Skip to contentSkip to conversation

uprobes/x86: Cleanups and fixes

PATCH Applied Peter Zijlstra 8 participants 31 messages Aug 21, 2025 – Sep 10, 2025

Subsystems (4)
lore
Share & tools
lore

Peter Zijlstra

author maintainer
#0 lore raw

cover uprobes/x86: Cleanups and fixes

Hi,

These are cleanups and fixes that I applied on top of Jiri's patches:

https://lkml.kernel.org/r/[email protected]

The combined lot sits in:

git://git.kernel.org/pub/scm/linux/kernel/git/peterz/queue.git perf/core

Jiri was going to send me some selftest updates that might mean rebasing that tree, but we'll see. If this all works we'll land it in -tip.

author maintainer
#1 lore raw

PATCH 1/6 uprobes/x86: Add struct uretprobe_syscall_args

Like uprobe_syscall_args; keep things consistent.

---
 arch/x86/kernel/uprobes.c |   33 ++++++++++++++++++++-------------
 1 file changed, 20 insertions(+), 13 deletions(-)
arch/x86/kernel/uprobes.c +20 -13
@@ -311,6 +311,12 @@ static int uprobe_init_insn(struct arch_
 
 #ifdef CONFIG_X86_64
 
+struct uretprobe_syscall_args {
+	unsigned long r11;
+	unsigned long cx;
+	unsigned long ax;
+};
+
 asm (
 	".pushsection .rodata\n"
 	".global uretprobe_trampoline_entry\n"
@@ -324,8 +330,8 @@ asm (
 	"uretprobe_syscall_check:\n"
 	"popq %r11\n"
 	"popq %rcx\n"
-
-	/* The uretprobe syscall replaces stored %rax value with final
+	/*
+	 * The uretprobe syscall replaces stored %rax value with final
 	 * return address, so we don't restore %rax in here and just
 	 * call ret.
 	 */
@@ -366,7 +372,8 @@ static unsigned long trampoline_check_ip
 SYSCALL_DEFINE0(uretprobe)
 {
 	struct pt_regs *regs = task_pt_regs(current);
-	unsigned long err, ip, sp, r11_cx_ax[3], tramp;
+	struct uretprobe_syscall_args args;
+	unsigned long err, ip, sp, tramp;
 
 	/* If there's no trampoline, we are called from wrong place. */
 	tramp = uprobe_get_trampoline_vaddr();
@@ -377,15 +384,15 @@ SYSCALL_DEFINE0(uretprobe)
 	if (unlikely(regs->ip != trampoline_check_ip(tramp)))
 		goto sigill;
 
-	err = copy_from_user(r11_cx_ax, (void __user *)regs->sp, sizeof(r11_cx_ax));
+	err = copy_from_user(&args, (void __user *)regs->sp, sizeof(args));
 	if (err)
 		goto sigill;
 
 	/* expose the "right" values of r11/cx/ax/sp to uprobe_consumer/s */
-	regs->r11 = r11_cx_ax[0];
-	regs->cx  = r11_cx_ax[1];
-	regs->ax  = r11_cx_ax[2];
-	regs->sp += sizeof(r11_cx_ax);
+	regs->r11 = args.r11;
+	regs->cx  = args.cx;
+	regs->ax  = args.ax;
+	regs->sp += sizeof(args);
 	regs->orig_ax = -1;
 
 	ip = regs->ip;
@@ -401,21 +408,21 @@ SYSCALL_DEFINE0(uretprobe)
 	 */
 	if (regs->sp != sp || shstk_is_enabled())
 		return regs->ax;
-	regs->sp -= sizeof(r11_cx_ax);
+	regs->sp -= sizeof(args);
 
 	/* for the case uprobe_consumer has changed r11/cx */
-	r11_cx_ax[0] = regs->r11;
-	r11_cx_ax[1] = regs->cx;
+	args.r11 = regs->r11;
+	args.cx  = regs->cx;
 
 	/*
 	 * ax register is passed through as return value, so we can use
 	 * its space on stack for ip value and jump to it through the
 	 * trampoline's ret instruction
 	 */
-	r11_cx_ax[2] = regs->ip;
+	args.ax  = regs->ip;
 	regs->ip = ip;
 
-	err = copy_to_user((void __user *)regs->sp, r11_cx_ax, sizeof(r11_cx_ax));
+	err = copy_to_user((void __user *)regs->sp, &args, sizeof(args));
 	if (err)
 		goto sigill;
 
merge bot Applied
#2 lore raw

[tip: perf/core] uprobes/x86: Add struct uretprobe_syscall_args

The following commit has been merged into the perf/core branch of tip:

The following commit has been merged into the perf/core branch of tip:

Commit-ID:     985e820b72e0a89746f51c072ed10caf78890244
Gitweb:        https://git.kernel.org/tip/985e820b72e0a89746f51c072ed10caf78890244
Author:        Peter Zijlstra <[email protected]>
AuthorDate:    Wed, 20 Aug 2025 12:28:36 +02:00
Committer:     Peter Zijlstra <[email protected]>
CommitterDate: Thu, 21 Aug 2025 20:09:21 +02:00

uprobes/x86: Add struct uretprobe_syscall_args

Like uprobe_syscall_args; keep things consistent.

---
 arch/x86/kernel/uprobes.c | 33 ++++++++++++++++++++-------------
 1 file changed, 20 insertions(+), 13 deletions(-)
arch/x86/kernel/uprobes.c +20 -13
index 209ce74..580989d 100644
@@ -311,6 +311,12 @@ static int uprobe_init_insn(struct arch_uprobe *auprobe, struct insn *insn, bool
 
 #ifdef CONFIG_X86_64
 
+struct uretprobe_syscall_args {
+	unsigned long r11;
+	unsigned long cx;
+	unsigned long ax;
+};
+
 asm (
 	".pushsection .rodata\n"
 	".global uretprobe_trampoline_entry\n"
@@ -324,8 +330,8 @@ asm (
 	"uretprobe_syscall_check:\n"
 	"popq %r11\n"
 	"popq %rcx\n"
-
-	/* The uretprobe syscall replaces stored %rax value with final
+	/*
+	 * The uretprobe syscall replaces stored %rax value with final
 	 * return address, so we don't restore %rax in here and just
 	 * call ret.
 	 */
@@ -366,7 +372,8 @@ static unsigned long trampoline_check_ip(unsigned long tramp)
 SYSCALL_DEFINE0(uretprobe)
 {
 	struct pt_regs *regs = task_pt_regs(current);
-	unsigned long err, ip, sp, r11_cx_ax[3], tramp;
+	struct uretprobe_syscall_args args;
+	unsigned long err, ip, sp, tramp;
 
 	/* If there's no trampoline, we are called from wrong place. */
 	tramp = uprobe_get_trampoline_vaddr();
@@ -377,15 +384,15 @@ SYSCALL_DEFINE0(uretprobe)
 	if (unlikely(regs->ip != trampoline_check_ip(tramp)))
 		goto sigill;
 
-	err = copy_from_user(r11_cx_ax, (void __user *)regs->sp, sizeof(r11_cx_ax));
+	err = copy_from_user(&args, (void __user *)regs->sp, sizeof(args));
 	if (err)
 		goto sigill;
 
 	/* expose the "right" values of r11/cx/ax/sp to uprobe_consumer/s */
-	regs->r11 = r11_cx_ax[0];
-	regs->cx  = r11_cx_ax[1];
-	regs->ax  = r11_cx_ax[2];
-	regs->sp += sizeof(r11_cx_ax);
+	regs->r11 = args.r11;
+	regs->cx  = args.cx;
+	regs->ax  = args.ax;
+	regs->sp += sizeof(args);
 	regs->orig_ax = -1;
 
 	ip = regs->ip;
@@ -401,21 +408,21 @@ SYSCALL_DEFINE0(uretprobe)
 	 */
 	if (regs->sp != sp || shstk_is_enabled())
 		return regs->ax;
-	regs->sp -= sizeof(r11_cx_ax);
+	regs->sp -= sizeof(args);
 
 	/* for the case uprobe_consumer has changed r11/cx */
-	r11_cx_ax[0] = regs->r11;
-	r11_cx_ax[1] = regs->cx;
+	args.r11 = regs->r11;
+	args.cx  = regs->cx;
 
 	/*
 	 * ax register is passed through as return value, so we can use
 	 * its space on stack for ip value and jump to it through the
 	 * trampoline's ret instruction
 	 */
-	r11_cx_ax[2] = regs->ip;
+	args.ax  = regs->ip;
 	regs->ip = ip;
 
-	err = copy_to_user((void __user *)regs->sp, r11_cx_ax, sizeof(r11_cx_ax));
+	err = copy_to_user((void __user *)regs->sp, &args, sizeof(args));
 	if (err)
 		goto sigill;
 
author maintainer
#3 lore raw

PATCH 2/6 uprobes/x86: Optimize is_optimize()

In reply to Peter Zijlstra

Make is_optimized() return a tri-state and avoid return through argument. This simplifies things a little.

---
 arch/x86/kernel/uprobes.c |   34 +++++++++++++---------------------
 1 file changed, 13 insertions(+), 21 deletions(-)
arch/x86/kernel/uprobes.c +13 -21
@@ -1047,7 +1047,7 @@ static bool __is_optimized(uprobe_opcode
 	return __in_uprobe_trampoline(vaddr + 5 + call->raddr);
 }
 
-static int is_optimized(struct mm_struct *mm, unsigned long vaddr, bool *optimized)
+static int is_optimized(struct mm_struct *mm, unsigned long vaddr)
 {
 	uprobe_opcode_t insn[5];
 	int err;
@@ -1055,8 +1055,7 @@ static int is_optimized(struct mm_struct
 	err = copy_from_vaddr(mm, vaddr, &insn, 5);
 	if (err)
 		return err;
-	*optimized = __is_optimized((uprobe_opcode_t *)&insn, vaddr);
-	return 0;
+	return __is_optimized((uprobe_opcode_t *)&insn, vaddr);
 }
 
 static bool should_optimize(struct arch_uprobe *auprobe)
@@ -1069,17 +1068,14 @@ int set_swbp(struct arch_uprobe *auprobe
 	     unsigned long vaddr)
 {
 	if (should_optimize(auprobe)) {
-		bool optimized = false;
-		int err;
-
 		/*
 		 * We could race with another thread that already optimized the probe,
 		 * so let's not overwrite it with int3 again in this case.
 		 */
-		err = is_optimized(vma->vm_mm, vaddr, &optimized);
-		if (err)
-			return err;
-		if (optimized)
+		int ret = is_optimized(vma->vm_mm, vaddr);
+		if (ret < 0)
+			return ret;
+		if (ret)
 			return 0;
 	}
 	return uprobe_write_opcode(auprobe, vma, vaddr, UPROBE_SWBP_INSN,
@@ -1090,17 +1086,13 @@ int set_orig_insn(struct arch_uprobe *au
 		  unsigned long vaddr)
 {
 	if (test_bit(ARCH_UPROBE_FLAG_CAN_OPTIMIZE, &auprobe->flags)) {
-		struct mm_struct *mm = vma->vm_mm;
-		bool optimized = false;
-		int err;
-
-		err = is_optimized(mm, vaddr, &optimized);
-		if (err)
-			return err;
-		if (optimized) {
-			err = swbp_unoptimize(auprobe, vma, vaddr);
-			WARN_ON_ONCE(err);
-			return err;
+		int ret = is_optimized(vma->vm_mm, vaddr);
+		if (ret < 0)
+			return ret;
+		if (ret) {
+			ret = swbp_unoptimize(auprobe, vma, vaddr);
+			WARN_ON_ONCE(ret);
+			return ret;
 		}
 	}
 	return uprobe_write_opcode(auprobe, vma, vaddr, *(uprobe_opcode_t *)&auprobe->insn,
merge bot Applied
#4 lore raw

[tip: perf/core] uprobes/x86: Optimize is_optimize()

The following commit has been merged into the perf/core branch of tip:

The following commit has been merged into the perf/core branch of tip:

Commit-ID:     fd54052b60cf6e73cf918fd8653cd7a5c84b0cc3
Gitweb:        https://git.kernel.org/tip/fd54052b60cf6e73cf918fd8653cd7a5c84b0cc3
Author:        Peter Zijlstra <[email protected]>
AuthorDate:    Wed, 20 Aug 2025 12:37:22 +02:00
Committer:     Peter Zijlstra <[email protected]>
CommitterDate: Thu, 21 Aug 2025 20:09:21 +02:00

uprobes/x86: Optimize is_optimize()

Make is_optimized() return a tri-state and avoid return through argument. This simplifies things a little.

---
 arch/x86/kernel/uprobes.c | 34 +++++++++++++---------------------
 1 file changed, 13 insertions(+), 21 deletions(-)
arch/x86/kernel/uprobes.c +13 -21
index 580989d..3b46a89 100644
@@ -1047,7 +1047,7 @@ static bool __is_optimized(uprobe_opcode_t *insn, unsigned long vaddr)
 	return __in_uprobe_trampoline(vaddr + 5 + call->raddr);
 }
 
-static int is_optimized(struct mm_struct *mm, unsigned long vaddr, bool *optimized)
+static int is_optimized(struct mm_struct *mm, unsigned long vaddr)
 {
 	uprobe_opcode_t insn[5];
 	int err;
@@ -1055,8 +1055,7 @@ static int is_optimized(struct mm_struct *mm, unsigned long vaddr, bool *optimiz
 	err = copy_from_vaddr(mm, vaddr, &insn, 5);
 	if (err)
 		return err;
-	*optimized = __is_optimized((uprobe_opcode_t *)&insn, vaddr);
-	return 0;
+	return __is_optimized((uprobe_opcode_t *)&insn, vaddr);
 }
 
 static bool should_optimize(struct arch_uprobe *auprobe)
@@ -1069,17 +1068,14 @@ int set_swbp(struct arch_uprobe *auprobe, struct vm_area_struct *vma,
 	     unsigned long vaddr)
 {
 	if (should_optimize(auprobe)) {
-		bool optimized = false;
-		int err;
-
 		/*
 		 * We could race with another thread that already optimized the probe,
 		 * so let's not overwrite it with int3 again in this case.
 		 */
-		err = is_optimized(vma->vm_mm, vaddr, &optimized);
-		if (err)
-			return err;
-		if (optimized)
+		int ret = is_optimized(vma->vm_mm, vaddr);
+		if (ret < 0)
+			return ret;
+		if (ret)
 			return 0;
 	}
 	return uprobe_write_opcode(auprobe, vma, vaddr, UPROBE_SWBP_INSN,
@@ -1090,17 +1086,13 @@ int set_orig_insn(struct arch_uprobe *auprobe, struct vm_area_struct *vma,
 		  unsigned long vaddr)
 {
 	if (test_bit(ARCH_UPROBE_FLAG_CAN_OPTIMIZE, &auprobe->flags)) {
-		struct mm_struct *mm = vma->vm_mm;
-		bool optimized = false;
-		int err;
-
-		err = is_optimized(mm, vaddr, &optimized);
-		if (err)
-			return err;
-		if (optimized) {
-			err = swbp_unoptimize(auprobe, vma, vaddr);
-			WARN_ON_ONCE(err);
-			return err;
+		int ret = is_optimized(vma->vm_mm, vaddr);
+		if (ret < 0)
+			return ret;
+		if (ret) {
+			ret = swbp_unoptimize(auprobe, vma, vaddr);
+			WARN_ON_ONCE(ret);
+			return ret;
 		}
 	}
 	return uprobe_write_opcode(auprobe, vma, vaddr, *(uprobe_opcode_t *)&auprobe->insn,
#5 lore raw

Re: [PATCH 2/6] uprobes/x86: Optimize is_optimize()

In reply to Peter Zijlstra
Quoted 39 lines · On Thu, 21 Aug 2025 14:28:24 +0200 Peter Zijlstra <[email protected]> wrote:

Make is_optimized() return a tri-state and avoid return through argument. This simplifies things a little.

---
 arch/x86/kernel/uprobes.c |   34 +++++++++++++---------------------
 1 file changed, 13 insertions(+), 21 deletions(-)
arch/x86/kernel/uprobes.c +2 -6
@@ -1047,7 +1047,7 @@ static bool __is_optimized(uprobe_opcode
 	return __in_uprobe_trampoline(vaddr + 5 + call->raddr);
 }
 
-static int is_optimized(struct mm_struct *mm, unsigned long vaddr, bool *optimized)
+static int is_optimized(struct mm_struct *mm, unsigned long vaddr)
 {
 	uprobe_opcode_t insn[5];
 	int err;
@@ -1055,8 +1055,7 @@ static int is_optimized(struct mm_struct
 	err = copy_from_vaddr(mm, vaddr, &insn, 5);
 	if (err)
 		return err;
-	*optimized = __is_optimized((uprobe_opcode_t *)&insn, vaddr);
-	return 0;
+	return __is_optimized((uprobe_opcode_t *)&insn, vaddr);
 }
 
 static bool should_optimize(struct arch_uprobe *auprobe)
@@ -1069,17 +1068,14 @@ int set_swbp(struct arch_uprobe *auprobe
 	     unsigned long vaddr)
 {
 	if (should_optimize(auprobe)) {
-		bool optimized = false;
-		int err;
-
 		/*
 		 * We could race with another thread that already optimized the probe,
 		 * so let's not overwrite it with int3 again in this case.

arch/x86/kernel/uprobes.c int set_swbp(struct arch_uprobe *auprobe · line 1079

Quoted 10 lines
arch/x86/kernel/uprobes.c +4 -4
 		 */
-		err = is_optimized(vma->vm_mm, vaddr, &optimized);
-		if (err)
-			return err;
-		if (optimized)
+		int ret = is_optimized(vma->vm_mm, vaddr);
+		if (ret < 0)
+			return ret;
+		if (ret)
 			return 0;

Looks like you should swap over 0 and 1.
That would then be: if (ret <= 0) return ret;

David

Quoted 30 lines
arch/x86/kernel/uprobes.c +7 -11
 	}
 	return uprobe_write_opcode(auprobe, vma, vaddr, UPROBE_SWBP_INSN,
@@ -1090,17 +1086,13 @@ int set_orig_insn(struct arch_uprobe *au
 		  unsigned long vaddr)
 {
 	if (test_bit(ARCH_UPROBE_FLAG_CAN_OPTIMIZE, &auprobe->flags)) {
-		struct mm_struct *mm = vma->vm_mm;
-		bool optimized = false;
-		int err;
-
-		err = is_optimized(mm, vaddr, &optimized);
-		if (err)
-			return err;
-		if (optimized) {
-			err = swbp_unoptimize(auprobe, vma, vaddr);
-			WARN_ON_ONCE(err);
-			return err;
+		int ret = is_optimized(vma->vm_mm, vaddr);
+		if (ret < 0)
+			return ret;
+		if (ret) {
+			ret = swbp_unoptimize(auprobe, vma, vaddr);
+			WARN_ON_ONCE(ret);
+			return ret;
 		}
 	}
 	return uprobe_write_opcode(auprobe, vma, vaddr, *(uprobe_opcode_t *)&auprobe->insn,
author maintainer
#6 lore raw

Re: [PATCH 2/6] uprobes/x86: Optimize is_optimize()

Quoted 15 lines · David Laight
Quoted 15 lines
arch/x86/kernel/uprobes.c +0 -7
@@ -1069,17 +1068,14 @@ int set_swbp(struct arch_uprobe *auprobe
 	     unsigned long vaddr)
 {
 	if (should_optimize(auprobe)) {
-		bool optimized = false;
-		int err;
-
 		/*
 		 * We could race with another thread that already optimized the probe,
 		 * so let's not overwrite it with int3 again in this case.
 		 */
-		err = is_optimized(vma->vm_mm, vaddr, &optimized);
-		if (err)
-			return err;
-		if (optimized)
Quoted 8 lines

arch/x86/kernel/uprobes.c int set_swbp(struct arch_uprobe *auprobe · line 1079

Quoted 5 lines
arch/x86/kernel/uprobes.c +4 -0
+		int ret = is_optimized(vma->vm_mm, vaddr);
+		if (ret < 0)
+			return ret;
+		if (ret)
 			return 0;

Looks like you should swap over 0 and 1.
That would then be: if (ret <= 0) return ret;

I considered that, but that was actually more confusing. Yes the return check is neat, but urgh.

The tri-state return is:

<0 -- error
0 -- false
1 -- true

and that is converted to the 'normal' convention:

<0 -- error
0 -- success

Making that intermediate:

<0 -- error
0 -- true
1 -- false

is just asking for trouble later.

#7 lore raw

Re: [PATCH 2/6] uprobes/x86: Optimize is_optimize()

Quoted 41 lines · On Tue, 26 Aug 2025 10:18:40 +0200 Peter Zijlstra <[email protected]> wrote:
Quoted 15 lines · David Laight
Quoted 15 lines
arch/x86/kernel/uprobes.c +0 -7
@@ -1069,17 +1068,14 @@ int set_swbp(struct arch_uprobe *auprobe
 	     unsigned long vaddr)
 {
 	if (should_optimize(auprobe)) {
-		bool optimized = false;
-		int err;
-
 		/*
 		 * We could race with another thread that already optimized the probe,
 		 * so let's not overwrite it with int3 again in this case.
 		 */
-		err = is_optimized(vma->vm_mm, vaddr, &optimized);
-		if (err)
-			return err;
-		if (optimized)
Quoted 8 lines

arch/x86/kernel/uprobes.c int set_swbp(struct arch_uprobe *auprobe · line 1079

Quoted 5 lines
arch/x86/kernel/uprobes.c +4 -0
+		int ret = is_optimized(vma->vm_mm, vaddr);
+		if (ret < 0)
+			return ret;
+		if (ret)
 			return 0;  

Looks like you should swap over 0 and 1.
That would then be: if (ret <= 0) return ret;

I considered that, but that was actually more confusing. Yes the return check is neat, but urgh.

The tri-state return is:

<0 -- error
0 -- false
1 -- true

and that is converted to the 'normal' convention:

<0 -- error
0 -- success

Quoted 7 lines

Making that intermediate:

<0 -- error
0 -- true
1 -- false

is just asking for trouble later.

I'm sure the function name could be changed to make it all work :-)

David

#8 lore raw

Re: [PATCH 2/6] uprobes/x86: Optimize is_optimize()

In reply to David Laight
Quoted 47 lines · David Laight
Quoted 44 lines · On Thu, 21 Aug 2025 14:28:24 +0200 Peter Zijlstra <[email protected]> wrote:

Make is_optimized() return a tri-state and avoid return through argument. This simplifies things a little.

---
 arch/x86/kernel/uprobes.c |   34 +++++++++++++---------------------
 1 file changed, 13 insertions(+), 21 deletions(-)
arch/x86/kernel/uprobes.c +2 -10
@@ -1047,7 +1047,7 @@ static bool __is_optimized(uprobe_opcode
 	return __in_uprobe_trampoline(vaddr + 5 + call->raddr);
 }
 
-static int is_optimized(struct mm_struct *mm, unsigned long vaddr, bool *optimized)
+static int is_optimized(struct mm_struct *mm, unsigned long vaddr)
 {
 	uprobe_opcode_t insn[5];
 	int err;
@@ -1055,8 +1055,7 @@ static int is_optimized(struct mm_struct
 	err = copy_from_vaddr(mm, vaddr, &insn, 5);
 	if (err)
 		return err;
-	*optimized = __is_optimized((uprobe_opcode_t *)&insn, vaddr);
-	return 0;
+	return __is_optimized((uprobe_opcode_t *)&insn, vaddr);
 }
 
 static bool should_optimize(struct arch_uprobe *auprobe)
@@ -1069,17 +1068,14 @@ int set_swbp(struct arch_uprobe *auprobe
 	     unsigned long vaddr)
 {
 	if (should_optimize(auprobe)) {
-		bool optimized = false;
-		int err;
-
 		/*
 		 * We could race with another thread that already optimized the probe,
 		 * so let's not overwrite it with int3 again in this case.
 		 */
-		err = is_optimized(vma->vm_mm, vaddr, &optimized);
-		if (err)
-			return err;
-		if (optimized)
Quoted 8 lines

arch/x86/kernel/uprobes.c int set_swbp(struct arch_uprobe *auprobe · line 1079

Quoted 5 lines
arch/x86/kernel/uprobes.c +4 -0
+		int ret = is_optimized(vma->vm_mm, vaddr);
+		if (ret < 0)
+			return ret;
+		if (ret)
 			return 0;

Looks like you should swap over 0 and 1.
That would then be: if (ret <= 0) return ret;

hum, but if it's not optimized (ret == 0) we need to follow up with installing breakpoint through following uprobe_write_opcode call

also I noticed we mix int/bool return, perhaps we could do fix below

jirka

---

arch/x86/kernel/uprobes.c +1 -1
index 0a8c0a4a5423..853abb2a5638 100644
@@ -1064,7 +1064,7 @@ static int is_optimized(struct mm_struct *mm, unsigned long vaddr)
 	err = copy_from_vaddr(mm, vaddr, &insn, 5);
 	if (err)
 		return err;
-	return __is_optimized((uprobe_opcode_t *)&insn, vaddr);
+	return __is_optimized((uprobe_opcode_t *)&insn, vaddr) ? 1 : 0;
 }
 
 static bool should_optimize(struct arch_uprobe *auprobe)
#9 lore raw

Re: [PATCH 2/6] uprobes/x86: Optimize is_optimize()

Quoted 57 lines · Jiri Olsa
Quoted 47 lines · David Laight
Quoted 44 lines · On Thu, 21 Aug 2025 14:28:24 +0200 Peter Zijlstra <[email protected]> wrote:

Make is_optimized() return a tri-state and avoid return through argument. This simplifies things a little.

---
 arch/x86/kernel/uprobes.c |   34 +++++++++++++---------------------
 1 file changed, 13 insertions(+), 21 deletions(-)
arch/x86/kernel/uprobes.c +2 -10
@@ -1047,7 +1047,7 @@ static bool __is_optimized(uprobe_opcode
 	return __in_uprobe_trampoline(vaddr + 5 + call->raddr);
 }
 
-static int is_optimized(struct mm_struct *mm, unsigned long vaddr, bool *optimized)
+static int is_optimized(struct mm_struct *mm, unsigned long vaddr)
 {
 	uprobe_opcode_t insn[5];
 	int err;
@@ -1055,8 +1055,7 @@ static int is_optimized(struct mm_struct
 	err = copy_from_vaddr(mm, vaddr, &insn, 5);
 	if (err)
 		return err;
-	*optimized = __is_optimized((uprobe_opcode_t *)&insn, vaddr);
-	return 0;
+	return __is_optimized((uprobe_opcode_t *)&insn, vaddr);
 }
 
 static bool should_optimize(struct arch_uprobe *auprobe)
@@ -1069,17 +1068,14 @@ int set_swbp(struct arch_uprobe *auprobe
 	     unsigned long vaddr)
 {
 	if (should_optimize(auprobe)) {
-		bool optimized = false;
-		int err;
-
 		/*
 		 * We could race with another thread that already optimized the probe,
 		 * so let's not overwrite it with int3 again in this case.
 		 */
-		err = is_optimized(vma->vm_mm, vaddr, &optimized);
-		if (err)
-			return err;
-		if (optimized)
Quoted 8 lines

arch/x86/kernel/uprobes.c int set_swbp(struct arch_uprobe *auprobe · line 1079

Quoted 5 lines
arch/x86/kernel/uprobes.c +4 -0
+		int ret = is_optimized(vma->vm_mm, vaddr);
+		if (ret < 0)
+			return ret;
+		if (ret)
 			return 0;

Looks like you should swap over 0 and 1.
That would then be: if (ret <= 0) return ret;

Quoted 2 lines

hum, but if it's not optimized (ret == 0) we need to follow up with installing breakpoint through following uprobe_write_opcode call

ah u meant to swap the whole thing.. got it

Quoted 20 lines

also I noticed we mix int/bool return, perhaps we could do fix below

jirka

---

arch/x86/kernel/uprobes.c +1 -1
index 0a8c0a4a5423..853abb2a5638 100644
@@ -1064,7 +1064,7 @@ static int is_optimized(struct mm_struct *mm, unsigned long vaddr)
 	err = copy_from_vaddr(mm, vaddr, &insn, 5);
 	if (err)
 		return err;
-	return __is_optimized((uprobe_opcode_t *)&insn, vaddr);
+	return __is_optimized((uprobe_opcode_t *)&insn, vaddr) ? 1 : 0;
 }
 
 static bool should_optimize(struct arch_uprobe *auprobe)
author maintainer
#10 lore raw

PATCH 3/6 uprobes/x86: Accept more NOP forms

In reply to Peter Zijlstra

Instead of only accepting the x86_64 nop5 chosen by the kernel, accept any x86_64 NOP or NOPL instruction that is 5 bytes.

Notably, the x86_64 nop5 pattern is valid in 32bit apps and could get compiler generated when build for i686 (which introduced NOPL). Since the trampoline is x86_64 only, make sure to limit to x86_64 code.

---
 arch/x86/kernel/uprobes.c |   37 ++++++++++++++++++++++++++++++++-----
 1 file changed, 32 insertions(+), 5 deletions(-)
arch/x86/kernel/uprobes.c +32 -5
@@ -1157,10 +1157,37 @@ void arch_uprobe_optimize(struct arch_up
 	mmap_write_unlock(mm);
 }
 
-static bool can_optimize(struct arch_uprobe *auprobe, unsigned long vaddr)
+static bool insn_is_nop(struct insn *insn)
 {
-	if (memcmp(&auprobe->insn, x86_nops[5], 5))
+	return insn->opcode.nbytes == 1 && insn->opcode.bytes[0] == 0x90;
+}
+
+static bool insn_is_nopl(struct insn *insn)
+{
+	if (insn->opcode.nbytes != 2)
+		return false;
+
+	if (insn->opcode.bytes[0] != 0x0f || insn->opcode.bytes[1] != 0x1f)
+		return false;
+
+	if (!insn->modrm.nbytes)
+		return false;
+
+	if (X86_MODRM_REG(insn->modrm.bytes[0]) != 0)
+		return false;
+
+	/* 0f 1f /0 - NOPL */
+	return true;
+}
+
+static bool can_optimize(struct insn *insn, unsigned long vaddr)
+{
+	if (!insn->x86_64 || insn->length != 5)
 		return false;
+
+	if (!insn_is_nop(insn) && !insn_is_nopl(insn))
+		return false;
+
 	/* We can't do cross page atomic writes yet. */
 	return PAGE_SIZE - (vaddr & ~PAGE_MASK) >= 5;
 }
@@ -1177,7 +1204,7 @@ static void riprel_pre_xol(struct arch_u
 static void riprel_post_xol(struct arch_uprobe *auprobe, struct pt_regs *regs)
 {
 }
-static bool can_optimize(struct arch_uprobe *auprobe, unsigned long vaddr)
+static bool can_optimize(struct insn *insn, unsigned long vaddr)
 {
 	return false;
 }
@@ -1539,15 +1566,15 @@ static int push_setup_xol_ops(struct arc
  */
 int arch_uprobe_analyze_insn(struct arch_uprobe *auprobe, struct mm_struct *mm, unsigned long addr)
 {
-	struct insn insn;
 	u8 fix_ip_or_call = UPROBE_FIX_IP;
+	struct insn insn;
 	int ret;
 
 	ret = uprobe_init_insn(auprobe, &insn, is_64bit_mm(mm));
 	if (ret)
 		return ret;
 
-	if (can_optimize(auprobe, addr))
+	if (can_optimize(&insn, addr))
 		set_bit(ARCH_UPROBE_FLAG_CAN_OPTIMIZE, &auprobe->flags);
 
 	ret = branch_setup_xol_ops(auprobe, &insn);
merge bot Applied
#11 lore raw

[tip: perf/core] uprobes/x86: Accept more NOP forms

The following commit has been merged into the perf/core branch of tip:

The following commit has been merged into the perf/core branch of tip:

Commit-ID:     7c2bfc183b05103287cc32ad68184f7d4312c06d
Gitweb:        https://git.kernel.org/tip/7c2bfc183b05103287cc32ad68184f7d4312c06d
Author:        Peter Zijlstra <[email protected]>
AuthorDate:    Wed, 20 Aug 2025 12:55:14 +02:00
Committer:     Peter Zijlstra <[email protected]>
CommitterDate: Thu, 21 Aug 2025 20:09:21 +02:00

uprobes/x86: Accept more NOP forms

Instead of only accepting the x86_64 nop5 chosen by the kernel, accept any x86_64 NOP or NOPL instruction that is 5 bytes.

Notably, the x86_64 nop5 pattern is valid in 32bit apps and could get compiler generated when build for i686 (which introduced NOPL). Since the trampoline is x86_64 only, make sure to limit to x86_64 code.

---
 arch/x86/kernel/uprobes.c | 37 ++++++++++++++++++++++++++++++++-----
 1 file changed, 32 insertions(+), 5 deletions(-)
arch/x86/kernel/uprobes.c +32 -5
index 3b46a89..d513c97 100644
@@ -1157,10 +1157,37 @@ unlock:
 	mmap_write_unlock(mm);
 }
 
-static bool can_optimize(struct arch_uprobe *auprobe, unsigned long vaddr)
+static bool insn_is_nop(struct insn *insn)
 {
-	if (memcmp(&auprobe->insn, x86_nops[5], 5))
+	return insn->opcode.nbytes == 1 && insn->opcode.bytes[0] == 0x90;
+}
+
+static bool insn_is_nopl(struct insn *insn)
+{
+	if (insn->opcode.nbytes != 2)
+		return false;
+
+	if (insn->opcode.bytes[0] != 0x0f || insn->opcode.bytes[1] != 0x1f)
+		return false;
+
+	if (!insn->modrm.nbytes)
+		return false;
+
+	if (X86_MODRM_REG(insn->modrm.bytes[0]) != 0)
+		return false;
+
+	/* 0f 1f /0 - NOPL */
+	return true;
+}
+
+static bool can_optimize(struct insn *insn, unsigned long vaddr)
+{
+	if (!insn->x86_64 || insn->length != 5)
 		return false;
+
+	if (!insn_is_nop(insn) && !insn_is_nopl(insn))
+		return false;
+
 	/* We can't do cross page atomic writes yet. */
 	return PAGE_SIZE - (vaddr & ~PAGE_MASK) >= 5;
 }
@@ -1177,7 +1204,7 @@ static void riprel_pre_xol(struct arch_uprobe *auprobe, struct pt_regs *regs)
 static void riprel_post_xol(struct arch_uprobe *auprobe, struct pt_regs *regs)
 {
 }
-static bool can_optimize(struct arch_uprobe *auprobe, unsigned long vaddr)
+static bool can_optimize(struct insn *insn, unsigned long vaddr)
 {
 	return false;
 }
@@ -1539,15 +1566,15 @@ static int push_setup_xol_ops(struct arch_uprobe *auprobe, struct insn *insn)
  */
 int arch_uprobe_analyze_insn(struct arch_uprobe *auprobe, struct mm_struct *mm, unsigned long addr)
 {
-	struct insn insn;
 	u8 fix_ip_or_call = UPROBE_FIX_IP;
+	struct insn insn;
 	int ret;
 
 	ret = uprobe_init_insn(auprobe, &insn, is_64bit_mm(mm));
 	if (ret)
 		return ret;
 
-	if (can_optimize(auprobe, addr))
+	if (can_optimize(&insn, addr))
 		set_bit(ARCH_UPROBE_FLAG_CAN_OPTIMIZE, &auprobe->flags);
 
 	ret = branch_setup_xol_ops(auprobe, &insn);
author maintainer
#12 lore raw

PATCH 4/6 uprobes/x86: Fix uprobe syscall vs shadow stack

In reply to Peter Zijlstra

The uprobe syscall stores and strips the trampoline stack frame from the user context, to make it appear similar to an exception at the original instruction. It then restores the trampoline stack when it can exit using sysexit.

Make sure to match the regular stack manipulation with shadow stack operations such that regular and shadow stack don't get out of sync and causes trouble.

This enables using the optimization when shadow stack is in use.

---
 arch/x86/include/asm/shstk.h |    4 ++++
 arch/x86/kernel/shstk.c      |   40 ++++++++++++++++++++++++++++++++++++++++
 arch/x86/kernel/uprobes.c    |   17 ++++++++---------
 3 files changed, 52 insertions(+), 9 deletions(-)
arch/x86/include/asm/shstk.h +4 -0
@@ -23,6 +23,8 @@ int setup_signal_shadow_stack(struct ksi
 int restore_signal_shadow_stack(void);
 int shstk_update_last_frame(unsigned long val);
 bool shstk_is_enabled(void);
+int shstk_pop(u64 *val);
+int shstk_push(u64 val);
 #else
 static inline long shstk_prctl(struct task_struct *task, int option,
 			       unsigned long arg2) { return -EINVAL; }
@@ -35,6 +37,8 @@ static inline int setup_signal_shadow_st
 static inline int restore_signal_shadow_stack(void) { return 0; }
 static inline int shstk_update_last_frame(unsigned long val) { return 0; }
 static inline bool shstk_is_enabled(void) { return false; }
+static inline int shstk_pop(u64 *val) { return -ENOTSUPP; }
+static inline int shstk_push(u64 val) { return -ENOTSUPP; }
 #endif /* CONFIG_X86_USER_SHADOW_STACK */
 
 #endif /* __ASSEMBLER__ */
arch/x86/kernel/shstk.c +40 -0
@@ -246,6 +246,46 @@ static unsigned long get_user_shstk_addr
 	return ssp;
 }
 
+int shstk_pop(u64 *val)
+{
+	int ret = 0;
+	u64 ssp;
+
+	if (!features_enabled(ARCH_SHSTK_SHSTK))
+		return -ENOTSUPP;
+
+	fpregs_lock_and_load();
+
+	rdmsrq(MSR_IA32_PL3_SSP, ssp);
+	if (val && get_user(*val, (__user u64 *)ssp))
+		ret = -EFAULT;
+	else
+		wrmsrq(MSR_IA32_PL3_SSP, ssp + SS_FRAME_SIZE);
+	fpregs_unlock();
+
+	return ret;
+}
+
+int shstk_push(u64 val)
+{
+	u64 ssp;
+	int ret;
+
+	if (!features_enabled(ARCH_SHSTK_SHSTK))
+		return -ENOTSUPP;
+
+	fpregs_lock_and_load();
+
+	rdmsrq(MSR_IA32_PL3_SSP, ssp);
+	ssp -= SS_FRAME_SIZE;
+	ret = write_user_shstk_64((__user void *)ssp, val);
+	if (!ret)
+		wrmsrq(MSR_IA32_PL3_SSP, ssp);
+	fpregs_unlock();
+
+	return ret;
+}
+
 #define SHSTK_DATA_BIT BIT(63)
 
 static int put_shstk_data(u64 __user *addr, u64 data)
arch/x86/kernel/uprobes.c +8 -9
@@ -804,7 +804,7 @@ SYSCALL_DEFINE0(uprobe)
 {
 	struct pt_regs *regs = task_pt_regs(current);
 	struct uprobe_syscall_args args;
-	unsigned long ip, sp;
+	unsigned long ip, sp, sret;
 	int err;
 
 	/* Allow execution only from uprobe trampolines. */
@@ -831,6 +831,10 @@ SYSCALL_DEFINE0(uprobe)
 
 	sp = regs->sp;
 
+	err = shstk_pop((u64 *)&sret);
+	if (err == -EFAULT || (!err && sret != args.retaddr))
+		goto sigill;
+
 	handle_syscall_uprobe(regs, regs->ip);
 
 	/*
@@ -855,6 +859,9 @@ SYSCALL_DEFINE0(uprobe)
 	if (args.retaddr - 5 != regs->ip)
 		args.retaddr = regs->ip;
 
+	if (shstk_push(args.retaddr) == -EFAULT)
+		goto sigill;
+
 	regs->ip = ip;
 
 	err = copy_to_user((void __user *)regs->sp, &args, sizeof(args));
@@ -1124,14 +1131,6 @@ void arch_uprobe_optimize(struct arch_up
 	struct mm_struct *mm = current->mm;
 	uprobe_opcode_t insn[5];
 
-	/*
-	 * Do not optimize if shadow stack is enabled, the return address hijack
-	 * code in arch_uretprobe_hijack_return_addr updates wrong frame when
-	 * the entry uprobe is optimized and the shadow stack crashes the app.
-	 */
-	if (shstk_is_enabled())
-		return;
-
 	if (!should_optimize(auprobe))
 		return;
 
#13 lore raw

Re: [PATCH 4/6] uprobes/x86: Fix uprobe syscall vs shadow stack

Quoted 46 lines · Peter Zijlstra

The uprobe syscall stores and strips the trampoline stack frame from the user context, to make it appear similar to an exception at the original instruction. It then restores the trampoline stack when it can exit using sysexit.

Make sure to match the regular stack manipulation with shadow stack operations such that regular and shadow stack don't get out of sync and causes trouble.

This enables using the optimization when shadow stack is in use.

---
 arch/x86/include/asm/shstk.h |    4 ++++
 arch/x86/kernel/shstk.c      |   40 ++++++++++++++++++++++++++++++++++++++++
 arch/x86/kernel/uprobes.c    |   17 ++++++++---------
 3 files changed, 52 insertions(+), 9 deletions(-)
arch/x86/include/asm/shstk.h +4 -0
@@ -23,6 +23,8 @@ int setup_signal_shadow_stack(struct ksi
 int restore_signal_shadow_stack(void);
 int shstk_update_last_frame(unsigned long val);
 bool shstk_is_enabled(void);
+int shstk_pop(u64 *val);
+int shstk_push(u64 val);
 #else
 static inline long shstk_prctl(struct task_struct *task, int option,
 			       unsigned long arg2) { return -EINVAL; }
@@ -35,6 +37,8 @@ static inline int setup_signal_shadow_st
 static inline int restore_signal_shadow_stack(void) { return 0; }
 static inline int shstk_update_last_frame(unsigned long val) { return 0; }
 static inline bool shstk_is_enabled(void) { return false; }
+static inline int shstk_pop(u64 *val) { return -ENOTSUPP; }
+static inline int shstk_push(u64 val) { return -ENOTSUPP; }
 #endif /* CONFIG_X86_USER_SHADOW_STACK */
 
 #endif /* __ASSEMBLER__ */
arch/x86/kernel/shstk.c +2 -0
@@ -246,6 +246,46 @@ static unsigned long get_user_shstk_addr
 	return ssp;
 }
 
+int shstk_pop(u64 *val)
+{

arch/x86/kernel/shstk.c static unsigned long get_user_shstk_addr · line 260

Quoted 10 lines
arch/x86/kernel/shstk.c +10 -0
+	int ret = 0;
+	u64 ssp;
+
+	if (!features_enabled(ARCH_SHSTK_SHSTK))
+		return -ENOTSUPP;
+
+	fpregs_lock_and_load();
+
+	rdmsrq(MSR_IA32_PL3_SSP, ssp);
+	if (val && get_user(*val, (__user u64 *)ssp))

It makes it so shstk_pop() can incssp without pushing anything to the shadow stack, but nothing uses this.

Also, since there is no read_user_shstk_64() it should probably check that the VMA is actually shadow stack, like how it does in shstk_pop_sigframe(). What this actually would expose, I'm not sure. It might be ok. There would just be a fault later during shstk_push(args.retaddr) I guess.

Hmm, I guess no strong objections, but I'm still not sure it's worth supporting the optimization.

Quoted 80 lines
arch/x86/kernel/shstk.c +28 -0
+		ret = -EFAULT;
+	else
+		wrmsrq(MSR_IA32_PL3_SSP, ssp + SS_FRAME_SIZE);
+	fpregs_unlock();
+
+	return ret;
+}
+
+int shstk_push(u64 val)
+{
+	u64 ssp;
+	int ret;
+
+	if (!features_enabled(ARCH_SHSTK_SHSTK))
+		return -ENOTSUPP;
+
+	fpregs_lock_and_load();
+
+	rdmsrq(MSR_IA32_PL3_SSP, ssp);
+	ssp -= SS_FRAME_SIZE;
+	ret = write_user_shstk_64((__user void *)ssp, val);
+	if (!ret)
+		wrmsrq(MSR_IA32_PL3_SSP, ssp);
+	fpregs_unlock();
+
+	return ret;
+}
+
 #define SHSTK_DATA_BIT BIT(63)
 
 static int put_shstk_data(u64 __user *addr, u64 data)
arch/x86/kernel/uprobes.c +8 -9
@@ -804,7 +804,7 @@ SYSCALL_DEFINE0(uprobe)
 {
 	struct pt_regs *regs = task_pt_regs(current);
 	struct uprobe_syscall_args args;
-	unsigned long ip, sp;
+	unsigned long ip, sp, sret;
 	int err;
 
 	/* Allow execution only from uprobe trampolines. */
@@ -831,6 +831,10 @@ SYSCALL_DEFINE0(uprobe)
 
 	sp = regs->sp;
 
+	err = shstk_pop((u64 *)&sret);
+	if (err == -EFAULT || (!err && sret != args.retaddr))
+		goto sigill;
+
 	handle_syscall_uprobe(regs, regs->ip);
 
 	/*
@@ -855,6 +859,9 @@ SYSCALL_DEFINE0(uprobe)
 	if (args.retaddr - 5 != regs->ip)
 		args.retaddr = regs->ip;
 
+	if (shstk_push(args.retaddr) == -EFAULT)
+		goto sigill;
+
 	regs->ip = ip;
 
 	err = copy_to_user((void __user *)regs->sp, &args, sizeof(args));
@@ -1124,14 +1131,6 @@ void arch_uprobe_optimize(struct arch_up
 	struct mm_struct *mm = current->mm;
 	uprobe_opcode_t insn[5];
 
-	/*
-	 * Do not optimize if shadow stack is enabled, the return address hijack
-	 * code in arch_uretprobe_hijack_return_addr updates wrong frame when
-	 * the entry uprobe is optimized and the shadow stack crashes the app.
-	 */
-	if (shstk_is_enabled())
-		return;
-
 	if (!should_optimize(auprobe))
 		return;
 
merge bot Applied
#14 lore raw

[tip: perf/core] uprobes/x86: Fix uprobe syscall vs shadow stack

In reply to Peter Zijlstra
The following commit has been merged into the perf/core branch of tip:

The following commit has been merged into the perf/core branch of tip:

Commit-ID:     f349ec80865dfeaf8a33feae6b4a500db039c69c
Gitweb:        https://git.kernel.org/tip/f349ec80865dfeaf8a33feae6b4a500db039c69c
Author:        Peter Zijlstra <[email protected]>
AuthorDate:    Wed, 20 Aug 2025 14:28:34 +02:00
Committer:     Peter Zijlstra <[email protected]>
CommitterDate: Thu, 21 Aug 2025 20:09:22 +02:00

uprobes/x86: Fix uprobe syscall vs shadow stack

The uprobe syscall stores and strips the trampoline stack frame from the user context, to make it appear similar to an exception at the original instruction. It then restores the trampoline stack when it can exit using sysexit.

Make sure to match the regular stack manipulation with shadow stack operations such that regular and shadow stack don't get out of sync and causes trouble.

This enables using the optimization when shadow stack is in use.

---
 arch/x86/include/asm/shstk.h |  4 ++++-
 arch/x86/kernel/shstk.c      | 40 +++++++++++++++++++++++++++++++++++-
 arch/x86/kernel/uprobes.c    | 17 +++++++--------
 3 files changed, 52 insertions(+), 9 deletions(-)
arch/x86/include/asm/shstk.h +4 -0
index ba6f2fe..6723dbf 100644
@@ -23,6 +23,8 @@ int setup_signal_shadow_stack(struct ksignal *ksig);
 int restore_signal_shadow_stack(void);
 int shstk_update_last_frame(unsigned long val);
 bool shstk_is_enabled(void);
+int shstk_pop(u64 *val);
+int shstk_push(u64 val);
 #else
 static inline long shstk_prctl(struct task_struct *task, int option,
 			       unsigned long arg2) { return -EINVAL; }
@@ -35,6 +37,8 @@ static inline int setup_signal_shadow_stack(struct ksignal *ksig) { return 0; }
 static inline int restore_signal_shadow_stack(void) { return 0; }
 static inline int shstk_update_last_frame(unsigned long val) { return 0; }
 static inline bool shstk_is_enabled(void) { return false; }
+static inline int shstk_pop(u64 *val) { return -ENOTSUPP; }
+static inline int shstk_push(u64 val) { return -ENOTSUPP; }
 #endif /* CONFIG_X86_USER_SHADOW_STACK */
 
 #endif /* __ASSEMBLER__ */
arch/x86/kernel/shstk.c +40 -0
index 2ddf233..2dd91a3 100644
@@ -246,6 +246,46 @@ static unsigned long get_user_shstk_addr(void)
 	return ssp;
 }
 
+int shstk_pop(u64 *val)
+{
+	int ret = 0;
+	u64 ssp;
+
+	if (!features_enabled(ARCH_SHSTK_SHSTK))
+		return -ENOTSUPP;
+
+	fpregs_lock_and_load();
+
+	rdmsrq(MSR_IA32_PL3_SSP, ssp);
+	if (val && get_user(*val, (__user u64 *)ssp))
+		ret = -EFAULT;
+	else
+		wrmsrq(MSR_IA32_PL3_SSP, ssp + SS_FRAME_SIZE);
+	fpregs_unlock();
+
+	return ret;
+}
+
+int shstk_push(u64 val)
+{
+	u64 ssp;
+	int ret;
+
+	if (!features_enabled(ARCH_SHSTK_SHSTK))
+		return -ENOTSUPP;
+
+	fpregs_lock_and_load();
+
+	rdmsrq(MSR_IA32_PL3_SSP, ssp);
+	ssp -= SS_FRAME_SIZE;
+	ret = write_user_shstk_64((__user void *)ssp, val);
+	if (!ret)
+		wrmsrq(MSR_IA32_PL3_SSP, ssp);
+	fpregs_unlock();
+
+	return ret;
+}
+
 #define SHSTK_DATA_BIT BIT(63)
 
 static int put_shstk_data(u64 __user *addr, u64 data)
arch/x86/kernel/uprobes.c +8 -9
index d513c97..ab6547b 100644
@@ -804,7 +804,7 @@ SYSCALL_DEFINE0(uprobe)
 {
 	struct pt_regs *regs = task_pt_regs(current);
 	struct uprobe_syscall_args args;
-	unsigned long ip, sp;
+	unsigned long ip, sp, sret;
 	int err;
 
 	/* Allow execution only from uprobe trampolines. */
@@ -831,6 +831,10 @@ SYSCALL_DEFINE0(uprobe)
 
 	sp = regs->sp;
 
+	err = shstk_pop((u64 *)&sret);
+	if (err == -EFAULT || (!err && sret != args.retaddr))
+		goto sigill;
+
 	handle_syscall_uprobe(regs, regs->ip);
 
 	/*
@@ -855,6 +859,9 @@ SYSCALL_DEFINE0(uprobe)
 	if (args.retaddr - 5 != regs->ip)
 		args.retaddr = regs->ip;
 
+	if (shstk_push(args.retaddr) == -EFAULT)
+		goto sigill;
+
 	regs->ip = ip;
 
 	err = copy_to_user((void __user *)regs->sp, &args, sizeof(args));
@@ -1124,14 +1131,6 @@ void arch_uprobe_optimize(struct arch_uprobe *auprobe, unsigned long vaddr)
 	struct mm_struct *mm = current->mm;
 	uprobe_opcode_t insn[5];
 
-	/*
-	 * Do not optimize if shadow stack is enabled, the return address hijack
-	 * code in arch_uretprobe_hijack_return_addr updates wrong frame when
-	 * the entry uprobe is optimized and the shadow stack crashes the app.
-	 */
-	if (shstk_is_enabled())
-		return;
-
 	if (!should_optimize(auprobe))
 		return;
 
author maintainer
#15 lore raw

PATCH 5/6 uprobes/x86: Make asm style consistent

In reply to Peter Zijlstra

The asm syntax in uretprobe_trampoline and uprobe_trampoline differs in the use of operand size suffixes. Make them consistent and remove all size suffixes.

---
 arch/x86/kernel/uprobes.c |   16 ++++++++--------
 1 file changed, 8 insertions(+), 8 deletions(-)
arch/x86/kernel/uprobes.c +8 -8
@@ -321,21 +321,21 @@ asm (
 	".pushsection .rodata\n"
 	".global uretprobe_trampoline_entry\n"
 	"uretprobe_trampoline_entry:\n"
-	"pushq %rax\n"
-	"pushq %rcx\n"
-	"pushq %r11\n"
-	"movq $" __stringify(__NR_uretprobe) ", %rax\n"
+	"push %rax\n"
+	"push %rcx\n"
+	"push %r11\n"
+	"mov $" __stringify(__NR_uretprobe) ", %rax\n"
 	"syscall\n"
 	".global uretprobe_syscall_check\n"
 	"uretprobe_syscall_check:\n"
-	"popq %r11\n"
-	"popq %rcx\n"
+	"pop %r11\n"
+	"pop %rcx\n"
 	/*
 	 * The uretprobe syscall replaces stored %rax value with final
 	 * return address, so we don't restore %rax in here and just
 	 * call ret.
 	 */
-	"retq\n"
+	"ret\n"
 	".global uretprobe_trampoline_end\n"
 	"uretprobe_trampoline_end:\n"
 	".popsection\n"
@@ -885,7 +885,7 @@ asm (
 	"push %rcx\n"
 	"push %r11\n"
 	"push %rax\n"
-	"movq $" __stringify(__NR_uprobe) ", %rax\n"
+	"mov $" __stringify(__NR_uprobe) ", %rax\n"
 	"syscall\n"
 	"pop %rax\n"
 	"pop %r11\n"
merge bot Applied
#16 lore raw

[tip: perf/core] uprobes/x86: Make asm style consistent

The following commit has been merged into the perf/core branch of tip:

The following commit has been merged into the perf/core branch of tip:

Commit-ID:     60ed85b7e46983725954103beaca5ca41816cd58
Gitweb:        https://git.kernel.org/tip/60ed85b7e46983725954103beaca5ca41816cd58
Author:        Peter Zijlstra <[email protected]>
AuthorDate:    Thu, 21 Aug 2025 13:48:48 +02:00
Committer:     Peter Zijlstra <[email protected]>
CommitterDate: Thu, 21 Aug 2025 20:09:22 +02:00

uprobes/x86: Make asm style consistent

The asm syntax in uretprobe_trampoline and uprobe_trampoline differs in the use of operand size suffixes. Make them consistent and remove all size suffixes.

---
 arch/x86/kernel/uprobes.c | 16 ++++++++--------
 1 file changed, 8 insertions(+), 8 deletions(-)
arch/x86/kernel/uprobes.c +8 -8
index ab6547b..643027e 100644
@@ -321,21 +321,21 @@ asm (
 	".pushsection .rodata\n"
 	".global uretprobe_trampoline_entry\n"
 	"uretprobe_trampoline_entry:\n"
-	"pushq %rax\n"
-	"pushq %rcx\n"
-	"pushq %r11\n"
-	"movq $" __stringify(__NR_uretprobe) ", %rax\n"
+	"push %rax\n"
+	"push %rcx\n"
+	"push %r11\n"
+	"mov $" __stringify(__NR_uretprobe) ", %rax\n"
 	"syscall\n"
 	".global uretprobe_syscall_check\n"
 	"uretprobe_syscall_check:\n"
-	"popq %r11\n"
-	"popq %rcx\n"
+	"pop %r11\n"
+	"pop %rcx\n"
 	/*
 	 * The uretprobe syscall replaces stored %rax value with final
 	 * return address, so we don't restore %rax in here and just
 	 * call ret.
 	 */
-	"retq\n"
+	"ret\n"
 	".global uretprobe_trampoline_end\n"
 	"uretprobe_trampoline_end:\n"
 	".popsection\n"
@@ -885,7 +885,7 @@ asm (
 	"push %rcx\n"
 	"push %r11\n"
 	"push %rax\n"
-	"movq $" __stringify(__NR_uprobe) ", %rax\n"
+	"mov $" __stringify(__NR_uprobe) ", %rax\n"
 	"syscall\n"
 	"pop %rax\n"
 	"pop %r11\n"
author maintainer
#17 lore raw

PATCH 6/6 uprobes/x86: Add SLS mitigation to the trampolines

In reply to Peter Zijlstra

It is trivial; no reason not to.

---
 arch/x86/kernel/uprobes.c |    2 ++
 1 file changed, 2 insertions(+)
arch/x86/kernel/uprobes.c +2 -0
@@ -336,6 +336,7 @@ asm (
 	 * call ret.
 	 */
 	"ret\n"
+	"int3\n"
 	".global uretprobe_trampoline_end\n"
 	"uretprobe_trampoline_end:\n"
 	".popsection\n"
@@ -891,6 +892,7 @@ asm (
 	"pop %r11\n"
 	"pop %rcx\n"
 	"ret\n"
+	"int3\n"
 	".balign " __stringify(PAGE_SIZE) "\n"
 	".popsection\n"
 );
merge bot Applied
#18 lore raw

[tip: perf/core] uprobes/x86: Add SLS mitigation to the trampolines

The following commit has been merged into the perf/core branch of tip:

The following commit has been merged into the perf/core branch of tip:

Commit-ID:     354492a0e1bc4a408e26ebe14166bd1064182439
Gitweb:        https://git.kernel.org/tip/354492a0e1bc4a408e26ebe14166bd1064182439
Author:        Peter Zijlstra <[email protected]>
AuthorDate:    Wed, 20 Aug 2025 19:49:56 +02:00
Committer:     Peter Zijlstra <[email protected]>
CommitterDate: Thu, 21 Aug 2025 20:09:22 +02:00

uprobes/x86: Add SLS mitigation to the trampolines

It is trivial; no reason not to.

---
 arch/x86/kernel/uprobes.c | 2 ++
 1 file changed, 2 insertions(+)
arch/x86/kernel/uprobes.c +2 -0
index 643027e..0a8c0a4 100644
@@ -336,6 +336,7 @@ asm (
 	 * call ret.
 	 */
 	"ret\n"
+	"int3\n"
 	".global uretprobe_trampoline_end\n"
 	"uretprobe_trampoline_end:\n"
 	".popsection\n"
@@ -891,6 +892,7 @@ asm (
 	"pop %r11\n"
 	"pop %rcx\n"
 	"ret\n"
+	"int3\n"
 	".balign " __stringify(PAGE_SIZE) "\n"
 	".popsection\n"
 );
maintainer Acked-by on v1 6/6
#19 lore raw

Re: [PATCH 6/6] uprobes/x86: Add SLS mitigation to the trampolines

In reply to Peter Zijlstra
Quoted 3 lines · On Thu, 21 Aug 2025 14:28:28 +0200 Peter Zijlstra <[email protected]> wrote:

It is trivial; no reason not to.

Looks good to me :)

Quoted 24 lines
---
 arch/x86/kernel/uprobes.c |    2 ++
 1 file changed, 2 insertions(+)
arch/x86/kernel/uprobes.c +2 -0
@@ -336,6 +336,7 @@ asm (
 	 * call ret.
 	 */
 	"ret\n"
+	"int3\n"
 	".global uretprobe_trampoline_end\n"
 	"uretprobe_trampoline_end:\n"
 	".popsection\n"
@@ -891,6 +892,7 @@ asm (
 	"pop %r11\n"
 	"pop %rcx\n"
 	"ret\n"
+	"int3\n"
 	".balign " __stringify(PAGE_SIZE) "\n"
 	".popsection\n"
 );
Signature · 1 line
Masami Hiramatsu (Google) <[email protected]>
#20 lore raw
In reply to Peter Zijlstra
Quoted 4 lines · Peter Zijlstra

Hi,

These are cleanups and fixes that I applied on top of Jiri's patches:

Quoted 9 lines

https://lkml.kernel.org/r/[email protected]

The combined lot sits in:

git://git.kernel.org/pub/scm/linux/kernel/git/peterz/queue.git perf/core

Jiri was going to send me some selftest updates that might mean rebasing that tree, but we'll see. If this all works we'll land it in -tip.

hi,
sent the selftest fix in here:
https://lore.kernel.org/bpf/[email protected]/T/#u

thanks,
jirka

#22 lore raw
Quoted 7 lines · Rick P
Quoted 3 lines · Jiri Olsa

hi,
sent the selftest fix in here:
  https://lore.kernel.org/bpf/[email protected]/T/#u

I tried for a bit to run the BPF selftests, and ran into various issues. Did you have access to CET HW to test?

yes,

  # cat /proc/cpuinfo  | grep user_shstk | head -1
  flags           : fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush dts acpi mmx fxsr sse sse2 ss ht tm pbe syscall nx pdpe1gb rdtscp lm constant_tsc art arch_perfmon pebs bts rep_good nopl xtopology nonstop_tsc cpuid aperfmperf tsc_known_freq pni pclmulqdq dtes64 monitor ds_cpl vmx smx est tm2 ssse3 sdbg fma cx16 xtpr pdcm pcid sse4_1 sse4_2 x2apic movbe popcnt tsc_deadline_timer aes xsave avx f16c rdrand lahf_lm abm 3dnowprefetch cpuid_fault epb ssbd ibrs ibpb stibp ibrs_enhanced tpr_shadow flexpriority ept vpid ept_ad fsgsbase tsc_adjust bmi1 avx2 smep bmi2 erms invpcid rdseed adx smap clflushopt clwb intel_pt sha_ni xsaveopt xsavec xgetbv1 xsaves split_lock_detect user_shstk avx_vnni dtherm ida arat pln pts hwp hwp_notify hwp_act_window hwp_epp hwp_pkg_req hfi vnmi umip pku ospke waitpkg gfni vaes vpclmulqdq rdpid bus_lock_detect movdiri movdir64b fsrm md_clear serialize pconfig arch_lbr ibt flush_l1d arch_capabilities

also I had to merge in bpf-next/master to be able to build it, what issues did you see?

jirka

#23 lore raw
Quoted 9 lines · Jiri Olsa

yes,

  # cat /proc/cpuinfo  | grep user_shstk | head -1
  flags           : fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush dts acpi mmx fxsr sse sse2 ss ht tm pbe syscall nx pdpe1gb rdtscp lm constant_tsc art arch_perfmon pebs bts rep_good nopl xtopology nonstop_tsc cpuid aperfmperf tsc_known_freq pni pclmulqdq dtes64 monitor ds_cpl vmx smx est tm2 ssse3 sdbg fma cx16 xtpr pdcm pcid sse4_1 sse4_2 x2apic movbe popcnt tsc_deadline_timer aes xsave avx f16c rdrand lahf_lm abm 3dnowprefetch cpuid_fault epb ssbd ibrs ibpb stibp ibrs_enhanced tpr_shadow flexpriority ept vpid ept_ad fsgsbase tsc_adjust bmi1 avx2 smep bmi2 erms invpcid rdseed adx smap clflushopt clwb intel_pt sha_ni xsaveopt xsavec xgetbv1 xsaves split_lock_detect user_shstk avx_vnni dtherm ida arat pln pts hwp hwp_notify hwp_act_window hwp_epp hwp_pkg_req hfi vnmi umip pku ospke waitpkg gfni vaes vpclmulqdq rdpid bus_lock_detect movdiri movdir64b fsrm md_clear serialize pconfig arch_lbr ibt flush_l1d arch_capabilities

also I had to merge in bpf-next/master to be able to build it, what issues did you see?

Ah. I had various compile issues. Some of which were I guess the common issue of needing the latest tools, but also some others. IIRC some spin lock definitions were missing. I guess it probably was the lack of merging that branch. But I was just going to run the test, so if you already have, I'll skip it.

#24 lore raw
In reply to Jiri Olsa
Quoted 11 lines · Jiri Olsa
Quoted 10 lines · Peter Zijlstra

Hi,

These are cleanups and fixes that I applied on top of Jiri's patches:

https://lkml.kernel.org/r/[email protected]

The combined lot sits in:

git://git.kernel.org/pub/scm/linux/kernel/git/peterz/queue.git perf/core

Quoted 7 lines
Quoted 3 lines

Jiri was going to send me some selftest updates that might mean rebasing that tree, but we'll see. If this all works we'll land it in -tip.

hi,
sent the selftest fix in here:
https://lore.kernel.org/bpf/[email protected]/T/#u

Andrii,
do we want any special logistic for the bpf/selftest changes or it could go through the tip tree?

thanks,
jirka

#25 lore raw
Quoted 17 lines · Jiri Olsa
Quoted 15 lines · Jiri Olsa
Quoted 13 lines · Peter Zijlstra

Hi,

These are cleanups and fixes that I applied on top of Jiri's patches:

https://lkml.kernel.org/r/[email protected]

The combined lot sits in:

git://git.kernel.org/pub/scm/linux/kernel/git/peterz/queue.git perf/core

Jiri was going to send me some selftest updates that might mean rebasing that tree, but we'll see. If this all works we'll land it in -tip.

Quoted 7 lines
Quoted 3 lines

hi,
sent the selftest fix in here:
https://lore.kernel.org/bpf/[email protected]/T/#u

Andrii,
do we want any special logistic for the bpf/selftest changes or it could go through the tip tree?

let's route selftest changes through tip together with the rest of uprobe changes, it's unlikely to conflict

Quoted 3 lines

thanks,
jirka

#26 lore raw
Quoted 22 lines · Andrii Nakryiko
Quoted 21 lines · Jiri Olsa
Quoted 18 lines · Jiri Olsa
Quoted 4 lines · Peter Zijlstra

Hi,

These are cleanups and fixes that I applied on top of Jiri's patches:

Quoted 9 lines

https://lkml.kernel.org/r/[email protected]

The combined lot sits in:

git://git.kernel.org/pub/scm/linux/kernel/git/peterz/queue.git perf/core

Jiri was going to send me some selftest updates that might mean rebasing that tree, but we'll see. If this all works we'll land it in -tip.

hi,
sent the selftest fix in here:
https://lore.kernel.org/bpf/[email protected]/T/#u

Quoted 6 lines
Quoted 3 lines

Andrii,
do we want any special logistic for the bpf/selftest changes or it could go through the tip tree?

let's route selftest changes through tip together with the rest of uprobe changes, it's unlikely to conflict

fyi, there's conflict now between tip/perf/core and bpf-next/master in the selftests.. due to usdt SIB argument support changes

please let me know if you need any help in resolving that

jirka

#27 lore raw
Quoted 28 lines · Jiri Olsa
Quoted 26 lines · Andrii Nakryiko
Quoted 24 lines · Jiri Olsa
Quoted 11 lines · Jiri Olsa
Quoted 10 lines · Peter Zijlstra

Hi,

These are cleanups and fixes that I applied on top of Jiri's patches:

https://lkml.kernel.org/r/[email protected]

The combined lot sits in:

git://git.kernel.org/pub/scm/linux/kernel/git/peterz/queue.git perf/core

Quoted 7 lines
Quoted 3 lines

Jiri was going to send me some selftest updates that might mean rebasing that tree, but we'll see. If this all works we'll land it in -tip.

hi,
sent the selftest fix in here:
https://lore.kernel.org/bpf/[email protected]/T/#u

Andrii,
do we want any special logistic for the bpf/selftest changes or it could go through the tip tree?

Quoted 7 lines
Quoted 2 lines

let's route selftest changes through tip together with the rest of uprobe changes, it's unlikely to conflict

fyi, there's conflict now between tip/perf/core and bpf-next/master in the selftests.. due to usdt SIB argument support changes

please let me know if you need any help in resolving that

so selftest change hasn't landed in tip/perf/core just yet, is that right? If there is a conflict, I guess that changes equation a bit. I'd land it in bpf-next and for now disable that test in BPF CI until the trees converge. WDYT?

Quoted 2 lines

jirka

#28 lore raw
Quoted 35 lines · Andrii Nakryiko
Quoted 34 lines · Jiri Olsa
Quoted 22 lines · Andrii Nakryiko
Quoted 21 lines · Jiri Olsa
Quoted 18 lines · Jiri Olsa
Quoted 4 lines · Peter Zijlstra

Hi,

These are cleanups and fixes that I applied on top of Jiri's patches:

Quoted 9 lines

https://lkml.kernel.org/r/[email protected]

The combined lot sits in:

git://git.kernel.org/pub/scm/linux/kernel/git/peterz/queue.git perf/core

Jiri was going to send me some selftest updates that might mean rebasing that tree, but we'll see. If this all works we'll land it in -tip.

hi,
sent the selftest fix in here:
https://lore.kernel.org/bpf/[email protected]/T/#u

Quoted 6 lines
Quoted 3 lines

Andrii,
do we want any special logistic for the bpf/selftest changes or it could go through the tip tree?

let's route selftest changes through tip together with the rest of uprobe changes, it's unlikely to conflict

fyi, there's conflict now between tip/perf/core and bpf-next/master in the selftests.. due to usdt SIB argument support changes

Quoted 6 lines
Quoted 1 line

please let me know if you need any help in resolving that

so selftest change hasn't landed in tip/perf/core just yet, is that right? If there is a conflict, I guess that changes equation a bit. I'd land it in bpf-next and for now disable that test in BPF CI until the trees converge. WDYT?

I can see the selftests changes in tip/perf/core already

jirka

16ed38922765 (HEAD -> tip/perf/core) perf: Skip user unwind if the task is a kernel thread d77e3319e310 perf: Simplify get_perf_callchain() user logic
90942f9fac05 perf: Use current->flags & PF_KTHREAD|PF_USER_WORKER instead of current->mm == NULL 153f9e74dec2 perf: Have get_perf_callchain() return NULL if crosstask and user are set e649bcda25b5 perf: Remove get_perf_callchain() init_nr argument
f49e1be19542 perf/x86: Print PMU counters bitmap in x86_pmu_show_pmu_cap() 2676dbf9f4fb perf/x86/intel: Add ICL_FIXED_0_ADAPTIVE bit into INTEL_FIXED_BITS_MASK 9b3e119784bc perf/x86/intel: Change macro GLOBAL_CTRL_EN_PERF_METRICS to BIT_ULL(48) 0c5caea762de perf/x86: Add PERF_CAP_PEBS_TIMING_INFO flag
43796f305078 perf/x86/intel: Fix IA32_PMC_x_CFG_B MSRs access error
d9cf9c6884d2 perf/x86/intel: Use early_initcall() to hook bts_init() e173287b5d21 uprobes: Remove redundant __GFP_NOWARN
9ffc7a635c35 selftests/seccomp: validate uprobe syscall passes through seccomp 89d1d8434d24 seccomp: passthrough uprobe systemcall without filtering 52718438af2a selftests/bpf: Fix uprobe syscall shadow stack test
3abf4298c613 selftests/bpf: Change test_uretprobe_regs_change for uprobe and uretprobe 275eae678986 selftests/bpf: Add uprobe_regs_equal test
875e1705ad99 selftests/bpf: Add optimized usdt variant for basic usdt test c11661bd9adf selftests/bpf: Add uprobe syscall sigill signal test
c8be59667cf1 selftests/bpf: Add hit/attach/detach race optimized uprobe test d5c86c337010 selftests/bpf: Add uprobe/usdt syscall tests
7932c4cf5771 selftests/bpf: Rename uprobe_syscall_executed prog to test_uretprobe_multi 4e7005223e6d selftests/bpf: Reorg the uprobe_syscall test function
17c3b0015764 selftests/bpf: Import usdt.h from libbpf/usdt project
354492a0e1bc uprobes/x86: Add SLS mitigation to the trampolines
60ed85b7e469 uprobes/x86: Make asm style consistent
...

#29 lore raw
Quoted 37 lines · Jiri Olsa
Quoted 35 lines · Andrii Nakryiko
Quoted 34 lines · Jiri Olsa
Quoted 22 lines · Andrii Nakryiko
Quoted 21 lines · Jiri Olsa
Quoted 18 lines · Jiri Olsa
Quoted 4 lines · Peter Zijlstra

Hi,

These are cleanups and fixes that I applied on top of Jiri's patches:

Quoted 9 lines

https://lkml.kernel.org/r/[email protected]

The combined lot sits in:

git://git.kernel.org/pub/scm/linux/kernel/git/peterz/queue.git perf/core

Jiri was going to send me some selftest updates that might mean rebasing that tree, but we'll see. If this all works we'll land it in -tip.

hi,
sent the selftest fix in here:
https://lore.kernel.org/bpf/[email protected]/T/#u

Quoted 6 lines
Quoted 3 lines

Andrii,
do we want any special logistic for the bpf/selftest changes or it could go through the tip tree?

let's route selftest changes through tip together with the rest of uprobe changes, it's unlikely to conflict

fyi, there's conflict now between tip/perf/core and bpf-next/master in the selftests.. due to usdt SIB argument support changes

Quoted 9 lines
Quoted 6 lines
Quoted 1 line

please let me know if you need any help in resolving that

so selftest change hasn't landed in tip/perf/core just yet, is that right? If there is a conflict, I guess that changes equation a bit. I'd land it in bpf-next and for now disable that test in BPF CI until the trees converge. WDYT?

I can see the selftests changes in tip/perf/core already

Ah, that's the shadow stack test. I thought we are talking about -ENXIO change to sys_uprobe and dropping that BPF selftests you had that makes sure the process is killed.

Well, I guess we'll have to handle that conflict then?..

Quoted 30 lines

jirka

16ed38922765 (HEAD -> tip/perf/core) perf: Skip user unwind if the task is a kernel thread d77e3319e310 perf: Simplify get_perf_callchain() user logic
90942f9fac05 perf: Use current->flags & PF_KTHREAD|PF_USER_WORKER instead of current->mm == NULL 153f9e74dec2 perf: Have get_perf_callchain() return NULL if crosstask and user are set e649bcda25b5 perf: Remove get_perf_callchain() init_nr argument
f49e1be19542 perf/x86: Print PMU counters bitmap in x86_pmu_show_pmu_cap() 2676dbf9f4fb perf/x86/intel: Add ICL_FIXED_0_ADAPTIVE bit into INTEL_FIXED_BITS_MASK 9b3e119784bc perf/x86/intel: Change macro GLOBAL_CTRL_EN_PERF_METRICS to BIT_ULL(48) 0c5caea762de perf/x86: Add PERF_CAP_PEBS_TIMING_INFO flag
43796f305078 perf/x86/intel: Fix IA32_PMC_x_CFG_B MSRs access error
d9cf9c6884d2 perf/x86/intel: Use early_initcall() to hook bts_init() e173287b5d21 uprobes: Remove redundant __GFP_NOWARN
9ffc7a635c35 selftests/seccomp: validate uprobe syscall passes through seccomp 89d1d8434d24 seccomp: passthrough uprobe systemcall without filtering 52718438af2a selftests/bpf: Fix uprobe syscall shadow stack test
3abf4298c613 selftests/bpf: Change test_uretprobe_regs_change for uprobe and uretprobe 275eae678986 selftests/bpf: Add uprobe_regs_equal test
875e1705ad99 selftests/bpf: Add optimized usdt variant for basic usdt test c11661bd9adf selftests/bpf: Add uprobe syscall sigill signal test
c8be59667cf1 selftests/bpf: Add hit/attach/detach race optimized uprobe test d5c86c337010 selftests/bpf: Add uprobe/usdt syscall tests
7932c4cf5771 selftests/bpf: Rename uprobe_syscall_executed prog to test_uretprobe_multi 4e7005223e6d selftests/bpf: Reorg the uprobe_syscall test function
17c3b0015764 selftests/bpf: Import usdt.h from libbpf/usdt project
354492a0e1bc uprobes/x86: Add SLS mitigation to the trampolines
60ed85b7e469 uprobes/x86: Make asm style consistent
...