2018-01-13 23:27:42

by Tom Lendacky

[permalink] [raw]
Subject: [PATCH v2] x86/retpoline: Add LFENCE to the retpoline/RSB filling RSB macros

The PAUSE instruction is currently used in the retpoline and RSB filling
macros as a speculation trap. The use of PAUSE was originally suggested
because it showed a very, very small difference in the amount of
cycles/time used to execute the retpoline as compared to LFENCE. On AMD,
the PAUSE instruction is not a serializing instruction, so the pause/jmp
loop will use excess power as it is speculated over waiting for return
to mispredict to the correct target.

The RSB filling macro is applicable to AMD, and, if software is unable to
verify that LFENCE is serializing on AMD (possible when running under a
hypervisor), the generic retpoline support will be used and, so, is also
applicable to AMD. Keep the current usage of PAUSE for Intel, but add an
LFENCE instruction to the speculation trap for AMD.

Signed-off-by: Tom Lendacky <[email protected]>
---
arch/x86/include/asm/nospec-branch.h | 6 +++++-
1 file changed, 5 insertions(+), 1 deletion(-)

diff --git a/arch/x86/include/asm/nospec-branch.h b/arch/x86/include/asm/nospec-branch.h
index 402a11c..7b45d84 100644
--- a/arch/x86/include/asm/nospec-branch.h
+++ b/arch/x86/include/asm/nospec-branch.h
@@ -11,7 +11,7 @@
* Fill the CPU return stack buffer.
*
* Each entry in the RSB, if used for a speculative 'ret', contains an
- * infinite 'pause; jmp' loop to capture speculative execution.
+ * infinite 'pause; lfence; jmp' loop to capture speculative execution.
*
* This is required in various cases for retpoline and IBRS-based
* mitigations for the Spectre variant 2 vulnerability. Sometimes to
@@ -38,11 +38,13 @@
call 772f; \
773: /* speculation trap */ \
pause; \
+ lfence; \
jmp 773b; \
772: \
call 774f; \
775: /* speculation trap */ \
pause; \
+ lfence; \
jmp 775b; \
774: \
dec reg; \
@@ -73,6 +75,7 @@
call .Ldo_rop_\@
.Lspec_trap_\@:
pause
+ lfence
jmp .Lspec_trap_\@
.Ldo_rop_\@:
mov \reg, (%_ASM_SP)
@@ -165,6 +168,7 @@
" .align 16\n" \
"901: call 903f;\n" \
"902: pause;\n" \
+ " lfence;\n" \
" jmp 902b;\n" \
" .align 16\n" \
"903: addl $4, %%esp;\n" \


2018-01-14 15:41:28

by Borislav Petkov

[permalink] [raw]
Subject: Re: [PATCH v2] x86/retpoline: Add LFENCE to the retpoline/RSB filling RSB macros

On Sat, Jan 13, 2018 at 05:27:30PM -0600, Tom Lendacky wrote:
> The PAUSE instruction is currently used in the retpoline and RSB filling
> macros as a speculation trap. The use of PAUSE was originally suggested
> because it showed a very, very small difference in the amount of
> cycles/time used to execute the retpoline as compared to LFENCE. On AMD,
> the PAUSE instruction is not a serializing instruction, so the pause/jmp
> loop will use excess power as it is speculated over waiting for return
> to mispredict to the correct target.
>
> The RSB filling macro is applicable to AMD, and, if software is unable to
> verify that LFENCE is serializing on AMD (possible when running under a
> hypervisor), the generic retpoline support will be used and, so, is also
> applicable to AMD. Keep the current usage of PAUSE for Intel, but add an
> LFENCE instruction to the speculation trap for AMD.
>
> Signed-off-by: Tom Lendacky <[email protected]>
> ---
> arch/x86/include/asm/nospec-branch.h | 6 +++++-
> 1 file changed, 5 insertions(+), 1 deletion(-)

Reviewed-by: Borislav Petkov <[email protected]>

--
Regards/Gruss,
Boris.

Good mailing practices for 400: avoid top-posting and trim the reply.

2018-01-14 17:09:27

by Woodhouse, David

[permalink] [raw]
Subject: Re: [PATCH v2] x86/retpoline: Add LFENCE to the retpoline/RSB filling RSB macros

On Sun, 2018-01-14 at 16:41 +0100, Borislav Petkov wrote:
> On Sat, Jan 13, 2018 at 05:27:30PM -0600, Tom Lendacky wrote:
> >
> > The PAUSE instruction is currently used in the retpoline and RSB filling
> > macros as a speculation trap.  The use of PAUSE was originally suggested
> > because it showed a very, very small difference in the amount of
> > cycles/time used to execute the retpoline as compared to LFENCE.  On AMD,
> > the PAUSE instruction is not a serializing instruction, so the pause/jmp
> > loop will use excess power as it is speculated over waiting for return
> > to mispredict to the correct target.
> >
> > The RSB filling macro is applicable to AMD, and, if software is unable to
> > verify that LFENCE is serializing on AMD (possible when running under a
> > hypervisor), the generic retpoline support will be used and, so, is also
> > applicable to AMD.  Keep the current usage of PAUSE for Intel, but add an
> > LFENCE instruction to the speculation trap for AMD.
> >
> > Signed-off-by: Tom Lendacky <[email protected]>
> > ---
> >  arch/x86/include/asm/nospec-branch.h |    6 +++++-
> >  1 file changed, 5 insertions(+), 1 deletion(-)
>
> Reviewed-by: Borislav Petkov <[email protected]>

Acked-by: Arjan van de Ven <[email protected]>
Acked-by: David Woodhouse <[email protected]>


Attachments:
smime.p7s (5.09 kB)
Subject: [tip:x86/pti] x86/retpoline: Add LFENCE to the retpoline/RSB filling RSB macros

Commit-ID: 607d93a2658a0075047e87e47fde5bcc626a1918
Gitweb: https://git.kernel.org/tip/607d93a2658a0075047e87e47fde5bcc626a1918
Author: Tom Lendacky <[email protected]>
AuthorDate: Sat, 13 Jan 2018 17:27:30 -0600
Committer: Thomas Gleixner <[email protected]>
CommitDate: Sun, 14 Jan 2018 18:22:15 +0100

x86/retpoline: Add LFENCE to the retpoline/RSB filling RSB macros

The PAUSE instruction is currently used in the retpoline and RSB filling
macros as a speculation trap. The use of PAUSE was originally suggested
because it showed a very, very small difference in the amount of
cycles/time used to execute the retpoline as compared to LFENCE. On AMD,
the PAUSE instruction is not a serializing instruction, so the pause/jmp
loop will use excess power as it is speculated over waiting for return
to mispredict to the correct target.

The RSB filling macro is applicable to AMD, and, if software is unable to
verify that LFENCE is serializing on AMD (possible when running under a
hypervisor), the generic retpoline support will be used and, so, is also
applicable to AMD. Keep the current usage of PAUSE for Intel, but add an
LFENCE instruction to the speculation trap for AMD.

The same sequence has been adopted by GCC for the GCC generated retpolines.

Signed-off-by: Tom Lendacky <[email protected]>
Signed-off-by: Thomas Gleixner <[email protected]>
Reviewed-by: Borislav Petkov <[email protected]>
Acked-by: David Woodhouse <[email protected]>
Acked-by: Arjan van de Ven <[email protected]>
Cc: Rik van Riel <[email protected]>
Cc: Andi Kleen <[email protected]>
Cc: Paul Turner <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: Tim Chen <[email protected]>
Cc: Jiri Kosina <[email protected]>
Cc: Dave Hansen <[email protected]>
Cc: Andy Lutomirski <[email protected]>
Cc: Josh Poimboeuf <[email protected]>
Cc: Dan Williams <[email protected]>
Cc: Linus Torvalds <[email protected]>
Cc: Greg Kroah-Hartman <[email protected]>
Cc: Kees Cook <[email protected]>
Link: https://lkml.kernel.org/r/[email protected]

---
arch/x86/include/asm/nospec-branch.h | 6 +++++-
1 file changed, 5 insertions(+), 1 deletion(-)

diff --git a/arch/x86/include/asm/nospec-branch.h b/arch/x86/include/asm/nospec-branch.h
index 402a11c..7b45d84 100644
--- a/arch/x86/include/asm/nospec-branch.h
+++ b/arch/x86/include/asm/nospec-branch.h
@@ -11,7 +11,7 @@
* Fill the CPU return stack buffer.
*
* Each entry in the RSB, if used for a speculative 'ret', contains an
- * infinite 'pause; jmp' loop to capture speculative execution.
+ * infinite 'pause; lfence; jmp' loop to capture speculative execution.
*
* This is required in various cases for retpoline and IBRS-based
* mitigations for the Spectre variant 2 vulnerability. Sometimes to
@@ -38,11 +38,13 @@
call 772f; \
773: /* speculation trap */ \
pause; \
+ lfence; \
jmp 773b; \
772: \
call 774f; \
775: /* speculation trap */ \
pause; \
+ lfence; \
jmp 775b; \
774: \
dec reg; \
@@ -73,6 +75,7 @@
call .Ldo_rop_\@
.Lspec_trap_\@:
pause
+ lfence
jmp .Lspec_trap_\@
.Ldo_rop_\@:
mov \reg, (%_ASM_SP)
@@ -165,6 +168,7 @@
" .align 16\n" \
"901: call 903f;\n" \
"902: pause;\n" \
+ " lfence;\n" \
" jmp 902b;\n" \
" .align 16\n" \
"903: addl $4, %%esp;\n" \

Subject: [tip:x86/pti] x86/retpoline: Add LFENCE to the retpoline/RSB filling RSB macros

Commit-ID: 28d437d550e1e39f805d99f9f8ac399c778827b7
Gitweb: https://git.kernel.org/tip/28d437d550e1e39f805d99f9f8ac399c778827b7
Author: Tom Lendacky <[email protected]>
AuthorDate: Sat, 13 Jan 2018 17:27:30 -0600
Committer: Thomas Gleixner <[email protected]>
CommitDate: Mon, 15 Jan 2018 00:32:55 +0100

x86/retpoline: Add LFENCE to the retpoline/RSB filling RSB macros

The PAUSE instruction is currently used in the retpoline and RSB filling
macros as a speculation trap. The use of PAUSE was originally suggested
because it showed a very, very small difference in the amount of
cycles/time used to execute the retpoline as compared to LFENCE. On AMD,
the PAUSE instruction is not a serializing instruction, so the pause/jmp
loop will use excess power as it is speculated over waiting for return
to mispredict to the correct target.

The RSB filling macro is applicable to AMD, and, if software is unable to
verify that LFENCE is serializing on AMD (possible when running under a
hypervisor), the generic retpoline support will be used and, so, is also
applicable to AMD. Keep the current usage of PAUSE for Intel, but add an
LFENCE instruction to the speculation trap for AMD.

The same sequence has been adopted by GCC for the GCC generated retpolines.

Signed-off-by: Tom Lendacky <[email protected]>
Signed-off-by: Thomas Gleixner <[email protected]>
Reviewed-by: Borislav Petkov <[email protected]>
Acked-by: David Woodhouse <[email protected]>
Acked-by: Arjan van de Ven <[email protected]>
Cc: Rik van Riel <[email protected]>
Cc: Andi Kleen <[email protected]>
Cc: Paul Turner <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: Tim Chen <[email protected]>
Cc: Jiri Kosina <[email protected]>
Cc: Dave Hansen <[email protected]>
Cc: Andy Lutomirski <[email protected]>
Cc: Josh Poimboeuf <[email protected]>
Cc: Dan Williams <[email protected]>
Cc: Linus Torvalds <[email protected]>
Cc: Greg Kroah-Hartman <[email protected]>
Cc: Kees Cook <[email protected]>
Link: https://lkml.kernel.org/r/[email protected]

---
arch/x86/include/asm/nospec-branch.h | 6 +++++-
1 file changed, 5 insertions(+), 1 deletion(-)

diff --git a/arch/x86/include/asm/nospec-branch.h b/arch/x86/include/asm/nospec-branch.h
index 402a11c..7b45d84 100644
--- a/arch/x86/include/asm/nospec-branch.h
+++ b/arch/x86/include/asm/nospec-branch.h
@@ -11,7 +11,7 @@
* Fill the CPU return stack buffer.
*
* Each entry in the RSB, if used for a speculative 'ret', contains an
- * infinite 'pause; jmp' loop to capture speculative execution.
+ * infinite 'pause; lfence; jmp' loop to capture speculative execution.
*
* This is required in various cases for retpoline and IBRS-based
* mitigations for the Spectre variant 2 vulnerability. Sometimes to
@@ -38,11 +38,13 @@
call 772f; \
773: /* speculation trap */ \
pause; \
+ lfence; \
jmp 773b; \
772: \
call 774f; \
775: /* speculation trap */ \
pause; \
+ lfence; \
jmp 775b; \
774: \
dec reg; \
@@ -73,6 +75,7 @@
call .Ldo_rop_\@
.Lspec_trap_\@:
pause
+ lfence
jmp .Lspec_trap_\@
.Ldo_rop_\@:
mov \reg, (%_ASM_SP)
@@ -165,6 +168,7 @@
" .align 16\n" \
"901: call 903f;\n" \
"902: pause;\n" \
+ " lfence;\n" \
" jmp 902b;\n" \
" .align 16\n" \
"903: addl $4, %%esp;\n" \