Re: [PATCH v2] neighbour: Don't let neigh_forced_gc() disable preemption for long

[David Ahern]

On 12/6/23 11:49 AM, Doug Anderson wrote:
> Hi,
> 
> On Wed, Dec 6, 2023 at 9:51 AM David Ahern <dsahern@xxxxxxxxxx> wrote:
>>
>> On 12/6/23 10:39 AM, Stephen Hemminger wrote:
>>> On Wed,  6 Dec 2023 03:38:33 +0000
>>> Judy Hsiao <judyhsiao@xxxxxxxxxxxx> wrote:
>>>
>>>> diff --git a/net/core/neighbour.c b/net/core/neighbour.c
>>>> index df81c1f0a570..552719c3bbc3 100644
>>>> --- a/net/core/neighbour.c
>>>> +++ b/net/core/neighbour.c
>>>> @@ -253,9 +253,11 @@ static int neigh_forced_gc(struct neigh_table *tbl)
>>>>  {
>>>>      int max_clean = atomic_read(&tbl->gc_entries) -
>>>>                      READ_ONCE(tbl->gc_thresh2);
>>>> +    u64 tmax = ktime_get_ns() + NSEC_PER_MSEC;
>>>>      unsigned long tref = jiffies - 5 * HZ;
>>>>      struct neighbour *n, *tmp;
>>>>      int shrunk = 0;
>>>> +    int loop = 0;
>>>>
>>>>      NEIGH_CACHE_STAT_INC(tbl, forced_gc_runs);
>>>>
>>>> @@ -278,11 +280,16 @@ static int neigh_forced_gc(struct neigh_table *tbl)
>>>>                              shrunk++;
>>>>                      if (shrunk >= max_clean)
>>>>                              break;
>>>> +                    if (++loop == 16) {
>>>
>>> Overall looks good.
>>> Minor comments:
>>>       - loop count should probably be unsigned
>>>         - the magic constant 16 should be a sysctl tuneable
>>
>> A tunable is needed here; the loop counter is just to keep the overhead
>> of the ktime_get_ns call in check.
> 
> From context, I'm going to assume you meant a tunable is _NOT_ needed here. ;-)
> 
> -Doug

yes, multitasking fail :-(