objtool: unreachable instruction #1148

nickdesaulniers · 2020-09-09T18:35:31Z

It looks like since Jul 30, I've seen a few reports from 0day bot about unreachable instructions. They looks randconfig based.

A bunch of reports aug 31, two on aug 6, two on Jul 30.

ihalip · 2020-09-10T07:19:37Z

https://lore.kernel.org/lkml/6653ad73c6b59c049211bd7c11ed3809c20ee9f5.1585761021.git.jpoimboe@redhat.com/

Actually that is already present in mainline since 0fe5f9c.

nickdesaulniers · 2020-09-15T22:23:14Z

@ihalip your previous version of your comment hinting at UBSAN was correct. Specifically, it seems this may be related to:
CONFIG_UBSAN_TRAP.

ihalip · 2020-09-16T06:43:40Z

W.r.t. the arch/x86/kernel/sev-es.o: warning: objtool: ist_exc_vmm_communication()+0x12: unreachable instruction case that's been talked about on the mailing list, this is what happens in short:

call panic
ud2

, where panic() is marked as noreturn. This could easily be fixed in objtool, since it already has a way to check if a function is a "dead end".

diff --git a/tools/objtool/check.c b/tools/objtool/check.c
index e034a8f24f46..9224e6565ba2 100644
--- a/tools/objtool/check.c
+++ b/tools/objtool/check.c
@@ -2612,9 +2612,10 @@ static bool is_ubsan_insn(struct instruction *insn)
                        "__ubsan_handle_builtin_unreachable"));
 }
 
-static bool ignore_unreachable_insn(struct instruction *insn)
+static bool ignore_unreachable_insn(struct objtool_file *file, struct instruction *insn)
 {
        int i;
+       struct instruction *prev_insn;
 
        if (insn->ignore || insn->type == INSN_NOP)
                return true;
@@ -2640,7 +2641,8 @@ static bool ignore_unreachable_insn(struct instruction *insn)
         * the UD2, which causes GCC's undefined trap logic to emit another UD2
         * (or occasionally a JMP to UD2).
         */
-       if (list_prev_entry(insn, list)->dead_end &&
+       prev_insn = list_prev_entry(insn, list);
+       if ((prev_insn->dead_end || dead_end_function(file, prev_insn->call_dest)) &&
            (insn->type == INSN_BUG ||
             (insn->type == INSN_JUMP_UNCONDITIONAL &&
              insn->jump_dest && insn->jump_dest->type == INSN_BUG)))
@@ -2767,7 +2769,7 @@ static int validate_reachable_instructions(struct objtool_file *file)
                return 0;
 
        for_each_insn(file, insn) {
-               if (insn->visited || ignore_unreachable_insn(insn))
+               if (insn->visited || ignore_unreachable_insn(file, insn))
                        continue;
 
                WARN_FUNC("unreachable instruction", insn->sec, insn->offset);

With CONFIG_UBSAN_TRAP enabled, the compiler may insert a trap instruction after a call to a noreturn function. In this case, objtool warns that the ud2 instruction is unreachable. objtool silences similar warnings (trap after dead end instructions), so expand that check to include dead end functions. Cc: Nick Desaulniers <ndesaulniers@google.com> Cc: Rong Chen <rong.a.chen@intel.com> Cc: Marco Elver <elver@google.com> Cc: Philip Li <philip.li@intel.com> Cc: Borislav Petkov <bp@alien8.de> Cc: kasan-dev@googlegroups.com Cc: x86@kernel.org Cc: clang-built-linux@googlegroups.com BugLink: ClangBuiltLinux#1148 Link: https://lore.kernel.org/lkml/CAKwvOdmptEpi8fiOyWUo=AiZJiX+Z+VHJOM2buLPrWsMTwLnyw@mail.gmail.com Suggested-by: Nick Desaulniers <ndesaulniers@google.com> Signed-off-by: Ilie Halip <ilie.halip@gmail.com>

ihalip · 2020-09-17T08:52:51Z

I sent this patch with an updated comment.

dileks · 2020-09-17T10:30:28Z

@ihalip

Shouldn't objtool maintainers be CCed?
Did you run ./scripts/get_maintainer.pl $patch?

[1] https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/MAINTAINERS#n12503

ihalip · 2020-09-17T11:13:21Z

I didn't add them in the commit message, but they are CCd on the actual e-mail.

nickdesaulniers · 2020-09-17T17:58:12Z

patch: https://lore.kernel.org/lkml/20200917084905.1647262-1-ilie.halip@gmail.com/

With CONFIG_UBSAN_TRAP enabled, the compiler may insert a trap instruction after a call to a noreturn function. In this case, objtool warns that the ud2 instruction is unreachable. objtool silences similar warnings (trap after dead end instructions), so expand that check to include dead end functions. Cc: Nick Desaulniers <ndesaulniers@google.com> Cc: Rong Chen <rong.a.chen@intel.com> Cc: Marco Elver <elver@google.com> Cc: Philip Li <philip.li@intel.com> Cc: Borislav Petkov <bp@alien8.de> Cc: kasan-dev@googlegroups.com Cc: x86@kernel.org Cc: clang-built-linux@googlegroups.com BugLink: ClangBuiltLinux#1148 Link: https://lore.kernel.org/lkml/CAKwvOdmptEpi8fiOyWUo=AiZJiX+Z+VHJOM2buLPrWsMTwLnyw@mail.gmail.com Suggested-by: Nick Desaulniers <ndesaulniers@google.com> Signed-off-by: Ilie Halip <ilie.halip@gmail.com>

dileks · 2020-09-18T00:47:10Z

@ihalip

Always good to point to the message-id of <lore.k.o> or give patchwork LKML link.

[1] https://lore.kernel.org/patchwork/patch/1307224/

dileks · 2020-09-18T01:15:04Z

Cool, first time using b4:

root# apt-get update && apt-get install -t testing b4

dileks$ git checkout tags/v5.9-rc5 -b for-5.9/objtool-unreachable-trap-ihalip

dileks$ b4 am https://lore.kernel.org/lkml/20200917084905.1647262-1-ilie.halip@gmail.com/ -o - | git am

[1] https://github.com/ClangBuiltLinux/linux/wiki/Command-line-tips-and-tricks > "For fetching a patch series from the mailing list"

With CONFIG_UBSAN_TRAP enabled, the compiler may insert a trap instruction after a call to a noreturn function. In this case, objtool warns that the ud2 instruction is unreachable. This is a behavior seen with clang, from the oldest version capable of building the mainline x64_64 kernel (9.0), to the latest experimental version (12). objtool silences similar warnings (trap after dead end instructions), so so expand that check to include dead end functions. This gets rid of 82 warnings in an x86_64 defconfig+UBSAN+UBSTAN_TRAP build. Cc: Nick Desaulniers <ndesaulniers@google.com> Cc: Rong Chen <rong.a.chen@intel.com> Cc: Marco Elver <elver@google.com> Cc: Philip Li <philip.li@intel.com> Cc: Borislav Petkov <bp@alien8.de> Cc: kasan-dev@googlegroups.com Cc: x86@kernel.org Cc: clang-built-linux@googlegroups.com BugLink: ClangBuiltLinux#1148 Link: https://lore.kernel.org/lkml/CAKwvOdmptEpi8fiOyWUo=AiZJiX+Z+VHJOM2buLPrWsMTwLnyw@mail.gmail.com Suggested-by: Nick Desaulniers <ndesaulniers@google.com> Reviewed-by: Nick Desaulniers <ndesaulniers@google.com> Tested-by: Nick Desaulniers <ndesaulniers@google.com> Reported-by: kbuild test robot <lkp@intel.com> Signed-off-by: Ilie Halip <ilie.halip@gmail.com>

nickdesaulniers · 2020-09-21T17:09:15Z

https://git.kernel.org/pub/scm/linux/kernel/git/tip/tip.git/commit/?id=14db1f0a93331d0958e90da522c429ff0890d2d6

With CONFIG_UBSAN_TRAP enabled, the compiler may insert a trap instruction after a call to a noreturn function. In this case, objtool warns that the UD2 instruction is unreachable. This is a behavior seen with Clang, from the oldest version capable of building the mainline x64_64 kernel (9.0), to the latest experimental version (12.0). Objtool silences similar warnings (trap after dead end instructions), so so expand that check to include dead end functions. Cc: Nick Desaulniers <ndesaulniers@google.com> Cc: Rong Chen <rong.a.chen@intel.com> Cc: Marco Elver <elver@google.com> Cc: Philip Li <philip.li@intel.com> Cc: Borislav Petkov <bp@alien8.de> Cc: kasan-dev@googlegroups.com Cc: x86@kernel.org Cc: clang-built-linux@googlegroups.com BugLink: ClangBuiltLinux#1148 Link: https://lore.kernel.org/lkml/CAKwvOdmptEpi8fiOyWUo=AiZJiX+Z+VHJOM2buLPrWsMTwLnyw@mail.gmail.com Suggested-by: Nick Desaulniers <ndesaulniers@google.com> Reviewed-by: Nick Desaulniers <ndesaulniers@google.com> Tested-by: Nick Desaulniers <ndesaulniers@google.com> Reported-by: kbuild test robot <lkp@intel.com> Signed-off-by: Ilie Halip <ilie.halip@gmail.com> Tested-by: Sedat Dilek <sedat.dilek@gmail.com> Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com>

nickdesaulniers · 2020-10-15T23:39:46Z

https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=14db1f0a93331d0958e90da522c429ff0890d2d6

With CONFIG_UBSAN_TRAP enabled, the compiler may insert a trap instruction after a call to a noreturn function. In this case, objtool warns that the UD2 instruction is unreachable. This is a behavior seen with Clang, from the oldest version capable of building the mainline x64_64 kernel (9.0), to the latest experimental version (12.0). Objtool silences similar warnings (trap after dead end instructions), so so expand that check to include dead end functions. Cc: Nick Desaulniers <ndesaulniers@google.com> Cc: Rong Chen <rong.a.chen@intel.com> Cc: Marco Elver <elver@google.com> Cc: Philip Li <philip.li@intel.com> Cc: Borislav Petkov <bp@alien8.de> Cc: kasan-dev@googlegroups.com Cc: x86@kernel.org Cc: clang-built-linux@googlegroups.com BugLink: ClangBuiltLinux/linux#1148 Link: https://lore.kernel.org/lkml/CAKwvOdmptEpi8fiOyWUo=AiZJiX+Z+VHJOM2buLPrWsMTwLnyw@mail.gmail.com Suggested-by: Nick Desaulniers <ndesaulniers@google.com> Reviewed-by: Nick Desaulniers <ndesaulniers@google.com> Tested-by: Nick Desaulniers <ndesaulniers@google.com> Reported-by: kbuild test robot <lkp@intel.com> Signed-off-by: Ilie Halip <ilie.halip@gmail.com> Tested-by: Sedat Dilek <sedat.dilek@gmail.com> Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com> Signed-off-by: Rapherion Rollerscaperers <rapherion@raphielgang.org>

nickdesaulniers added [BUG] Untriaged Something isn't working [ARCH] x86_64 This bug impacts ARCH=x86_64 [TOOL] objtool warning is produced by the kernel's objtool [Reported-by] kbuild test robot Reported-by: kbuild test robot <lkp@intel.com> labels Sep 9, 2020

ihalip added [PATCH] Submitted A patch has been submitted for review and removed [BUG] Untriaged Something isn't working labels Sep 11, 2020

nickdesaulniers removed the [PATCH] Submitted A patch has been submitted for review label Sep 15, 2020

ihalip added the [PATCH] Submitted A patch has been submitted for review label Sep 17, 2020

ihalip added [PATCH] Accepted A submitted patch has been accepted upstream and removed [PATCH] Submitted A patch has been submitted for review labels Sep 21, 2020

nickdesaulniers assigned ihalip Sep 21, 2020

ihalip mentioned this issue Sep 30, 2020

fs/gfs2/util.o: warning: objtool: gfs2_assert_warn_i()+0x184: unreachable instruction #1167

Closed

ihalip closed this as completed Oct 15, 2020

nickdesaulniers added [BUG] linux A bug that should be fixed in the mainline kernel. and removed [PATCH] Accepted A submitted patch has been accepted upstream labels Oct 15, 2020

nickdesaulniers added the [FIXED][LINUX] 5.10 This bug was fixed in Linux 5.10 label Oct 15, 2020

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

objtool: unreachable instruction #1148

objtool: unreachable instruction #1148

nickdesaulniers commented Sep 9, 2020

ihalip commented Sep 10, 2020 •

edited

Loading

nickdesaulniers commented Sep 15, 2020

ihalip commented Sep 16, 2020

ihalip commented Sep 17, 2020

dileks commented Sep 17, 2020

ihalip commented Sep 17, 2020

nickdesaulniers commented Sep 17, 2020

dileks commented Sep 18, 2020 •

edited

Loading

dileks commented Sep 18, 2020

nickdesaulniers commented Sep 21, 2020

nickdesaulniers commented Oct 15, 2020

objtool: unreachable instruction #1148

objtool: unreachable instruction #1148

Comments

nickdesaulniers commented Sep 9, 2020

ihalip commented Sep 10, 2020 • edited Loading

nickdesaulniers commented Sep 15, 2020

ihalip commented Sep 16, 2020

ihalip commented Sep 17, 2020

dileks commented Sep 17, 2020

ihalip commented Sep 17, 2020

nickdesaulniers commented Sep 17, 2020

dileks commented Sep 18, 2020 • edited Loading

dileks commented Sep 18, 2020

nickdesaulniers commented Sep 21, 2020

nickdesaulniers commented Oct 15, 2020

ihalip commented Sep 10, 2020 •

edited

Loading

dileks commented Sep 18, 2020 •

edited

Loading