Re: 5.4 and 4.19 fix for LLVM_IAS/clang-12

[Nick Desaulniers <ndesaulniers@google.com>]

[-- Attachment #1: Type: text/plain, Size: 2290 bytes --]

On Tue, Dec 8, 2020 at 11:21 PM Jiri Slaby <jirislaby@kernel.org> wrote:
>
> On 09. 12. 20, 8:20, Jiri Slaby wrote:
> > On 09. 12. 20, 1:12, Nick Desaulniers wrote:
> >> Dear stable kernel maintainers,
> >> Please consider accepting the following backport to 5.4 and 4.19 of
> >> commit 4d6ffa27b8e5 ("x86/lib: Change .weak to SYM_FUNC_START_WEAK for
> >> arch/x86/lib/mem*_64.S"), attached.

Also, first landed in v5.10-rc3. Exists in v5.9.7 as
305da744c138487864a46b2fb0bd7cf54e1e03b4.

> >>
> >> The patch to 5.4 had a conflict due to 5.4 missing upstream commit
> >> e9b9d020c487 ("x86/asm: Annotate aliases") which first landed in
> >> v5.5-rc1.
> >>
> >> The patch to 4.19 had a conflict due to 4.19 missing the above commit
> >> and ffedeeb780dc ("linkage: Introduce new macros for assembler
> >> symbols") which also first landed in v5.5-rc1 but was backported to
> >> linux-5.4.y as commit 840d8c9b3e5f ("linkage: Introduce new macros for
> >> assembler symbols") which shipped in v5.4.76.
> >>
> >> This patch fixes a build error from clang's assembler when building
> >> with Clang-12, which now errors when symbols are redeclared with
> >> different bindings.  We're using clang's assembler in Android and
> >> ChromeOS for 4.19+.
> >>
> >> Jiri, would you mind reviewing the 4.19 patch (or both)?  It simply
> >> open codes what the upstream macros would expand to; this can be and
> >> was observed from running:
> >
> > You don't have to touch (expand) __memcpy, __memmove, and __memset, right?

Sure, attached a v2.  It's actually a little cleaner (smaller
diffstat) that way.

>
> Also, no need for doubled p2align.

eh, maybe, but
1. it's what SYM_FUNC_START expands to precisely; removing it would be
(too) different from what upstream is doing IMO.
2. it's obviously doubled when in v1 I was expanding the __ prefixed
symbols as well; not so obvious in v2.
3. the order of __ prefixed symbols differs between these three; I'd
rather not rely on the order or have to reorder them to get the
initial symbol realigned; specifying the alignment for both seems
safer to me.

So I haven't included that suggestion in v2, but please feel free to
convince me otherwise if you feel strongly.

Thanks for taking the time to review.  I appreciate it.
-- 
Thanks,
~Nick Desaulniers

[-- Attachment #2: 4d6ffa27b8e5.4.19.v2.patch.txt --]
[-- Type: text/plain, Size: 3525 bytes --]

From 5707bea0fcef6263d3b23c5f0b61e6a289d54c22 Mon Sep 17 00:00:00 2001
From: Fangrui Song <maskray@google.com>
Date: Mon, 2 Nov 2020 17:23:58 -0800
Subject: [PATCH] x86/lib: Change .weak to SYM_FUNC_START_WEAK for
 arch/x86/lib/mem*_64.S

commit 4d6ffa27b8e5116c0abb318790fd01d4e12d75e6 upstream.

Commit

  393f203f5fd5 ("x86_64: kasan: add interceptors for memset/memmove/memcpy functions")

added .weak directives to arch/x86/lib/mem*_64.S instead of changing the
existing ENTRY macros to WEAK. This can lead to the assembly snippet

  .weak memcpy
  ...
  .globl memcpy

which will produce a STB_WEAK memcpy with GNU as but STB_GLOBAL memcpy
with LLVM's integrated assembler before LLVM 12. LLVM 12 (since
https://reviews.llvm.org/D90108) will error on such an overridden symbol
binding.

Commit

  ef1e03152cb0 ("x86/asm: Make some functions local")

changed ENTRY in arch/x86/lib/memcpy_64.S to SYM_FUNC_START_LOCAL, which
was ineffective due to the preceding .weak directive.

Use the appropriate SYM_FUNC_START_WEAK instead.

Fixes: 393f203f5fd5 ("x86_64: kasan: add interceptors for memset/memmove/memcpy functions")
Fixes: ef1e03152cb0 ("x86/asm: Make some functions local")
Reported-by: Sami Tolvanen <samitolvanen@google.com>
Signed-off-by: Fangrui Song <maskray@google.com>
Signed-off-by: Borislav Petkov <bp@suse.de>
Reviewed-by: Nick Desaulniers <ndesaulniers@google.com>
Tested-by: Nathan Chancellor <natechancellor@gmail.com>
Tested-by: Nick Desaulniers <ndesaulniers@google.com>
Cc: <stable@vger.kernel.org>
Link: https://lkml.kernel.org/r/20201103012358.168682-1-maskray@google.com
[nd: backport due to missing
  commit e9b9d020c487 ("x86/asm: Annotate aliases")
  commit ffedeeb780dc ("linkage: Introduce new macros for assembler symbols")]
Signed-off-by: Nick Desaulniers <ndesaulniers@google.com>
---
 arch/x86/lib/memcpy_64.S  | 6 +++---
 arch/x86/lib/memmove_64.S | 4 ++--
 arch/x86/lib/memset_64.S  | 6 +++---
 3 files changed, 8 insertions(+), 8 deletions(-)

diff --git a/arch/x86/lib/memcpy_64.S b/arch/x86/lib/memcpy_64.S
index 9d05572370ed..84b0078272d1 100644
--- a/arch/x86/lib/memcpy_64.S
+++ b/arch/x86/lib/memcpy_64.S
@@ -14,8 +14,6 @@
  * to a jmp to memcpy_erms which does the REP; MOVSB mem copy.
  */
 
-.weak memcpy
-
 /*
  * memcpy - Copy a memory block.
  *
@@ -28,7 +26,9 @@
  * rax original destination
  */
 ENTRY(__memcpy)
-ENTRY(memcpy)
+.weak memcpy
+.p2align 4, 0x90
+memcpy:
 	ALTERNATIVE_2 "jmp memcpy_orig", "", X86_FEATURE_REP_GOOD, \
 		      "jmp memcpy_erms", X86_FEATURE_ERMS
 
diff --git a/arch/x86/lib/memmove_64.S b/arch/x86/lib/memmove_64.S
index bbec69d8223b..e1cfc880f42d 100644
--- a/arch/x86/lib/memmove_64.S
+++ b/arch/x86/lib/memmove_64.S
@@ -25,8 +25,8 @@
  * rax: dest
  */
 .weak memmove
-
-ENTRY(memmove)
+.p2align 4, 0x90
+memmove:
 ENTRY(__memmove)
 
 	/* Handle more 32 bytes in loop */
diff --git a/arch/x86/lib/memset_64.S b/arch/x86/lib/memset_64.S
index 9bc861c71e75..084189acdcd0 100644
--- a/arch/x86/lib/memset_64.S
+++ b/arch/x86/lib/memset_64.S
@@ -6,8 +6,6 @@
 #include <asm/alternative-asm.h>
 #include <asm/export.h>
 
-.weak memset
-
 /*
  * ISO C memset - set a memory block to a byte value. This function uses fast
  * string to get better performance than the original function. The code is
@@ -19,7 +17,9 @@
  *
  * rax   original destination
  */
-ENTRY(memset)
+.weak memset
+.p2align 4, 0x90
+memset:
 ENTRY(__memset)
 	/*
 	 * Some CPUs support enhanced REP MOVSB/STOSB feature. It is recommended
-- 
2.29.2.576.ga3fc446d84-goog


[-- Attachment #3: 4d6ffa27b8e5.5.4.v2.patch.txt --]
[-- Type: text/plain, Size: 3431 bytes --]

From 4985e5d070135c56ce091bd1601996ec47419d07 Mon Sep 17 00:00:00 2001
From: Fangrui Song <maskray@google.com>
Date: Mon, 2 Nov 2020 17:23:58 -0800
Subject: [PATCH] x86/lib: Change .weak to SYM_FUNC_START_WEAK for
 arch/x86/lib/mem*_64.S

commit 4d6ffa27b8e5116c0abb318790fd01d4e12d75e6 upstream.

Commit

  393f203f5fd5 ("x86_64: kasan: add interceptors for memset/memmove/memcpy functions")

added .weak directives to arch/x86/lib/mem*_64.S instead of changing the
existing ENTRY macros to WEAK. This can lead to the assembly snippet

  .weak memcpy
  ...
  .globl memcpy

which will produce a STB_WEAK memcpy with GNU as but STB_GLOBAL memcpy
with LLVM's integrated assembler before LLVM 12. LLVM 12 (since
https://reviews.llvm.org/D90108) will error on such an overridden symbol
binding.

Commit

  ef1e03152cb0 ("x86/asm: Make some functions local")

changed ENTRY in arch/x86/lib/memcpy_64.S to SYM_FUNC_START_LOCAL, which
was ineffective due to the preceding .weak directive.

Use the appropriate SYM_FUNC_START_WEAK instead.

Fixes: 393f203f5fd5 ("x86_64: kasan: add interceptors for memset/memmove/memcpy functions")
Fixes: ef1e03152cb0 ("x86/asm: Make some functions local")
Reported-by: Sami Tolvanen <samitolvanen@google.com>
Signed-off-by: Fangrui Song <maskray@google.com>
Signed-off-by: Borislav Petkov <bp@suse.de>
Reviewed-by: Nick Desaulniers <ndesaulniers@google.com>
Tested-by: Nathan Chancellor <natechancellor@gmail.com>
Tested-by: Nick Desaulniers <ndesaulniers@google.com>
Cc: <stable@vger.kernel.org>
Link: https://lkml.kernel.org/r/20201103012358.168682-1-maskray@google.com
[nd: backport due to missing commit e9b9d020c487 ("x86/asm: Annotate aliases")]
Signed-off-by: Nick Desaulniers <ndesaulniers@google.com>
---
 arch/x86/lib/memcpy_64.S  | 4 +---
 arch/x86/lib/memmove_64.S | 4 +---
 arch/x86/lib/memset_64.S  | 4 +---
 3 files changed, 3 insertions(+), 9 deletions(-)

diff --git a/arch/x86/lib/memcpy_64.S b/arch/x86/lib/memcpy_64.S
index 92748660ba51..dc2fb886db2b 100644
--- a/arch/x86/lib/memcpy_64.S
+++ b/arch/x86/lib/memcpy_64.S
@@ -15,8 +15,6 @@
  * to a jmp to memcpy_erms which does the REP; MOVSB mem copy.
  */
 
-.weak memcpy
-
 /*
  * memcpy - Copy a memory block.
  *
@@ -29,7 +27,7 @@
  * rax original destination
  */
 ENTRY(__memcpy)
-ENTRY(memcpy)
+SYM_FUNC_START_WEAK(memcpy)
 	ALTERNATIVE_2 "jmp memcpy_orig", "", X86_FEATURE_REP_GOOD, \
 		      "jmp memcpy_erms", X86_FEATURE_ERMS
 
diff --git a/arch/x86/lib/memmove_64.S b/arch/x86/lib/memmove_64.S
index bbec69d8223b..b292445467b6 100644
--- a/arch/x86/lib/memmove_64.S
+++ b/arch/x86/lib/memmove_64.S
@@ -24,9 +24,7 @@
  * Output:
  * rax: dest
  */
-.weak memmove
-
-ENTRY(memmove)
+SYM_FUNC_START_WEAK(memmove)
 ENTRY(__memmove)
 
 	/* Handle more 32 bytes in loop */
diff --git a/arch/x86/lib/memset_64.S b/arch/x86/lib/memset_64.S
index 9bc861c71e75..e3376c7d4c97 100644
--- a/arch/x86/lib/memset_64.S
+++ b/arch/x86/lib/memset_64.S
@@ -6,8 +6,6 @@
 #include <asm/alternative-asm.h>
 #include <asm/export.h>
 
-.weak memset
-
 /*
  * ISO C memset - set a memory block to a byte value. This function uses fast
  * string to get better performance than the original function. The code is
@@ -19,7 +17,7 @@
  *
  * rax   original destination
  */
-ENTRY(memset)
+SYM_FUNC_START_WEAK(memset)
 ENTRY(__memset)
 	/*
 	 * Some CPUs support enhanced REP MOVSB/STOSB feature. It is recommended
-- 
2.29.2.576.ga3fc446d84-goog