Re: Build regressions/improvements in v5.17-rc4

[Geert Uytterhoeven]

Hi Randy,

On Tue, Feb 15, 2022 at 10:14 PM Randy Dunlap <rdunlap@xxxxxxxxxxxxx> wrote:
> On 2/14/22 01:37, Geert Uytterhoeven wrote:
> > Below is the list of build error/warning regressions/improvements in
> > v5.17-rc4[1] compared to v5.16[2].
> >
> > Summarized:
> >   - build errors: +14/-2
> >   - build warnings: +23/-25

> > 23 warning regressions:
> >   + modpost: WARNING: modpost: drivers/net/ethernet/qlogic/qed/qed.o(.data+0x14410): Section mismatch in reference from the variable qed_mfw_legacy_maps to the variable .init.rodata:qed_mfw_legacy_bb_100g:  => N/A
>
> This is a little confusing. Was there a problem with qed and qede network drivers?
> Is it still a problem?

Yes there is.

> Many of the lines above and below are very similar.

Sorry, my script to filter out errors/warnings that just moved does
not handle modpost output yet.

> If there is still a problem, what $ARCH and config(s) cause it?
> I can't reproduce it on 11 $ARCH tries.
> Or maybe it is compiler-specific also.

xtensa-gcc11/xtensa-allmodconfig
http://kisskb.ellerman.id.au/kisskb/buildresult/14697521/
I don't know why it doesn't show up in the other gcc11 builds.

I had a look at the problem before, but fixing it is non-trivial:
the driver uses a non-init table of structures that contains pointers
to init data (only used early) and non-init data (used later, too).
Sprinkling __ref over the code may help, but won't protect against
introducing new bugs.

Gr{oetje,eeting}s,

                        Geert

--
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@xxxxxxxxxxxxxx

In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
                                -- Linus Torvalds