From: "Burakov, Anatoly" <anatoly.burakov@intel.com>
To: oulijun <oulijun@huawei.com>,
"Hunt, David" <david.hunt@intel.com>,
David Marchand <david.marchand@redhat.com>
Cc: Van Haaren Harry <harry.van.haaren@intel.com>, dev <dev@dpdk.org>,
"Pattan, Reshma" <reshma.pattan@intel.com>
Subject: Re: [dpdk-dev] 【BUG REPORT】l3fwd-power can not exit by ctrl+c
Date: Wed, 27 May 2020 09:57:43 +0100 [thread overview]
Message-ID: <09a046a5-c04a-33f2-cf91-b78160e4443a@intel.com> (raw)
In-Reply-To: <db078078-6326-f4c1-24b5-4fbc05481e0b@huawei.com>
On 27-May-20 7:04 AM, oulijun wrote:
>
>
> 在 2020/5/26 17:24, Burakov, Anatoly 写道:
>> On 26-May-20 4:50 AM, oulijun wrote:
>>> Hi,
>>> I have update the code into 20.05-rc2. However, the l3fwd-power
>>> startup fail.
>>>
>>> [root@centos-C3 build]# l3fwd-power -w 0000:7d:00.1 -c 0xc000000 -n
>>> 4 -- -P -p 0x01 --config '(0,0,27)' --parse-ptype
>>> EAL: Detected 128 lcore(s)
>>> EAL: Detected 4 NUMA nodes
>>> EAL: Multi-process socket /var/run/dpdk/rte/mp_socket
>>> EAL: Selected IOVA mode 'PA'
>>> EAL: No available hugepages reported in hugepages-2048kB
>>> EAL: No available hugepages reported in hugepages-32768kB
>>> EAL: No available hugepages reported in hugepages-64kB
>>> EAL: Probing VFIO support...
>>> EAL: VFIO support initialized
>>> EAL: using IOMMU type 1 (Type 1)
>>> EAL: Ignore mapping IO port bar(1)
>>> EAL: Ignore mapping IO port bar(3)
>>> EAL: Probe PCI driver: net_hns3 (19e5:a221) device: 0000:7d:00.1
>>> (socket 0)
>>> No telemetry legacy support - No legacy callbacks, legacy socket not
>>> created
>>> Promiscuous mode selected
>>> soft parse-ptype is enabled
>>> POWER: Env isn't set yet!
>>> POWER: Attempting to initialise ACPI cpufreq power management...
>>> POWER: Power management governor of lcore 26 has been set to user
>>> space successfully
>>> POWER: File not opened
>>> POWER: Cannot get available frequencies of lcore 26
>>> POWER: Attempting to initialise PSTAT power management...
>>> POWER: Power management governor of lcore 26 has been set to
>>> performance successfully
>>> POWER: Error opening '/dev/cpu/26/msr': No such file or directory
>>> POWER: Cannot init for setting frequency for lcore 26
>>> POWER: Attempting to initialise VM power management...
>>> GUEST_CHANNEL: Opening channel
>>> '/dev/virtio-ports/virtio.serial.port.poweragent.26' for lcore 26
>>> GUEST_CHANNEL: Unable to to connect to
>>> '/dev/virtio-ports/virtio.serial.port.poweragent.26' with error No
>>> such file or directory
>>> POWER: Unable to set Power Management Environment for lcore 26
>>> POWER: Library initialization failed on core 26
>>> EAL: Error - exiting with code: 1
>>> Cause: init_power_library failed
>>>
>>> Thanks
>>> Lijun Ou
>>>
>>
>> Hi,
>>
>> Previously, l3fwd-power was very lax in what it was allowing. Now, if
>> it can't enable power management, it will not run, because it is more
>> strict in what it allows.
>>
>> As is shown in the log, it tries to initialize ACPI power management,
>> but fails. Then, it tries to initialize pstate mode, and it appears
>> that MSR driver is not loaded, so it doesn't work either. It also
>> tries to check if a KVM channel exists, which doesn't, and that fails
>> as well. So, no power management environment can be enabled, and the
>> application fails to start.
>>
>> In order to make it work, you should either boot in ACPI mode (kernel
>> parameter: "intel_pstate=disable"), or load the MSR driver (root
>> command: modprobe msr).
>>
>> This more strict power library initialization was added because we saw
>> people misusing l3fwd-power to attempt to run without power management
>> environment being initialized (such as inside a VM while not having
>> telemetry mode enabled), which it was not designed to do and which
>> would be an invalid test.
>>
> thanks. I will try it
>
In another email, you said that your system is arm64. The pstate mode
won't work on that (it's an Intel-specific driver, after all), and i'm
not sure if ACPI would work either, we never tested it on arm64 and i'm
not even sure it's _supposed to_ work. The librte_power library has
always been IA-specific.
I have a sneaking suspicion that you too are attempting to use
l3fwd-power to test interrupts, rather than the librte_power library. Is
that the case?
--
Thanks,
Anatoly
next prev parent reply other threads:[~2020-05-27 8:57 UTC|newest]
Thread overview: 19+ messages / expand[flat|nested] mbox.gz Atom feed top
2020-05-20 3:18 oulijun
2020-05-20 7:22 ` David Marchand
2020-05-20 8:41 ` oulijun
2020-05-20 8:49 ` David Marchand
2020-05-20 11:37 ` Burakov, Anatoly
2020-05-22 8:17 ` oulijun
2020-05-22 9:00 ` Burakov, Anatoly
2020-05-22 9:02 ` Hunt, David
2020-05-26 3:50 ` oulijun
2020-05-26 8:36 ` Hunt, David
2020-05-26 9:11 ` oulijun
2020-05-27 6:10 ` oulijun
2020-05-27 8:21 ` Hunt, David
2020-05-26 9:24 ` Burakov, Anatoly
2020-05-27 6:04 ` oulijun
2020-05-27 8:57 ` Burakov, Anatoly [this message]
2020-05-20 9:45 ` Burakov, Anatoly
2020-05-21 1:26 ` oulijun
2020-05-21 10:46 ` Burakov, Anatoly
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=09a046a5-c04a-33f2-cf91-b78160e4443a@intel.com \
--to=anatoly.burakov@intel.com \
--cc=david.hunt@intel.com \
--cc=david.marchand@redhat.com \
--cc=dev@dpdk.org \
--cc=harry.van.haaren@intel.com \
--cc=oulijun@huawei.com \
--cc=reshma.pattan@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).