[ovs-discuss] pci_map_resource(): cannot mmap(..): Invalid argument (0xffffffffffffffff)

[Charlie Li]

Thanks Ian,

That was exactly my problem.
After commenting out "pci_request_regions(dev, "igb_uio");" in igb_uio.c,
the issue was resolved.

Charlie

On Fri, Jul 1, 2016 at 6:07 AM, Stokes, Ian <ian.stokes at intel.com> wrote:

> Hi,
>
>
>
> This sounds similar to an issue discussed on the DPDK mailing list, essentially if CONFIG_IO_STRICT_DEVMEM is turned on in the kernel this behavior occurs. More details in the link below.
>
>
>
> http://dpdk.org/ml/archives/dev/2016-July/043122.html
>
>
>
> Are you using a kernel 4.5+ with fedora 23?
>
>
>
> Thanks
>
> Ian
>
>
>
> *From:* discuss [mailto:discuss-bounces at openvswitch.org] *On Behalf Of *Charlie
> Li
> *Sent:* Wednesday, June 29, 2016 3:51 PM
> *To:* Fischetti, Antonio
>
> *Cc:* discuss at openvswitch.org
> *Subject:* Re: [ovs-discuss] pci_map_resource(): cannot mmap(..): Invalid
> argument (0xffffffffffffffff)
>
>
>
> Good point Antonio. I will try DPDK without OVS first.
>
>
>
> Thanks,
>
> Charlie
>
>
>
> On Wed, Jun 29, 2016 at 7:21 AM, Fischetti, Antonio <
> antonio.fischetti at intel.com> wrote:
>
> It’s also strange you get -1 as the NUMA socket.
>
> EAL: PCI device 0000:02:00.0 on NUMA socket -1
>
>
>
> Maybe you could have the same issue if you keep OVS off and just run some
> simple DPDK test application?
>
> For ex by running testpmd, so no OVS process is involved.
>
>
>
> Also, as this is an error message coming from DPDK code – and especially
> if you still have the same issue with a DPDK app without OVS - it could be
> a good idea to post also to the DPDK Mailing-list.
>
>
>
>
>
> Antonio
>
>
>
> *From:* Charlie Li [mailto:charlie.li at gmail.com]
> *Sent:* Tuesday, June 28, 2016 8:36 PM
>
>
> *To:* Fischetti, Antonio <antonio.fischetti at intel.com>
> *Cc:* discuss at openvswitch.org
> *Subject:* Re: [ovs-discuss] pci_map_resource(): cannot mmap(..): Invalid
> argument (0xffffffffffffffff)
>
>
>
> It does not make any difference.
>
>
>
> $ sudo ./ovs-vswitchd --dpdk -c 0x1 -n 4 --socket-mem 1024,0 --
> unix:$DB_SOCK --pidfile --detach
>
> 2016-06-28T19:34:17Z|00001|dpdk|INFO|No -vhost_sock_dir provided -
> defaulting to /usr/local/var/run/openvswitch
>
> EAL: Detected lcore 0 as core 0 on socket 0
>
> EAL: Detected lcore 1 as core 1 on socket 0
>
> EAL: Detected lcore 2 as core 2 on socket 0
>
> EAL: Detected lcore 3 as core 3 on socket 0
>
> EAL: Support maximum 128 logical core(s) by configuration.
>
> EAL: Detected 4 lcore(s)
>
> EAL: VFIO modules not all loaded, skip VFIO support...
>
> EAL: Setting up physically contiguous memory...
>
> EAL: Ask a virtual area of 0x40000000 bytes
>
> EAL: Virtual area found at 0x7fcc80000000 (size = 0x40000000)
>
> EAL: Requesting 1 pages of size 1024MB from socket 0
>
> EAL: TSC frequency is ~2096069 KHz
>
> EAL: Master lcore 0 is ready (tid=542beac0;cpuset=[0])
>
> EAL: PCI device 0000:02:00.0 on NUMA socket -1
>
> EAL:   probe driver: 8086:10fb rte_ixgbe_pmd
>
> EAL: pci_map_resource(): cannot mmap(10, 0x7fccc0000000, 0x80000, 0x0):
> Invalid argument (0xffffffffffffffff)
>
> EAL: Error - exiting with code: 1
>
>   Cause: Requested device 0000:02:00.0 cannot be used
>
>
>
> On Tue, Jun 28, 2016 at 11:38 AM, Fischetti, Antonio <
> antonio.fischetti at intel.com> wrote:
>
> Can you try to specify the socket memory parameter
>
> --socket-mem 1024,0
>
> ?
>
>
>
> So the command
>
> sudo ./ovs-vswitchd --dpdk -c 0x1 -n 4 -- unix:$DB_SOCK --pidfile --detach
>
>
>
> becomes
>
> sudo ./ovs-vswitchd --dpdk -c 0x1 -n 4 --socket-mem 1024,0
>
> -- unix:$DB_SOCK --pidfile –detach
>
>
>
>
>
> *From:* Charlie Li [mailto:charlie.li at gmail.com]
> *Sent:* Tuesday, June 28, 2016 4:41 PM
> *To:* Fischetti, Antonio <antonio.fischetti at intel.com>
> *Cc:* discuss at openvswitch.org
> *Subject:* Re: [ovs-discuss] pci_map_resource(): cannot mmap(..): Invalid
> argument (0xffffffffffffffff)
>
>
>
> Hi Antonio,
>
>
>
> Please see the dumps in the attached file.
>
>
>
> Thanks,
>
> Charlie
>
>
>
> On Tue, Jun 28, 2016 at 2:39 AM, Fischetti, Antonio <
> antonio.fischetti at intel.com> wrote:
>
> Hi Charlie,
>
> while this issue is happening with Fed23 can you check what are the 2
> outputs of
>
>
>
> 1.       sudo $DPDK_DIR/tools/dpdk_nic_bind.py --status
>
>
>
> and also
>
>
>
> 2.       lspci | grep 02:00
>
>
>
> ?
>
>
>
> Antonio
>
>
>
> *From:* discuss [mailto:discuss-bounces at openvswitch.org] *On Behalf Of *Charlie
> Li
> *Sent:* Friday, June 24, 2016 4:36 PM
> *To:* discuss at openvswitch.org
> *Subject:* [ovs-discuss] pci_map_resource(): cannot mmap(..): Invalid
> argument (0xffffffffffffffff)
>
>
>
> Hello,
>
>
>
> I am trying to set up openvswitch-2.5.0 with DPDK-2.2.0 by following "
> INSTALL.DPDK.md".
>
>
>
> First tried with Ubuntu server 16.04 and it worked as expected.
>
>
>
> But when tried with Fedora server 23, it hit the following error
>
>
>
> EAL: pci_map_resource(): cannot mmap(10, 0x7f1800000000, 0x80000, 0x0):
> Invalid argument (0xffffffffffffffff)
>
>
>
> when invoking
>
>
>
> sudo ./ovs-vswitchd --dpdk -c 0x1 -n 4 -- unix:$DB_SOCK --pidfile --detach
>
>
>
> I tested it on the same hardware platform and followed the same
> instructions. Why it only failed with Fedora?
>
>
>
> I am also attaching the detailed logs from both Ubuntu (working case) and
> Fedora (failing case).
>
>
>
> Any help is appreciated.
>
>
>
> Thanks,
>
> Charlie
>
>
>
>
>
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://openvswitch.org/pipermail/ovs-discuss/attachments/20160701/c86f938e/attachment-0002.html>