From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from mga05.intel.com (mga05.intel.com [192.55.52.43]) by dpdk.org (Postfix) with ESMTP id E12A23572 for ; Thu, 17 Jan 2019 11:30:18 +0100 (CET) X-Amp-Result: SKIPPED(no attachment in message) X-Amp-File-Uploaded: False Received: from orsmga007.jf.intel.com ([10.7.209.58]) by fmsmga105.fm.intel.com with ESMTP/TLS/DHE-RSA-AES256-GCM-SHA384; 17 Jan 2019 02:30:18 -0800 X-ExtLoop1: 1 X-IronPort-AV: E=Sophos;i="5.56,488,1539673200"; d="scan'208";a="107288034" Received: from irsmsx154.ger.corp.intel.com ([163.33.192.96]) by orsmga007.jf.intel.com with ESMTP; 17 Jan 2019 02:30:17 -0800 Received: from irsmsx105.ger.corp.intel.com ([169.254.7.116]) by IRSMSX154.ger.corp.intel.com ([169.254.12.253]) with mapi id 14.03.0415.000; Thu, 17 Jan 2019 10:30:16 +0000 From: "Ananyev, Konstantin" To: "Iremonger, Bernard" , "dev@dpdk.org" Thread-Topic: [PATCH] test/ipsec: fix test suite setup function Thread-Index: AQHUrMRNVcZQp3tOgkm/NjnU7fqUBKWwNbjggAAN8gCAAAJQYIAC+KwAgAAF4GA= Date: Thu, 17 Jan 2019 10:30:15 +0000 Message-ID: <2601191342CEEE43887BDE71AB977258010D904F1F@irsmsx105.ger.corp.intel.com> References: <1547551191-28863-1-git-send-email-bernard.iremonger@intel.com> <2601191342CEEE43887BDE71AB977258010D90422F@irsmsx105.ger.corp.intel.com> <8CEF83825BEC744B83065625E567D7C260D4ECBD@IRSMSX108.ger.corp.intel.com> <2601191342CEEE43887BDE71AB977258010D904263@irsmsx105.ger.corp.intel.com> <8CEF83825BEC744B83065625E567D7C260D4F7FD@IRSMSX108.ger.corp.intel.com> In-Reply-To: <8CEF83825BEC744B83065625E567D7C260D4F7FD@IRSMSX108.ger.corp.intel.com> Accept-Language: en-IE, en-US Content-Language: en-US X-MS-Has-Attach: X-MS-TNEF-Correlator: x-titus-metadata-40: eyJDYXRlZ29yeUxhYmVscyI6IiIsIk1ldGFkYXRhIjp7Im5zIjoiaHR0cDpcL1wvd3d3LnRpdHVzLmNvbVwvbnNcL0ludGVsMyIsImlkIjoiMjFhNjIyMWMtNDBkYS00YWNjLWI2OTEtZTNmNzE1YTZmNzgyIiwicHJvcHMiOlt7Im4iOiJDVFBDbGFzc2lmaWNhdGlvbiIsInZhbHMiOlt7InZhbHVlIjoiQ1RQX05UIn1dfV19LCJTdWJqZWN0TGFiZWxzIjpbXSwiVE1DVmVyc2lvbiI6IjE3LjEwLjE4MDQuNDkiLCJUcnVzdGVkTGFiZWxIYXNoIjoiYnJVeXBMNXVWRjQ3Mmp5UGpkd3d1YSt3QU9EVmpOclpad3RCS3EyK1dLcjZEYUthYitSQlR6MFpmelM4Y05hNCJ9 x-ctpclassification: CTP_NT dlp-product: dlpe-windows dlp-version: 11.0.400.15 dlp-reaction: no-action x-originating-ip: [163.33.239.181] Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: quoted-printable MIME-Version: 1.0 Subject: Re: [dpdk-dev] [PATCH] test/ipsec: fix test suite setup function X-BeenThere: dev@dpdk.org X-Mailman-Version: 2.1.15 Precedence: list List-Id: DPDK patches and discussions List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Thu, 17 Jan 2019 10:30:19 -0000 Hi Bernard, > > > > > Subject: [PATCH] test/ipsec: fix test suite setup function > > > > > > > > > > Check for valid crypto_null devices before continuing. > > > > > > > > > > Fixes: 05fe65eb66b2 ("test/ipsec: introduce functional test") > > > > > Signed-off-by: Bernard Iremonger > > > > > --- > > > > > test/test/test_ipsec.c | 17 +++++++++++++++-- > > > > > 1 file changed, 15 insertions(+), 2 deletions(-) > > > > > > > > > > diff --git a/test/test/test_ipsec.c b/test/test/test_ipsec.c inde= x > > > > > ff1a1c4..4dfc55b 100644 > > > > > --- a/test/test/test_ipsec.c > > > > > +++ b/test/test/test_ipsec.c > > > > > @@ -46,6 +46,8 @@ > > > > > #define BURST_SIZE 32 > > > > > #define REORDER_PKTS 1 > > > > > > > > > > +static int gbl_driver_id; > > > > > + > > > > > > > > Why do you need that global here? > > > > > > test_ipsec.c is based on test_cryptodev.c. > > > gbl_driver_id used to store the ID of the required driver. > > > > Sorry but referencing someone else code is not an answer. > > Why do *you* need it *here*? >=20 > The global is not needed. > I have renamed it to driver_id and added it as a local variable where it = is used. >=20 > > > > > struct user_params { > > > > > enum rte_crypto_sym_xform_type auth; > > > > > enum rte_crypto_sym_xform_type cipher; @@ -218,7 +220,7 @@ > > > > > testsuite_setup(void) { > > > > > struct ipsec_testsuite_params *ts_params =3D &testsuite_params; > > > > > struct rte_cryptodev_info info; > > > > > - uint32_t nb_devs, dev_id; > > > > > + uint32_t i, nb_devs, dev_id; > > > > > size_t sess_sz; > > > > > > > > > > memset(ts_params, 0, sizeof(*ts_params)); @@ -251,7 +253,18 @@ > > > > > testsuite_setup(void) > > > > > return TEST_FAILED; > > > > > } > > > > > > > > > > - ts_params->valid_devs[ts_params->valid_dev_count++] =3D 0; > > > > > + gbl_driver_id =3D rte_cryptodev_driver_id_get( > > > > > + RTE_STR(CRYPTODEV_NAME_NULL_PMD)); > > > > > > These tests only work with the crypto_null PMD's, gbl_driver_id is = set to the > > crypto_null PMD id here. > > > > > > > > + > > > > > + /* Create list of valid crypto devs */ > > > > > + for (i =3D 0; i < nb_devs; i++) { > > > > > + rte_cryptodev_info_get(i, &info); > > > > > + if (info.driver_id =3D=3D gbl_driver_id) > > > > > + ts_params->valid_devs[ts_params->valid_dev_count++] > > > > =3D i; > > > > > + } > > > > > > > > I think you need to check driver capabilities, instead of relying o= n driver > > name. > > > > > > I don't think it is necessary to check the driver capabilities. > > > > I still think that the valid way to check supported algorithms is to ch= eck device > > capabilities, not the driver name. >=20 > In the testsuite_setup() function the parameters for the check_cryptodev_= capability() are not setup. They are setup in the test functions of > the testsuite. Ok, so what prevents us to setup them earlier? >=20 > > > This is how it is done in test_cryptodev.c. > > > I think it makes sense to mirror the test_cryptodev.c implementation= . > > > > > > > > + > > > > > + if (ts_params->valid_dev_count < 1) > > > > > + return TEST_FAILED; > > > > > > > > > > /* Set up all the qps on the first of the valid devices found *= / > > > > > dev_id =3D ts_params->valid_devs[0]; > > > > > > > > If we always use just valid_dev[0] to determine private session > > > > size, why do you keep going though all devs in the loop above? > > > > > > There may be several crypto devs present for example, crypto_aesni_mb= 0, > > crypto_aseni_mb1, crypto_null0 and crypto_null1. > > > > Yes. > > > > > The valid_dev[] array will contain all devs of the requested type, in= this case > > crypto_null0 and crypto_null1. > > > > But we need/use only one. >=20 > I will change the code to replace the valid_devs[] with one valid_dev. >=20 > > > > Another thing, as I mentioned off-line - later you still use all > > > > vald_devs[] to init > > > > session: > > > > s =3D rte_cryptodev_sym_session_create(qp->mp_session); > > > > if (s =3D=3D NULL) > > > > return -ENOMEM; > > > > > > > > /* initiliaze SA crypto session for all supported devices *= / > > > > for (i =3D 0; i !=3D devnum; i++) { > > > > rc =3D rte_cryptodev_sym_session_init(devid[i], s, > > > > ut->crypto_xforms, qp->mp_session_private); > > > > if (rc !=3D 0) > > > > break; > > > > } > > > > > > > > I think we need either to determine max private session size based > > > > on *all* valid_devs[], or just use one device that can do NULL algo= rithm. > > > > > > The valid_devs[] array only contains crypto_null PMD's The code is > > > using the crypto_null PMD only. > > > > In fact there is no reason to be crypto_null only. > > I think it could be any crypto-dev that does support NULL auth/cipher. >=20 > As discussed offline it should be sufficient to test with the crypto_dev = NULL PMD. As we discussed offline - yes, I don't think it's too excessive to verify = ipsec_autotest with each existing driver that supports _NULL algs, but I don't see the rea= son why it shouldn't support anything except crypto_null. Konstantin=20 >=20 > > > > As we always enqueue/dequeuer into valid_devs[0] - I think there is > > > > no point to have an arrays here, just single valid_dev should be su= fficient. > > > > > > The test program may be started with several crypto_dev PMD's for exa= mple: > > > > > > test -c f -n 4 --vdev crypto_aesni_mb0 --vdev crypto_null0 --vdev > > > crypto_aesni_mb1 --vdev crypto_dev_null1 > > > > > > In this case the valid_devs[] array will contain crypto_dev_null0 and > > crypto_dev_null1. >=20 > I will replace the valid_devs[] with valid_dev which contains the first c= rypto_null device found. >=20 > I will send a v2 patch >=20 > Regards, >=20 > Bernard.