On 2019-08-24 19:09, Michael Butler wrote:
> On 2019-08-24 14:04, Konstantin Belousov wrote:
>> On Sat, Aug 24, 2019 at 11:02:20AM -0600, Warner Losh wrote:
>>> forward declaring struct pcpu; in md_var.h "fixes" this, but I'm not sure
>>> that's the right fix.
>> More correct way to fix it is to include sys/pcpu.h before machine/md_var.h,
>> same as all in-tree consumers of the header do, apparently.
>>
>> But another question is why the driver needs md_var.h, there are no
>> externally usable definitions there.
>
> There are uses of other variables from machine/md_var.h, notably
> cpu_feature, in linux_compat.c.
>
> Including sys/pcpu.h allows the build to continue .. as in ..
>
> *** linuxkpi/gplv2/src/linux_compat.c~ Wed Aug 7 14:36:56 2019
> --- linuxkpi/gplv2/src/linux_compat.c Sat Aug 24 18:58:08 2019
> ***************
> *** 2,7 ****
> --- 2,8 ----
> #include <sys/kernel.h>
> #if defined(__i386__) || defined(__amd64__)
> #include <machine/specialreg.h>
> + #include <sys/pcpu.h>
> #include <machine/md_var.h>
> #endif
> #include <linux/bitops.h>
>
> Locally, I've put this patch into graphics/drm-current-kmod/files so I
> don't trip over it on subsequent builds,
This is now resolved in-tree by ports SVN r510009 - thanks to all,
imb
_______________________________________________
freebsd-current@freebsd.org mailing list
https://lists.freebsd.org/mailman/listinfo/freebsd-current
To unsubscribe, send any mail to "freebsd-current-unsubscr...@freebsd.org"
