Re: [PATCH -next] iwlwifi: dbg: work around clang bug by marking debug strings static

From: Sedat Dilek
Date: Wed Jul 24 2019 - 11:38:48 EST


On Tue, Jul 16, 2019 at 11:15 PM Luca Coelho <luca@xxxxxxxxx> wrote:
>
> On Tue, 2019-07-16 at 10:28 -0700, Nick Desaulniers wrote:
> > On Thu, Jul 11, 2019 at 7:15 PM Joe Perches <joe@xxxxxxxxxxx> wrote:
> > > On Thu, 2019-07-11 at 17:17 -0700, Nick Desaulniers wrote:
> > > > Commit r353569 in prerelease Clang-9 is producing a linkage failure:
> > > >
> > > > ld: drivers/net/wireless/intel/iwlwifi/fw/dbg.o:
> > > > in function `_iwl_fw_dbg_apply_point':
> > > > dbg.c:(.text+0x827a): undefined reference to `__compiletime_assert_2387'
> > > >
> > > > when the following configs are enabled:
> > > > - CONFIG_IWLWIFI
> > > > - CONFIG_IWLMVM
> > > > - CONFIG_KASAN
> > > >
> > > > Work around the issue for now by marking the debug strings as `static`,
> > > > which they probably should be any ways.
> > > >
> > > > Link: https://bugs.llvm.org/show_bug.cgi?id=42580
> > > > Link: https://github.com/ClangBuiltLinux/linux/issues/580
> > > > Reported-by: Arnd Bergmann <arnd@xxxxxxxx>
> > > > Reported-by: Nathan Chancellor <natechancellor@xxxxxxxxx>
> > > > Signed-off-by: Nick Desaulniers <ndesaulniers@xxxxxxxxxx>
> > > > ---
> > > > drivers/net/wireless/intel/iwlwifi/fw/dbg.c | 4 ++--
> > > > 1 file changed, 2 insertions(+), 2 deletions(-)
> > > >
> > > > diff --git a/drivers/net/wireless/intel/iwlwifi/fw/dbg.c b/drivers/net/wireless/intel/iwlwifi/fw/dbg.c
> > > > index e411ac98290d..f8c90ea4e9b4 100644
> > > > --- a/drivers/net/wireless/intel/iwlwifi/fw/dbg.c
> > > > +++ b/drivers/net/wireless/intel/iwlwifi/fw/dbg.c
> > > > @@ -2438,7 +2438,7 @@ static void iwl_fw_dbg_info_apply(struct iwl_fw_runtime *fwrt,
> > > > {
> > > > u32 img_name_len = le32_to_cpu(dbg_info->img_name_len);
> > > > u32 dbg_cfg_name_len = le32_to_cpu(dbg_info->dbg_cfg_name_len);
> > > > - const char err_str[] =
> > > > + static const char err_str[] =
> > > > "WRT: ext=%d. Invalid %s name length %d, expected %d\n";
> > >
> > > Better still would be to use the format string directly
> > > in both locations instead of trying to deduplicate it
> > > via storing it into a separate pointer.
> > >
> > > Let the compiler/linker consolidate the format.
> > > It's smaller object code, allows format/argument verification,
> > > and is simpler for humans to understand.
> >
> > Whichever Kalle prefers, I just want my CI green again.
>
> We already changed this in a later internal patch, which will reach the
> mainline (-next) soon. So let's skip this for now.
>

Do you have a link to your internal Git or patchwork for this?

I am interested in testing it.

- Sedat -