[Cake] cake infinite loop(?) with hfsc on one-armed router

Sat Jan 5 05:59:18 EST 2019

> On Jan 5, 2019, at 11:06 AM, Toke Høiland-Jørgensen <toke at toke.dk> wrote:
> 
> Hmm, try this version for 3.16 - probably doesn't work on later kernels.
> I'll look into a proper backport once you've confirmed that it works :)

Thanks! Quick reminder, I’ve only seen this happen with hfsc, not when cake is a leaf below htb, for whatever reason, but that aside...

After the patch I was able to do an iperf3 upload through the one-armed router (receive on default VLAN and send on tagged VLAN), but when I ran iperf3 in reverse mode (receive on tagged VLAN and send on default VLAN), this happened right away (also see compile warnings below):

root at apu1a:~# [  341.268556] BUG: unable to handle kernel NULL pointer dereference at 00000008
[  341.275801] IP: [<fa0e8834>] adjust_parent_qlen+0x37/0xf1a [sch_cake]
[  341.282290] *pde = 00000000 
[  341.285203] Oops: 0000 [#1] SMP 
[  341.288496] Modules linked in: em_meta cls_basic sch_hfsc sch_cake(O) xt_ACCOUNT(O) ipt_REJECT xt_recent iptable_mangle iptable_nat nfi
[  341.339568] CPU: 1 PID: 0 Comm: swapper/1 Tainted: G           O  3.16.7-ckt9-voyage #1
[  341.347576] Hardware name: PC Engines APU/APU, BIOS 4.0 09/08/2014
[  341.353765] task: f5c811d0 ti: f5c86000 task.ti: f5c86000
[  341.359173] EIP: 0060:[<fa0e8834>] EFLAGS: 00210206 CPU: 1
[  341.364669] EIP is at adjust_parent_qlen+0x37/0xf1a [sch_cake]
[  341.370508] EAX: f5d65000 EBX: ffffffe8 ECX: 00000000 EDX: 00000003
[  341.376774] ESI: 00010000 EDI: f2900000 EBP: f5c99cf0 ESP: f5c99ce8
[  341.383041]  DS: 007b ES: 007b FS: 00d8 GS: 0000 SS: 0068
[  341.388447] CR0: 8005003b CR2: 00000008 CR3: 35cde000 CR4: 00000790
[  341.394713] Stack:
[  341.396734]  000001e1 f56f0000 f5c99d84 fa0e913c e0b706f9 0209090a f2907840 000017a8
[  341.404678]  00000000 00000004 000001e0 000001e1 f56f0100 f2a90000 000017a8 c131996d
[  341.412626]  ca4edbd8 0000004d f54fca00 f54fca00 f5c99dbc f5c99d54 0209090a 0209480a
[  341.420573] Call Trace:
[  341.423037]  [<fa0e913c>] adjust_parent_qlen+0x93f/0xf1a [sch_cake]
[  341.429322]  [<c131996d>] ? _raw_spin_unlock_bh+0x13/0x15
[  341.434734]  [<c129f91c>] ? tc_classify_compat+0x2f/0x5f
[  341.440054]  [<c12a091e>] ? tc_classify+0x1a/0x8b
[  341.444767]  [<fa0f4a8f>] 0xfa0f4a8e
[  341.448361]  [<c128b628>] __dev_queue_xmit+0x210/0x35f
[  341.453504]  [<c12af74e>] ? ip_fragment+0x79f/0x79f
[  341.458392]  [<c128b78b>] dev_queue_xmit+0xa/0xc
[  341.463023]  [<c12916f9>] neigh_resolve_output+0x12f/0x145
[  341.468517]  [<c12afaa0>] ip_finish_output+0x352/0x73d
[  341.473664]  [<c12b0e19>] ip_output+0x73/0xaf
[  341.478033]  [<c12ad986>] ip_forward_finish+0x66/0x6b
[  341.483091]  [<c12adc3b>] ip_forward+0x2b0/0x36d
[  341.487720]  [<c12ac467>] ip_rcv_finish+0x267/0x29a
[  341.492607]  [<c12aca4c>] ip_rcv+0x2b4/0x338
[  341.496894]  [<c12895dd>] __netif_receive_skb_core+0x467/0x4b6
[  341.502741]  [<c1289674>] __netif_receive_skb+0x48/0x59
[  341.507975]  [<c1289cb9>] netif_receive_skb_internal+0x59/0x85
[  341.513818]  [<c1289d6c>] napi_gro_complete+0x87/0x8c
[  341.518878]  [<c128a020>] napi_gro_flush+0x3e/0x53
[  341.523680]  [<c128a04c>] napi_complete+0x17/0x27
[  341.528394]  [<f80361a3>] 0xf80361a2
[  341.531985]  [<c128a0b2>] net_rx_action+0x56/0x10e
[  341.536785]  [<c102d689>] __do_softirq+0x91/0x175
[  341.541501]  [<c102d5f8>] ? __hrtimer_tasklet_trampoline+0x1a/0x1a
[  341.547685]  [<c10033c3>] do_softirq_own_stack+0x1d/0x23
[  341.553002]  <IRQ> 
[  341.554933]  [<c102d8a9>] irq_exit+0x34/0x75
[  341.559445]  [<c1002f30>] do_IRQ+0x92/0xa6
[  341.563554]  [<c131a4ec>] common_interrupt+0x2c/0x40
[  341.568531]  [<c126add1>] ? cpuidle_enter_state+0x37/0x96
[  341.573936]  [<c126aee8>] cpuidle_enter+0xf/0x12
[  341.578567]  [<c1051e54>] cpu_startup_entry+0x135/0x1e1
[  341.583802]  [<c101d553>] start_secondary+0x1a6/0x1ab
[  341.588856] Code: 70 24 85 f6 74 5e 66 31 f6 81 fe 00 00 ff ff 74 53 8b 40 34 8b 00 f6 40 08 01 75 05 39 70 20 74 16 8b 58 18 83 eb 18e
[  341.609292] EIP: [<fa0e8834>] adjust_parent_qlen+0x37/0xf1a [sch_cake] SS:ESP 0068:f5c99ce8
[  341.617689] CR2: 0000000000000008
[  341.621012] ---[ end trace db8ecd998020cc49 ]---
[  341.625639] Kernel panic - not syncing: Fatal exception in interrupt
[  341.632073] Kernel Offset: 0x0 from 0xc1000000 (relocation range: 0xc0000000-0xf7ffdfff)
[  341.640215] Rebooting in 30 seconds..

Probably less critically, some compile warnings:

root at apu1a:~/src/sch_cake# make clean
make[1]: Entering directory '/usr/src/linux-headers-3.16.7-ckt9-voyage'
  CLEAN   /root/src/sch_cake/.tmp_versions
  CLEAN   /root/src/sch_cake/Module.symvers
make[1]: Leaving directory '/usr/src/linux-headers-3.16.7-ckt9-voyage'
root at apu1a:~/src/sch_cake# make
make[1]: Entering directory '/usr/src/linux-headers-3.16.7-ckt9-voyage'
  CC [M]  /root/src/sch_cake/sch_cake.o
/root/src/sch_cake/sch_cake.c: In function ‘adjust_parent_qlen’:
/root/src/sch_cake/sch_cake.c:1753:31: warning: passing argument 1 of ‘qdisc_match_from_root’ from incompatible pointer type
   sch = qdisc_match_from_root(qdisc_dev(sch), TC_H_MAJ(parentid));
                               ^
/root/src/sch_cake/sch_cake.c:1727:22: note: expected ‘struct Qdisc *’ but argument is of type ‘struct net_device *’
 static struct Qdisc *qdisc_match_from_root(struct Qdisc *root, u32 handle)
                      ^
  Building modules, stage 2.
  MODPOST 1 modules
  CC      /root/src/sch_cake/sch_cake.mod.o
  LD [M]  /root/src/sch_cake/sch_cake.ko
make[1]: Leaving directory '/usr/src/linux-headers-3.16.7-ckt9-voyage'