[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-ID: <SJ0PR18MB52165D46518A63E214F4FBBFDBCE9@SJ0PR18MB5216.namprd18.prod.outlook.com>
Date: Tue, 17 May 2022 04:39:08 +0000
From: Suman Ghosh <sumang@...vell.com>
To: Jakub Kicinski <kuba@...nel.org>
CC: "davem@...emloft.net" <davem@...emloft.net>,
"edumazet@...gle.com" <edumazet@...gle.com>,
"pabeni@...hat.com" <pabeni@...hat.com>,
Sunil Kovvuri Goutham <sgoutham@...vell.com>,
Subbaraya Sundeep Bhatta <sbhatta@...vell.com>,
Geethasowjanya Akula <gakula@...vell.com>,
"Sunil.Goutham@...ium.com" <Sunil.Goutham@...ium.com>,
Hariprasad Kelam <hkelam@...vell.com>,
"colin.king@...el.com" <colin.king@...el.com>,
"netdev@...r.kernel.org" <netdev@...r.kernel.org>
Subject: RE: [EXT] Re: [net-next PATCH V2] octeontx2-pf: Add support for
adaptive interrupt coalescing
>> @@ -1230,7 +1262,7 @@ static int otx2_set_link_ksettings(struct
>> net_device *netdev,
>>
>> static const struct ethtool_ops otx2_ethtool_ops = {
>> .supported_coalesce_params = ETHTOOL_COALESCE_USECS |
>> - ETHTOOL_COALESCE_MAX_FRAMES,
>> + ETHTOOL_COALESCE_MAX_FRAMES |
>ETHTOOL_COALESCE_USE_ADAPTIVE,
>
>Ah there it is. Now you actually tested it with upstream :/
>
>Please wrap the line.
[Suman] Done.
>
>Also please reject user trying to set asymmetric adaptive mode since you
>don't seem to support it:
[Suman] Done
>
>> + /* Check and update coalesce status */
>> + if ((pfvf->flags & OTX2_FLAG_ADPTV_INT_COAL_ENABLED) ==
>> + OTX2_FLAG_ADPTV_INT_COAL_ENABLED) {
>> + priv_coalesce_status = 1;
>> + if (!ec->use_adaptive_rx_coalesce || !ec-
>>use_adaptive_tx_coalesce)
>> + pfvf->flags &= ~OTX2_FLAG_ADPTV_INT_COAL_ENABLED;
>> + } else {
>> + priv_coalesce_status = 0;
>> + if (ec->use_adaptive_rx_coalesce || ec-
>>use_adaptive_tx_coalesce)
>> + pfvf->flags |= OTX2_FLAG_ADPTV_INT_COAL_ENABLED;
>
>If I'm reading this right user doing:
>
>ethtool -C eth0 adaptive-rx on adaptive-tx off
>
>multiple times will keep switching between adaptive and non-adaptive
>mode. This will be super confusing to automation which may assume
>configuration commands are idempotent.
[Suman] I see your point now. Updated the code to make sure adaptive-rx is same as adaptive-tx.
Powered by blists - more mailing lists