Date:	Fri, 16 May 2014 00:41:58 -0600
From:	Kelly Anderson <kelly@...ka.com>
To:	Daniel Borkmann <dborkman@...hat.com>
Cc:	Network Development <netdev@...r.kernel.org>
Subject: Re: 3.14 tc oops

Cong,

Just checked 3.14.4 and the problem still exists.  I have better information 
now (i ran the tc script with bash -x to see which line caused the oops.

*********************************************
Initial section of tc script
*********************************************

# ================================ Device eth1 
================================

tc qdisc add dev eth1 handle 1:0 root dsmark indices 4 default_index 0
tc qdisc add dev eth1 handle 2:0 parent 1:0 htb r2q 20
tc class add dev eth1 parent 2:0 classid 2:1 htb rate 112500bps
tc class add dev eth1 parent 2:1 classid 2:2 htb rate 112500bps prio 1
tc class add dev eth1 parent 2:2 classid 2:3 htb rate 62500bps ceil 112500bps 
prio 1
tc qdisc add dev eth1 handle 3:0 parent 2:3 sfq perturb 10
tc class add dev eth1 parent 2:2 classid 2:4 htb rate 37500bps ceil 87500bps 
prio 2
tc qdisc add dev eth1 handle 4:0 parent 2:4 sfq perturb 10
tc class add dev eth1 parent 2:2 classid 2:5 htb rate 12500bps ceil 25000bps 
prio 3
tc qdisc add dev eth1 handle 5:0 parent 2:5 sfq perturb 10
tc filter add dev eth1 parent 2:0 protocol all prio 1 tcindex mask 0x3 shift 0
tc filter add dev eth1 parent 2:0 protocol all prio 1 handle 3 tcindex classid 
2:5
tc filter add dev eth1 parent 2:0 protocol all prio 1 handle 2 tcindex classid 
2:3
tc filter add dev eth1 parent 2:0 protocol all prio 1 handle 1 tcindex classid 
2:4
tc filter add dev eth1 parent 1:0 protocol all prio 1 handle 20 fw classid 1:1
tc filter add dev eth1 parent 1:0 protocol all prio 2 handle 1:0:0 u32 divisor 
1
tc filter add dev eth1 parent 1:0 protocol all prio 2 u32 match u8 0x6 0xff at 9 
offset at 0 mask 0f00 shift 6 eat link 1:0:0
tc filter add dev eth1 parent 1:0 protocol all prio 2 handle 1:0:1 u32 ht 1:0:0 
match u16 0x50 0xffff at 0 classid 1:1
tc filter add dev eth1 parent 1:0 protocol all prio 2 u32 match u8 0x29 0xff at 
9 match u16 0xbb81 0xffff at 60 classid 1:1
tc filter add dev eth1 parent 1:0 protocol all prio 3 handle 10 fw classid 1:2
tc filter add dev eth1 parent 1:0 protocol all prio 4 u32 match u8 0x1 0xff at 9 
classid 1:2
tc filter add dev eth1 parent 1:0 protocol all prio 4 handle 2:0:0 u32 divisor 
1
tc filter add dev eth1 parent 1:0 protocol all prio 4 u32 match u8 0x6 0xff at 9 
offset at 0 mask 0f00 shift 6 eat link 2:0:0

*********************************************
The oops is caused by the first u32 match.
*********************************************

+ tc filter add dev eth1 parent 1:0 protocol all prio 1 handle 20 fw classid 
1:1
+ tc filter add dev eth1 parent 1:0 protocol all prio 2 handle 1:0:0 u32 
divisor 1
+ tc filter add dev eth1 parent 1:0 protocol all prio 2 u32 match u8 0x6 0xff at 
9 offset at 0 mUnable to handle kernel NULL pointer dereference at virtual 
address 00000024
ask 0f00 shift 6pgd = 93438000
[00000024] *pgd=14a25831
                        + tc filter add, *pte=00000000 dev eth1 parent, 
*ppte=00000000 1:0 protocol al
l prio 2 handle Internal error: Oops: 17 [#1] PREEMPT ARM
Modules linked in: cls_fw cls_tcindex sch_sfq sch_dsmark nf_conntrack_netlink 
xt_LOG xt_limit xt_set ip6table_filter ip6_tables xt_nat ip_set_hash_net ip_set 
xt_tcpudp xt_multiport xt_iprange xt_mark xt_connmark xt_CLASSIFY xfrm_user 
cls_u32 sch_htb sch_cbq xfrm4_tunnel iptable_nat nf_conntrack_ipv4 
nf_defrag_ipv4 nf_nat_ipv4 nf_nat ipcomp nf_conntrack xfrm_ipcomp 
iptable_filter esp4 ah4 iptable_mangle ip_tables x_tables af_key xfrm_algo 
snd_soc_kirkwood zram snd_hrtimer sha1_arm nfnetlink asix usbnet mii aes_arm
CPU: 0 PID: 765 Comm: bash Not tainted 3.14.4 #1
task: bc027180 ti: be4ee000 task.ti: be4ee000
PC is at tcf_action_exec+0x34/0x94
LR is at tc_classify_compat+0x50/0x7c
pc : [<8040f8a0>]    lr : [<8040b368>]    psr: 200e0113
sp : be4efbc8  ip : 00000000  fp : 00000000
r10: bd807780  r9 : 00000020  r8 : be110000
r7 : 9a978804  r6 : be4efc08  r5 : 93443bac  r4 : bd807780
r3 : 00000000  r2 : be4efc08  r1 : 9a978804  r0 : bd807780
Flags: nzCv  IRQs on  FIQs on  Mode SVC_32  ISA ARM  Segment user
Control: 10c5387d  Table: 13438019  DAC: 00000015
Process bash (pid: 765, stack limit = 0xbe4ee248)
Stack: (0xbe4efbc8 to 0xbe4f0000)
fbc0:                   9a978800 93b45300 00000008 bd807780 be4efc08 8040b368
fbe0: 00000000 bd807780 93b45300 be4efc08 bef9e200 8040c588 9a979800 bd807780
fc00: 00000007 7f0a5c18 00000000 00000000 bef9e200 bd807780 bc360100 9a979800
fc20: bef9e200 7f110778 00000000 bc271a00 bc360100 00000000 be4ee000 803f2cb8
fc40: 00000000 0028bb70 00000000 bc271a00 bc271a7c 0000000e bd807780 00000000
fc60: 00000020 00000000 be4ee000 80420ca0 80000000 0100a8c0 be110000 bd807780
fc80: 80698784 bee7a000 00000008 80698798 00000000 803f00b4 bee7a548 be4efcd8
fca0: 00000000 806c1ea0 c12b0170 bee7a480 00000001 bd807780 00000010 80698798
fcc0: 00000000 00000003 bd807780 6fa55fee c12b0170 bee7a480 00000001 bd807780
fce0: 00000010 803f0e6c bd807780 bee7a5bc bee7a000 80318548 00000000 0002923e
fd00: 8d28ac43 00000040 8069c9dc bee7a548 bee7a480 805bdf2c 805e0c3c 805e0888
fd20: 805fb7e8 bee7a5b8 bee7a548 00000000 806c75c8 00000001 bee7a548 00000040
fd40: 0000012c 806c7640 000004e0 806c7648 806a0c98 803f0934 00000000 806c7836
fd60: 806a353c 00000008 806c844c 806c8440 be4ee000 40000003 00000102 00000003
fd80: 00000004 800222b0 00000000 00000141 806a2078 0000000a 806c8400 806a0c98
fda0: 000004df be4ee008 804e3c84 00404100 be4ee000 be4ee018 0000001d 00000000
fdc0: be4efe10 00000000 bf003200 bf00321c be4ee000 80022678 806af770 8000e92c
fde0: 00000000 806d4b00 00000001 80008610 8001a814 400e0013 ffffffff be4efe44
fe00: 98e5e3b0 76eed000 94a27580 80011d80 98e5ebb0 17a0e38d 00000800 00000000
fe20: 17a0e3cd 76eec000 00000009 93be43b4 98e5e3b0 76eed000 94a27580 be4ee000
fe40: 0003fffb be4efe58 800a9f8c 8001a814 400e0013 ffffffff aed47dfd fffa3942
fe60: 6d9d6122 00100073 17a0e3cf 93be43b0 bebfbc70 00000000 be3b3d80 93449db8
fe80: 93439db8 8069ab80 806d0f54 be3b3600 00100100 00000000 00000002 00000000
fea0: bd98cfa0 76eed000 93449db8 93439db8 76eed000 94a27580 be3b3600 be3b3d80
fec0: 76eecfff 800aa3b0 94a27580 76eea000 76eed000 94a27580 00000000 94a27898
fee0: be4ee000 806c7f80 be3b3d80 be83f67c 00000000 00000000 94a27580 8001cc88
ff00: bc0261b4 be83f65c 00000000 be3b3600 bc026134 00000000 00000003 bc025f80
ff20: 01200011 76f80278 be4ee000 94a27abc 94a27aa8 94a27ab8 806c7f80 be3b3634
ff40: be3b3db4 94a27ab0 00004bd9 fffffff4 00000000 01200011 00000000 00000000
ff60: 00000000 00000000 be4ee000 00020000 7ef6e844 8001d76c 00000000 00000000
ff80: 0000075b 00000000 00000001 00000000 76f80278 00000000 76ee9000 00000078
ffa0: 8000e1e4 8000e060 76f80278 00000000 01200011 00000000 00000000 00000000
ffc0: 76f80278 00000000 76ee9000 00000078 76f80210 000e0d88 000002fd 7ef6e844
ffe0: 76f806d0 7ef6e818 00000000 76e5589c 600e0010 01200011 ffffffff ffffffff
[<8040f8a0>] (tcf_action_exec) from [<8040b368>] 
(tc_classify_compat+0x50/0x7c)
[<8040b368>] (tc_classify_compat) from [<8040c588>] (tc_classify+0x28/0x90)
[<8040c588>] (tc_classify) from [<7f0a5c18>] (htb_enqueue+0x90/0x320 
[sch_htb])
[<7f0a5c18>] (htb_enqueue [sch_htb]) from [<7f110778>] 
(dsmark_enqueue+0x118/0x250 [sch_dsmark])
[<7f110778>] (dsmark_enqueue [sch_dsmark]) from [<803f2cb8>] 
(__dev_queue_xmit+0x2bc/0x52c)
[<803f2cb8>] (__dev_queue_xmit) from [<80420ca0>] 
(ip_finish_output+0x21c/0x490)
[<80420ca0>] (ip_finish_output) from [<803f00b4>] 
(__netif_receive_skb_core+0x21c/0x77c)
[<803f00b4>] (__netif_receive_skb_core) from [<803f0e6c>] 
(napi_gro_receive+0x60/0x8c)
[<803f0e6c>] (napi_gro_receive) from [<80318548>] 
(mv643xx_eth_poll+0x5f0/0x6ec)
[<80318548>] (mv643xx_eth_poll) from [<803f0934>] (net_rx_action+0xa8/0x164)
[<803f0934>] (net_rx_action) from [<800222b0>] (__do_softirq+0xd4/0x224)
[<800222b0>] (__do_softirq) from [<80022678>] (irq_exit+0xa8/0xf0)
[<80022678>] (irq_exit) from [<8000e92c>] (handle_IRQ+0x3c/0x84)
[<8000e92c>] (handle_IRQ) from [<80008610>] (orion_handle_irq+0x7c/0x9c)
[<80008610>] (orion_handle_irq) from [<80011d80>] (__irq_svc+0x40/0x70)
Exception stack(0xbe4efe10 to 0xbe4efe58)
fe00:                                     98e5ebb0 17a0e38d 00000800 00000000
fe20: 17a0e3cd 76eec000 00000009 93be43b4 98e5e3b0 76eed000 94a27580 be4ee000
fe40: 0003fffb be4efe58 800a9f8c 8001a814 400e0013 ffffffff
[<80011d80>] (__irq_svc) from [<8001a814>] (cpu_v7_set_pte_ext+0x54/0x58)
[<8001a814>] (cpu_v7_set_pte_ext) from [<8069ab80>] (0x8069ab80)
Code: e2455010 0a000017 e5953004 e1a00004 (e5933024) 
1:0:1 u32 ht 1:0---[ end trace a0acc7192639a6ae ]---
:0 match u16 0x5Kernel panic - not syncing: Fatal exception in interrupt






The patch did not fix the problem.

For your convenience, I've attached tc filter show for eth0 and eth1, as well 
as a mangled version of my tcng input file.



On Sunday, April 06, 2014 22:56:27 you wrote:
> On Thu, Apr 3, 2014 at 9:24 AM, Cong Wang <xiyou.wangcong@...il.com> wrote:
> > On Wed, Apr 2, 2014 at 7:24 AM, Kelly Anderson <kelly@...ka.com> wrote:
> >> Hi,
> >> 
> >> I hit a kernel oops when starting traffic control on my armv7 router, I
> >> don't think the architecture is related, the same tc code worked
> >> perfectly with earlier kernel versions, i.e. 3.13.x.
> >> 
> >> I also attached an object dump with line numbers to make identifying the
> >> problem a bit easier.
> >> 
> >> Please cc me to keep me in the loop, I can test patches.
> > 
> > Thanks for the report! Looks like it is caused by one of my patches,
> > I will look into this shortly.
> 
> Which tc filter are you using on that htb qdisc? (tc filter show dev ....)
> 
> I suspect it's tcindex filter, if so please try the following patch:
> 
> diff --git a/net/sched/cls_tcindex.c b/net/sched/cls_tcindex.c
> index eed8404..14618cc 100644
> --- a/net/sched/cls_tcindex.c
> +++ b/net/sched/cls_tcindex.c
> @@ -298,8 +298,10 @@ tcindex_set_parms(struct net *net, struct
> tcf_proto *tp, unsigned long base,
>         tcf_exts_change(tp, &cr.exts, &e);
> 
>         tcf_tree_lock(tp);
> -       if (old_r && old_r != r)
> +       if (old_r && old_r != r) {
>                 memset(old_r, 0, sizeof(*old_r));
> +               tcf_exts_init(&old_r->exts, TCA_TCINDEX_ACT,
> TCA_TCINDEX_POLICE);
> +       }
> 
>         memcpy(p, &cp, sizeof(cp));
>         memcpy(r, &cr, sizeof(cr));
View attachment "tc-filter-show-eth0.txt" of type "text/plain" (9391 bytes)

View attachment "tc-filter-show-eth1.txt" of type "text/plain" (7828 bytes)

View attachment "idle-mangle.tcc" of type "text/x-csrc" (5061 bytes)

Powered by blists - more mailing lists