netdev - Re: [PATCH net-next 4/6] ionic: batch rx buffer refilling

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <712741a2-e51c-cd6d-34fa-d9163e776cfe@pensando.io>
Date:   Thu, 5 Nov 2020 10:49:28 -0800
From:   Shannon Nelson <snelson@...sando.io>
To:     Saeed Mahameed <saeed@...nel.org>, netdev@...r.kernel.org,
        davem@...emloft.net, kuba@...nel.org
Subject: Re: [PATCH net-next 4/6] ionic: batch rx buffer refilling

On 11/4/20 5:08 PM, Saeed Mahameed wrote:
> On Wed, 2020-11-04 at 14:33 -0800, Shannon Nelson wrote:
>> We don't need to refill the rx descriptors on every napi
>> if only a few were handled.  Waiting until we can batch up
>> a few together will save us a few Rx cycles.
>>
>> Signed-off-by: Shannon Nelson <snelson@...sando.io>
>> ---
>>   .../net/ethernet/pensando/ionic/ionic_dev.h    |  4 +++-
>>   .../net/ethernet/pensando/ionic/ionic_txrx.c   | 18 ++++++++++----
>> ----
>>   2 files changed, 13 insertions(+), 9 deletions(-)
>>
>> diff --git a/drivers/net/ethernet/pensando/ionic/ionic_dev.h
>> b/drivers/net/ethernet/pensando/ionic/ionic_dev.h
>> index 6c243b17312c..9064222a087a 100644
>> --- a/drivers/net/ethernet/pensando/ionic/ionic_dev.h
>> +++ b/drivers/net/ethernet/pensando/ionic/ionic_dev.h
>> @@ -12,8 +12,10 @@
>>   
>>   #define IONIC_MAX_TX_DESC		8192
>>   #define IONIC_MAX_RX_DESC		16384
>> -#define IONIC_MIN_TXRX_DESC		16
>> +#define IONIC_MIN_TXRX_DESC		64
>>   #define IONIC_DEF_TXRX_DESC		4096
>> +#define IONIC_RX_FILL_THRESHOLD		64
> isn't 64 a bit high ? 64 is the default napi budget
>
> Many drivers do this with bulks of 8/16 but I couldn't find any with
> higher numbers.
>
> also, just for a reference, GRO and XDP they bulk up to 8. but not
> more.
>
> IMHO i think you need to relax the re-fill threshold a bit.

Yeah, the work being done internally on tuning has been a bit 
aggressive, I can dial this back a bit.
sln

>
>> +#define IONIC_RX_FILL_DIV		8
>>
> ...
>
>> -	if (work_done)
>> +	rx_fill_threshold = min_t(u16, IONIC_RX_FILL_THRESHOLD,
>> +				  cq->num_descs / IONIC_RX_FILL_DIV);
>> +	if (work_done && ionic_q_space_avail(cq->bound_q) >=
>> rx_fill_threshold)
>>   		ionic_rx_fill(cq->bound_q);
>>   
>