RE: rdma_rxe and the loopback network interface

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 




> -----Original Message-----
> From: linux-rdma-owner@xxxxxxxxxxxxxxx <linux-rdma-
> owner@xxxxxxxxxxxxxxx> On Behalf Of Zhu Yanjun
> Sent: Friday, March 1, 2019 7:02 PM
> To: Bart Van Assche <bvanassche@xxxxxxx>; linux-rdma@xxxxxxxxxxxxxxx
> Subject: Re: rdma_rxe and the loopback network interface
> 
> 
> On 2019/3/2 上午8:26, Bart Van Assche wrote:
> > On Sat, 2019-03-02 at 07:39 +0800, Yanjun Zhu wrote:
> >> On 2019/2/27 6:48, Bart Van Assche wrote:
> >>> Because of security reasons I would like to run some RDMA tests in a
> >>> virtual machine with the rdma_rxe driver attached to the loopback
> >>> ("lo") interface and with no other network interfaces configured.
> >>> Although it is possible to associate the rdma_rxe network interface
> >>> I have not yet found a way to let the RDMA/CM set up a connection
> from address ::1 to ::1.
> >> Sorry. Can you use ipv4 address instead of ipv6 address?
> > Hi Yanjun,
> >
> > Login with IPv4 also fails. The logs that appear if I run blktests on
> > top of a v5.0-rc6 kernel with several debug patches applied are as follows:
> >
> > kernel: device-mapper: multipath service-time: version 0.3.0 loaded
> > kernel: ib_srp:srp_parse_in: ib_srp: 127.0.0.1:5555 -> 127.0.0.1:5555
> > kernel: ib_srp:srp_create_target: ib_srp: max_sectors = 1024;
> > max_pages_per_mr = 512; mr_page_size = 4096; max_sectors_per_mr =
> > 4096; mr_per_cmd = 2
> > kernel: ib_srp:srp_create_ch_ib: ib_srp: 1; dev 00000000a502471d
> > (rxe0) <> 0000000061eb80ae (rxe1)
> > kernel: ib_srp: QP creation failed for dev rxe1: -22
> > multipathd[850]: mpatho: load table [0 65536 multipath 1
> > queue_if_no_path 0 1 1 service-time 0 1 1 8:32 1]
> > kernel: ib_srp:srp_parse_in: ib_srp: [::1]:5555 -> [::1]:5555/0%0
> > kernel: ib_srp:srp_create_target: ib_srp: max_sectors = 1024;
> > max_pages_per_mr = 512; mr_page_size = 4096; max_sectors_per_mr =
> > 4096; mr_per_cmd = 2
> > kernel: ib_srp:srp_create_ch_ib: ib_srp: 1; dev 00000000a502471d
> > (rxe0) <> 0000000061eb80ae (rxe1)
> > multipathd[850]: mpatho: event checker started
> > multipathd[850]: sdc [8:32]: path added to devmap mpatho
> > kernel: ib_srp: QP creation failed for dev rxe1: -22
> >
> > rxe0 is associated with an Ethernet network interface and rxe1 is
> > associated with the loopback interface. QP creation fails because the
> > RDMA CM/ID is associated with another RDMA device than the RDMA
> > protection domain. So the root cause here is that calling
> > rdma_resolve_route() with either 127.0.0.1 or ::1 as source and/or
> > destination addresses associates the RDMA CM/ID with the wrong RDMA
> device.
> 
> Thanks a lot. Let me try to reproduce this problem.
> 
Its not a rdma_rxe issue.

This patch [1] should help for IPv4 127.0.0.1 to make progress by binding rxe to lo device.
[1] https://patchwork.kernel.org/patch/10831261/

I have been using it with loopback experimental driver with nvme-fabrics, rping, perftests.

> Zhu Yanjun
> 
> 
> > Bart.




[Index of Archives]     [Linux USB Devel]     [Video for Linux]     [Linux Audio Users]     [Photo]     [Yosemite News]     [Yosemite Photos]     [Linux Kernel]     [Linux SCSI]     [XFree86]

  Powered by Linux