[dpdk-users] wrong vermagic in igb_uio.ko
O Mahony, Billy
billy.o.mahony at intel.com
Thu Oct 5 10:57:37 CEST 2017
Hat-tip to Ferruh for the fix on this.
> From: O Mahony, Billy
> Sent: Wednesday, October 4, 2017 6:19 PM
> To: Yigit, Ferruh <ferruh.yigit at intel.com>;
> Subject: RE: wrong vermagic in igb_uio.ko
>
> Excellent. Thanks, Ferruh.
>
> On my system /lib/modules/$(uname -r)/build and source were both pointing to
> the src/headers for the other kernel src. Manually changing those fixed the
> issue.
>
> I wonder what sets up those links in the first place?
>
> > -----Original Message-----
> > From: Yigit, Ferruh
> > Sent: Wednesday, October 4, 2017 5:51 PM
> > To: O Mahony, Billy <billy.o.mahony at intel.com>;
> > Subject: RE: wrong vermagic in igb_uio.ko
> >
> > By default kernel modules uses " /lib/modules/$(shell uname -r)/build"
> > as RTE_KERNELDIR.
> >
> > But you can provide this RTE_KERNELDIR variable externally to the
> > kernel source you want to compile modules against.
> -----Original Message-----
> From: O Mahony, Billy
> Sent: Tuesday, October 3, 2017 5:18 PM
> To: users at dpdk.org
> Subject: wrong vermagic in igb_uio.ko
>
> Hi,
>
> I have two kernels installed on my system. One with some local patches applied.
>
> 4.9.30-rt20
> And
> 4.9.30-rt20-timers-patch
>
> But regardless of which kernel I boot with when I build DPDK igb_uio always has
> vermagic from just one kernel. 4.9.30-rt20. Even though the build puts it in a
> correctly named path.
>
> # modinfo ./install/lib/modules/4.9.30-rt20-timers-
> patch/extra/dpdk/igb_uio.ko | grep -i magic
> vermagic: 4.9.30-rt20 SMP preempt mod_unload
> # uname -r
> 4.9.30-rt20-timers-patch
>
> I can't see any kernel/vermagic configuration in the various config files. I'm
> pretty sure I'd figured out before that the dpdk build was pulling this info from
> /usr/src/linux/.config CONFIG_LOCALVERSION setting but it seems to be
> ignoring anything I put in that now.
>
> Where /does/ the dpdk build get the kernel version info from?
>
> Cheers,
> Billy
>
>
>
>
>
>
>
More information about the users
mailing list