Re: [sparc64] unable to build kernel after bbfceba15f8 / "kdb: Get rid of confusing diag msg from "rd" if current task has no regs"

[Daniel Thompson <daniel.thompson@xxxxxxxxxx>]

On Wed, Feb 05, 2020 at 12:41:46PM +0300, Anatoly Pugachev wrote:
> On Wed, Feb 5, 2020 at 1:13 AM Doug Anderson <dianders@xxxxxxxxxxxx> wrote:
> >
> > Hi,
> >
> > On Tue, Feb 4, 2020 at 1:14 PM Anatoly Pugachev <matorola@xxxxxxxxx> wrote:
> > >
> > > Hello!
> > >
> > > After (bisected) bbfceba15f8d1260c328a254efc2b3f2deae4904 got
> > > mainline, i'm unable to build kernel with the following messages:
> > >
> > > ~/linux-2.6$ make -j olddefconfig; make kernel/debug/kdb/
> > > scripts/kconfig/conf  --olddefconfig Kconfig
> > > #
> > > # No change to .config
> > > #
> > >   CALL    scripts/atomic/check-atomics.sh
> > >   CALL    scripts/checksyscalls.sh
> > > <stdin>:1511:2: warning: #warning syscall clone3 not implemented [-Wcpp]
> > >   CC      kernel/debug/kdb/kdb_main.o
> > > kernel/debug/kdb/kdb_main.c: In function ‘kdb_rd’:
> > > kernel/debug/kdb/kdb_main.c:1857:11: error: ‘dbg_reg_def’ undeclared
> > > (first use in this function); did you mean ‘dbg_reg_def_t’?
> > >  1857 |   rsize = dbg_reg_def[i].size * 2;
> > >       |           ^~~~~~~~~~~
> > >       |           dbg_reg_def_t
> > > kernel/debug/kdb/kdb_main.c:1857:11: note: each undeclared identifier
> > > is reported only once for each function it appears in
> > > kernel/debug/kdb/kdb_main.c:1868:12: error: implicit declaration of
> > > function ‘dbg_get_reg’ [-Werror=implicit-function-declaration]
> > >  1868 |    rname = dbg_get_reg(i, &reg8, kdb_current_regs);
> > >       |            ^~~~~~~~~~~
> > > kernel/debug/kdb/kdb_main.c:1868:10: warning: assignment to ‘char *’
> > > from ‘int’ makes pointer from integer without a cast
> > > [-Wint-conversion]
> > >  1868 |    rname = dbg_get_reg(i, &reg8, kdb_current_regs);
> > >       |          ^
> > > kernel/debug/kdb/kdb_main.c:1874:10: warning: assignment to ‘char *’
> > > from ‘int’ makes pointer from integer without a cast
> > > [-Wint-conversion]
> > >  1874 |    rname = dbg_get_reg(i, &reg16, kdb_current_regs);
> > >       |          ^
> > > kernel/debug/kdb/kdb_main.c:1880:10: warning: assignment to ‘char *’
> > > from ‘int’ makes pointer from integer without a cast
> > > [-Wint-conversion]
> > >  1880 |    rname = dbg_get_reg(i, &reg32, kdb_current_regs);
> > >       |          ^
> > > kernel/debug/kdb/kdb_main.c:1886:10: warning: assignment to ‘char *’
> > > from ‘int’ makes pointer from integer without a cast
> > > [-Wint-conversion]
> > >  1886 |    rname = dbg_get_reg(i, &reg64, kdb_current_regs);
> > >       |          ^
> > > cc1: some warnings being treated as errors
> > > make[3]: *** [scripts/Makefile.build:266: kernel/debug/kdb/kdb_main.o] Error 1
> > > make[2]: *** [scripts/Makefile.build:503: kernel/debug/kdb] Error 2
> > > make[1]: *** [scripts/Makefile.build:503: kernel/debug] Error 2
> > > make: *** [Makefile:1681: kernel] Error 2
> > >
> > >
> > > reversing bbfceba15f8d with current git master branch - makes kernel
> > > compilable again.
> > >
> > > kernel config attached
> >
> > Argh.  Sorry about this.  :(  I'm not setup with a sparc compiler, but
> > I've simulated your results on arm64 by just commenting out the
> > definition of "DBG_MAX_REG_NUM" in "arch/arm64/include/asm/kgdb.h".
> > Hopefully:
> >
> > https://lore.kernel.org/r/20200204141219.1.Ief3f3a7edbbd76165901b14813e90381c290786d@changeid
> >
> > ...will fix you?
> 
> Fixes issue for me. Thank you.

Thanks for the update. I'll pass it along shortly.

Somewhat ironically I don't think I ever enabled pre-release sparc
build testing because of conflicts between the way the sparc kernel
adds -Werror to some of the architecture specific files and the tools I
use to detect new warnings. I might have to look at that as well!


Daniel.