linux-kernel - Re: objtool segfault with ORC unwinder enabled

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives
Hash Suite: Windows password security audit tool. GUI, reports in PDF.
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Date:   Thu, 4 Jan 2018 11:32:03 -0600
From:   Josh Poimboeuf <jpoimboe@...hat.com>
To:     Markus <M4rkusXXL@....de>
Cc:     Greg Kroah-Hartman <gregkh@...uxfoundation.org>,
        lkml <linux-kernel@...r.kernel.org>,
        Ingo Molnar <mingo@...nel.org>
Subject: Re: objtool segfault with ORC unwinder enabled

On Thu, Jan 04, 2018 at 05:56:30PM +0100, Markus wrote:
> On Thursday, 4 January 2018 16:46:13 CET Josh Poimboeuf wrote:
> > On Wed, Jan 03, 2018 at 06:26:19PM +0100, Markus wrote:
> > > > > > I'm unable to recreate.  Can you attach one of the .o files (like
> > > > > > the
> > > > > > above irq.o)?
> > > > > 
> > > > > Sure, see attached. (From vanilla linux-4.14.11.)
> > > > 
> > > > There's something weird with the toolchain.  The object file doesn't
> > > > have an ELF section symbol for the .irqentry.text section.
> > > > 
> > > > Are there any special KCFLAGS being added?  Can you build the object
> > > > with V=1 to show the full gcc command line?
> > > 
> > > I have not added anything. There is no env variable set like $KCFLAGS or
> > > $CFLAGS. (If that was the question.)
> > > 
> > > I think you mean this line from output:
> > > gcc -Wp,-MD,arch/x86/kernel/.irq.o.d  -nostdinc -isystem
> > > /usr/lib/gcc/x86_64- pc-linux-gnu/6.4.0/include -I./arch/x86/include
> > > -I./arch/x86/include/generated -I./include -I./arch/x86/include/uapi
> > > -I./arch/x86/include/generated/uapi -I./ include/uapi
> > > -I./include/generated/uapi -include ./include/linux/kconfig.h -
> > > D__KERNEL__ -Wall -Wundef -Wstrict-prototypes -Wno-trigraphs -fno-strict-
> > > aliasing -fno-common -fshort-wchar -Werror-implicit-function-declaration
> > > -Wno- format-security -std=gnu89 -fno-PIE -mno-sse -mno-mmx -mno-sse2
> > > -mno-3dnow - mno-avx -m64 -falign-jumps=1 -falign-loops=1 -mno-80387
> > > -mno-fp-ret-in-387 - mpreferred-stack-boundary=3 -mskip-rax-setup
> > > -mtune=generic -mno-red-zone - mcmodel=kernel -funit-at-a-time
> > > -DCONFIG_AS_CFI=1 -
> > > DCONFIG_AS_CFI_SIGNAL_FRAME=1 -DCONFIG_AS_CFI_SECTIONS=1
> > > -DCONFIG_AS_FXSAVEQ=1 -DCONFIG_AS_SSSE3=1 -DCONFIG_AS_CRC32=1
> > > -DCONFIG_AS_AVX=1 -DCONFIG_AS_AVX2=1 - DCONFIG_AS_AVX512=1
> > > -DCONFIG_AS_SHA1_NI=1 -DCONFIG_AS_SHA256_NI=1 -pipe -Wno- sign-compare
> > > -fno-asynchronous-unwind-tables -fno-delete-null-pointer-checks -
> > > Wno-frame-address -O2 --param=allow-store-data-races=0 -DCC_HAVE_ASM_GOTO
> > > - Wframe-larger-than=2048 -fno-stack-protector
> > > -Wno-unused-but-set-variable - Wno-unused-const-variable
> > > -fomit-frame-pointer -fno-var-tracking-assignments -
> > > Wdeclaration-after-statement -Wno-pointer-sign -fno-strict-overflow -fno-
> > > stack-check -fconserve-stack -Werror=implicit-int
> > > -Werror=strict-prototypes - Werror=date-time
> > > -Werror=incompatible-pointer-types -Werror=designated-init -
> > > Iarch/x86/kernel/../include/asm/trace    -DKBUILD_BASENAME='"irq"'  -
> > > DKBUILD_MODNAME='"irq"' -c -o arch/x86/kernel/.tmp_irq.o
> > > arch/x86/kernel/irq.c
> > > 
> > > The next line is the objtool that segfaults.
> > 
> > I don't see anything unusual there.  Are there any Gentoo patches
> > against either the kernel or GCC which would strip unused symbols?
> 
> The kernel is the vanilla kernel. (4.14.11 and also 4.15-rc6)
> Its not a gentoo specific gcc patch. (Then every gentoo user would be 
> affected?)
> 
> But I enabled ld.gold as default linker like 5 years ago. Never had a problem 
> with this.
> 
> Is ld.gold supposed to fail here?
> 
> I switched back to ld.bfd and it seems to work.

Ah, that explains it.  With CONFIG_MODVERSIONS, the linker does some
work after gcc, but before objtool.  Can you try this patch?  (Note this
isn't the final patch, as this breaks the CONFIG_MODVERSIONS=n case.)

diff --git a/scripts/Makefile.build b/scripts/Makefile.build
index cb8997ed0149..3cf3cc6077ea 100644
--- a/scripts/Makefile.build
+++ b/scripts/Makefile.build
@@ -270,7 +270,7 @@ endif
 # 'OBJECT_FILES_NON_STANDARD_foo.o := 'n': override directory skip for a file
 cmd_objtool = $(if $(patsubst y%,, \
 	$(OBJECT_FILES_NON_STANDARD_$(basetarget).o)$(OBJECT_FILES_NON_STANDARD)n), \
-	$(__objtool_obj) $(objtool_args) "$(@)";)
+	$(__objtool_obj) $(objtool_args) "$(@D)/.tmp_$(@F)";)
 objtool_obj = $(if $(patsubst y%,, \
 	$(OBJECT_FILES_NON_STANDARD_$(basetarget).o)$(OBJECT_FILES_NON_STANDARD)n), \
 	$(__objtool_obj))
@@ -286,16 +286,16 @@ objtool_dep = $(objtool_obj)					\
 define rule_cc_o_c
 	$(call echo-cmd,checksrc) $(cmd_checksrc)			  \
 	$(call cmd_and_fixdep,cc_o_c)					  \
+	$(call echo-cmd,objtool) $(cmd_objtool)				  \
 	$(cmd_modversions_c)						  \
 	$(cmd_checkdoc)							  \
-	$(call echo-cmd,objtool) $(cmd_objtool)				  \
 	$(call echo-cmd,record_mcount) $(cmd_record_mcount)
 endef
 
 define rule_as_o_S
 	$(call cmd_and_fixdep,as_o_S)					  \
-	$(cmd_modversions_S)						  \
-	$(call echo-cmd,objtool) $(cmd_objtool)
+	$(call echo-cmd,objtool) $(cmd_objtool)				  \
+	$(cmd_modversions_S)
 endef
 
 # List module undefined symbols (or empty line if not enabled)