[PATCH v2 0/5] alpha: perf register sampling and user stack dumps

[PATCH v2 0/5] alpha: perf register sampling and user stack dumps

[Matt Turner]

This lets perf sample registers and dump the user stack on alpha, which
perf record --call-graph dwarf needs to unwind user space.

r9-r15 are not in pt_regs, so they cannot be sampled. An unwinder
therefore cannot get past a first frame that uses $15 as its frame
pointer (-O0 code or alloca).

The perf tool side is a separate patch for perf-tools-next:

  https://lore.kernel.org/r/20261011-alpha-perf-tools-regs-v1-1-2a5940e22548@gmail.com

libdw also needs alpha unwinding support, which I sent to
elfutils-devel:

  https://inbox.sourceware.org/elfutils-devel/20261012-alpha-v1-0-46760216dfb7@gmail.com/

---
Changes in v2:
- Patch 4: report sp as 0 for software event samples, whose pt_regs is
  a per-CPU buffer rather than a trap frame (Magnus).
- New patch 5: fill in pc for those samples via
  perf_arch_fetch_caller_regs() (Magnus).
- Collect trailers. Magnus's Tested-by on patch 4 dropped until he
  retests the change.
- Link the perf and elfutils postings.
- Link to v1: https://lore.kernel.org/r/20261011-alpha-perf-regs-v3-v1-0-eaa625a39712@gmail.com

---
Matt Turner (5):
      alpha: fix stale user_stack_pointer()
      alpha: stop storing the user stack pointer in pt_regs
      alpha: use user_stack_pointer() in KSTK_ESP() and dump_elf_thread()
      alpha: add HAVE_PERF_REGS and HAVE_PERF_USER_STACK_DUMP support
      alpha: fill in pc for software event samples

 .../features/perf/perf-regs/arch-support.txt       |  2 +-
 .../features/perf/perf-stackdump/arch-support.txt  |  2 +-
 arch/alpha/Kconfig                                 |  2 +
 arch/alpha/include/asm/perf_event.h                |  7 +++
 arch/alpha/include/asm/processor.h                 |  3 +-
 arch/alpha/include/asm/ptrace.h                    |  7 +--
 arch/alpha/include/uapi/asm/perf_regs.h            | 45 +++++++++++++++
 arch/alpha/include/uapi/asm/ptrace.h               |  2 +-
 arch/alpha/kernel/Makefile                         |  1 +
 arch/alpha/kernel/asm-offsets.c                    |  1 -
 arch/alpha/kernel/entry.S                          |  5 --
 arch/alpha/kernel/perf_regs.c                      | 64 ++++++++++++++++++++++
 arch/alpha/kernel/process.c                        |  2 +-
 arch/alpha/kernel/ptrace.c                         | 12 ++++
 arch/alpha/kernel/traps.c                          |  8 ---
 15 files changed, 137 insertions(+), 26 deletions(-)
---
base-commit: 8f5dbfb93440d1aba216a1a34369a65654b94be9
change-id: 20261001-alpha-perf-regs-v3-7e23508c1ca2

Best regards,
-- 
Matt Turner <mattst88@gmail.com>

[PATCH v2 1/5] alpha: fix stale user_stack_pointer()

[Matt Turner]

PALcode keeps the user stack pointer in the PCB, not the trap frame, so
pt_regs->usp is only valid where entry code stores it: entSys,
do_entArith() and do_entIF(). Interrupts, page faults and unaligned
traps leave it stale, and ptrace and /proc/<pid>/syscall read it
through user_stack_pointer().

A user-mode pt_regs is at the top of its task's stack, so find the
thread_info from it and use rdusp() for current or pcb.usp otherwise,
as KSTK_ESP() and the regset code already do.

Fixes: 8a2857bc2b30 ("alpha: enable regset-based ptrace and core dumps")
Assisted-by: Claude:claude-opus-5-5
Reviewed-by: Magnus Lindholm <linmag7@gmail.com>
Tested-by: Magnus Lindholm <linmag7@gmail.com>
Acked-by: Ian Rogers <irogers@google.com>
Signed-off-by: Matt Turner <mattst88@gmail.com>
---
 arch/alpha/include/asm/ptrace.h |  7 +------
 arch/alpha/kernel/ptrace.c      | 12 ++++++++++++
 2 files changed, 13 insertions(+), 6 deletions(-)

diff --git a/arch/alpha/include/asm/ptrace.h b/arch/alpha/include/asm/ptrace.h
index 430e8dc27ff7..45d1229344b6 100644
--- a/arch/alpha/include/asm/ptrace.h
+++ b/arch/alpha/include/asm/ptrace.h
@@ -26,12 +26,7 @@ static inline unsigned long regs_return_value(struct pt_regs *regs)
 	return regs->r0;
 }
 
-/* Helpers for working with the user stack pointer */
-static inline unsigned long user_stack_pointer(struct pt_regs *regs)
-{
-	/* Valid for user-mode regs */
-	return regs->usp;
-}
+unsigned long user_stack_pointer(struct pt_regs *regs);
 
 static __always_inline bool regs_irqs_disabled(struct pt_regs *regs)
 {
diff --git a/arch/alpha/kernel/ptrace.c b/arch/alpha/kernel/ptrace.c
index e1d8c2f69688..dba7e9905cfa 100644
--- a/arch/alpha/kernel/ptrace.c
+++ b/arch/alpha/kernel/ptrace.c
@@ -440,6 +440,18 @@ ptrace_cancel_bpt(struct task_struct * child)
 	return (nsaved != 0);
 }
 
+/*
+ * PALcode keeps the user stack pointer in the PCB, not the trap frame.
+ * A user-mode pt_regs sits at the top of its task's stack, so the
+ * thread_info is at the base.
+ */
+unsigned long user_stack_pointer(struct pt_regs *regs)
+{
+	struct thread_info *ti = (void *)((unsigned long)regs & -THREAD_SIZE);
+
+	return ti == current_thread_info() ? rdusp() : ti->pcb.usp;
+}
+
 void user_enable_single_step(struct task_struct *child)
 {
 	/* Mark single stepping.  */

-- 
2.55.0

[PATCH v2 2/5] alpha: stop storing the user stack pointer in pt_regs

[Matt Turner]

Nothing reads pt_regs->usp now that user_stack_pointer() uses the PCB.
Drop the stores, which saves a PAL_rdusp call on every syscall, and
give the field back its old name.

Assisted-by: Claude:claude-opus-5-5
Reviewed-by: Magnus Lindholm <linmag7@gmail.com>
Tested-by: Magnus Lindholm <linmag7@gmail.com>
Acked-by: Ian Rogers <irogers@google.com>
Signed-off-by: Matt Turner <mattst88@gmail.com>
---
 arch/alpha/include/uapi/asm/ptrace.h | 2 +-
 arch/alpha/kernel/asm-offsets.c      | 1 -
 arch/alpha/kernel/entry.S            | 5 -----
 arch/alpha/kernel/traps.c            | 8 --------
 4 files changed, 1 insertion(+), 15 deletions(-)

diff --git a/arch/alpha/include/uapi/asm/ptrace.h b/arch/alpha/include/uapi/asm/ptrace.h
index 9d86b2a1526e..72ed913a910f 100644
--- a/arch/alpha/include/uapi/asm/ptrace.h
+++ b/arch/alpha/include/uapi/asm/ptrace.h
@@ -43,7 +43,7 @@ struct pt_regs {
 	unsigned long trap_a1;
 	unsigned long trap_a2;
 /* This makes the stack 16-byte aligned as GCC expects */
-	unsigned long usp;
+	unsigned long __pad0;
 /* These are saved by PAL-code: */
 	unsigned long ps;
 	unsigned long pc;
diff --git a/arch/alpha/kernel/asm-offsets.c b/arch/alpha/kernel/asm-offsets.c
index c89792c86044..22ed3c975b1c 100644
--- a/arch/alpha/kernel/asm-offsets.c
+++ b/arch/alpha/kernel/asm-offsets.c
@@ -30,7 +30,6 @@ static void __used foo(void)
 
 	DEFINE(HAE_CACHE, offsetof(struct alpha_machine_vector, hae_cache));
 	DEFINE(HAE_REG, offsetof(struct alpha_machine_vector, hae_register));
-	DEFINE(PT_REGS_USP, offsetof(struct pt_regs, usp));
 
 	DEFINE(TI_FLAGS,        offsetof(struct thread_info, flags));
 	DEFINE(TI_SYSCALL_WORK, offsetof(struct thread_info, syscall_work));
diff --git a/arch/alpha/kernel/entry.S b/arch/alpha/kernel/entry.S
index bfa116d455fb..790389a15c3c 100644
--- a/arch/alpha/kernel/entry.S
+++ b/arch/alpha/kernel/entry.S
@@ -476,11 +476,6 @@ entSys:
 	ldq	$1, 0($sp)	/* syscall nr from saved r0 */
 	stq	$1, 8($sp)	/* regs->r1 = shadow syscall nr */
 
-	mov	$0, $8
-	call_pal PAL_rdusp
-	stq	$0, PT_REGS_USP($sp)
-	mov	$8, $0
-
 	lda	$8, 0x3fff
 	bic	$sp, $8, $8
 
diff --git a/arch/alpha/kernel/traps.c b/arch/alpha/kernel/traps.c
index cfaef96c4ce8..5882c65b7ef1 100644
--- a/arch/alpha/kernel/traps.c
+++ b/arch/alpha/kernel/traps.c
@@ -30,12 +30,6 @@
 
 #include "proto.h"
 
-static __always_inline void alpha_snapshot_usp(struct pt_regs *regs)
-{
-	if (user_mode(regs))
-		regs->usp = rdusp();
-}
-
 void
 dik_show_regs(struct pt_regs *regs, unsigned long *r9_15)
 {
@@ -186,7 +180,6 @@ do_entArith(unsigned long summary, unsigned long write_mask,
 {
 	long si_code = FPE_FLTINV;
 
-	alpha_snapshot_usp(regs);
 	if (summary & 1) {
 		/* Software-completion summary bit is set, so try to
 		   emulate the instruction.  If the processor supports
@@ -208,7 +201,6 @@ do_entIF(unsigned long type, struct pt_regs *regs)
 {
 	int signo, code;
 
-	alpha_snapshot_usp(regs);
 	if (type == 3) { /* FEN fault */
 		/* Irritating users can call PAL_clrfen to disable the
 		   FPU for the process.  The kernel will then trap in

-- 
2.55.0

[PATCH v2 3/5] alpha: use user_stack_pointer() in KSTK_ESP() and dump_elf_thread()

[Matt Turner]

Both open-code the same PCB lookup.

Assisted-by: Claude:claude-opus-5-5
Tested-by: Magnus Lindholm <linmag7@gmail.com>
Reviewed-by: Magnus Lindholm <linmag7@gmail.com>
Acked-by: Ian Rogers <irogers@google.com>
Signed-off-by: Matt Turner <mattst88@gmail.com>
---
 arch/alpha/include/asm/processor.h | 3 +--
 arch/alpha/kernel/process.c        | 2 +-
 2 files changed, 2 insertions(+), 3 deletions(-)

diff --git a/arch/alpha/include/asm/processor.h b/arch/alpha/include/asm/processor.h
index c461a26d47b3..73251cd671e0 100644
--- a/arch/alpha/include/asm/processor.h
+++ b/arch/alpha/include/asm/processor.h
@@ -43,8 +43,7 @@ unsigned long __get_wchan(struct task_struct *p);
 
 #define KSTK_EIP(tsk) (task_pt_regs(tsk)->pc)
 
-#define KSTK_ESP(tsk) \
-  ((tsk) == current ? rdusp() : task_thread_info(tsk)->pcb.usp)
+#define KSTK_ESP(tsk) user_stack_pointer(task_pt_regs(tsk))
 
 #define cpu_relax()	barrier()
 
diff --git a/arch/alpha/kernel/process.c b/arch/alpha/kernel/process.c
index 06522451f018..7861c9e767b7 100644
--- a/arch/alpha/kernel/process.c
+++ b/arch/alpha/kernel/process.c
@@ -318,7 +318,7 @@ dump_elf_thread(elf_greg_t *dest, struct pt_regs *pt, struct thread_info *ti)
 	dest[27] = pt->r27;
 	dest[28] = pt->r28;
 	dest[29] = pt->gp;
-	dest[30] = ti == current_thread_info() ? rdusp() : ti->pcb.usp;
+	dest[30] = user_stack_pointer(pt);
 	dest[31] = pt->pc;
 
 	/* Once upon a time this was the PS value.  Which is stupid

-- 
2.55.0

[PATCH v2 4/5] alpha: add HAVE_PERF_REGS and HAVE_PERF_USER_STACK_DUMP support

[Matt Turner]

Let perf sample registers and dump the user stack, which is what the
perf tool needs for --call-graph dwarf.

The register indices follow the hardware numbering, so r0-r30 map
directly to DWARF register numbers, with pc after them. r9-r15
are callee-saved and live in switch_stack rather than pt_regs, so they
cannot be sampled. Without r15 an unwinder cannot recover the CFA of a
first frame that uses $15 as its frame pointer.

sp is not in pt_regs either: for user frames it comes from
user_stack_pointer(), and for kernel frames (PERF_SAMPLE_REGS_INTR) it
is the end of the trap frame, as in stacktrace.c.

Software events such as context switches and tracepoints hand perf a
per-CPU pt_regs filled in by perf_fetch_caller_regs() instead of a
trap frame. It is not on the stack and has no room for sp, so report
sp as 0 there rather than the address of the per-CPU buffer.

Assisted-by: Claude:claude-opus-5-5
Acked-by: Ian Rogers <irogers@google.com>
Signed-off-by: Matt Turner <mattst88@gmail.com>
---
 .../features/perf/perf-regs/arch-support.txt       |  2 +-
 .../features/perf/perf-stackdump/arch-support.txt  |  2 +-
 arch/alpha/Kconfig                                 |  2 +
 arch/alpha/include/uapi/asm/perf_regs.h            | 45 +++++++++++++++
 arch/alpha/kernel/Makefile                         |  1 +
 arch/alpha/kernel/perf_regs.c                      | 64 ++++++++++++++++++++++
 6 files changed, 114 insertions(+), 2 deletions(-)

diff --git a/Documentation/features/perf/perf-regs/arch-support.txt b/Documentation/features/perf/perf-regs/arch-support.txt
index 4c9a5a012075..7d736d678fd4 100644
--- a/Documentation/features/perf/perf-regs/arch-support.txt
+++ b/Documentation/features/perf/perf-regs/arch-support.txt
@@ -6,7 +6,7 @@
     -----------------------
     |         arch |status|
     -----------------------
-    |       alpha: | TODO |
+    |       alpha: |  ok  |
     |         arc: | TODO |
     |         arm: |  ok  |
     |       arm64: |  ok  |
diff --git a/Documentation/features/perf/perf-stackdump/arch-support.txt b/Documentation/features/perf/perf-stackdump/arch-support.txt
index dd362b5cb638..17b84ea7769d 100644
--- a/Documentation/features/perf/perf-stackdump/arch-support.txt
+++ b/Documentation/features/perf/perf-stackdump/arch-support.txt
@@ -6,7 +6,7 @@
     -----------------------
     |         arch |status|
     -----------------------
-    |       alpha: | TODO |
+    |       alpha: |  ok  |
     |         arc: | TODO |
     |         arm: |  ok  |
     |       arm64: |  ok  |
diff --git a/arch/alpha/Kconfig b/arch/alpha/Kconfig
index b2c359b947f6..8fafbfabd1e5 100644
--- a/arch/alpha/Kconfig
+++ b/arch/alpha/Kconfig
@@ -41,6 +41,8 @@ config ALPHA
 	select HAVE_PAGE_SIZE_8KB
 	select HAVE_PCSPKR_PLATFORM
 	select HAVE_PERF_EVENTS
+	select HAVE_PERF_REGS
+	select HAVE_PERF_USER_STACK_DUMP
 	select NEED_DMA_MAP_STATE
 	select NEED_SG_DMA_LENGTH
 	select GENERIC_IRQ_PROBE
diff --git a/arch/alpha/include/uapi/asm/perf_regs.h b/arch/alpha/include/uapi/asm/perf_regs.h
new file mode 100644
index 000000000000..da96a31b12fb
--- /dev/null
+++ b/arch/alpha/include/uapi/asm/perf_regs.h
@@ -0,0 +1,45 @@
+/* SPDX-License-Identifier: GPL-2.0 WITH Linux-syscall-note */
+#ifndef _UAPI_ASM_ALPHA_PERF_REGS_H
+#define _UAPI_ASM_ALPHA_PERF_REGS_H
+
+/*
+ * Indices 0-30 match the hardware (and DWARF) register numbers.
+ * r9-r15 are not saved in pt_regs and cannot be sampled.
+ */
+enum perf_event_alpha_regs {
+	PERF_REG_ALPHA_R0,
+	PERF_REG_ALPHA_R1,
+	PERF_REG_ALPHA_R2,
+	PERF_REG_ALPHA_R3,
+	PERF_REG_ALPHA_R4,
+	PERF_REG_ALPHA_R5,
+	PERF_REG_ALPHA_R6,
+	PERF_REG_ALPHA_R7,
+	PERF_REG_ALPHA_R8,
+	PERF_REG_ALPHA_R9,
+	PERF_REG_ALPHA_R10,
+	PERF_REG_ALPHA_R11,
+	PERF_REG_ALPHA_R12,
+	PERF_REG_ALPHA_R13,
+	PERF_REG_ALPHA_R14,
+	PERF_REG_ALPHA_R15,
+	PERF_REG_ALPHA_R16,
+	PERF_REG_ALPHA_R17,
+	PERF_REG_ALPHA_R18,
+	PERF_REG_ALPHA_R19,
+	PERF_REG_ALPHA_R20,
+	PERF_REG_ALPHA_R21,
+	PERF_REG_ALPHA_R22,
+	PERF_REG_ALPHA_R23,
+	PERF_REG_ALPHA_R24,
+	PERF_REG_ALPHA_R25,
+	PERF_REG_ALPHA_R26,
+	PERF_REG_ALPHA_R27,
+	PERF_REG_ALPHA_R28,
+	PERF_REG_ALPHA_GP,
+	PERF_REG_ALPHA_SP,
+	PERF_REG_ALPHA_PC,
+	PERF_REG_ALPHA_MAX,
+};
+
+#endif /* _UAPI_ASM_ALPHA_PERF_REGS_H */
diff --git a/arch/alpha/kernel/Makefile b/arch/alpha/kernel/Makefile
index 4ea5c189e60e..0c04b89676ed 100644
--- a/arch/alpha/kernel/Makefile
+++ b/arch/alpha/kernel/Makefile
@@ -18,6 +18,7 @@ obj-$(CONFIG_PCI)	+= pci.o pci_iommu.o pci-sysfs.o
 obj-$(CONFIG_SRM_ENV)	+= srm_env.o
 obj-$(CONFIG_MODULES)	+= module.o
 obj-$(CONFIG_PERF_EVENTS) += perf_event.o
+obj-$(CONFIG_HAVE_PERF_REGS) += perf_regs.o
 obj-$(CONFIG_RTC_DRV_ALPHA) += rtc.o
 obj-$(CONFIG_AUDIT)	+= audit.o
 
diff --git a/arch/alpha/kernel/perf_regs.c b/arch/alpha/kernel/perf_regs.c
new file mode 100644
index 000000000000..33ea2e831ba0
--- /dev/null
+++ b/arch/alpha/kernel/perf_regs.c
@@ -0,0 +1,64 @@
+// SPDX-License-Identifier: GPL-2.0
+
+#include <linux/bits.h>
+#include <linux/perf_event.h>
+#include <linux/sched/task_stack.h>
+#include <asm/perf_regs.h>
+
+#define REG(idx, field) \
+	[PERF_REG_ALPHA_##idx] = offsetof(struct pt_regs, field)
+
+#define REG_RESERVED	(~GENMASK_ULL(PERF_REG_ALPHA_MAX - 1, 0) | \
+			 GENMASK_ULL(PERF_REG_ALPHA_R15, PERF_REG_ALPHA_R9))
+
+static const unsigned int pt_regs_offset[PERF_REG_ALPHA_MAX] = {
+	REG(R0, r0),	REG(R1, r1),	REG(R2, r2),	REG(R3, r3),
+	REG(R4, r4),	REG(R5, r5),	REG(R6, r6),	REG(R7, r7),
+	REG(R8, r8),	REG(R16, r16),	REG(R17, r17),	REG(R18, r18),
+	REG(R19, r19),	REG(R20, r20),	REG(R21, r21),	REG(R22, r22),
+	REG(R23, r23),	REG(R24, r24),	REG(R25, r25),	REG(R26, r26),
+	REG(R27, r27),	REG(R28, r28),	REG(GP, gp),	REG(PC, pc),
+};
+
+u64 perf_reg_value(struct pt_regs *regs, int idx)
+{
+	if (idx == PERF_REG_ALPHA_SP) {
+		if (user_mode(regs))
+			return user_stack_pointer(regs);
+		/*
+		 * Software events built by perf_fetch_caller_regs() use a
+		 * per-CPU pt_regs rather than a trap frame, and pt_regs has
+		 * nowhere to record sp, so report none.
+		 */
+		if (!object_is_on_stack(regs))
+			return 0;
+		/* Same kernel sp approximation as stacktrace.c. */
+		return (unsigned long)(regs + 1);
+	}
+
+	if (WARN_ON_ONCE((u32)idx >= PERF_REG_ALPHA_MAX ||
+			 BIT_ULL(idx) & REG_RESERVED))
+		return 0;
+
+	return *(unsigned long *)((void *)regs + pt_regs_offset[idx]);
+}
+
+int perf_reg_validate(u64 mask)
+{
+	if (!mask || mask & REG_RESERVED)
+		return -EINVAL;
+
+	return 0;
+}
+
+u64 perf_reg_abi(struct task_struct *task)
+{
+	return PERF_SAMPLE_REGS_ABI_64;
+}
+
+void perf_get_regs_user(struct perf_regs *regs_user,
+			struct pt_regs *regs)
+{
+	regs_user->regs = task_pt_regs(current);
+	regs_user->abi = perf_reg_abi(current);
+}

-- 
2.55.0

[PATCH v2 5/5] alpha: fill in pc for software event samples

[Matt Turner]

Context switch and tracepoint events have no trap frame, so perf builds
one with perf_fetch_caller_regs(). Alpha does not define
perf_arch_fetch_caller_regs(), which leaves that pt_regs zeroed, and
these samples report pc as 0.

Record the caller's address as pc. The zero ps already marks the frame
as kernel mode, and pt_regs has nowhere to keep sp, so pc is all that
can be filled in.

Reported-by: Magnus Lindholm <linmag7@gmail.com>
Closes: https://lore.kernel.org/all/CA%2B%3DFv5S8-Zk8-LUpZnWiBfTGSQXthepmRc_-a9O%2BpEPmVW0mRA@mail.gmail.com/
Assisted-by: Claude:claude-opus-5-5
Signed-off-by: Matt Turner <mattst88@gmail.com>
---
 arch/alpha/include/asm/perf_event.h | 7 +++++++
 1 file changed, 7 insertions(+)

diff --git a/arch/alpha/include/asm/perf_event.h b/arch/alpha/include/asm/perf_event.h
index 5996e7a6757e..72536566014a 100644
--- a/arch/alpha/include/asm/perf_event.h
+++ b/arch/alpha/include/asm/perf_event.h
@@ -1,4 +1,11 @@
 #ifndef __ASM_ALPHA_PERF_EVENT_H
 #define __ASM_ALPHA_PERF_EVENT_H
 
+/*
+ * Software events have no trap frame, so perf_fetch_caller_regs() fills
+ * in a zeroed pt_regs. A zero ps already means kernel mode; only pc is
+ * missing.
+ */
+#define perf_arch_fetch_caller_regs(regs, __ip)	((regs)->pc = (__ip))
+
 #endif /* __ASM_ALPHA_PERF_EVENT_H */

-- 
2.55.0

Re: [PATCH v2 5/5] alpha: fill in pc for software event samples

[sashiko-bot]

> Context switch and tracepoint events have no trap frame, so perf builds
> one with perf_fetch_caller_regs(). Alpha does not define
> perf_arch_fetch_caller_regs(), which leaves that pt_regs zeroed, and
> these samples report pc as 0.
> 
> Record the caller's address as pc. The zero ps already marks the frame
> as kernel mode, and pt_regs has nowhere to keep sp, so pc is all that
> can be filled in.
> 
> Reported-by: Magnus Lindholm <linmag7@gmail.com>
> Closes: https://lore.kernel.org/all/CA%2B%3DFv5S8-Zk8-LUpZnWiBfTGSQXthepmRc_-a9O%2BpEPmVW0mRA@mail.gmail.com/
> Assisted-by: Claude:claude-opus-5-5
> Signed-off-by: Matt Turner <mattst88@gmail.com>

Sashiko has reviewed this patch and found no issues.

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20261016-alpha-perf-regs-v3-v2-0-17654742226a@gmail.com?part=5

Re: [PATCH v2 3/5] alpha: use user_stack_pointer() in KSTK_ESP() and dump_elf_thread()

[sashiko-bot]

> Both open-code the same PCB lookup.
> 
> Assisted-by: Claude:claude-opus-5-5
> Tested-by: Magnus Lindholm <linmag7@gmail.com>
> Reviewed-by: Magnus Lindholm <linmag7@gmail.com>
> Acked-by: Ian Rogers <irogers@google.com>
> Signed-off-by: Matt Turner <mattst88@gmail.com>

Sashiko has reviewed this patch and found no issues.

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20261016-alpha-perf-regs-v3-v2-0-17654742226a@gmail.com?part=3

Re: [PATCH v2 2/5] alpha: stop storing the user stack pointer in pt_regs

[sashiko-bot]

> Nothing reads pt_regs->usp now that user_stack_pointer() uses the PCB.
> Drop the stores, which saves a PAL_rdusp call on every syscall, and
> give the field back its old name.
> 
> Assisted-by: Claude:claude-opus-5-5
> Reviewed-by: Magnus Lindholm <linmag7@gmail.com>
> Tested-by: Magnus Lindholm <linmag7@gmail.com>
> Acked-by: Ian Rogers <irogers@google.com>
> Signed-off-by: Matt Turner <mattst88@gmail.com>

Sashiko has reviewed this patch and found no issues.

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20261016-alpha-perf-regs-v3-v2-0-17654742226a@gmail.com?part=2

Re: [PATCH v2 1/5] alpha: fix stale user_stack_pointer()

[sashiko-bot]

> PALcode keeps the user stack pointer in the PCB, not the trap frame, so
> pt_regs->usp is only valid where entry code stores it: entSys,
> do_entArith() and do_entIF(). Interrupts, page faults and unaligned
> traps leave it stale, and ptrace and /proc/<pid>/syscall read it
> through user_stack_pointer().
> 
> A user-mode pt_regs is at the top of its task's stack, so find the
> thread_info from it and use rdusp() for current or pcb.usp otherwise,
> as KSTK_ESP() and the regset code already do.
> 
> Fixes: 8a2857bc2b30 ("alpha: enable regset-based ptrace and core dumps")
> Assisted-by: Claude:claude-opus-5-5
> Reviewed-by: Magnus Lindholm <linmag7@gmail.com>
> Tested-by: Magnus Lindholm <linmag7@gmail.com>
> Acked-by: Ian Rogers <irogers@google.com>
> Signed-off-by: Matt Turner <mattst88@gmail.com>

Sashiko has reviewed this patch and found no issues.

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20261016-alpha-perf-regs-v3-v2-0-17654742226a@gmail.com?part=1

Re: [PATCH v2 4/5] alpha: add HAVE_PERF_REGS and HAVE_PERF_USER_STACK_DUMP support

[sashiko-bot]

> Let perf sample registers and dump the user stack, which is what the
> perf tool needs for --call-graph dwarf.
> 
> The register indices follow the hardware numbering, so r0-r30 map
> directly to DWARF register numbers, with pc after them. r9-r15
> are callee-saved and live in switch_stack rather than pt_regs, so they
> cannot be sampled. Without r15 an unwinder cannot recover the CFA of a
> first frame that uses $15 as its frame pointer.
> 
> sp is not in pt_regs either: for user frames it comes from
> user_stack_pointer(), and for kernel frames (PERF_SAMPLE_REGS_INTR) it
> is the end of the trap frame, as in stacktrace.c.
> 
> Software events such as context switches and tracepoints hand perf a
> per-CPU pt_regs filled in by perf_fetch_caller_regs() instead of a
> trap frame. It is not on the stack and has no room for sp, so report
> sp as 0 there rather than the address of the per-CPU buffer.
> 
> Assisted-by: Claude:claude-opus-5-5
> Acked-by: Ian Rogers <irogers@google.com>
> Signed-off-by: Matt Turner <mattst88@gmail.com>

Sashiko has reviewed this patch and found no issues.

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20261016-alpha-perf-regs-v3-v2-0-17654742226a@gmail.com?part=4