[PATCH] net: deal wrong skb and failure ret from __tcp_retransmit_skb

[PATCH] net: deal wrong skb and failure ret from __tcp_retransmit_skb

[Liu, Changcheng]

Hit below panic due to skb is NULL, WARN wrong skb first.
if __tcp_retransmit_skb return failure e.g. -EAGAIN, it
needn't do further action in tcp_retransmit_skb.

gdb vmlinux
Reading symbols from vmlinux...done.
(gdb) p &((struct tcp_skb_cb *) \
&(((struct sk_buff *)0)->cb[0]))->tcp_gso_segs
$1 = (u16 *) 0x30 <irq_stack_union+48>

[ 9040.917533] BUG: unable to handle kernel NULL pointer dereference at 0000000000000030
[ 9040.926279] IP: tcp_retransmit_skb+0x5c/0xc0
[ 9040.931043] PGD 0 P4D 0
[ 9040.933865] Oops: 0000 [#1] PREEMPT SMP PTI
[ 9040.972151] RIP: 0010:tcp_retransmit_skb+0x5c/0xc0
[ 9040.977496] RSP: 0018:ffff8802bec83e40 EFLAGS: 00010202
[ 9041.062527] Call Trace:
[ 9041.065250]  <IRQ>
[ 9041.067489]  tcp_retransmit_timer+0x481/0x820
[ 9041.077697]  tcp_write_timer_handler+0xe9/0x230
[ 9041.082751]  tcp_write_timer+0x75/0x80
[ 9041.086932]  call_timer_fn+0x29/0x150
[ 9041.091018]  run_timer_softirq+0x411/0x460
[ 9041.105017]  __do_softirq+0x115/0x311
[ 9041.109103]  irq_exit+0xb0/0xc0
[ 9041.112605]  smp_apic_timer_interrupt+0x67/0x140

Signed-off-by: Liu Changcheng <changcheng.liu@intel.com>

diff --git a/net/ipv4/tcp_output.c b/net/ipv4/tcp_output.c
index 383cac0..545b9b3 100644
--- a/net/ipv4/tcp_output.c
+++ b/net/ipv4/tcp_output.c
@@ -2920,7 +2920,10 @@ int __tcp_retransmit_skb(struct sock *sk, struct sk_buff *skb, int segs)
 int tcp_retransmit_skb(struct sock *sk, struct sk_buff *skb, int segs)
 {
 	struct tcp_sock *tp = tcp_sk(sk);
-	int err = __tcp_retransmit_skb(sk, skb, segs);
+	int err = 0;
+
+	WARN_ONCE(!skb, "sk_buff is NULL\n");
+	err = __tcp_retransmit_skb(sk, skb, segs);
 
 	if (err == 0) {
 #if FASTRETRANS_DEBUG > 0
@@ -2935,6 +2938,8 @@ int tcp_retransmit_skb(struct sock *sk, struct sk_buff *skb, int segs)
 		if (!tp->retrans_stamp)
 			tp->retrans_stamp = tcp_skb_timestamp(skb);
 
+	} else {
+		return err;
 	}
 
 	if (tp->undo_retrans < 0)
-- 
2.7.4

Re: [PATCH] net: deal wrong skb and failure ret from __tcp_retransmit_skb

[Eric Dumazet]

On 04/19/2018 04:26 AM, Liu, Changcheng wrote:
> Hit below panic due to skb is NULL, WARN wrong skb first.
> if __tcp_retransmit_skb return failure e.g. -EAGAIN, it
> needn't do further action in tcp_retransmit_skb.
> 
> gdb vmlinux
> Reading symbols from vmlinux...done.
> (gdb) p &((struct tcp_skb_cb *) \
> &(((struct sk_buff *)0)->cb[0]))->tcp_gso_segs
> $1 = (u16 *) 0x30 <irq_stack_union+48>
> 
> [ 9040.917533] BUG: unable to handle kernel NULL pointer dereference at 0000000000000030
> [ 9040.926279] IP: tcp_retransmit_skb+0x5c/0xc0
> [ 9040.931043] PGD 0 P4D 0
> [ 9040.933865] Oops: 0000 [#1] PREEMPT SMP PTI
> [ 9040.972151] RIP: 0010:tcp_retransmit_skb+0x5c/0xc0
> [ 9040.977496] RSP: 0018:ffff8802bec83e40 EFLAGS: 00010202
> [ 9041.062527] Call Trace:
> [ 9041.065250]  <IRQ>
> [ 9041.067489]  tcp_retransmit_timer+0x481/0x820
> [ 9041.077697]  tcp_write_timer_handler+0xe9/0x230
> [ 9041.082751]  tcp_write_timer+0x75/0x80
> [ 9041.086932]  call_timer_fn+0x29/0x150
> [ 9041.091018]  run_timer_softirq+0x411/0x460
> [ 9041.105017]  __do_softirq+0x115/0x311
> [ 9041.109103]  irq_exit+0xb0/0xc0
> [ 9041.112605]  smp_apic_timer_interrupt+0x67/0x140
> 
> Signed-off-by: Liu Changcheng <changcheng.liu@intel.com>
> 

On which kernel have you seen this problem ?

In networking, we want patches with a 'Fixes:' tag to clearly identify which patch
added a regression.


Anyway your patch comes too late I guess

https://git.kernel.org/pub/scm/linux/kernel/git/davem/net.git/commit/?id=bffd168c3fc5cc7d2bad4c668fa90e7a9010db4b

Re: [PATCH] net: deal wrong skb and failure ret from __tcp_retransmit_skb

[Liu, Changcheng]

On 06:40 Thu 19 Apr, Eric Dumazet wrote:
> 
> 
> On 04/19/2018 04:26 AM, Liu, Changcheng wrote:
> > Hit below panic due to skb is NULL, WARN wrong skb first.
> > if __tcp_retransmit_skb return failure e.g. -EAGAIN, it
> > needn't do further action in tcp_retransmit_skb.
> > 
> > gdb vmlinux
> > Reading symbols from vmlinux...done.
> > (gdb) p &((struct tcp_skb_cb *) \
> > &(((struct sk_buff *)0)->cb[0]))->tcp_gso_segs
> > $1 = (u16 *) 0x30 <irq_stack_union+48>
> > 
> > [ 9040.917533] BUG: unable to handle kernel NULL pointer dereference at 0000000000000030
> > [ 9040.926279] IP: tcp_retransmit_skb+0x5c/0xc0
> > [ 9040.931043] PGD 0 P4D 0
> > [ 9040.933865] Oops: 0000 [#1] PREEMPT SMP PTI
> > [ 9040.972151] RIP: 0010:tcp_retransmit_skb+0x5c/0xc0
> > [ 9040.977496] RSP: 0018:ffff8802bec83e40 EFLAGS: 00010202
> > [ 9041.062527] Call Trace:
> > [ 9041.065250]  <IRQ>
> > [ 9041.067489]  tcp_retransmit_timer+0x481/0x820
> > [ 9041.077697]  tcp_write_timer_handler+0xe9/0x230
> > [ 9041.082751]  tcp_write_timer+0x75/0x80
> > [ 9041.086932]  call_timer_fn+0x29/0x150
> > [ 9041.091018]  run_timer_softirq+0x411/0x460
> > [ 9041.105017]  __do_softirq+0x115/0x311
> > [ 9041.109103]  irq_exit+0xb0/0xc0
> > [ 9041.112605]  smp_apic_timer_interrupt+0x67/0x140
> > 
> > Signed-off-by: Liu Changcheng <changcheng.liu@intel.com>
> > 
> 
> On which kernel have you seen this problem ?
[Changcheng]: My kernel is 4.16.0 Fearless Coyote

> 
> In networking, we want patches with a 'Fixes:' tag to clearly identify which patch
> added a regression.
> 
> 
> Anyway your patch comes too late I guess
> 
> https://git.kernel.org/pub/scm/linux/kernel/git/davem/net.git/commit/?id=bffd168c3fc5cc7d2bad4c668fa90e7a9010db4b
[Changcheng]: Check upstream master kernel(head: 87ef12027b9b1), patch
isn't merged yet. I'll check it in my side.

> 
>