Re: [PATCH] sparc32: fix a braino in fault handling in csum_and_copy_..._user()

[Sam Ravnborg <sam@ravnborg.org>]

Hi Al,

> 
> Looking at the csum_and_copy_... instances on various architectures,
> noticing that and going "how the fuck could it possibly work and
> what moron had broken it?  Oh, lovely - it couldn't, it doesn't
> and that moron had been myself ;-/"

:-)

> 
> > Do you actually use sparc32 these
> > days?
> 
> qemu image only, TBH - I have an SS20 box, but it hadn't booted for
> about 10 years...

Looked at my "museum". At least 5 pcs of sparc32 boxes but I guess
the NVRAM needs repair on all of them.

I have a LEON4 board somewhere on the shelf too.
None of this powered up for years either - but I cannot make myself
getting rid of it. Because maybe one day I find time...

All that said - I have also once suggested to drop sun4m support,
keeping only the LEON parts. Back then there was some resistance,
but all for sentimental reasons and I can relate to that, since
I kept the sparc32 boxes around.

> > You could also kill the EX2 define while touchign the file,
> > it is no longer used after ab5e8b331244.
> 
> Er?  No EX2 in checksum_32.S...  There is one in copy_user.S,
> but that one _is_ used -
> 
> copy_user_table_end:
>         be      copy_user_last7
>          andcc  %g1, 4, %g0
> 
>         EX(ldd  [%o1], %g2, and %g1, 0xf)
>         add     %o0, 8, %o0
>         add     %o1, 8, %o1
>         EX(st   %g2, [%o0 - 0x08], and %g1, 0xf)
>         EX2(st  %g3, [%o0 - 0x04], and %g1, 0xf, %g1, sub %g1, 4)

Yeah, I'm blind. Somehow the grep output tricked me.

	Sam