From: "Tu, Lijuan" <lijuan.tu@intel.com>
To: "Ma, LihongX" <lihongx.ma@intel.com>, "dts@dpdk.org" <dts@dpdk.org>
Cc: "Ma, LihongX" <lihongx.ma@intel.com>
Subject: Re: [dts] [PATCH V1] tests/loopback_virtio_user_server_mode: update cases according to testplan
Date: Wed, 7 Aug 2019 03:37:58 +0000 [thread overview]
Message-ID: <8CE3E05A3F976642AAB0F4675D0AD20E0BAFEC4D@SHSMSX101.ccr.corp.intel.com> (raw)
In-Reply-To: <1563154750-2532-1-git-send-email-lihongx.ma@intel.com>
Applied, thanks
> -----Original Message-----
> From: dts [mailto:dts-bounces@dpdk.org] On Behalf Of lihong
> Sent: Monday, July 15, 2019 9:39 AM
> To: dts@dpdk.org
> Cc: Ma, LihongX <lihongx.ma@intel.com>
> Subject: [dts] [PATCH V1] tests/loopback_virtio_user_server_mode: update
> cases according to testplan
>
> 1. add path for virtio user server mode
> 2. all reconnet case add behavior of reconnet from vhost, reconnet from
> virtio and port restart
>
> Signed-off-by: lihong <lihongx.ma@intel.com>
> ---
> .../TestSuite_loopback_virtio_user_server_mode.py | 235
> ++++++++++++++++++---
> 1 file changed, 200 insertions(+), 35 deletions(-)
>
> diff --git a/tests/TestSuite_loopback_virtio_user_server_mode.py
> b/tests/TestSuite_loopback_virtio_user_server_mode.py
> index d6bff6a..fd4b807 100644
> --- a/tests/TestSuite_loopback_virtio_user_server_mode.py
> +++ b/tests/TestSuite_loopback_virtio_user_server_mode.py
> @@ -68,7 +68,8 @@ class TestLoopbackVirtioUserServerMode(TestCase):
> self.dut.send_expect("rm -rf ./vhost-net*", "#")
> self.dut.send_expect("killall -s INT testpmd", "#")
> # Prepare the result table
> - self.table_header = ["Mode", "Throughput(Mpps)", "Cycle"]
> + self.table_header = ["Mode", "Pkt_size", "Throughput(Mpps)",
> + "Queue Number", "Cycle"]
> self.result_table_create(self.table_header)
>
> self.vhost = self.dut.new_session(suite="vhost")
> @@ -115,10 +116,10 @@ class
> TestLoopbackVirtioUserServerMode(TestCase):
> command_line_user = self.dut.target + "/app/testpmd -n %d -c %s --
> socket-mem 1024,1024 " + \
> "--legacy-mem --no-pci --file-prefix=virtio " + \
> "--
> vdev=net_virtio_user0,mac=00:01:02:03:04:05,path=vhost-
> net,server=1,queues=%d,%s " + \
> - " -- -i --tx-offloads=0x0 --enable-hw-vlan-strip --rss-ip --nb-
> cores=%d --rxq=%d --txq=%d %s"
> + " -- -i %s --nb-cores=%d --rxq=%d --txq=%d"
> command_line_user = command_line_user %
> (self.dut.get_memory_channels(),
> - self.core_mask_user, self.queue_number, mode,
> self.nb_cores,
> - self.queue_number, self.queue_number, extern_params)
> + self.core_mask_user, self.queue_number, mode,
> extern_params,
> + self.nb_cores, self.queue_number,
> + self.queue_number)
> self.virtio_user.send_expect(command_line_user, "testpmd> ", 120)
> self.virtio_user.send_expect("set fwd mac", "testpmd> ", 120)
>
> @@ -202,7 +203,9 @@ class TestLoopbackVirtioUserServerMode(TestCase):
> self.verify(Mpps > 0, "port can not receive packets")
>
> results_row.append(case_info)
> + results_row.append('64')
> results_row.append(Mpps)
> + results_row.append(self.queue_number)
> results_row.append(cycle)
> self.result_table_add(results_row)
>
> @@ -262,85 +265,247 @@ class
> TestLoopbackVirtioUserServerMode(TestCase):
> self.result_table_print()
> self.close_all_testpmd()
>
> - def
> test_server_mode_multi_queue_reconnect_from_vhost_side_with_virtio11(s
> elf):
> + def test_server_mode_reconnect_with_virtio11_mergeable_path(self):
> """
> - reconnect virtio-user from the vhost side with multi_queues
> + reconnect test with virtio 1.1 mergeable path and server mode
> """
> self.queue_number = 2
> self.nb_cores = 2
> + case_info = 'virtio1.1 mergeable path'
> mode = "packed_vq=1,in_order=0,mrg_rxbuf=1"
> + extern_params = '--tx-offloads=0x0 --enable-hw-vlan-strip --rss-ip'
> self.lanuch_vhost_testpmd_with_multi_queue()
> - self.lanuch_virtio_user_testpmd_with_multi_queue(mode=mode)
> + self.lanuch_virtio_user_testpmd_with_multi_queue(mode=mode,
> + extern_params=extern_params)
> self.start_to_send_packets(self.vhost, self.virtio_user)
> - self.calculate_avg_throughput("reconnet from vhost with virtio1.1",
> - "before reconnet")
> + self.calculate_avg_throughput(case_info, "before reconnet")
>
> + # reconnect from vhost
> + self.logger.info('now reconnet from vhost')
> self.relanuch_vhost_testpmd_with_multi_queue()
> self.start_to_send_packets(self.virtio_user, self.vhost)
> - self.calculate_avg_throughput("reconnet from vhost with virtio1.1",
> - "after reconnet")
> + self.calculate_avg_throughput(case_info, "reconnet from vhost")
> +
> + # reconnet from virtio
> + self.logger.info('now reconnet from virtio_user')
> + self.relanuch_virtio_testpmd_with_multi_queue(mode=mode,
> extern_params=extern_params)
> + self.start_to_send_packets(self.vhost, self.virtio_user)
> + self.calculate_avg_throughput(case_info, "reconnet from virtio
> + user")
> +
> + # port restart
> + self.logger.info('now vhost port restart')
> + self.port_restart()
> + self.calculate_avg_throughput(case_info, "after port restart")
> +
> self.result_table_print()
> self.check_packets_of_each_queue()
> self.close_all_testpmd()
>
> - def
> test_server_mode_multi_queue_reconnect_from_vhost_side_with_virtio10(s
> elf):
> + def test_server_mode_reconnect_with_virtio11_normal_path(self):
> """
> - reconnect virtio-user from the vhost side with multi_queues
> + reconnect test with virtio 1.1 normal path and server mode
> """
> self.queue_number = 2
> self.nb_cores = 2
> - mode = "packed_vq=0,in_order=1,mrg_rxbuf=1"
> + case_info = 'virtio1.1 normal path'
> + mode = "packed_vq=1,in_order=0,mrg_rxbuf=0"
> + extern_params = '--tx-offloads=0x0 --enable-hw-vlan-strip --rss-ip'
> self.lanuch_vhost_testpmd_with_multi_queue()
> - self.lanuch_virtio_user_testpmd_with_multi_queue(mode=mode)
> + self.lanuch_virtio_user_testpmd_with_multi_queue(mode=mode,
> + extern_params=extern_params)
> self.start_to_send_packets(self.vhost, self.virtio_user)
> - self.calculate_avg_throughput("reconnet from vhost with virtio1.0",
> - "before reconnet")
> + self.calculate_avg_throughput(case_info, "before reconnet")
>
> + # reconnect from vhost
> + self.logger.info('now reconnet from vhost')
> self.relanuch_vhost_testpmd_with_multi_queue()
> self.start_to_send_packets(self.virtio_user, self.vhost)
> - self.calculate_avg_throughput("reconnet from vhost with virtio1.0",
> - "after reconnet")
> + self.calculate_avg_throughput(case_info, "reconnet from vhost")
> +
> + # reconnet from virtio
> + self.logger.info('now reconnet from virtio_user')
> + self.relanuch_virtio_testpmd_with_multi_queue(mode=mode,
> extern_params=extern_params)
> + self.start_to_send_packets(self.vhost, self.virtio_user)
> + self.calculate_avg_throughput(case_info, "reconnet from
> + virtio_user")
> +
> + # port restart
> + self.logger.info('now vhost port restart')
> + self.port_restart()
> + self.calculate_avg_throughput(case_info, "after port restart")
> +
> self.result_table_print()
> self.check_packets_of_each_queue()
> self.close_all_testpmd()
>
> - def
> test_server_mode_multi_queue_reconnect_from_virito_side_with_virtio10(s
> elf):
> + def
> test_server_mode_reconnect_with_virtio10_inorder_mergeable_path(self):
> """
> - reconnect vhost-user from the virtio side with vhost/virtio1.0 loopback
> multi_queues
> + reconnect test with virtio 1.0 inorder mergeable path and
> + server mode
> """
> self.queue_number = 2
> self.nb_cores = 2
> - mode = "packed_vq=0,in_order=1,mrg_rxbuf=1"
> + case_info = 'virtio1.0 inorder mergeable path'
> + mode = "in_order=1,mrg_rxbuf=1"
> + extern_params = '--tx-offloads=0x0 --enable-hw-vlan-strip --rss-ip'
> self.lanuch_vhost_testpmd_with_multi_queue()
> - self.lanuch_virtio_user_testpmd_with_multi_queue(mode=mode)
> + self.lanuch_virtio_user_testpmd_with_multi_queue(mode=mode,
> + extern_params=extern_params)
> self.start_to_send_packets(self.vhost, self.virtio_user)
> - self.calculate_avg_throughput("reconnet from virito with virtio1.0",
> - "before reconnet")
> + self.calculate_avg_throughput(case_info, "before reconnet")
>
> - self.relanuch_virtio_testpmd_with_multi_queue(mode=mode)
> + # reconnet from vhost
> + self.logger.info('now reconnet from vhost')
> + self.relanuch_vhost_testpmd_with_multi_queue()
> + self.start_to_send_packets(self.virtio_user, self.vhost)
> + self.calculate_avg_throughput(case_info, "reconnet from vhost")
> +
> + # reconnet from virtio
> + self.logger.info('now reconnet from virtio_user')
> + self.relanuch_virtio_testpmd_with_multi_queue(mode=mode,
> + extern_params=extern_params)
> self.start_to_send_packets(self.vhost, self.virtio_user)
> - self.calculate_avg_throughput("reconnet from virito with virtio1.0",
> - "after reconnet")
> + self.calculate_avg_throughput(case_info, "reconnet from
> + virtio_user")
> +
> + # port restart
> + self.logger.info('now vhost port restart')
> + self.port_restart()
> + self.calculate_avg_throughput(case_info, "after port restart")
> +
> self.result_table_print()
> + self.check_packets_of_each_queue()
> self.close_all_testpmd()
>
> - def test_server_mode_port_start_stop_at_vhost(self):
> + def
> test_server_mode_reconnect_with_virtio10_inorder_no_mergeable_path(sel
> f):
> """
> - port start/stop at vhost side with server mode multi queues
> + reconnect test with virtio 1.0 inorder no_mergeable path and
> + server mode
> """
> self.queue_number = 2
> self.nb_cores = 2
> - mode = "packed_vq=1,in_order=0,mrg_rxbuf=1"
> - extern_params = "--txd=1024 --rxd=1024"
> -
> self.lanuch_vhost_testpmd_with_multi_queue(extern_params=extern_para
> ms)
> + case_info = 'virtio1.0 inorder no_mergeable path'
> + mode = "in_order=1,mrg_rxbuf=0"
> + extern_params = '--tx-offloads=0x0 --enable-hw-vlan-strip --rss-ip'
> + self.lanuch_vhost_testpmd_with_multi_queue()
> self.lanuch_virtio_user_testpmd_with_multi_queue(mode=mode,
> extern_params=extern_params)
> self.start_to_send_packets(self.vhost, self.virtio_user)
> - self.calculate_avg_throughput("vhost port restart", "before restart")
> + self.calculate_avg_throughput(case_info, "before reconnet")
> +
> + # reconnet from vhost
> + self.logger.info('now reconnet from vhost')
> + self.relanuch_vhost_testpmd_with_multi_queue()
> + self.start_to_send_packets(self.virtio_user, self.vhost)
> + self.calculate_avg_throughput(case_info, "reconnet from vhost")
>
> + # reconnet from virtio
> + self.logger.info('now reconnet from virtio_user')
> + self.relanuch_virtio_testpmd_with_multi_queue(mode=mode,
> extern_params=extern_params)
> + self.start_to_send_packets(self.vhost, self.virtio_user)
> + self.calculate_avg_throughput(case_info, "reconnet from
> + virtio_user")
> +
> + # port restart
> + self.logger.info('now vhost port restart')
> self.port_restart()
> - self.calculate_avg_throughput("vhost port restart", "after restart")
> + self.calculate_avg_throughput(case_info, "after port restart")
> +
> self.result_table_print()
> + self.check_packets_of_each_queue()
> + self.close_all_testpmd()
> +
> + def test_server_mode_reconnect_with_virtio10_mergeable_path(self):
> + """
> + reconnect test with virtio 1.0 mergeable path and server mode
> + """
> + self.queue_number = 2
> + self.nb_cores = 2
> + case_info = 'virtio1.0 mergeable path'
> + mode = "in_order=0,mrg_rxbuf=1"
> + extern_params = '--tx-offloads=0x0 --enable-hw-vlan-strip --rss-ip'
> + self.lanuch_vhost_testpmd_with_multi_queue()
> + self.lanuch_virtio_user_testpmd_with_multi_queue(mode=mode,
> extern_params=extern_params)
> + self.start_to_send_packets(self.vhost, self.virtio_user)
> + self.calculate_avg_throughput(case_info, "before reconnet")
> +
> + # reconnet from vhost
> + self.logger.info('now reconnet from vhost')
> + self.relanuch_vhost_testpmd_with_multi_queue()
> + self.start_to_send_packets(self.virtio_user, self.vhost)
> + self.calculate_avg_throughput(case_info, "reconnet from vhost")
>
> + # reconnet from virtio
> + self.logger.info('now reconnet from virtio_user')
> + self.relanuch_virtio_testpmd_with_multi_queue(mode=mode,
> extern_params=extern_params)
> + self.start_to_send_packets(self.vhost, self.virtio_user)
> + self.calculate_avg_throughput(case_info, "reconnet from
> + virtio_user")
> +
> + # port restart
> + self.logger.info('now vhost port restart')
> + self.port_restart()
> + self.calculate_avg_throughput(case_info, "after port restart")
> +
> + self.result_table_print()
> + self.check_packets_of_each_queue()
> + self.close_all_testpmd()
> +
> + def test_server_mode_reconnect_with_virtio10_normal_path(self):
> + """
> + reconnect test with virtio 1.0 normal path and server mode
> + """
> + self.queue_number = 2
> + self.nb_cores = 2
> + case_info = 'virtio1.0 normal path'
> + mode = "in_order=0,mrg_rxbuf=0"
> + extern_params = '--tx-offloads=0x0 --enable-hw-vlan-strip --rss-ip'
> + self.lanuch_vhost_testpmd_with_multi_queue()
> + self.lanuch_virtio_user_testpmd_with_multi_queue(mode=mode,
> extern_params=extern_params)
> + self.start_to_send_packets(self.vhost, self.virtio_user)
> + self.calculate_avg_throughput(case_info, "before reconnet")
> +
> + # reconnet from vhost
> + self.logger.info('now reconnet from vhost')
> + self.relanuch_vhost_testpmd_with_multi_queue()
> + self.start_to_send_packets(self.virtio_user, self.vhost)
> + self.calculate_avg_throughput(case_info, "reconnet from vhost")
> +
> + # reconnet from virtio
> + self.logger.info('now reconnet from virtio_user')
> + self.relanuch_virtio_testpmd_with_multi_queue(mode=mode,
> extern_params=extern_params)
> + self.start_to_send_packets(self.vhost, self.virtio_user)
> + self.calculate_avg_throughput(case_info, "reconnet from
> + virtio_user")
> +
> + # port restart
> + self.logger.info('now vhost port restart')
> + self.port_restart()
> + self.calculate_avg_throughput(case_info, "after port restart")
> +
> + self.result_table_print()
> + self.check_packets_of_each_queue()
> + self.close_all_testpmd()
> +
> + def test_server_mode_reconnect_with_virtio10_vector_rx_path(self):
> + """
> + reconnect test with virtio 1.0 vector_rx path and server mode
> + """
> + self.queue_number = 2
> + self.nb_cores = 2
> + case_info = 'virtio1.0 vector_rx path'
> + mode = "in_order=0,mrg_rxbuf=0"
> + self.lanuch_vhost_testpmd_with_multi_queue()
> + self.lanuch_virtio_user_testpmd_with_multi_queue(mode=mode)
> + self.start_to_send_packets(self.vhost, self.virtio_user)
> + self.calculate_avg_throughput(case_info, "before reconnet")
> +
> + # reconnet from vhost
> + self.logger.info('now reconnet from vhost')
> + self.relanuch_vhost_testpmd_with_multi_queue()
> + self.start_to_send_packets(self.virtio_user, self.vhost)
> + self.calculate_avg_throughput(case_info, "reconnet from vhost")
> +
> + # reconnet from virtio
> + self.logger.info('now reconnet from virtio_user')
> + self.relanuch_virtio_testpmd_with_multi_queue(mode=mode)
> + self.start_to_send_packets(self.vhost, self.virtio_user)
> + self.calculate_avg_throughput(case_info, "reconnet from
> + virtio_user")
> +
> + # port restart
> + self.logger.info('now vhost port restart')
> + self.port_restart()
> + self.calculate_avg_throughput(case_info, "after port restart")
> +
> + self.result_table_print()
> self.check_packets_of_each_queue()
> self.close_all_testpmd()
>
> --
> 2.7.4
prev parent reply other threads:[~2019-08-07 3:38 UTC|newest]
Thread overview: 2+ messages / expand[flat|nested] mbox.gz Atom feed top
2019-07-15 1:39 lihong
2019-08-07 3:37 ` Tu, Lijuan [this message]
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=8CE3E05A3F976642AAB0F4675D0AD20E0BAFEC4D@SHSMSX101.ccr.corp.intel.com \
--to=lijuan.tu@intel.com \
--cc=dts@dpdk.org \
--cc=lihongx.ma@intel.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).