On 8/12/20 1:00 AM, pet...@infradead.org wrote:
> On Tue, Aug 11, 2020 at 06:45:23PM -0700, Maciej Żenczykowski wrote:
>> On Tue, Aug 11, 2020 at 6:34 PM Eric Dumazet <eduma...@google.com> wrote:
>>
>>> We must accept an empty mask in store_rps_map(), or we are not able
>>> to disable RPS on a queue.
>>>
>>> Fixes: 07bbecb34106 ("net: Restrict receive packets queuing to
>>> housekeeping CPUs")
>>> Signed-off-by: Eric Dumazet <eduma...@google.com>
>>> Reported-by: Maciej Żenczykowski <m...@google.com>
>>> Cc: Alex Belits <abel...@marvell.com>
>>> Cc: Nitesh Narayan Lal <nit...@redhat.com>
>>> Cc: Peter Zijlstra (Intel) <pet...@infradead.org>
>>> ---
>>> net/core/net-sysfs.c | 12 +++++++-----
>>> 1 file changed, 7 insertions(+), 5 deletions(-)
>>>
>>> diff --git a/net/core/net-sysfs.c b/net/core/net-sysfs.c
>>> index
>>> 9de33b594ff2693c054022a42703c90084122444..efec66fa78b70b2fe5b0a5920317eb1d0415d9e3
>>> 100644
>>> --- a/net/core/net-sysfs.c
>>> +++ b/net/core/net-sysfs.c
>>> @@ -757,11 +757,13 @@ static ssize_t store_rps_map(struct netdev_rx_queue
>>> *queue,
>>> return err;
>>> }
>>>
>>> - hk_flags = HK_FLAG_DOMAIN | HK_FLAG_WQ;
>>> - cpumask_and(mask, mask, housekeeping_cpumask(hk_flags));
>>> - if (cpumask_empty(mask)) {
>>> - free_cpumask_var(mask);
>>> - return -EINVAL;
>>> + if (!cpumask_empty(mask)) {
>>> + hk_flags = HK_FLAG_DOMAIN | HK_FLAG_WQ;
>>> + cpumask_and(mask, mask, housekeeping_cpumask(hk_flags));
>>> + if (cpumask_empty(mask)) {
>>> + free_cpumask_var(mask);
>>> + return -EINVAL;
>>> + }
>>> }
>
> Ah indeed! Sorry about that.
>
> Acked-by: Peter Zijlstra (Intel) <pet...@infradead.org>
>
No worries, thanks for the review guys.
We probably should add a test doing something like
echo ffffffff >/sys/class/net/lo/queues/rx-0/rps_cpus
echo 0 >/sys/class/net/lo/queues/rx-0/rps_cpus
