[U-Boot] [PATCH v2 00/67] net: net subsystem ops cleanup

Tomas Hlavacek tmshlvck at gmail.com
Thu Nov 22 01:40:36 CET 2012


Hello Thomas!

On Wed, Nov 21, 2012 at 7:45 PM, Langer Thomas (LQDE RD ST PON SW)
<thomas.langer at lantiq.com> wrote:
> Hello Tomas,
>
> sorry for not replying to V1 series already. I have a remark:
>
>> 2) Replace dynamic init of ops function pointers by static struct.
>
> Some ARCHs don't have full relocation until now, which breaks the static initialization of pointers!
> Please see this:
> $ grep -r "#define.*CONFIG_NEEDS_MANUAL_RELOC" *
> arch/nds32/include/asm/config.h:#define CONFIG_NEEDS_MANUAL_RELOC
> arch/sparc/include/asm/config.h:#define CONFIG_NEEDS_MANUAL_RELOC
> arch/m68k/include/asm/config.h:#define CONFIG_NEEDS_MANUAL_RELOC
> arch/avr32/include/asm/config.h:#define CONFIG_NEEDS_MANUAL_RELOC
> arch/mips/include/asm/config.h:#define CONFIG_NEEDS_MANUAL_RELOC
>
> I don't see any fixups for these cases. I assume, you haven't done any runtime test on any
> of the relevant platforms?

No... With only two exceptions and both were tested in QEMU, not on real HW.

>
> Adding this manual relocation to eth_register() should be no problem.
> But to be bisectable, please be careful to detect the difference of statically or dynamically initialized structures, because  you may have both while applying the series.

Thanks for the remark. I will do that in the next version.

Tomas


More information about the U-Boot mailing list