WARNING: CPU: 3 PID: 261 at kernel/bpf/memalloc.c:342

[Björn Töpel]

Hou Tao <houtao at huaweicloud.com> writes:

> Hi,
>
> On 8/25/2023 11:28 PM, Yonghong Song wrote:
>>
>>
>> On 8/25/23 3:32 AM, Björn Töpel wrote:
>>> I'm chasing a workqueue hang on RISC-V/qemu (TCG), using the bpf
>>> selftests on bpf-next 9e3b47abeb8f.
>>>
>>> I'm able to reproduce the hang by multiple runs of:
>>>   | ./test_progs -a link_api -a linked_list
>>> I'm currently investigating that.
>>>
>>> But! Sometimes (every blue moon) I get a warn_on_once hit:
>>>   | ------------[ cut here ]------------
>>>   | WARNING: CPU: 3 PID: 261 at kernel/bpf/memalloc.c:342
>>> bpf_mem_refill+0x1fc/0x206
>>>   | Modules linked in: bpf_testmod(OE)
>>>   | CPU: 3 PID: 261 Comm: test_progs-cpuv Tainted: G           OE   
>>> N 6.5.0-rc5-01743-gdcb152bb8328 #2
>>>   | Hardware name: riscv-virtio,qemu (DT)
>>>   | epc : bpf_mem_refill+0x1fc/0x206
>>>   |  ra : irq_work_single+0x68/0x70
>>>   | epc : ffffffff801b1bc4 ra : ffffffff8015fe84 sp : ff2000000001be20
>>>   |  gp : ffffffff82d26138 tp : ff6000008477a800 t0 : 0000000000046600
>>>   |  t1 : ffffffff812b6ddc t2 : 0000000000000000 s0 : ff2000000001be70
>>>   |  s1 : ff5ffffffffe8998 a0 : ff5ffffffffe8998 a1 : ff600003fef4b000
>>>   |  a2 : 000000000000003f a3 : ffffffff80008250 a4 : 0000000000000060
>>>   |  a5 : 0000000000000080 a6 : 0000000000000000 a7 : 0000000000735049
>>>   |  s2 : ff5ffffffffe8998 s3 : 0000000000000022 s4 : 0000000000001000
>>>   |  s5 : 0000000000000007 s6 : ff5ffffffffe8570 s7 : ffffffff82d6bd30
>>>   |  s8 : 000000000000003f s9 : ffffffff82d2c5e8 s10: 000000000000ffff
>>>   |  s11: ffffffff82d2c5d8 t3 : ffffffff81ea8f28 t4 : 0000000000000000
>>>   |  t5 : ff6000008fd28278 t6 : 0000000000040000
>>>   | status: 0000000200000100 badaddr: 0000000000000000 cause:
>>> 0000000000000003
>>>   | [<ffffffff801b1bc4>] bpf_mem_refill+0x1fc/0x206
>>>   | [<ffffffff8015fe84>] irq_work_single+0x68/0x70
>>>   | [<ffffffff8015feb4>] irq_work_run_list+0x28/0x36
>>>   | [<ffffffff8015fefa>] irq_work_run+0x38/0x66
>>>   | [<ffffffff8000828a>] handle_IPI+0x3a/0xb4
>>>   | [<ffffffff800a5c3a>] handle_percpu_devid_irq+0xa4/0x1f8
>>>   | [<ffffffff8009fafa>] generic_handle_domain_irq+0x28/0x36
>>>   | [<ffffffff800ae570>] ipi_mux_process+0xac/0xfa
>>>   | [<ffffffff8000a8ea>] sbi_ipi_handle+0x2e/0x88
>>>   | [<ffffffff8009fafa>] generic_handle_domain_irq+0x28/0x36
>>>   | [<ffffffff807ee70e>] riscv_intc_irq+0x36/0x4e
>>>   | [<ffffffff812b5d3a>] handle_riscv_irq+0x54/0x86
>>>   | [<ffffffff812b6904>] do_irq+0x66/0x98
>>>   | ---[ end trace 0000000000000000 ]---
>>>
>>> Code:
>>>   | static void free_bulk(struct bpf_mem_cache *c)
>>>   | {
>>>   |     struct bpf_mem_cache *tgt = c->tgt;
>>>   |     struct llist_node *llnode, *t;
>>>   |     unsigned long flags;
>>>   |     int cnt;
>>>   |
>>>   |     WARN_ON_ONCE(tgt->unit_size != c->unit_size);
>>>   | ...
>>>
>>> I'm not well versed in the memory allocator; Before I dive into it --
>>> has anyone else hit it? Ideas on why the warn_on_once is hit?
>>
>> Maybe take a look at the patch
>>   822fb26bdb55  bpf: Add a hint to allocated objects.
>>
>> In the above patch, we have
>>
>> +       /*
>> +        * Remember bpf_mem_cache that allocated this object.
>> +        * The hint is not accurate.
>> +        */
>> +       c->tgt = *(struct bpf_mem_cache **)llnode;
>>
>> I suspect that the warning may be related to the above.
>> I tried the above ./test_progs command line (running multiple
>> at the same time) and didn't trigger the issue.
>
> The extra 8-bytes before the freed pointer is used to save the pointer
> of the original bpf memory allocator where the freed pointer came from,
> so unit_free() could free the pointer back to the original allocator to
> prevent alloc-and-free unbalance.
>
> I suspect that a wrong pointer was passed to bpf_obj_drop, but do not
> find anything suspicious after checking linked_list. Another possibility
> is that there is write-after-free problem which corrupts the extra
> 8-bytes before the freed pointer. Could you please apply the following
> debug patch to check whether or not the extra 8-bytes are corrupted ?

Thanks for getting back!

I took your patch for a run, and there's a hit:
  | bad cache ff5ffffffffe8570: got size 96 work ffffffff801b19c8, cache ff5ffffffffe8980 exp size 128 work ffffffff801b19c8
  | WARNING: CPU: 3 PID: 265 at kernel/bpf/memalloc.c:847 bpf_mem_free+0xb2/0xe0
  | ---[ end trace 0000000000000000 ]---
  | 00000000: 80 e3 dc 85 00 00 60 ff 00 00 00 00 00 00 00 00
  | 00000010: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
  | 00000020: 20 00 00 00 00 00 00 00 c8 19 1b 80 ff ff ff ff
  | 00000030: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
  | 00000040: 60 00 00 00 33 00 00 00 20 00 00 00 60 00 00 00
  | 00000050: 30 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
  | 00000060: 70 85 fe ff ff ff 5f ff 00 00 00 00 00 00 00 00
  | 00000070: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
  | 00000080: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
  | 00000090: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
  | 000000a0: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
  | 000000b0: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
  | 000000c0: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
  | WARNING: CPU: 2 PID: 265 at kernel/bpf/memalloc.c:342 bpf_mem_refill+0x1fc/0x206
  | ---[ end trace 0000000000000000 ]---


Björn