svn commit: r315333 - in stable/10/sys: conf dev/ixgbe modules/ix modules/ixv

[Eric Joyner]

I fixed the netmap build for amd64 at least with r317711. I'm running a
tinderbox build on it now, though, so I'll see if that fixes it.

On Fri, Apr 28, 2017 at 1:40 PM Eric Joyner <erj at freebsd.org> wrote:

> I notified Jeb (the original patch submitter) about that netmap symbol
> issue, but we've both been distracted with other development. I'll work on
> updating it.
>
> - Eric
>
> On Thu, Apr 27, 2017 at 3:06 PM Ngie Cooper <yaneurabeya at gmail.com> wrote:
>
>> On Thu, Apr 27, 2017 at 1:07 PM, Ed Maste <emaste at freebsd.org> wrote:
>> > On 15 March 2017 at 17:20, Eric Joyner <erj at freebsd.org> wrote:
>> >> Author: erj
>> >> Date: Wed Mar 15 21:20:17 2017
>> >> New Revision: 315333
>> >> URL: https://svnweb.freebsd.org/changeset/base/315333
>> >>
>> >> Log:
>> >>   ixgbe(4): Update to 3.2.11-k
>> >
>> > This broke tinderbox on many architectures:
>> >
>> > ia64 GENERIC and powerpc GENERIC64:
>> >
>> > /scratch/tmp/emaste/freebsd/sys/dev/ixgbe/ixv_osdep.c:39: warning: no
>> > previous prototype for 'ixv_read_pci_cfg' [-Wmissing-prototypes]
>> > /scratch/tmp/emaste/freebsd/sys/dev/ixgbe/ixv_osdep.c:45: warning: no
>> > previous prototype for 'ixv_write_pci_cfg' [-Wmissing-prototypes]
>> >
>> > sparc64 LINT:
>> >
>> > /scratch/tmp/emaste/freebsd/sys/dev/ixgbe/ix_txrx.c:43: warning:
>> > redundant redeclaration of 'ix_crcstrip' [-Wredundant-decls]
>> > /scratch/tmp/emaste/freebsd/sys/dev/ixgbe/ixgbe_netmap.h:45: warning:
>> > previous declaration of 'ix_crcstrip' was here
>> >
>> > amd64 LINT:
>> >
>> > /scratch/tmp/emaste/freebsd/sys/dev/ixgbe/ixv_netmap.c:(.text+0x0):
>> > multiple definition of `ixgbe_netmap_attach'
>> >
>> ixgbe_netmap.o:/scratch/tmp/emaste/freebsd/sys/dev/ixgbe/ixgbe_netmap.c:(.text+0x0):
>> > first defined here
>>
>> Hi Ed,
>>
>> It has to do with netmap(4) refactoring on head not being backported,
>> in combination with ixgbe/ix being MFCed in a refactored state back to
>> ^/stable/10 (it wasn't easy to backport due to PCI-IOV only being on
>> ^/stable/10 -- I tried starting this work and failed because I lacked
>> the hardware to test this out with).
>>
>> This has been known to be broken for almost two months -- a surgical
>> fix should probably be applied to do what's required to make the
>> symbol appear in the appropriate places, since this basically was a
>> direct commit to ^/stable/10 in some regards.
>>
>> Thanks,
>> -Ngie
>>
>