On 2021/9/6 9:46, Li, Xiaoyun wrote: > Hi > >> -----Original Message----- >> From: fengchengwen <fengcheng...@huawei.com> >> Sent: Saturday, September 4, 2021 15:02 >> To: Maxime Coquelin <maxime.coque...@redhat.com>; Richardson, Bruce >> <bruce.richard...@intel.com>; Li, Xiaoyun <xiaoyun...@intel.com> >> Cc: tho...@monjalon.net; Yigit, Ferruh <ferruh.yi...@intel.com>; >> jer...@marvell.com; jerinjac...@gmail.com; andrew.rybche...@oktetlabs.ru; >> dev@dpdk.org; m...@smartsharesystems.com; nipun.gu...@nxp.com; >> hemant.agra...@nxp.com; honnappa.nagaraha...@arm.com; >> david.march...@redhat.com; sbu...@marvell.com; pkap...@marvell.com; >> Ananyev, Konstantin <konstantin.anan...@intel.com>; Walsh, Conor >> <conor.wa...@intel.com>; Xia, Chenbo <chenbo....@intel.com> >> Subject: Re: [PATCH v18 8/8] maintainers: add for dmadev >> >> On 2021/9/3 20:59, Maxime Coquelin wrote: >>> Hi, >>> >>> On 9/2/21 3:39 PM, fengchengwen wrote: >>>> Fix in v19 >>>> >>>> I think there many patches wait for dmadev framework upstream, so >>>> could you help review unreviewd patches (like dma/skeleton and app/test)? >>> >>> Thanks for all the work, it looks really promising! >>> >>>> Also, we have developed the dma driver for hisilicon, and >>>> corresponding test app (like examples/vhost and testpmd) is being >>>> developed. >>>> >>>> examples/vhost: will extend support dmadev type. >>> >>> I think Sunil has posted RFC for Vhost lib and example, you might want >>> to have a look to avoid duplicate work. >> >> Got it, thanks >> >>> >>>> testpmd: will introduces some dma control commands and the dma >> forwarding mode. >>>> the dma forwarding mode process: >>>> // 1st: call rte_eth_rx_burst >>>> // 2st: post received packet to dma, move data to another >>>> packet >>>> // 3st: set newcopyed rte_mbuf header >>>> // 4st: free received packet >>>> // 5st: get dma completed request and associate it rte_mbuf >>>> // 6st: send 5st's rte_mbuf to nic >>>> >>>> @Maxime @Chenbo @Xiaoyun Like to hear your opinion. > > What're some dma control commands? Can you give some examples here?
It is only a plan. Our test team reports that some test points, such as configuring the queue depth, are required. Based on the current discussion, we are not prepared to support this feature, this thread could be closed. > > And the fwding mode in testpmd sounds like a function test to verify > functions of dmadev. > Why don't you just test in app/test like test_eventdev or test_cryptodev? > What's the purpose of this dma forwarding mode since it's just dma-copy from > one mbuf to another? > >>> >>> We might also think of adding async support to Vhost PMD, that would >>> be another way to test dmadev with testpmd. > > +1 on this. This makes more sense. > >> >> OK >> >>> >>> Thanks, >>> Maxime >>> >>>> >>>> On 2021/9/2 19:51, Bruce Richardson wrote: >>>>> On Thu, Sep 02, 2021 at 06:54:17PM +0800, Chengwen Feng wrote: >>>>>> This patch add myself as dmadev's maintainer and update release notes. >>>>>> >>>>>> Signed-off-by: Chengwen Feng <fengcheng...@huawei.com> --- >>>>> >>>>> Just in case you are doing any further revisions of this patchset, >>>>> the maintainers entry, and RN entry, generally is added in the first >>>>> patch, so squash this patch into patch #1. >>>>> . >>>>> >>>> >>> >>> . >>>
