Re: [PATCH] ftrace: x86: Fix a compile error about get_kernel_nofault()

[Peter Zijlstra <peterz@infradead.org>]

On Thu, Feb 13, 2025 at 11:08:36AM +0100, Peter Zijlstra wrote:
> On Wed, Feb 12, 2025 at 08:52:27AM -0800, Sami Tolvanen wrote:
> > On Tue, Feb 11, 2025 at 7:49 AM Steven Rostedt <rostedt@goodmis.org> wrote:
> > >
> > > On Tue, 11 Feb 2025 11:09:14 +0100
> > > Peter Zijlstra <peterz@infradead.org> wrote:
> > >
> > > > I was aiming my patch for x86/core, but if there's a reason to expedite
> > > > them, I can stick it in x86/urgent I suppose.
> > > >
> > > > Just need a reason -- what's this compile error nonsense about, my
> > > > kernels build just fine?
> > >
> > > Masami,
> > >
> > > Do you have a config that fails to build without this fix? If so, can you
> > > please reply with it, and then this can go in as a quick fix.
> > 
> > x86 builds with both CONFIG_GENDWARFKSYMS and CONFIG_FUNCTION_TRACER
> > are broken without this fix. Here's how to reproduce:
> > 
> > $ make defconfig
> > $ ./scripts/config -e DEBUG_INFO -e DEBUG_INFO_DWARF5 -e MODVERSIONS
> > -e GENDWARFKSYMS -e FUNCTION_TRACER
> > $ make olddefconfig && make -j
> > ...
> > In file included from ./arch/x86/include/asm/asm-prototypes.h:2,
> >                  from <stdin>:3:
> > ./arch/x86/include/asm/ftrace.h: In function ‘arch_ftrace_get_symaddr’:
> > ./arch/x86/include/asm/ftrace.h:46:21: error: implicit declaration of
> > function ‘get_kernel_nofault’ [-Wimplicit-function-declaration]
> >    46 |                 if (get_kernel_nofault(instr, (u32
> > *)(fentry_ip - ENDBR_INSN_SIZE)))
> > ...
> 
> It breaks much sooner, complaining about not having dwarf.h.. let me go
> figure out what package provides that :/

Bah, ofcourse there's libdwarf-dev and libdw-dev, both providing
dwarf.h. Obviously I installed libdwarf-dev and instead I need libdw-dev
*hate*

Anyway, yes, compile now fails as advertised.

And patch fixes it -- now I need to figure out what to do about urgent,
because applying it on top of Linus' tree will create conflicts with
patches already in tip/x86/mm *sigh*.