Re: [PATCH v4 15/17] module: Introduce hash-based integrity checking

[Nicolas Schier <nsc@kernel.org>]

On Tue, Feb 03, 2026 at 01:55:05PM +0100, Thomas Weißschuh wrote:
> On 2026-01-30 18:06:20+0100, Petr Pavlu wrote:
> > On 1/13/26 1:28 PM, Thomas Weißschuh wrote:
> > > Normally the .ko module files depend on a fully built vmlinux to be
> > > available for modpost validation and BTF generation. With
> > > CONFIG_MODULE_HASHES, vmlinux now depends on the modules
> > > to build a merkle tree. This introduces a dependency cycle which is
> > > impossible to satisfy. Work around this by building the modules during
> > > link-vmlinux.sh, after vmlinux is complete enough for modpost and BTF
> > > but before the final module hashes are
> > 
> > I wonder if this dependency cycle could be resolved by utilizing the
> > split into vmlinux.unstripped and vmlinux that occurred last year.
> > 
> > The idea is to create the following ordering: vmlinux.unstripped ->
> > modules -> vmlinux, and to patch in .module_hashes only when building
> > the final vmlinux.
> > 
> > This would require the following:
> > * Split scripts/Makefile.vmlinux into two Makefiles, one that builds the
> >   current vmlinux.unstripped and the second one that builds the final
> >   vmlinux from it.
> > * Modify the top Makefile to recognize vmlinux.unstripped and update the
> >   BTF generation rule 'modules: vmlinux' to
> >   'modules: vmlinux.unstripped'.
> > * Add the 'vmlinux: modules' ordering in the top Makefile for
> >   CONFIG_MODULE_HASHES=y.
> > * Remove the patching of vmlinux.unstripped in scripts/link-vmlinux.sh
> >   and instead move it into scripts/Makefile.vmlinux when running objcopy
> >   to produce the final vmlinux.
> > 
> > I think this approach has two main advantages:
> > * CONFIG_MODULE_HASHES can be made orthogonal to
> >   CONFIG_DEBUG_INFO_BTF_MODULES.
> > * All dependencies are expressed at the Makefile level instead of having
> >   scripts/link-vmlinux.sh invoke 'make -f Makefile modules'.
> > 
> > Below is a rough prototype that applies on top of this series. It is a
> > bit verbose due to the splitting of part of scripts/Makefile.vmlinux
> > into scripts/Makefile.vmlinux_unstripped.
> 
> That looks like a feasible alternative. Before adopting it, I'd like to
> hear the preference of the kbuild folks.

After the first run-through, the proposed alternative sounds good.
Unfortunately, I ran out of time for this week.  I can give a more
founded reply in a few days.

Kind regards,
Nicolas



> > diff --git a/Makefile b/Makefile
> > index 841772a5a260..19a3beb82fa7 100644
> > --- a/Makefile
> > +++ b/Makefile
> > @@ -1259,7 +1259,7 @@ vmlinux_o: vmlinux.a $(KBUILD_VMLINUX_LIBS)
> >  vmlinux.o modules.builtin.modinfo modules.builtin: vmlinux_o
> >  	@:
> >  
> > -PHONY += vmlinux
> > +PHONY += vmlinux.unstripped vmlinux
> >  # LDFLAGS_vmlinux in the top Makefile defines linker flags for the top vmlinux,
> >  # not for decompressors. LDFLAGS_vmlinux in arch/*/boot/compressed/Makefile is
> >  # unrelated; the decompressors just happen to have the same base name,
> > @@ -1270,9 +1270,11 @@ PHONY += vmlinux
> >  #   https://savannah.gnu.org/bugs/?61463
> >  # For Make > 4.4, the following simple code will work:
> >  #  vmlinux: private export LDFLAGS_vmlinux := $(LDFLAGS_vmlinux)
> > -vmlinux: private _LDFLAGS_vmlinux := $(LDFLAGS_vmlinux)
> > -vmlinux: export LDFLAGS_vmlinux = $(_LDFLAGS_vmlinux)
> > -vmlinux: vmlinux.o $(KBUILD_LDS) modpost
> > +vmlinux.unstripped: private _LDFLAGS_vmlinux := $(LDFLAGS_vmlinux)
> > +vmlinux.unstripped: export LDFLAGS_vmlinux = $(_LDFLAGS_vmlinux)
> > +vmlinux.unstripped: vmlinux.o $(KBUILD_LDS) modpost
> > +	$(Q)$(MAKE) -f $(srctree)/scripts/Makefile.vmlinux_unstripped
> > +vmlinux: vmlinux.unstripped
> >  	$(Q)$(MAKE) -f $(srctree)/scripts/Makefile.vmlinux
> 
> Maybe we could keep them together in a single Makefile,
> and instead have different targets in it.
> 
> (...)
> 
> > @@ -98,70 +44,15 @@ remove-symbols := -w --strip-unneeded-symbol='__mod_device_table__*'
> >  # To avoid warnings: "empty loadable segment detected at ..." from GNU objcopy,
> >  # it is necessary to remove the PT_LOAD flag from the segment.
> >  quiet_cmd_strip_relocs = OBJCOPY $@
> > -      cmd_strip_relocs = $(OBJCOPY) $(patsubst %,--set-section-flags %=noload,$(remove-section-y)) $< $@; \
> > -                         $(OBJCOPY) $(addprefix --remove-section=,$(remove-section-y)) $(remove-symbols) $@
> > +      cmd_script_relocs = $(OBJCOPY) $(patsubst %,--set-section-flags %=noload,$(remove-section-y)) $< $@; \
> > +                          $(OBJCOPY) $(addprefix --remove-section=,$(remove-section-y)) \
> > +                                     $(remove-symbols) \
> > +                                     $(patch-module-hashes) $@
> 
> cmd_script_relocs -> cmd_strip_relocs
> 
> (...)

-- 
Nicolas