From mboxrd@z Thu Jan 1 00:00:00 1970 From: Ferruh Yigit Subject: Re: net/failsafe: add default Tx mbuf fast free capability Date: Fri, 21 Dec 2018 13:31:47 +0000 Message-ID: <000e1ce9-819e-a75c-cc51-0f6ddcbbb5be@intel.com> References: <1539344187-21481-1-git-send-email-arybchenko@solarflare.com> <585c9670-07b6-abfa-027d-e4d07febb7d4@intel.com> <969cb7e3-6e1a-f5ac-ed1b-e4334f928b17@solarflare.com> Mime-Version: 1.0 Content-Type: text/plain; charset=utf-8 Content-Transfer-Encoding: 8bit Cc: dev@dpdk.org, Ivan Malov To: Andrew Rybchenko , =?UTF-8?Q?Ga=c3=abtan_Rivet?= Return-path: Received: from mga05.intel.com (mga05.intel.com [192.55.52.43]) by dpdk.org (Postfix) with ESMTP id 89F091BDE1 for ; Fri, 21 Dec 2018 14:31:49 +0100 (CET) In-Reply-To: <969cb7e3-6e1a-f5ac-ed1b-e4334f928b17@solarflare.com> Content-Language: en-US List-Id: DPDK patches and discussions List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Errors-To: dev-bounces@dpdk.org Sender: "dev" On 12/21/2018 12:52 PM, Andrew Rybchenko wrote: > Hi Ferruh, > > On 12/21/18 3:43 PM, Ferruh Yigit wrote: >> On 12/21/2018 12:28 PM, Andrew Rybchenko wrote: >>> On 12/21/18 3:12 PM, Ferruh Yigit wrote: >>>> On 10/12/2018 12:36 PM, Andrew Rybchenko wrote: >>>>> From: Ivan Malov >>>>> >>>>> This capability is reported when supported by the current emitting >>>>> sub-device. Failsafe PMD itself does not excercise fast free logic. >>>> I think overlay device capability reporting already discussed a few times, the >>>> question is if an overlay devices should claim a feature when it depends on >>>> underlay devices? >>> The capability may be reported by the failsafe since it is transparent from >>> fast free logic point of view. >> Why it is transparent? If one of the underlying device doesn't support/claim >> this feature, application can't use this feature with failsafe, isn't it? > > tx_offload_capa in failsafe is a mask to apply on sub-device capabilities. I missed this one, I see why it is transparent. Why failsafe doesn't set a full tx_offload_capa MASK but maintain a list? > So, if the capability is not supported by any sub-device it will not be > reported. > As well if there is the capability bit in the mask, it will not be > reported regardless > sub-devices capabilities. The description for the patch above tries to > explain it - > it looks like not that successful. > >>>> Given that no ack/review given to the patch, I am updating it as rejected. >>> Is it a new policy? I thought that it was vice versa before. >> Hi Andrew, >> >> Yes policy is other-way around indeed, when there is no comment at all default >> behavior is accept, but please take above paragraph as my comment to the patch. > > Got it. > >> And I was thinking it is a little controversial and there is no support to have >> it, so lets don't get it. What do you think? > > I see you motivation. > >>>>> Signed-off-by: Ivan Malov >>>>> Signed-off-by: Andrew Rybchenko >>>>> --- >>>>> doc/guides/nics/features/failsafe.ini | 1 + >>>>> drivers/net/failsafe/failsafe_ops.c | 1 + >>>>> 2 files changed, 2 insertions(+) >>>>> >>>>> diff --git a/doc/guides/nics/features/failsafe.ini b/doc/guides/nics/features/failsafe.ini >>>>> index e3c4c08f2..b6f3dcee6 100644 >>>>> --- a/doc/guides/nics/features/failsafe.ini >>>>> +++ b/doc/guides/nics/features/failsafe.ini >>>>> @@ -7,6 +7,7 @@ >>>>> Link status = Y >>>>> Link status event = Y >>>>> Rx interrupt = Y >>>>> +Fast mbuf free = Y >>>>> Queue start/stop = Y >>>>> Runtime Rx queue setup = Y >>>>> Runtime Tx queue setup = Y >>>>> diff --git a/drivers/net/failsafe/failsafe_ops.c b/drivers/net/failsafe/failsafe_ops.c >>>>> index 7f8bcd4c6..e3add404b 100644 >>>>> --- a/drivers/net/failsafe/failsafe_ops.c >>>>> +++ b/drivers/net/failsafe/failsafe_ops.c >>>>> @@ -78,6 +78,7 @@ static struct rte_eth_dev_info default_infos = { >>>>> DEV_RX_OFFLOAD_SECURITY, >>>>> .tx_offload_capa = >>>>> DEV_TX_OFFLOAD_MULTI_SEGS | >>>>> + DEV_TX_OFFLOAD_MBUF_FAST_FREE | >>>>> DEV_TX_OFFLOAD_IPV4_CKSUM | >>>>> DEV_TX_OFFLOAD_UDP_CKSUM | >>>>> DEV_TX_OFFLOAD_TCP_CKSUM | >>>>> > >