Re: [PATCH 1/1] Cn10K crypto Tests: Added Marvell Cn10K specific crypto Tests to TestCryptoPerfCryptodevPerf

[Patrick Robb <probb@iol.unh.edu>]

[-- Attachment #1: Type: text/plain, Size: 29989 bytes --]

Hi Gnanesh, thanks for sending this. My name is Patrick Robb, and I'm a CI
Testing Lab manager at the DPDK Community Lab hosted at UNH, and a
maintainer for the "new" DTS framework which exists within the DPDK repo
and should be replacing the legacy DTS project in the long term(what this
patch is for). We do have a Marvell CN10K board at the lab, and I would
love to begin running some crypto testing from that if possible (so, thanks
again for this series).

So, most likely we will want to provide a review from UNH and test this, so
that it can be merged to DTS (technically I was granted the authority to
apply patches to the legacy repo).

I am adding Cody to this thread as I think he can assist me on this. Cody I
am going to provide some comments inline as an initial review for this
series, and then we should apply this series locally on the traffic
generator/test engine for the CN10K (Intel-1 in our lab), dry run the test
which Gnanesh has added, and report back to him. If all goes well, it
should be possible to merge.

Long term, we will want to circle back and add this test to the new DTS
framework, but it is not ready for this test yet, as currently new DTS is
testpmd only - there is not cryptodev support. Although, adding such
support to new DTS is something I would love to work with you on in the
next year if you have time available for development towards DPDK testing!
Anyways, comments inline:

On Wed, Oct 30, 2024 at 12:44 PM Gnanesh <gpalanethra@marvell.com> wrote:

>     Changes included in this patch:
>     -- New TestCases are added for Cn10K CPT Hardware crypto accelerator.
>     -- when Testing newly added Testcases it is found Necessary to make
> Below Changes
>           1> tests/cryptodev_common.py
>                --> bind_qat_device function is updated to generate VFs for
> the given crypto_dev_id
>                --> added New function bind_mrvl_devices to accept list of
> PCI Ids and bind it to vfio_pci driver.
>           2> framework/crb.py
>                --> updated pci_devices_information_uncached_linux Method
> to handle additional Nic Speeds for Cavium,
>                    as it was restricting only to 1 GIG NIC speed
>

Thanks, I also made this change on our local DTS project for cn10k, but the
patch was not submitted to the "legacy DTS" mailing list here as it was
unmaintained (and still is, to David's point). But, if we can do a
review it is good to get these things into mainline. Maybe I can rebase our
local patch on yours if there were any other changes, and get feedback for
cn10k specific questions.

I have a hard time understanding how this 10G NIC speed requirement got
added in the first place as it was quite peculiar and confused me for a
time as it was dropping the ports list! I'm sure you had a similar
experience... :)


>           3> framework/settings.py
>                --> Changed the default Cavium NIC driver to rvu_nicpf.
>

Thanks, we are also running from this change at the Community Lab. But, I
thought you said above you are creating VFs from the bind_qat_device
function. So, does rvu_nicvf need to be added as well?


>           4> nics/net_device.py
>                --> added Missing expect object
>           5> conf/crypto_perf_cryptodev_perf.cfg
>                --> added Marvell Cn10K configs for
> crypto_perf_cryptodev_perf
>           6> framework/ssh_pexpect.py
>                --> output.replace command's output is not set before
> returning from get_output_all method
>                    hence Unnecessarily returning Prompt along with the
> output
>           7> tests/TestSuite_crypto_perf_cryptodev_perf.py
>                -->  added Marvell CN10K Testcases
>

Again we will apply these, give them a run, and report back.


>              Updated Below Private functions
>                 6.1> _run_crypto_perf() --> receives additional KW
> arguments
>                 6.2> _parse_output()  ---> Updated to handle Marvell Cn10K
> app command output
>                 6.3> _run_crypto_perf_throughput() --> receives additional
> KW arguments
>


I don't see any modifications for the build process. Is there a need to
include "-Dplatform=cn10k" when setting up the build from meson, or is the
default build fine?



>
>
> Signed-off-by: Gnanesh <gpalanethra@marvell.com>
> ---
>  conf/crypto_perf_cryptodev_perf.cfg           | 849 ++++++++++++++++++
>  framework/crb.py                              |   6 +-
>  framework/settings.py                         |   4 +-
>  framework/ssh_pexpect.py                      |   2 +-
>  nics/net_device.py                            |   3 +
>  tests/TestSuite_crypto_perf_cryptodev_perf.py | 322 ++++++-
>  tests/cryptodev_common.py                     |  11 +-
>  7 files changed, 1172 insertions(+), 25 deletions(-)
>
> diff --git a/framework/crb.py b/framework/crb.py
> index 9e3b0a58..2fe9d147 100644
> --- a/framework/crb.py
> +++ b/framework/crb.py
> @@ -374,15 +374,15 @@ class Crb(object):
>                  pass
>
>          for i in range(len(match)):
> -            # check if device is cavium and check its linkspeed, append
> only if it is 10G
> +            # for cavium supported link speed specfied under linkspeeds
>              if "177d:" in match[i][1]:
> -                linkspeed = "10000"
> +                linkspeeds = ["40000", "10000", "25000", "50000",
> "100000"]
>

Honestly I don't understand why we have this condition at all. Other
devices in DTS do not have this check where the portlinks can be dropped in
crbs.py. What do you think about updating the accepted linkspeeds vs just
removing this step entirely? Can we expect additional linkspeeds to be
added for 177d in the future, and have this problem crop up again for a new
linkspeed, or is this list of speeds "final" for 177d?


>                  nic_linkspeed = self.send_expect(
>                      "cat /sys/bus/pci/devices/%s/net/*/speed" %
> match[i][0],
>                      "# ",
>                      alt_session=True,
>                  )
> -                if nic_linkspeed.split()[0] == linkspeed:
> +                if nic_linkspeed.split()[0] in  linkspeeds:
>                      self.pci_devices_info.append((match[i][0],
> match[i][1]))
>              else:
>                  self.pci_devices_info.append((match[i][0], match[i][1]))
> diff --git a/framework/settings.py b/framework/settings.py
> index 1a561dda..b2c371f8 100644
> --- a/framework/settings.py
> +++ b/framework/settings.py
> @@ -166,8 +166,8 @@ DRIVERS = {
>      "cavium_a034": "thunder-nicvf",
>      "cavium_0011": "thunder-nicvf",
>      "IXGBE_10G-X550EM_X_SFP": "ixgbe",
> -    "cavium_a063": "octeontx2-nicpf",
> -    "cavium_a064": "octeontx2-nicvf",
> +    "cavium_a063": "rvu_nicpf",
> +    "cavium_a064": "rvu_nicvf",
>

Okay disregard my comment from above about nicvf, I see you've added it.
Please do update the commit message to reflect that this is included. And,
while you're at it, you can update the commit message to accord with the
style guidelines for submitting DPDK patches as seen here on sections 9.6
and 9.7 https://doc.dpdk.org/guides/contributing/patches.html

:)

     "ICE_100G-E810C_QSFP": "ice",
>      "ICE_25G-E810C_SFP": "ice",
>      "ICE_25G-E823C_QSFP": "ice",
> diff --git a/framework/ssh_pexpect.py b/framework/ssh_pexpect.py
> index 2132c066..a96087d2 100644
> --- a/framework/ssh_pexpect.py
> +++ b/framework/ssh_pexpect.py
> @@ -186,7 +186,7 @@ class SSHPexpect:
>
>      def get_output_all(self):
>          output = self.session.before
> -        output.replace("[PEXPECT]", "")
> +        output = output.replace("[PEXPECT]#", "")
>

So, this is just removing the leading "#" from the output? I'm not sure we
should be making styling changes for a "deprecated" repository. If I look
at this example output, from one of the CI runs of this testsuite on a
quickassist card:

# Device 1 on lcore 40

# total operations: 30000000
# Buffer size: 1024
# Burst size: 32
#     Number of bursts: 937500
#
#                 Total   Average   Maximum   Minimum
#  enqueued    30000000        32        32        32
#  dequeued    30000000        32        32         1
#    cycles 91892876610      3063    304625       485
# time [us]  3675715064   122.520 12185.000    19.400

I imagine the DTS maintainers were aware of this and thought it was the
most clear way to print the logs. I don't think we need to override that
decision now.


>          return output
>
>      def close(self, force=False):
> diff --git a/nics/net_device.py b/nics/net_device.py
> index 0f9c1af4..dd0d0e92 100644
> --- a/nics/net_device.py
> +++ b/nics/net_device.py
> @@ -43,6 +43,7 @@ class NetDevice(object):
>          if not isinstance(crb, Crb):
>              raise Exception("  Please input the instance of Crb!!!")
>          self.crb = crb
> +        self.__send_expect = self.crb.send_expect
>          self.domain_id = domain_id
>          self.bus_id = bus_id
>          self.devfun_id = devfun_id
> @@ -727,6 +728,8 @@ class NetDevice(object):
>              "/sys/bus/pci/devices/%s:%s:%s" % (domain_id, bus_id,
> devfun_id),
>              vf_reg_file,
>          )
> +        self.__send_expect("echo 0 > %s" %
> +                           (vf_reg_path), "# ")
>

So, the pre-existing VFs need to be destroyed before a new set of VFs can
be created? Do you know if this is Marvell specific, or a universal
behavior?


>          self.__send_expect("echo %d > %s" % (int(vf_num), vf_reg_path),
> "# ")
>
>      def generate_sriov_vfs_linux_igb_uio(self, domain_id, bus_id,
> devfun_id, vf_num):
> diff --git a/tests/TestSuite_crypto_perf_cryptodev_perf.py
> b/tests/TestSuite_crypto_perf_cryptodev_perf.py
> index a3f48eee..b435cefa 100644
> --- a/tests/TestSuite_crypto_perf_cryptodev_perf.py
> +++ b/tests/TestSuite_crypto_perf_cryptodev_perf.py
> @@ -59,14 +59,74 @@ class TestCryptoPerfCryptodevPerf(TestCase):
>                  "# ",
>                  5,
>              )
> -
> -        cc.bind_qat_device(self, "vfio-pci")
> +        if self.nic == "cavium_a063":
> +            count = len(self.get_suite_cfg().get('l').split(",")) - 1
> +            cc.bind_qat_device(self, "vfio-pci", generate_vfs=True,
> vf_count=count)
> +        else:
> +            cc.bind_qat_device(self, "vfio-pci")
>

If we are going to use this function for non-QAT devices, can we please
update the function name to prevent confusion? It looks like it's a short
list:

probb@d121003:~/repos/dts> grep -r "bind_qat_d" .
./tests/TestSuite_compressdev_qat_pmd.py:        cc.bind_qat_device(self,
self.drivername)
./tests/TestSuite_crypto_perf_cryptodev_perf.py:
 cc.bind_qat_device(self, "vfio-pci")
./tests/TestSuite_fips_cryptodev.py:        cc.bind_qat_device(self,
"vfio-pci")
./tests/TestSuite_ipsec_gw_cryptodev_func.py:
 cc.bind_qat_device(self, self.drivername)
./tests/TestSuite_l2fwd_cryptodev_func.py:        cc.bind_qat_device(self,
self.drivername)
./tests/TestSuite_virtio_ipsec_cryptodev_func.py:
 cc.bind_qat_device(self, self.drivername)
./tests/TestSuite_virtio_perf_cryptodev_func.py:
 cc.bind_qat_device(self, self.drivername)
./tests/TestSuite_virtio_unit_cryptodev_func.py:
 cc.bind_qat_device(self, self.drivername)
./tests/compress_common.py:def bind_qat_device(test_case, driver="igb_uio"):
./tests/cryptodev_common.py:def bind_qat_device(test_case,
driver="igb_uio"):


>          src_files = ["dep/test_aes_cbc.data", "dep/test_aes_gcm.data"]
>          self.dut_file_dir = "/tmp"
>          for file in src_files:
>              self.dut.session.copy_file_to(file, self.dut_file_dir)
>
>      def tear_down_all(self):
> +        if self.nic == "cavium_a063":
> +            cc.bind_qat_device(self, "vfio-pci", generate_vfs=True,
> +                               vf_count=2)
> +            import pandas as pd
>

I don't have a problem with this, but I'm just wondering why this is
imported here as opposed to the top of the file?


> +            tuples = [('TestCase',), ('performance', 'failed_enq'),
> +                      ('performance', 'failed_deq'),
> +                      ('performance', 'throughput_mops'),
> +                      ('performance', 'cycle_buf'),
> +                      ('performance', 'throughput', 'value'),
> +                      ('performance', 'throughput', 'delta'),
> +                      ('parameters', 'core_num/thread_num'),
> +                      ('parameters', 'frame_size'),
> +                      ('parameters', 'burst_size'),
> +                      ('parameters', 'total_ops'), ('status',)]
> +            mindex = pd.MultiIndex.from_tuples(tuples)
> +            index = 0
> +            df = pd.DataFrame([], columns=mindex)
> +            for tname, data in self._perf_result.items():
> +                for dt in data:
> +                    for key, value in dt.items():
> +                        df.loc[index, 'TestCase'] = tname
> +                        if key == 'status':
> +                            df.loc[index, (key)] = value
> +                            continue
> +                        for item in value:
> +                            if item['name'] == 'throughput':
> +                                df.loc[index, (key, item['name'],
> 'value')] = item['value']
> +                                df.loc[index, (key, item['name'],
> 'delta')] = item['delta']
> +                            else:
> +                                df.loc[index, (key, item['name'])] =
> item['value']
> +                    index += 1
> +            perf_xl = self.logger.log_path + "/" +
> "perf_cryptodev_result.xls"
> +            writer = pd.ExcelWriter(perf_xl, engine='xlsxwriter')
> +            wb  = writer.book
> +            df.to_excel(writer, sheet_name='Sheet1')
> +            ws = writer.sheets["Sheet1"]
> +            merge_format = wb.add_format({"bold": 1,"border": 1,"align":
> "center",
> +                                          "valign": "vcenter"})
> +            ws.merge_range('B1:B3', "TestCase", merge_format)
> +            ws.merge_range('C2:C3', "failed_enq", merge_format)
> +            ws.merge_range('D2:D3', "failed_deq", merge_format)
> +            ws.merge_range('E2:E3', "throughput_mops", merge_format)
> +            ws.merge_range('F2:F3', "throughput_mops", merge_format)
> +            ws.merge_range('I2:I3', "core_num/thread_num", merge_format)
> +            ws.merge_range('J2:J3', "frame_size", merge_format)
> +            ws.merge_range('K2:K3', "burst_size", merge_format)
> +            ws.merge_range('L2:L3', "total_ops", merge_format)
> +            ws.merge_range('M1:M3', "status", merge_format)
> +
> +            format1 = wb.add_format({"bg_color": "#C6EFCE", "font_color":
> "#006100", "border": 1})
> +            format2 = wb.add_format({"bg_color": "#FFC7CE", "font_color":
> "#9C0006"})
> +            ws.conditional_format("B1:M3", {"type": "cell", "criteria":
> ">=", "value": 0 , 'format': format2})
> +            condition = 'B5:M%s'%(5+len(df)-1)
> +            ws.conditional_format(condition, {'type': 'no_blanks',
> 'format':format1})
> +            ws.autofit()
> +
> +            wb.close()
>          if self._perf_result:
>              with open(
>                  self.logger.log_path + "/" +
> "perf_cryptodev_result.json", "a"
> @@ -175,12 +235,197 @@ class TestCryptoPerfCryptodevPerf(TestCase):
>      def test_scheduler_zuc(self):
>          self._run_crypto_perf_throughput()
>
> +    # Marvell CN10K Testcases
> +    def _crypto_func_cn10k(self):
> +        """Functional tests Helper
> +        Returns:
> +           None
> +        """
> +        self._run_crypto_func()
> +
> +    def _crypto_perf_cn10k(self):
> +        """Perf tests Helper function
> +        Returns:
> +           None
> +        """
> +        self._run_crypto_perf_throughput(expected="# ",
> +                                         trim_whitespace=False)
> +    def test_perf_mrvl_aes_cbc(self):
> +        self._crypto_perf_cn10k()
> +
> +    def test_perf_mrvl_aes_gcm_encrypt(self):
> +        self._crypto_perf_cn10k()
> +
> +    def test_perf_mrvl_aes_cbc_cipher_then_auth(self):
> +        self._crypto_perf_cn10k()
> +
> +    def test_perf_mrvl_aes_sha2_256_cipher_then_auth(self):
> +        self._crypto_perf_cn10k()
> +
> +    def test_perf_mrvl_zuc_eea3_cipher_only(self):
> +        self._crypto_perf_cn10k()
> +
> +    def test_perf_mrvl_zuc_eia3_auth_only(self):
> +        self._crypto_perf_cn10k()
> +
> +    def test_perf_mrvl_aes_gmac_auth_only(self):
> +        self._crypto_perf_cn10k()
> +
> +    def test_perf_mrvl_null_cipher_only(self):
> +        self._crypto_perf_cn10k()
> +
> +    def test_perf_mrvl_snow3g_uea2_cipher_only(self):
> +        self._crypto_perf_cn10k()
> +
> +    def test_perf_mrvl_kasumi_f8_cipher_only(self):
> +        self._crypto_perf_cn10k() #F
> +
> +    def test_perf_mrvl_snow3g_uia2_auth_only(self):
> +        self._crypto_perf_cn10k()
> +
> +    def test_perf_mrvl_kasumi_f9_auth_only(self):
> +        self._crypto_perf_cn10k()
> +
> +    def test_perf_mrvl_sha1_hmac_auth_only(self):
> +        self._crypto_perf_cn10k()
> +
> +    def test_perf_mrvl_sha1_auth_only(self):
> +        self._crypto_perf_cn10k()
> +
> +    def test_perf_mrvl_aes_ctr_cipher_then_auth(self):
> +        self._crypto_perf_cn10k()
> +
> +    def test_perf_mrvl_aes_ctr_snow_3g_uia2_cipher_then_auth(self):
> +        self._crypto_perf_cn10k()
> +
> +    def test_perf_mrvl_aes_cmac_cipher_then_auth(self):
> +        self._crypto_perf_cn10k()
> +
> +    def test_perf_mrvl_3des_cbc_cipher_then_auth(self):
> +        self._crypto_perf_cn10k()
> +
> +    def test_perf_mrvl_sha1_hmac_cipher_then_auth(self):
> +        self._crypto_perf_cn10k()
> +
> +    def test_perf_mrvl_sha2_224_hmac_cipher_then_auth(self):
> +        self._crypto_perf_cn10k()
> +
> +    def test_perf_mrvl_sha2_256_hmac_cipher_then_auth(self):
> +        self._crypto_perf_cn10k()
> +
> +    def test_perf_mrvl_sha2_384_hmac_cipher_then_auth(self):
> +        self._crypto_perf_cn10k()
> +
> +    def test_perf_mrvl_aes_ctr_snow_3g_uia2_cipher_then_auth(self):
> +        self._crypto_perf_cn10k()
> +
> +    def test_perf_mrvl_aes_ctr_zuc_eia3_cipher_then_auth(self):
> +        self._crypto_perf_cn10k()
> +
> +    def test_perf_mrvl_aes_ctr_aes_cmac_cipher_then_auth(self):
> +        self._crypto_perf_cn10k()
> +
> +    def test_perf_mrvl_zuc_eea3_snow3g_uia2_cipher_then_auth(self):
> +        self._crypto_perf_cn10k()
> +
> +    def test_perf_mrvl_zuc_eea3_zuc_eia3_cipher_then_auth(self):
> +        self._crypto_perf_cn10k()
> +
> +    def test_perf_mrvl_zuc_eea3_aes_cmac_cipher_then_auth(self):
> +        self._crypto_perf_cn10k()
> +
> +    def test_perf_mrvl_snow3g_uea2_snow3g_uia2_cipher_then_auth(self):
> +        self._crypto_perf_cn10k()
> +
> +    def test_perf_mrvl_snow3g_uea2_zuc_eia3_cipher_then_auth(self):
> +        self._crypto_perf_cn10k()
> +
> +    def test_perf_mrvl_snow3g_uea2_aes_cmac_cipher_then_auth(self):
> +        self._crypto_perf_cn10k()
> +
> +    def test_perf_mrvl_md5_hmac_auth_only(self):
> +        self._crypto_perf_cn10k()
> +
> +    def test_perf_mrvl_aes_ctr_cipher_only(self):
> +        self._crypto_perf_cn10k()
> +
> +    def test_perf_mrvl_3des_cbc_cipher_only(self):
> +        self._crypto_perf_cn10k()
> +
> +    def test_perf_mrvl_sha2_224_hmac_auth_only(self):
> +        self._crypto_perf_cn10k()
> +
> +    def test_perf_mrvl_sha2_256_hmac_auth_only(self):
> +        self._crypto_perf_cn10k()
> +
> +    def test_perf_mrvl_sha2_384_hmac_auth_only(self):
> +        self._crypto_perf_cn10k()
> +
> +    def test_perf_mrvl_sha2_512_hmac_auth_only(self):
> +        self._crypto_perf_cn10k()
> +
> +    def test_perf_mrvl_aes_xts__cipher_only(self):
> +        self._crypto_perf_cn10k()
> +
> +    def test_perf_mrvl_des_cbc_cipher_only(self):
> +        self._crypto_perf_cn10k()
> +
> +    def test_perf_mrvl_md5_auth_only(self):
> +        self._crypto_perf_cn10k()
> +
> +    def test_perf_mrvl_sha512_auth_only(self):
> +        self._crypto_perf_cn10k()
> +
> +    def test_perf_mrvl_sha384_auth_only(self):
> +        self._crypto_perf_cn10k()
> +
> +    def test_perf_mrvl_sha256_auth_only(self):
> +        self._crypto_perf_cn10k()
> +
> +    def test_perf_mrvl_sha224_auth_only(self):
> +        self._crypto_perf_cn10k()
> +
> +    def test_perf_mrvl_snow3g_uea2_snow3g_uia2(self):
> +        self._crypto_perf_cn10k()
> +
> +    def test_perf_mrvl_sha1_hmac_armv8(self):
> +        self._crypto_perf_cn10k()
> +
> +    def test_perf_mrvl_sha2_hmac_armv8(self):
> +        self._crypto_perf_cn10k()
> +
> +    def test_mrvl_ptest(self):
> +        self._crypto_func_cn10k()
> +
> +    def test_mrvl_burst(self):
> +        self._crypto_func_cn10k()
> +
> +    def test_mrvl_devtype(self):
> +        self._crypto_func_cn10k()
> +
> +    def test_mrvl_crypto_sha1_hmac_armv8(self):
> +        self._crypto_func_cn10k()
> +
> +    def test_mrvl_chacha20_poly1305_aead(self):
> +        self._crypto_func_cn10k()
> +
> +    def test_mrvl_crypto_sha2_hmac_armv8(self):
> +        self._crypto_func_cn10k()
> +
> +    def test_mrvl_buffer(self):
> +        self._crypto_func_cn10k()
> +
>      # Private functions
>      def _run_crypto_func(self):
> +        """ Runs Crypto functional test.
> +        Returns:
> +            None
> +        """
> +
>          if cc.is_test_skip(self):
>              return
> -
> -        cores = ",".join(self.dut.get_core_list("1S/2C/1T"))
> +        core_list = self.get_suite_cfg().get('core_list',"1S/2C/1T")
> +        cores = ",".join(self.dut.get_core_list(core_list))
>          config = {"l": cores}
>          devices = self._get_crypto_device(1)
>          if not devices:
> @@ -206,12 +451,22 @@ class TestCryptoPerfCryptodevPerf(TestCase):
>
>          out = self.dut.send_command(
>              "cat %s/%s.txt" % (self.dut_file_dir, self.running_case), 30
> -        )
> -
> +        ).strip()
>          self.verify("Error" not in out, "Test function failed")
>          self.verify("failed" not in out, "Test function failed")
> -
> -    def _run_crypto_perf(self):
> +        assert (out != ""), "No output"
> +
> +    def _run_crypto_perf(self, **kwargs):
> +        """ Runs Crypto Performance test.
> +        Args:
> +          **kwargs:
> +              expected: send_expect match prompt
> +              trim_whitespace: trime whitespace from command output
> +        Returns:
> +            None
> +        """
> +        expected = kwargs.get("expected", "#")
> +        trim_whitespace = kwargs.get("trim_whitespace", True)
>          if cc.is_test_skip(self):
>              return "skip"
>
> @@ -222,13 +477,14 @@ class TestCryptoPerfCryptodevPerf(TestCase):
>              return "skip"
>
>          eal_opt_str = cc.get_eal_opt_str(self, devices)
> -        crypto_perf_opt_str = self._get_crypto_perf_opt_str()
> +        crypto_perf_opt_str = self._get_crypto_perf_opt_str(**kwargs)
>
>          cmd_str = cc.get_dpdk_app_cmd_str(
>              self._app_path, eal_opt_str, crypto_perf_opt_str
>          )
>          try:
> -            out = self.dut.send_expect(cmd_str, "#", 600)
> +            out = self.dut.send_expect(cmd_str, expected, 600,
> +                                       trim_whitespace=trim_whitespace)
>          except Exception as ex:
>              self.logger.error(ex)
>              raise ex
> @@ -237,25 +493,44 @@ class TestCryptoPerfCryptodevPerf(TestCase):
>
>          return results
>
> -    def _get_crypto_perf_opt_str(self, override_crypto_perf_opts={}):
> +    def _get_crypto_perf_opt_str(self, **kwargs):
> +        """get crypto perf app option string
> +        args:
> +            **kwargs:
> +                override_crypto_perf_opts- Suite/Case specfic perf
> +                app config to overwrite defaults
> +        Returns:
> +            crypto perf option string
> +        """
> +        override_crypto_perf_opts =
> kwargs.get("override_crypto_perf_opts", {})
>          return cc.get_opt_str(
>              self, self._default_crypto_perf_opts,
> override_crypto_perf_opts
>          )
>
>      def _parse_output(self, output):
>          try:
> +            dtype = self.get_case_cfg().get('devtype')
> +            match_str = r"    lcore id|#lcore id"
>              lines = output.split("\r\n")
>              line_nb = len(lines)
>              self.logger.debug("Total output lines: " + str(line_nb))
>
>              for line_index in range(line_nb):
> -                if lines[line_index].startswith("    lcore id"):
> +                if re.match(match_str, lines[line_index]):
>                      self.logger.debug("data output line from: " +
> str(line_index))
>                      break
> -            data_line = line_index - 1
> +            if dtype == 'crypto_cn10k':
> +                data_line = line_index + 2
> +            else:
> +                data_line = line_index - 1
> +
> +            if len(lines[data_line].split(","))>1:
> +                pattern = re.compile(r',')
> +            else:
> +                pattern = re.compile(r'\s+')
>
>              results = []
> -            pattern = re.compile(r"\s+")
> +
>              for line in lines[data_line:-1]:
>                  print(line)
>                  result = {}
> @@ -334,9 +609,15 @@ class TestCryptoPerfCryptodevPerf(TestCase):
>              cpu_info[key] = value.strip()
>          core, thread = 0, 0
>          lcores = self.get_case_cfg()["l"].split(",")
> +        if 'Core(s) per cluster' in out:
> +            cl_soc1 = 'Core(s) per cluster'
> +            cl_soc2 = 'Cluster(s)'
> +        else:
> +            cl_soc1 = 'Core(s) per socket'
> +            cl_soc2 = 'Socket(s)'
>          for lcore in lcores[1:]:
> -            if int(lcore.strip()) < int(cpu_info["Core(s) per socket"]) *
> int(
> -                cpu_info["Socket(s)"]
> +            if int(lcore.strip()) < int(cpu_info[cl_soc1]) * int(
> +                cpu_info[cl_soc2]
>              ):
>                  core += 1
>                  thread += 1
> @@ -362,6 +643,11 @@ class TestCryptoPerfCryptodevPerf(TestCase):
>              dev = "crypto_snow3g"
>          elif self.get_case_cfg()["devtype"] == "crypto_zuc":
>              dev = "crypto_zuc"
> +        elif self.get_case_cfg()["devtype"] == "crypto_cn10k":
> +            dev = "crypto_cn10k"
> +            vf = cc.get_qat_devices(self, cpm_num=1, num=num)
> +            device["a"] = ' -a '.join(vf)
> +            device["vdev"] = None
>          elif self.get_case_cfg()["devtype"] == "crypto_scheduler":
>              dev = "crypto_scheduler"
>              w = cc.get_qat_devices(self, cpm_num=3, num=num * 3)
> @@ -388,8 +674,8 @@ class TestCryptoPerfCryptodevPerf(TestCase):
>
>          return device
>
> -    def _run_crypto_perf_throughput(self):
> -        results = self._run_crypto_perf()
> +    def _run_crypto_perf_throughput(self, **kwargs):
> +        results = self._run_crypto_perf(**kwargs)
>          if results == "skip":
>              return
>          self.verify(results, "test results is none, Test Failed")
> diff --git a/tests/cryptodev_common.py b/tests/cryptodev_common.py
> index b550b468..91067b05 100644
> --- a/tests/cryptodev_common.py
> +++ b/tests/cryptodev_common.py
> @@ -8,7 +8,9 @@ from nics.net_device import GetNicObj
>  conf = SuiteConf("cryptodev_sample")
>
>
> -def bind_qat_device(test_case, driver="igb_uio"):
> +def bind_qat_device(test_case, driver="igb_uio", generate_vfs=False,
> +                    vf_count=2, drvr_type='generic'):
> +
>      if driver == "vfio-pci":
>          test_case.dut.send_expect("modprobe vfio", "#", 10)
>          test_case.dut.send_expect("modprobe vfio-pci", "#", 10)
> @@ -38,6 +40,8 @@ def bind_qat_device(test_case, driver="igb_uio"):
>          devfun_id = addr_array[2]
>          pf_port = GetNicObj(test_case.dut, domain_id, bus_id, devfun_id)
>
> +        if generate_vfs:
> +            pf_port.generate_sriov_vfs_linux(domain_id, bus_id,
> devfun_id, vf_count, drvr_type)
>

Is this different from calling generate_sriov_vfs(), which I assume is the
"correct" behavior. There appears to be a common practice in legacy DTS of
templating out function names to call instead of using inheritance, so you
will have a base function like bind_port(), and then that templates out
calls to bind_port_linux(). bind_port_freebsd(), bind_port_windows() etc.
It's a little strange but it is what it is and we should probably stick
with the existing system if we are going to make additions. But, if there
is a practical reason why you did it this way, then okay. :)


>          sriov_vfs_pci = pf_port.get_sriov_vfs_pci()
>          if not sriov_vfs_pci:
>              raise Exception("can not get vf pci")
> @@ -51,6 +55,11 @@ def bind_qat_device(test_case, driver="igb_uio"):
>
>      test_case.dev = dev
>
> +def bind_mrvl_devices(test_case, driver='vfio-pci', **kwargs):
> +    pci_list = kwargs.get("pci_list").split()
> +    for pci in pci_list:
> +        test_case.dut.bind_eventdev_port(driver, pci)
> +    return pci_list
>
>  def get_qat_devices(test_case, cpm_num=None, num=1):
>      if not cpm_num:
> --
> 2.25.1
>
>

[-- Attachment #2: Type: text/html, Size: 38359 bytes --]