averin
2021-08-02 8:52 ` [PATCH NET v3 5/7] vrf: use skb_expand_head in vrf_finish_output Vasily Averin
2021-08-05 11:55 ` Julian Wiedmann
2021-08-05 12:55 ` Vasily Averin
2021-08-06 7:49 ` [PATCH NET v4 0/7] skbuff: introduce skb_expand_head() Vasily Averin
2021-08-06 10:14 ` David Miller
2021-08-06 12:53 ` [PATCH NET] vrf: fix null pointer dereference in vrf_finish_output() Vasily Averin
2021-08-06 22:42 ` Jakub Kicinski
2021-08-07 6:41 ` Vasily Averin
[not found] ` <cover.1628235065.git.vvs@virtuozzo.com>
2021-08-06 7:49 ` [PATCH NET v4 1/7] skbuff: introduce skb_expand_head() Vasily Averin
2021-08-06 7:50 ` [PATCH NET v4 2/7] ipv6: use skb_expand_head in ip6_finish_output2 Vasily Averin
2021-08-06 7:50 ` [PATCH NET v4 3/7] ipv6: use skb_expand_head in ip6_xmit Vasily Averin
[not found] ` <CALMXkpaay1y=0tkbnskr4gf-HTMjJJsVryh4Prnej_ws-hJvBg@mail.gmail.com>
2021-08-20 22:44 ` Christoph Paasch
2021-08-21 6:21 ` Vasily Averin
2021-08-22 17:04 ` Christoph Paasch
2021-08-22 17:13 ` Christoph Paasch
2021-08-23 5:44 ` Vasily Averin
2021-08-23 5:59 ` Vasily Averin
2021-08-23 7:56 ` [PATCH NET-NEXT] ipv6: skb_expand_head() adjust skb->truesize incorrectly Vasily Averin
2021-08-23 17:25 ` Christoph Paasch
2021-08-23 21:45 ` Eric Dumazet
2021-08-23 21:51 ` Eric Dumazet
2021-08-23 22:23 ` Eric Dumazet
2021-08-24 8:50 ` Vasily Averin
2021-08-24 17:21 ` Vasily Averin
2021-08-25 17:49 ` Christoph Paasch [this message]
2021-08-29 12:59 ` [PATCH v2] " Vasily Averin
2021-08-30 5:52 ` [PATCH net-next " Vasily Averin
2021-08-30 16:01 ` [PATCH " Eric Dumazet
2021-08-30 18:09 ` Vasily Averin
2021-08-30 18:37 ` Vasily Averin
2021-08-30 19:58 ` Eric Dumazet
2021-08-31 14:34 ` [PATCH net-next v3 RFC] " Vasily Averin
2021-08-31 19:38 ` Eric Dumazet
2021-09-01 6:20 ` Vasily Averin
2021-09-01 8:11 ` [PATCH net-next v4] " Vasily Averin
2021-09-01 16:58 ` Christoph Paasch
2021-09-01 19:17 ` Eric Dumazet
2021-09-02 3:59 ` Vasily Averin
2021-09-02 4:32 ` Eric Dumazet
2021-09-02 4:48 ` Eric Dumazet
2021-09-02 7:13 ` Vasily Averin
2021-09-02 7:33 ` Vasily Averin
2021-09-02 8:31 ` Vasily Averin
2021-09-02 11:12 ` [PATCH net-next v5] " Vasily Averin
2021-09-02 15:53 ` Christoph Paasch
2021-09-02 16:32 ` Vasily Averin
2021-09-06 18:01 ` [PATCH net v6] " Vasily Averin
2021-09-06 18:03 ` Vasily Averin
2021-08-27 15:23 ` [PATCH NET-NEXT] ipv6: " Vasily Averin
2021-08-27 16:47 ` Eric Dumazet
2021-08-28 8:01 ` Vasily Averin
2021-08-06 7:50 ` [PATCH NET v4 4/7] ipv4: use skb_expand_head in ip_finish_output2 Vasily Averin
2021-08-06 7:50 ` [PATCH NET v4 5/7] vrf: use skb_expand_head in vrf_finish_output Vasily Averin
2021-08-06 7:50 ` [PATCH NET v4 6/7] ax25: use skb_expand_head Vasily Averin
2021-08-06 7:50 ` [PATCH NET v4 7/7] bpf: use skb_expand_head in bpf_out_neigh_v4/6 Vasily Averin
2021-08-02 8:52 ` [PATCH NET v3 6/7] ax25: use skb_expand_head Vasily Averin
2021-08-02 8:52 ` [PATCH NET v3 7/7] bpf: use skb_expand_head in bpf_out_neigh_v4/6 Vasily Averin
[not found] ` <cover.1626206993.git.vvs@virtuozzo.com>
2021-07-13 20:57 ` [PATCH NET v2 1/7] skbuff: introduce skb_expand_head() Vasily Averin
2021-07-13 20:58 ` [PATCH NET v2 2/7] ipv6: use skb_expand_head in ip6_finish_output2 Vasily Averin
2021-07-13 20:58 ` [PATCH NET v2 3/7] ipv6: use skb_expand_head in ip6_xmit Vasily Averin
2021-07-13 20:58 ` [PATCH NET v2 4/7] ipv4: use skb_expand_head in ip_finish_output2 Vasily Averin
2021-07-13 20:58 ` [PATCH NET v2 5/7] vrf: use skb_expand_head in vrf_finish_output Vasily Averin
2021-07-13 20:58 ` [PATCH NET v2 6/7] ax25: use skb_expand_head Vasily Averin
2021-07-13 20:58 ` [PATCH NET v2 7/7] bpf: use skb_expand_head in bpf_out_neigh_v4/6 Vasily Averin
2021-07-12 13:26 ` [PATCH NET 2/7] ipv6: use pskb_realloc_headroom in ip6_finish_output2 Vasily Averin
2021-07-12 13:26 ` [PATCH NET 3/7] ipv6: use pskb_realloc_headroom in ip6_xmit refactoring Vasily Averin
2021-07-12 13:27 ` [PATCH NET 4/7] ipv4: use pskb_realloc_headroom in ip_finish_output2 Vasily Averin
2021-07-12 13:27 ` [PATCH NET 5/7] vrf: use pskb_realloc_headroom in vrf_finish_output Vasily Averin
2021-07-12 13:27 ` [PATCH NET 6/7] ax25: use pskb_realloc_headroom Vasily Averin
2021-07-12 13:27 ` [PATCH NET 7/7] bpf: use pskb_realloc_headroom in bpf_out_neigh_v4/6 Vasily Averin
[not found] ` <cover.1625818825.git.vvs@virtuozzo.com>
2021-07-09 9:04 ` [PATCH IPV6 v2 1/4] ipv6: allocate enough headroom in ip6_finish_output2() Vasily Averin
2021-07-09 17:58 ` David Miller
2021-07-10 2:53 ` Vasily Averin
2021-07-09 9:04 ` [PATCH IPV6 v2 2/4] ipv6: use new helper skb_expand_head() in ip6_xmit() Vasily Averin
2021-07-09 9:05 ` [PATCH IPV6 v2 3/4] ipv6: ip6_finish_output2 refactoring Vasily Averin
2021-07-09 9:05 ` [PATCH IPV6 v2 4/4] ipv6: ip6_xmit refactoring Vasily Averin
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=CALMXkpZYGC5HNkJAi4wCuawC-9CVNjN1LqO073YJvUF5ONwupA@mail.gmail.com \
--to=christoph.paasch@gmail.com \
--cc=davem@davemloft.net \
--cc=dsahern@kernel.org \
--cc=eric.dumazet@gmail.com \
--cc=jwi@linux.ibm.com \
--cc=kernel@openvz.org \
--cc=kuba@kernel.org \
--cc=linux-kernel@vger.kernel.org \
--cc=netdev@vger.kernel.org \
--cc=vvs@virtuozzo.com \
--cc=yoshfuji@linux-ipv6.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
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).