Skip to contentSkip to conversation

Fix CFI failures with GCOV_PROFILE_ALL

PATCH In review Sami Tolvanen 5 participants 8 messages Jun 23, 2023 – Jun 24, 2023 Kbuild

lore
Share & tools
lore
#0 lore raw

cover Fix CFI failures with GCOV_PROFILE_ALL

Hi folks,

The following two patches fix CFI failures with GCOV_PROFILE_ALL, where the compiler injects indirectly called functions to object files that otherwise contain no executable code, and are not processed by objtool or don't have CFI enabled. This results in missing or incorrect type hashes during boot and when modules are loaded.

Sami Tolvanen (2):
kbuild: Fix CFI failures with GCOV
kbuild: Disable GCOV for *.mod.o

 init/Makefile             | 1 +
 scripts/Makefile.modfinal | 2 +-
 scripts/Makefile.vmlinux  | 1 +
 3 files changed, 3 insertions(+), 1 deletion(-)

base-commit: 007034977130b49b618a5206aad54f634d9f169c

Signature · 1 line
2.41.0.162.gfafddb0af9-goog
#1 lore raw

PATCH 1/2 kbuild: Fix CFI failures with GCOV

With GCOV_PROFILE_ALL, Clang injects __llvm_gcov_* functions to each object file, and the functions are indirectly called during boot. However, when code is injected to object files that are not part of vmlinux.o, it's also not processed by objtool, which breaks CFI hash randomization as the hashes in these files won't be included in the .cfi_sites section and thus won't be randomized.

Similarly to commit 42633ed852de ("kbuild: Fix CFI hash randomization with KASAN"), disable GCOV for .vmlinux.export.o and init/version-timestamp.o to avoid emitting unnecessary functions to object files that don't otherwise have executable code.

Fixes 0c3e806ec0f9 ("x86/cfi: Add boot time hash randomization")
---
 init/Makefile            | 1 +
 scripts/Makefile.vmlinux | 1 +
 2 files changed, 2 insertions(+)
init/Makefile +1 -0
index 26de459006c4..ec557ada3c12 100644
@@ -60,3 +60,4 @@ include/generated/utsversion.h: FORCE
 $(obj)/version-timestamp.o: include/generated/utsversion.h
 CFLAGS_version-timestamp.o := -include include/generated/utsversion.h
 KASAN_SANITIZE_version-timestamp.o := n
+GCOV_PROFILE_version-timestamp.o := n
scripts/Makefile.vmlinux +1 -0
index 10176dec97ea..3cd6ca15f390 100644
@@ -19,6 +19,7 @@ quiet_cmd_cc_o_c = CC      $@
 
 ifdef CONFIG_MODULES
 KASAN_SANITIZE_.vmlinux.export.o := n
+GCOV_PROFILE_.vmlinux.export.o := n
 targets += .vmlinux.export.o
 vmlinux: .vmlinux.export.o
 endif
Signature · 1 line
2.41.0.162.gfafddb0af9-goog
#2 lore raw

PATCH 2/2 kbuild: Disable GCOV for *.mod.o

In reply to Sami Tolvanen

With GCOV_PROFILE_ALL, Clang injects __llvm_gcov_* functions to each object file, including the *.mod.o. As we filter out CC_FLAGS_CFI for *.mod.o, the compiler won't generate type hashes for the injected functions, and therefore indirectly calling them during module loading trips indirect call checking.

Enabling CFI for *.mod.o isn't sufficient to fix this issue after commit 0c3e806ec0f9 ("x86/cfi: Add boot time hash randomization"), as *.mod.o aren't processed by objtool, which means any hashes emitted there won't be randomized. Therefore, in addition to disabling CFI for *.mod.o, also disable GCOV, as the object files don't otherwise contain any executable code.

Fixes cf68fffb66d6 ("add support for Clang CFI")
---
 scripts/Makefile.modfinal | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
scripts/Makefile.modfinal +1 -1
index 4703f652c009..fc19f67039bd 100644
@@ -23,7 +23,7 @@ modname = $(notdir $(@:.mod.o=))
 part-of-module = y
 
 quiet_cmd_cc_o_c = CC [M]  $@
-      cmd_cc_o_c = $(CC) $(filter-out $(CC_FLAGS_CFI), $(c_flags)) -c -o $@ $<
+      cmd_cc_o_c = $(CC) $(filter-out $(CC_FLAGS_CFI) $(CFLAGS_GCOV), $(c_flags)) -c -o $@ $<
 
 %.mod.o: %.mod.c FORCE
 	$(call if_changed_dep,cc_o_c)
Signature · 1 line
2.41.0.162.gfafddb0af9-goog
Acked-by on the v1 series
#3 lore raw
In reply to Sami Tolvanen
Quoted 9 lines · Sami Tolvanen

Hi folks,

The following two patches fix CFI failures with GCOV_PROFILE_ALL, where the compiler injects indirectly called functions to object files that otherwise contain no executable code, and are not processed by objtool or don't have CFI enabled. This results in missing or incorrect type hashes during boot and when modules are loaded.

Quoted 9 lines

Sami Tolvanen (2):
kbuild: Fix CFI failures with GCOV
kbuild: Disable GCOV for *.mod.o

 init/Makefile             | 1 +
 scripts/Makefile.modfinal | 2 +-
 scripts/Makefile.vmlinux  | 1 +
 3 files changed, 3 insertions(+), 1 deletion(-)

Urgh, tricky stuff this.

And yes, objtool essentially assumes vmlinux.o is complete and does LTO like passes. Is there something kbuild can do to ensure noting else gets linked in after this?

Reviewed-by on the v1 series
#4 lore raw
In reply to Sami Tolvanen
Quoted 9 lines · Sami Tolvanen

Hi folks,

The following two patches fix CFI failures with GCOV_PROFILE_ALL, where the compiler injects indirectly called functions to object files that otherwise contain no executable code, and are not processed by objtool or don't have CFI enabled. This results in missing or incorrect type hashes during boot and when modules are loaded.

Quoted 8 lines

Sami Tolvanen (2):
kbuild: Fix CFI failures with GCOV
kbuild: Disable GCOV for *.mod.o

 init/Makefile             | 1 +
 scripts/Makefile.modfinal | 2 +-
 scripts/Makefile.vmlinux  | 1 +
 3 files changed, 3 insertions(+), 1 deletion(-)

Nice hunting!

Should these get Cc: stable tags maybe?

Signature · 1 line
Kees Cook
#5 lore raw
Quoted 20 lines · Kees Cook
Quoted 9 lines · Sami Tolvanen

Hi folks,

The following two patches fix CFI failures with GCOV_PROFILE_ALL, where the compiler injects indirectly called functions to object files that otherwise contain no executable code, and are not processed by objtool or don't have CFI enabled. This results in missing or incorrect type hashes during boot and when modules are loaded.

Quoted 8 lines

Sami Tolvanen (2):
kbuild: Fix CFI failures with GCOV
kbuild: Disable GCOV for *.mod.o

 init/Makefile             | 1 +
 scripts/Makefile.modfinal | 2 +-
 scripts/Makefile.vmlinux  | 1 +
 3 files changed, 3 insertions(+), 1 deletion(-)
Quoted 5 lines

Nice hunting!

Should these get Cc: stable tags maybe?

I was under the impression that Fixes: tags would be sufficient these days, but agreed, explicit Cc: probably wouldn't hurt.

Sami

#6 lore raw
Quoted 24 lines · Sami Tolvanen
Quoted 22 lines · Kees Cook
Quoted 9 lines · Sami Tolvanen

Hi folks,

The following two patches fix CFI failures with GCOV_PROFILE_ALL, where the compiler injects indirectly called functions to object files that otherwise contain no executable code, and are not processed by objtool or don't have CFI enabled. This results in missing or incorrect type hashes during boot and when modules are loaded.

Quoted 8 lines

Sami Tolvanen (2):
kbuild: Fix CFI failures with GCOV
kbuild: Disable GCOV for *.mod.o

 init/Makefile             | 1 +
 scripts/Makefile.modfinal | 2 +-
 scripts/Makefile.vmlinux  | 1 +
 3 files changed, 3 insertions(+), 1 deletion(-)

Nice hunting!

Quoted 8 lines
Quoted 3 lines

Should these get Cc: stable tags maybe?

I was under the impression that Fixes: tags would be sufficient these days, but agreed, explicit Cc: probably wouldn't hurt.

Sami

Both applied to linux-kbuild.

I also think Fixes: tags would be enough
to make them back-ported.

Signature · 2 lines
Best Regards
Masahiro Yamada
Reviewed-by on the v1 series
#7 lore raw
In reply to Sami Tolvanen
Quoted 10 lines · Sami Tolvanen

Hi folks,

The following two patches fix CFI failures with GCOV_PROFILE_ALL, where the compiler injects indirectly called functions to object files that otherwise contain no executable code, and are not processed by objtool or don't have CFI enabled. This results in missing or incorrect type hashes during boot and when modules are loaded.

Quoted 8 lines

Sami Tolvanen (2):
kbuild: Fix CFI failures with GCOV
kbuild: Disable GCOV for *.mod.o

 init/Makefile             | 1 +
 scripts/Makefile.modfinal | 2 +-
 scripts/Makefile.vmlinux  | 1 +
 3 files changed, 3 insertions(+), 1 deletion(-)

Thanks for the patches!

Quoted 7 lines

base-commit: 007034977130b49b618a5206aad54f634d9f169c --
2.41.0.162.gfafddb0af9-goog

Signature · 2 lines
Thanks,
~Nick Desaulniers