1. Patchew
  2. linux
  3. [PATCH 0/4] x86/alternatives: Do NOPs optimization on a temporary buffer
  4. View patch

[PATCH 2/4] x86/alternatives: Get rid of __optimize_nops()

Borislav Petkov posted 4 patches 1 year, 10 months ago
There is a newer version of this series
[PATCH 2/4] x86/alternatives: Get rid of __optimize_nops()
Posted by Borislav Petkov 1 year, 10 months ago 
From: "Borislav Petkov (AMD)" <bp@alien8.de>

There's no need to carve out bits of the NOP optimization functionality
and look at JMP opcodes - simply do one more NOPs optimization pass
at the end of the patching.

A lot simpler code.

Signed-off-by: Borislav Petkov (AMD) <bp@alien8.de>
---
 arch/x86/kernel/alternative.c | 52 +++++++----------------------------
 1 file changed, 10 insertions(+), 42 deletions(-)

diff --git a/arch/x86/kernel/alternative.c b/arch/x86/kernel/alternative.c
index d633eb59f2b6..2dd1c7fe0949 100644
--- a/arch/x86/kernel/alternative.c
+++ b/arch/x86/kernel/alternative.c
@@ -216,47 +216,12 @@ static int skip_nops(u8 *buf, int offset, int len)
 	return offset;
 }
 
-/*
- * Optimize a sequence of NOPs, possibly preceded by an unconditional jump
- * to the end of the NOP sequence into a single NOP.
- */
-static bool __init_or_module
-__optimize_nops(const u8 * const instr, u8 *buf, size_t len, struct insn *insn, int *next, int *prev, int *target)
-{
-	int i = *next - insn->length;
-
-	switch (insn->opcode.bytes[0]) {
-	case JMP8_INSN_OPCODE:
-	case JMP32_INSN_OPCODE:
-		*prev = i;
-		*target = *next + insn->immediate.value;
-		return false;
-	}
-
-	if (insn_is_nop(insn)) {
-		int nop = i;
-
-		*next = skip_nops(buf, *next, len);
-		if (*target && *next == *target)
-			nop = *prev;
-
-		add_nop(buf + nop, *next - nop);
-		DUMP_BYTES(ALT, buf, len, "%px: [%d:%d) optimized NOPs: ", instr, nop, *next);
-		return true;
-	}
-
-	*target = 0;
-	return false;
-}
-
 /*
  * "noinline" to cause control flow change and thus invalidate I$ and
  * cause refetch after modification.
  */
 static void __init_or_module noinline optimize_nops(const u8 * const instr, u8 *buf, size_t len)
 {
-	int prev, target = 0;
-
 	for (int next, i = 0; i < len; i = next) {
 		struct insn insn;
 
@@ -265,7 +230,14 @@ static void __init_or_module noinline optimize_nops(const u8 * const instr, u8 *
 
 		next = i + insn.length;
 
-		__optimize_nops(instr, buf, len, &insn, &next, &prev, &target);
+		if (insn_is_nop(&insn)) {
+			int nop = i;
+
+			next = skip_nops(buf, next, len);
+
+			add_nop(buf + nop, next - nop);
+			DUMP_BYTES(ALT, buf, len, "%px: [%d:%d) optimized NOPs: ", instr, nop, next);
+		}
 	}
 }
 
@@ -342,8 +314,6 @@ bool need_reloc(unsigned long offset, u8 *src, size_t src_len)
 static void __init_or_module noinline
 apply_relocation(const u8 * const instr, u8 *buf, size_t len, u8 *src, size_t src_len)
 {
-	int prev, target = 0;
-
 	for (int next, i = 0; i < len; i = next) {
 		struct insn insn;
 
@@ -352,9 +322,6 @@ apply_relocation(const u8 * const instr, u8 *buf, size_t len, u8 *src, size_t sr
 
 		next = i + insn.length;
 
-		if (__optimize_nops(instr, buf, len, &insn, &next, &prev, &target))
-			continue;
-
 		switch (insn.opcode.bytes[0]) {
 		case 0x0f:
 			if (insn.opcode.bytes[1] < 0x80 ||
@@ -533,7 +500,8 @@ void __init_or_module noinline apply_alternatives(struct alt_instr *start,
 		for (; insn_buff_sz < a->instrlen; insn_buff_sz++)
 			insn_buff[insn_buff_sz] = 0x90;
 
-		apply_relocation(instr, insn_buff, a->instrlen, replacement, a->replacementlen);
+		apply_relocation(instr, insn_buff, a->instrlen, replacement, insn_buff_sz);
+		optimize_nops(instr, insn_buff, insn_buff_sz);
 
 		DUMP_BYTES(ALT, instr, a->instrlen, "%px:   old_insn: ", instr);
 		DUMP_BYTES(ALT, replacement, a->replacementlen, "%px:   rpl_insn: ", replacement);
-- 
2.43.0
[tip: x86/alternatives] x86/alternatives: Get rid of __optimize_nops()
Posted by tip-bot2 for Borislav Petkov (AMD) 1 year, 8 months ago 
The following commit has been merged into the x86/alternatives branch of tip:

Commit-ID:     da8f9cf7e721c690ca169fa88641a6c4cee5cae4
Gitweb:        https://git.kernel.org/tip/da8f9cf7e721c690ca169fa88641a6c4cee5cae4
Author:        Borislav Petkov (AMD) <bp@alien8.de>
AuthorDate:    Tue, 30 Jan 2024 11:59:39 +01:00
Committer:     Borislav Petkov (AMD) <bp@alien8.de>
CommitterDate: Tue, 09 Apr 2024 18:12:53 +02:00

x86/alternatives: Get rid of __optimize_nops()

There's no need to carve out bits of the NOP optimization functionality
and look at JMP opcodes - simply do one more NOPs optimization pass
at the end of patching.

A lot simpler code.

Signed-off-by: Borislav Petkov (AMD) <bp@alien8.de>
Link: https://lore.kernel.org/r/20240130105941.19707-3-bp@alien8.de
---
 arch/x86/kernel/alternative.c | 59 +++++++++-------------------------
 1 file changed, 16 insertions(+), 43 deletions(-)

diff --git a/arch/x86/kernel/alternative.c b/arch/x86/kernel/alternative.c
index ec94f13..4b3378c 100644
--- a/arch/x86/kernel/alternative.c
+++ b/arch/x86/kernel/alternative.c
@@ -217,46 +217,11 @@ static int skip_nops(u8 *buf, int offset, int len)
 }
 
 /*
- * Optimize a sequence of NOPs, possibly preceded by an unconditional jump
- * to the end of the NOP sequence into a single NOP.
- */
-static bool
-__optimize_nops(const u8 * const instr, u8 *buf, size_t len, struct insn *insn, int *next, int *prev, int *target)
-{
-	int i = *next - insn->length;
-
-	switch (insn->opcode.bytes[0]) {
-	case JMP8_INSN_OPCODE:
-	case JMP32_INSN_OPCODE:
-		*prev = i;
-		*target = *next + insn->immediate.value;
-		return false;
-	}
-
-	if (insn_is_nop(insn)) {
-		int nop = i;
-
-		*next = skip_nops(buf, *next, len);
-		if (*target && *next == *target)
-			nop = *prev;
-
-		add_nop(buf + nop, *next - nop);
-		DUMP_BYTES(ALT, buf, len, "%px: [%d:%d) optimized NOPs: ", instr, nop, *next);
-		return true;
-	}
-
-	*target = 0;
-	return false;
-}
-
-/*
  * "noinline" to cause control flow change and thus invalidate I$ and
  * cause refetch after modification.
  */
-static void __init_or_module noinline optimize_nops(const u8 * const instr, u8 *buf, size_t len)
+static void noinline optimize_nops(const u8 * const instr, u8 *buf, size_t len)
 {
-	int prev, target = 0;
-
 	for (int next, i = 0; i < len; i = next) {
 		struct insn insn;
 
@@ -265,7 +230,14 @@ static void __init_or_module noinline optimize_nops(const u8 * const instr, u8 *
 
 		next = i + insn.length;
 
-		__optimize_nops(instr, buf, len, &insn, &next, &prev, &target);
+		if (insn_is_nop(&insn)) {
+			int nop = i;
+
+			next = skip_nops(buf, next, len);
+
+			add_nop(buf + nop, next - nop);
+			DUMP_BYTES(ALT, buf, len, "%px: [%d:%d) optimized NOPs: ", instr, nop, next);
+		}
 	}
 }
 
@@ -339,10 +311,8 @@ bool need_reloc(unsigned long offset, u8 *src, size_t src_len)
 	return (target < src || target > src + src_len);
 }
 
-void apply_relocation(u8 *buf, const u8 * const instr, size_t instrlen, u8 *repl, size_t repl_len)
+static void __apply_relocation(u8 *buf, const u8 * const instr, size_t instrlen, u8 *repl, size_t repl_len)
 {
-	int prev, target = 0;
-
 	for (int next, i = 0; i < instrlen; i = next) {
 		struct insn insn;
 
@@ -351,9 +321,6 @@ void apply_relocation(u8 *buf, const u8 * const instr, size_t instrlen, u8 *repl
 
 		next = i + insn.length;
 
-		if (__optimize_nops(instr, buf, instrlen, &insn, &next, &prev, &target))
-			continue;
-
 		switch (insn.opcode.bytes[0]) {
 		case 0x0f:
 			if (insn.opcode.bytes[1] < 0x80 ||
@@ -398,6 +365,12 @@ void apply_relocation(u8 *buf, const u8 * const instr, size_t instrlen, u8 *repl
 	}
 }
 
+void apply_relocation(u8 *buf, const u8 * const instr, size_t instrlen, u8 *repl, size_t repl_len)
+{
+	__apply_relocation(buf, instr, instrlen, repl, repl_len);
+	optimize_nops(instr, buf, repl_len);
+}
+
 /* Low-level backend functions usable from alternative code replacements. */
 DEFINE_ASM_FUNC(nop_func, "", .entry.text);
 EXPORT_SYMBOL_GPL(nop_func);
[tip: x86/alternatives] x86/alternatives: Get rid of __optimize_nops()
Posted by tip-bot2 for Borislav Petkov (AMD) 1 year, 10 months ago 
The following commit has been merged into the x86/alternatives branch of tip:

Commit-ID:     f9265b75648165f61467a7ea3c9bbae33be7ce27
Gitweb:        https://git.kernel.org/tip/f9265b75648165f61467a7ea3c9bbae33be7ce27
Author:        Borislav Petkov (AMD) <bp@alien8.de>
AuthorDate:    Tue, 30 Jan 2024 11:59:39 +01:00
Committer:     Borislav Petkov (AMD) <bp@alien8.de>
CommitterDate: Tue, 13 Feb 2024 16:25:46 +01:00

x86/alternatives: Get rid of __optimize_nops()

There's no need to carve out bits of the NOP optimization functionality
and look at JMP opcodes - simply do one more NOPs optimization pass
at the end of patching.

A lot simpler code.

Signed-off-by: Borislav Petkov (AMD) <bp@alien8.de>
Link: https://lore.kernel.org/r/20240130105941.19707-3-bp@alien8.de
---
 arch/x86/kernel/alternative.c | 52 ++++++----------------------------
 1 file changed, 10 insertions(+), 42 deletions(-)

diff --git a/arch/x86/kernel/alternative.c b/arch/x86/kernel/alternative.c
index 835e343..cdbece3 100644
--- a/arch/x86/kernel/alternative.c
+++ b/arch/x86/kernel/alternative.c
@@ -217,46 +217,11 @@ static int skip_nops(u8 *buf, int offset, int len)
 }
 
 /*
- * Optimize a sequence of NOPs, possibly preceded by an unconditional jump
- * to the end of the NOP sequence into a single NOP.
- */
-static bool __init_or_module
-__optimize_nops(const u8 * const instr, u8 *buf, size_t len, struct insn *insn, int *next, int *prev, int *target)
-{
-	int i = *next - insn->length;
-
-	switch (insn->opcode.bytes[0]) {
-	case JMP8_INSN_OPCODE:
-	case JMP32_INSN_OPCODE:
-		*prev = i;
-		*target = *next + insn->immediate.value;
-		return false;
-	}
-
-	if (insn_is_nop(insn)) {
-		int nop = i;
-
-		*next = skip_nops(buf, *next, len);
-		if (*target && *next == *target)
-			nop = *prev;
-
-		add_nop(buf + nop, *next - nop);
-		DUMP_BYTES(ALT, buf, len, "%px: [%d:%d) optimized NOPs: ", instr, nop, *next);
-		return true;
-	}
-
-	*target = 0;
-	return false;
-}
-
-/*
  * "noinline" to cause control flow change and thus invalidate I$ and
  * cause refetch after modification.
  */
 static void __init_or_module noinline optimize_nops(const u8 * const instr, u8 *buf, size_t len)
 {
-	int prev, target = 0;
-
 	for (int next, i = 0; i < len; i = next) {
 		struct insn insn;
 
@@ -265,7 +230,14 @@ static void __init_or_module noinline optimize_nops(const u8 * const instr, u8 *
 
 		next = i + insn.length;
 
-		__optimize_nops(instr, buf, len, &insn, &next, &prev, &target);
+		if (insn_is_nop(&insn)) {
+			int nop = i;
+
+			next = skip_nops(buf, next, len);
+
+			add_nop(buf + nop, next - nop);
+			DUMP_BYTES(ALT, buf, len, "%px: [%d:%d) optimized NOPs: ", instr, nop, next);
+		}
 	}
 }
 
@@ -342,8 +314,6 @@ bool need_reloc(unsigned long offset, u8 *src, size_t src_len)
 static void __init_or_module noinline
 apply_relocation(const u8 * const instr, u8 *buf, size_t len, u8 *src, size_t src_len)
 {
-	int prev, target = 0;
-
 	for (int next, i = 0; i < len; i = next) {
 		struct insn insn;
 
@@ -352,9 +322,6 @@ apply_relocation(const u8 * const instr, u8 *buf, size_t len, u8 *src, size_t sr
 
 		next = i + insn.length;
 
-		if (__optimize_nops(instr, buf, len, &insn, &next, &prev, &target))
-			continue;
-
 		switch (insn.opcode.bytes[0]) {
 		case 0x0f:
 			if (insn.opcode.bytes[1] < 0x80 ||
@@ -533,7 +500,8 @@ void __init_or_module noinline apply_alternatives(struct alt_instr *start,
 		for (; insn_buff_sz < a->instrlen; insn_buff_sz++)
 			insn_buff[insn_buff_sz] = 0x90;
 
-		apply_relocation(instr, insn_buff, a->instrlen, replacement, a->replacementlen);
+		apply_relocation(instr, insn_buff, a->instrlen, replacement, insn_buff_sz);
+		optimize_nops(instr, insn_buff, insn_buff_sz);
 
 		DUMP_BYTES(ALT, instr, a->instrlen, "%px:   old_insn: ", instr);
 		DUMP_BYTES(ALT, replacement, a->replacementlen, "%px:   rpl_insn: ", replacement);

Patchew 2.1-devel-b67e4c2

© 2016 - 2025 Red Hat, Inc.