Re: [PATCH bpf] xdp: Handle device unregister for devmap_hash map type

["Toke Høiland-Jørgensen" <toke@redhat.com>]

Alexei Starovoitov <ast@fb.com> writes:

> On 10/17/19 3:28 AM, Toke Høiland-Jørgensen wrote:
>> Tetsuo Handa <penguin-kernel@i-love.sakura.ne.jp> writes:
>> 
>>> On 2019/10/16 22:28, Toke Høiland-Jørgensen wrote:
>>>> It seems I forgot to add handling of devmap_hash type maps to the device
>>>> unregister hook for devmaps. This omission causes devices to not be
>>>> properly released, which causes hangs.
>>>>
>>>> Fix this by adding the missing handler.
>>>>
>>>> Fixes: 6f9d451ab1a3 ("xdp: Add devmap_hash map type for looking up devices by hashed index")
>>>> Reported-by: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp>
>>>> Signed-off-by: Toke Høiland-Jørgensen <toke@redhat.com>
>>>
>>> Well, regarding 6f9d451ab1a3, I think that we want explicit "(u64)" cast
>>>
>>> @@ -97,6 +123,14 @@ static int dev_map_init_map(struct bpf_dtab *dtab, union bpf_attr *attr)
>>>          cost = (u64) dtab->map.max_entries * sizeof(struct bpf_dtab_netdev *);
>>>          cost += sizeof(struct list_head) * num_possible_cpus();
>>>
>>> +       if (attr->map_type == BPF_MAP_TYPE_DEVMAP_HASH) {
>>> +               dtab->n_buckets = roundup_pow_of_two(dtab->map.max_entries);
>>> +
>>> +               if (!dtab->n_buckets) /* Overflow check */
>>> +                       return -EINVAL;
>>> +               cost += sizeof(struct hlist_head) * dtab->n_buckets;
>>>
>>>                                                      ^here
>>>
>>> +       }
>>> +
>>>          /* if map size is larger than memlock limit, reject it */
>>>          err = bpf_map_charge_init(&dtab->map.memory, cost);
>>>          if (err)
>>>
>>> like "(u64) dtab->map.max_entries * sizeof(struct bpf_dtab_netdev *)" does.
>>> Otherwise, on 32bits build, "sizeof(struct hlist_head) * dtab->n_buckets" can become 0.
>> 
>> Oh, right. I kinda assumed the compiler would be smart enough to figure
>> that out based on the type of the LHS; will send a separate fix for this.
>
> compiler smart enough?! you must be kidding.
> It's a C standard. Compiler has to do 32 bit multiply because n_buckets
> is u32 and sizeof is 32 bit in 32bit arches as Tetsuo explained.

Sure, I can see that now that Tetsuo pointed it out (thanks for that,
BTW!).

I'm just saying that since it's being assigned to a u64, the fact that
the calculation is not automatically promoted to 64-bit is somewhat
unintuitive (to me), regardless of whether it's in the standard or not.

-Toke