On Sat, Mar 30, 2024 at 10:46:27AM +0100, gregkh@xxxxxxxxxxxxxxxxxxx wrote:
> The patch below does not apply to the 6.8-stable tree.
> If someone wants it applied there, or to any other stable or longterm
> tree, then please email the backport, including the original git commit
> id to <stable@xxxxxxxxxxxxxxx>.
Here it is:
---
From: "Borislav Petkov (AMD)" <bp@xxxxxxxxx>
Date: Thu, 28 Mar 2024 13:59:05 +0100
Subject: [PATCH] x86/bugs: Fix the SRSO mitigation on Zen3/4
Commit 4535e1a4174c4111d92c5a9a21e542d232e0fcaa upstream.
The original version of the mitigation would patch in the calls to the
untraining routines directly. That is, the alternative() in UNTRAIN_RET
will patch in the CALL to srso_alias_untrain_ret() directly.
However, even if commit e7c25c441e9e ("x86/cpu: Cleanup the untrain
mess") meant well in trying to clean up the situation, due to micro-
architectural reasons, the untraining routine srso_alias_untrain_ret()
must be the target of a CALL instruction and not of a JMP instruction as
it is done now.
Reshuffle the alternative macros to accomplish that.
Fixes: e7c25c441e9e ("x86/cpu: Cleanup the untrain mess")
Signed-off-by: Borislav Petkov (AMD) <bp@xxxxxxxxx>
Reviewed-by: Ingo Molnar <mingo@xxxxxxxxxx>
Cc: stable@xxxxxxxxxx
Signed-off-by: Linus Torvalds <torvalds@xxxxxxxxxxxxxxxxxxxx>
---
arch/x86/include/asm/asm-prototypes.h | 1 +
arch/x86/include/asm/nospec-branch.h | 21 ++++++++++++++++-----
arch/x86/lib/retpoline.S | 11 ++++++-----
3 files changed, 23 insertions(+), 10 deletions(-)
diff --git a/arch/x86/include/asm/asm-prototypes.h b/arch/x86/include/asm/asm-prototypes.h
index b1a98fa38828..0e82074517f6 100644
--- a/arch/x86/include/asm/asm-prototypes.h
+++ b/arch/x86/include/asm/asm-prototypes.h
@@ -13,6 +13,7 @@
#include <asm/preempt.h>
#include <asm/asm.h>
#include <asm/gsseg.h>
+#include <asm/nospec-branch.h>
#ifndef CONFIG_X86_CMPXCHG64
extern void cmpxchg8b_emu(void);
diff --git a/arch/x86/include/asm/nospec-branch.h b/arch/x86/include/asm/nospec-branch.h
index 2aa52cab1e46..93c3e28dd8e0 100644
--- a/arch/x86/include/asm/nospec-branch.h
+++ b/arch/x86/include/asm/nospec-branch.h
@@ -271,11 +271,20 @@
.Lskip_rsb_\@:
.endm
+/*
+ * The CALL to srso_alias_untrain_ret() must be patched in directly at
+ * the spot where untraining must be done, ie., srso_alias_untrain_ret()
+ * must be the target of a CALL instruction instead of indirectly
+ * jumping to a wrapper which then calls it. Therefore, this macro is
+ * called outside of __UNTRAIN_RET below, for the time being, before the
+ * kernel can support nested alternatives with arbitrary nesting.
+ */
+.macro CALL_UNTRAIN_RET
#if defined(CONFIG_CPU_UNRET_ENTRY) || defined(CONFIG_CPU_SRSO)
-#define CALL_UNTRAIN_RET "call entry_untrain_ret"
-#else
-#define CALL_UNTRAIN_RET ""
+ ALTERNATIVE_2 "", "call entry_untrain_ret", X86_FEATURE_UNRET, \
+ "call srso_alias_untrain_ret", X86_FEATURE_SRSO_ALIAS
#endif
+.endm
/*
* Mitigate RETBleed for AMD/Hygon Zen uarch. Requires KERNEL CR3 because the
@@ -291,8 +300,8 @@
.macro __UNTRAIN_RET ibpb_feature, call_depth_insns
#if defined(CONFIG_RETHUNK) || defined(CONFIG_CPU_IBPB_ENTRY)
VALIDATE_UNRET_END
- ALTERNATIVE_3 "", \
- CALL_UNTRAIN_RET, X86_FEATURE_UNRET, \
+ CALL_UNTRAIN_RET
+ ALTERNATIVE_2 "", \
"call entry_ibpb", \ibpb_feature, \
__stringify(\call_depth_insns), X86_FEATURE_CALL_DEPTH
#endif
@@ -351,6 +360,8 @@ extern void retbleed_return_thunk(void);
static inline void retbleed_return_thunk(void) {}
#endif
+extern void srso_alias_untrain_ret(void);
+
#ifdef CONFIG_CPU_SRSO
extern void srso_return_thunk(void);
extern void srso_alias_return_thunk(void);
diff --git a/arch/x86/lib/retpoline.S b/arch/x86/lib/retpoline.S
index 7b2589877d06..1e59367b4681 100644
--- a/arch/x86/lib/retpoline.S
+++ b/arch/x86/lib/retpoline.S
@@ -163,6 +163,7 @@ SYM_CODE_START_NOALIGN(srso_alias_untrain_ret)
lfence
jmp srso_alias_return_thunk
SYM_FUNC_END(srso_alias_untrain_ret)
+__EXPORT_THUNK(srso_alias_untrain_ret)
.popsection
.pushsection .text..__x86.rethunk_safe
@@ -224,10 +225,12 @@ SYM_CODE_START(srso_return_thunk)
SYM_CODE_END(srso_return_thunk)
#define JMP_SRSO_UNTRAIN_RET "jmp srso_untrain_ret"
-#define JMP_SRSO_ALIAS_UNTRAIN_RET "jmp srso_alias_untrain_ret"
#else /* !CONFIG_CPU_SRSO */
#define JMP_SRSO_UNTRAIN_RET "ud2"
-#define JMP_SRSO_ALIAS_UNTRAIN_RET "ud2"
+/* Dummy for the alternative in CALL_UNTRAIN_RET. */
+SYM_CODE_START(srso_alias_untrain_ret)
+ RET
+SYM_FUNC_END(srso_alias_untrain_ret)
#endif /* CONFIG_CPU_SRSO */
#ifdef CONFIG_CPU_UNRET_ENTRY
@@ -319,9 +322,7 @@ SYM_FUNC_END(retbleed_untrain_ret)
#if defined(CONFIG_CPU_UNRET_ENTRY) || defined(CONFIG_CPU_SRSO)
SYM_FUNC_START(entry_untrain_ret)
- ALTERNATIVE_2 JMP_RETBLEED_UNTRAIN_RET, \
- JMP_SRSO_UNTRAIN_RET, X86_FEATURE_SRSO, \
- JMP_SRSO_ALIAS_UNTRAIN_RET, X86_FEATURE_SRSO_ALIAS
+ ALTERNATIVE JMP_RETBLEED_UNTRAIN_RET, JMP_SRSO_UNTRAIN_RET, X86_FEATURE_SRSO
SYM_FUNC_END(entry_untrain_ret)
__EXPORT_THUNK(entry_untrain_ret)
--
2.43.0
--
Regards/Gruss,
Boris.
https://people.kernel.org/tglx/notes-about-netiquette
