From: Ferruh Yigit <ferruh.yigit@intel.com>
To: Jerin Jacob <jerin.jacob@caviumnetworks.com>
Cc: dev@dpdk.org, Angela Czubak <aczubak@caviumnetworks.com>,
Thomas Monjalon <thomas@monjalon.net>
Subject: Re: [dpdk-dev] [PATCH] net/thunderx: manage PCI device mapping for SQS VFs
Date: Tue, 6 Jun 2017 15:54:35 +0100 [thread overview]
Message-ID: <b5b09625-db48-9f8e-cf53-ce79b5b32dc4@intel.com> (raw)
In-Reply-To: <20170606140506.GA31583@jerin>
On 6/6/2017 3:05 PM, Jerin Jacob wrote:
> -----Original Message-----
>> Date: Tue, 6 Jun 2017 14:36:09 +0100
>> From: Ferruh Yigit <ferruh.yigit@intel.com>
>> To: Jerin Jacob <jerin.jacob@caviumnetworks.com>, dev@dpdk.org
>> CC: Angela Czubak <aczubak@caviumnetworks.com>, Thomas Monjalon
>> <thomas@monjalon.net>
>> Subject: Re: [dpdk-dev] [PATCH] net/thunderx: manage PCI device mapping for
>> SQS VFs
>> User-Agent: Mozilla/5.0 (Windows NT 10.0; WOW64; rv:52.0) Gecko/20100101
>> Thunderbird/52.1.1
>>
>> On 6/1/2017 2:05 PM, Jerin Jacob wrote:
>>> Since the commit e84ad157b7bc ("pci: unmap resources if probe fails"),
>>> EAL unmaps the PCI device if ethdev probe returns positive or
>>> negative value.
>>>
>>> nicvf thunderx PMD needs special treatment for Secondary queue set(SQS)
>>> PCIe VF devices, where, it expects to not unmap or free the memory
>>> without registering the ethdev subsystem.
>>>
>>> To keep the same behavior, moved the PCI map function inside
>>> the driver without using the EAL services.
>>
>> What do you think adding a flag something like
>> RTE_PCI_DRV_FIXED_MAPPING? Does mapping but not unmap on error.
>> This would be more generic solution.
>>
>> I am concerned about calling eal level API from PMD.
>
> Understood.
>
> Another option is to unmap only on ERROR(ie, when probe return <0 value)
>
> ret = dr->probe(dr, dev);
> if (ret) { // change to if (ret < 0)
> dev->driver = NULL;
> if (dr->drv_flags & RTE_PCI_DRV_NEED_MAPPING)
> rte_pci_unmap_device(dev);
> }
>
> I am fine with either way. Let me know, what you prefer. I will
> change accordingly.
"unmap only on ERROR" looks simpler, but it needs to be documented -in
the code, otherwise easy to miss in the future:
probe() return:
0 : success
< 0 : error, unmap resources
> 0 : error, no unmap
And requires to check that all existing drivers return negative error on
probe()
Adding new flag is more explicit, and no need to concern about what
other PMDs return, but can be overkill.
I would go with second option, but I guess both are OK, -as long as
behavior change in first one is commented in the code. Please pick one.
Thanks,
ferruh
>
>>
>>>
>>> Signed-off-by: Jerin Jacob <jerin.jacob@caviumnetworks.com>
>>> Signed-off-by: Angela Czubak <aczubak@caviumnetworks.com>
>>> ---
>>> drivers/net/thunderx/nicvf_ethdev.c | 9 ++++++++-
>>> 1 file changed, 8 insertions(+), 1 deletion(-)
>>>
>>> diff --git a/drivers/net/thunderx/nicvf_ethdev.c b/drivers/net/thunderx/nicvf_ethdev.c
>>> index 796701b0f..6ec2f9266 100644
>>> --- a/drivers/net/thunderx/nicvf_ethdev.c
>>> +++ b/drivers/net/thunderx/nicvf_ethdev.c
>>> @@ -2025,6 +2025,13 @@ nicvf_eth_dev_init(struct rte_eth_dev *eth_dev)
>>> }
>>>
>>> pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
>>> +
>>> + ret = rte_pci_map_device(pci_dev);
>>> + if (ret) {
>>> + PMD_INIT_LOG(ERR, "Failed to map pci device");
>>> + goto fail;
>>> + }
>>> +
>>> rte_eth_copy_pci_info(eth_dev, pci_dev);
>>>
>>> nic->device_id = pci_dev->id.device_id;
>>> @@ -2171,7 +2178,7 @@ static int nicvf_eth_pci_remove(struct rte_pci_device *pci_dev)
>>>
>>> static struct rte_pci_driver rte_nicvf_pmd = {
>>> .id_table = pci_id_nicvf_map,
>>> - .drv_flags = RTE_PCI_DRV_NEED_MAPPING | RTE_PCI_DRV_INTR_LSC,
>>> + .drv_flags = RTE_PCI_DRV_INTR_LSC,
>>> .probe = nicvf_eth_pci_probe,
>>> .remove = nicvf_eth_pci_remove,
>>> };
>>>
>>
next prev parent reply other threads:[~2017-06-06 14:54 UTC|newest]
Thread overview: 21+ messages / expand[flat|nested] mbox.gz Atom feed top
2017-06-01 13:05 Jerin Jacob
2017-06-06 13:36 ` Ferruh Yigit
2017-06-06 14:05 ` Jerin Jacob
2017-06-06 14:54 ` Ferruh Yigit [this message]
2017-06-08 11:44 ` [dpdk-dev] [PATCH v2 1/2] eal/pci: introduce a PCI driver flag Jerin Jacob
2017-06-08 11:44 ` [dpdk-dev] [PATCH v2 2/2] net/thunderx: manage PCI device mapping for SQS VFs Jerin Jacob
2017-06-08 14:40 ` [dpdk-dev] [PATCH v2 1/2] eal/pci: introduce a PCI driver flag Ferruh Yigit
2017-06-08 17:15 ` Jerin Jacob
2017-06-08 19:44 ` Ferruh Yigit
2017-06-09 4:35 ` Jerin Jacob
2017-06-09 9:13 ` Ferruh Yigit
2017-06-09 9:27 ` Jerin Jacob
2017-06-09 10:27 ` [dpdk-dev] [PATCH v3 " Jerin Jacob
2017-06-09 10:27 ` [dpdk-dev] [PATCH v3 2/2] net/thunderx: manage PCI device mapping for SQS VFs Jerin Jacob
2017-06-09 10:46 ` [dpdk-dev] [PATCH v3 1/2] eal/pci: introduce a PCI driver flag Ferruh Yigit
2017-06-12 16:02 ` Ferruh Yigit
2017-06-12 16:21 ` Thomas Monjalon
2017-06-13 4:43 ` Jerin Jacob
2017-06-13 7:03 ` Thomas Monjalon
2017-06-13 7:24 ` Jerin Jacob
2017-06-13 8:18 ` Thomas Monjalon
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=b5b09625-db48-9f8e-cf53-ce79b5b32dc4@intel.com \
--to=ferruh.yigit@intel.com \
--cc=aczubak@caviumnetworks.com \
--cc=dev@dpdk.org \
--cc=jerin.jacob@caviumnetworks.com \
--cc=thomas@monjalon.net \
/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).