Live Patching
 help / color / mirror / Atom feed
From: Dylan Hatch <dylanbhatch@google.com>
To: Roman Gushchin <roman.gushchin@linux.dev>,
	Weinan Liu <wnliu@google.com>,  Will Deacon <will@kernel.org>,
	Josh Poimboeuf <jpoimboe@kernel.org>,
	 Indu Bhagat <ibhagatgnu@gmail.com>,
	Peter Zijlstra <peterz@infradead.org>,
	 Steven Rostedt <rostedt@goodmis.org>,
	Catalin Marinas <catalin.marinas@arm.com>,
	 Jiri Kosina <jikos@kernel.org>,
	Mark Rutland <mark.rutland@arm.com>,
	 Jens Remus <jremus@linux.ibm.com>
Cc: Dylan Hatch <dylanbhatch@google.com>,
	Prasanna Kumar T S M <ptsm@linux.microsoft.com>,
	 Puranjay Mohan <puranjay@kernel.org>, Song Liu <song@kernel.org>,
	joe.lawrence@redhat.com,  linux-toolchains@vger.kernel.org,
	linux-kernel@vger.kernel.org,  live-patching@vger.kernel.org,
	linux-arm-kernel@lists.infradead.org,
	 Randy Dunlap <rdunlap@infradead.org>,
	Mostafa Saleh <smostafa@google.com>,
	 Herbert Xu <herbert@gondor.apana.org.au>,
	"David S. Miller" <davem@davemloft.net>
Subject: [PATCH v6 9/9] unwind: arm64: Use sframe to unwind interrupt frames
Date: Tue, 19 May 2026 06:49:50 +0000	[thread overview]
Message-ID: <20260519064950.493949-10-dylanbhatch@google.com> (raw)
In-Reply-To: <20260519064950.493949-1-dylanbhatch@google.com>

Add unwind_next_frame_sframe() function to unwind by sframe info if
present. Use this method at exception boundaries, falling back to
frame-pointer unwind only on failure. In such failure cases, the
stacktrace is considered unreliable.

During normal unwind, prefer frame pointer unwind (for better
performance) with sframe as a backup.

This change restores the LR behavior originally introduced in commit
c2c6b27b5aa14fa2 ("arm64: stacktrace: unwind exception boundaries"),
But later removed in commit 32ed1205682e ("arm64: stacktrace: Skip
reporting LR at exception boundaries")

This can be done because the sframe data can be used to determine
whether the LR is current for the PC value recovered from pt_regs at the
exception boundary.

Signed-off-by: Weinan Liu <wnliu@google.com>
Reviewed-by: Prasanna Kumar T S M <ptsm@linux.microsoft.com>
Reviewed-by: Jens Remus <jremus@linux.ibm.com>
Signed-off-by: Dylan Hatch <dylanbhatch@google.com>
---
 arch/arm64/kernel/stacktrace.c | 222 ++++++++++++++++++++++++++++++---
 1 file changed, 202 insertions(+), 20 deletions(-)

diff --git a/arch/arm64/kernel/stacktrace.c b/arch/arm64/kernel/stacktrace.c
index 3ebcf8c53fb0..cee860ca8ce5 100644
--- a/arch/arm64/kernel/stacktrace.c
+++ b/arch/arm64/kernel/stacktrace.c
@@ -13,6 +13,7 @@
 #include <linux/sched.h>
 #include <linux/sched/debug.h>
 #include <linux/sched/task_stack.h>
+#include <linux/sframe.h>
 #include <linux/stacktrace.h>
 
 #include <asm/efi.h>
@@ -26,6 +27,7 @@ enum kunwind_source {
 	KUNWIND_SOURCE_CALLER,
 	KUNWIND_SOURCE_TASK,
 	KUNWIND_SOURCE_REGS_PC,
+	KUNWIND_SOURCE_REGS_LR,
 };
 
 union unwind_flags {
@@ -45,6 +47,7 @@ union unwind_flags {
  * @kr_cur:      When KRETPROBES is selected, holds the kretprobe instance
  *               associated with the most recently encountered replacement lr
  *               value.
+ * @unreliable:  Stacktrace is unreliable.
  */
 struct kunwind_state {
 	struct unwind_state common;
@@ -56,6 +59,7 @@ struct kunwind_state {
 	enum kunwind_source source;
 	union unwind_flags flags;
 	struct pt_regs *regs;
+	bool unreliable;
 };
 
 static __always_inline void
@@ -181,7 +185,6 @@ int kunwind_next_regs_pc(struct kunwind_state *state)
 	state->regs = regs;
 	state->common.pc = regs->pc;
 	state->common.fp = regs->regs[29];
-	state->regs = NULL;
 	state->source = KUNWIND_SOURCE_REGS_PC;
 	return 0;
 }
@@ -244,6 +247,168 @@ kunwind_next_frame_record(struct kunwind_state *state)
 	return 0;
 }
 
+#ifdef CONFIG_HAVE_UNWIND_KERNEL_SFRAME
+
+static __always_inline struct stack_info *
+get_word(struct unwind_state *state, unsigned long *word)
+{
+	unsigned long addr = *word;
+	struct stack_info *info;
+
+	info = unwind_find_stack(state, addr, sizeof(addr));
+	if (!info)
+		return info;
+
+	*word = READ_ONCE(*(unsigned long *)addr);
+
+	return info;
+}
+
+static __always_inline int
+get_consume_word(struct unwind_state *state, unsigned long *word)
+{
+	struct stack_info *info;
+	unsigned long addr = *word;
+
+	info = get_word(state, word);
+	if (!info)
+		return -EINVAL;
+
+	unwind_consume_stack(state, info, addr, sizeof(addr));
+	return 0;
+}
+
+/*
+ * Unwind from a pt_regs according to sframe.
+ */
+static __always_inline int
+kunwind_next_regs_sframe(struct kunwind_state *state)
+{
+	struct unwind_frame frame;
+	unsigned long cfa, fp, ra;
+	enum kunwind_source source = KUNWIND_SOURCE_FRAME;
+	struct pt_regs *regs = state->regs;
+
+	int err;
+
+	if (WARN_ON_ONCE(state->source != KUNWIND_SOURCE_REGS_PC))
+		return -EINVAL;
+	if (WARN_ON_ONCE(!state->regs))
+		return -EINVAL;
+
+	/* FP/SP alignment 8 bytes */
+	if (state->common.fp & 0x7)
+		return -EINVAL;
+
+	err = sframe_find_kernel(state->common.pc, &frame);
+	if (err)
+		return -EINVAL;
+
+	/*
+	 * A kernel unwind should always end at a FRAME_META_TYPE_FINAL
+	 * frame. There should be no outermost frames within the kernel.
+	 */
+	if (frame.outermost)
+		return -EINVAL;
+
+	/* Get the Canonical Frame Address (CFA) */
+	switch (frame.cfa.rule) {
+	case UNWIND_CFA_RULE_SP_OFFSET:
+		cfa = state->regs->sp;
+		break;
+	case UNWIND_CFA_RULE_FP_OFFSET:
+		if (state->common.fp < state->regs->sp)
+			return -EINVAL;
+		cfa = state->common.fp;
+		break;
+	/*
+	 * UNWIND_CFA_RULE_REG_OFFSET and UNWIND_CFA_RULE_REG_OFFSET_DEREF not
+	 * implemented -- flexible FDEs are not currently generated by assembler
+	 * for arm64.
+	 */
+	default:
+		WARN_ON_ONCE(1);
+		return -EINVAL;
+	}
+	cfa += frame.cfa.offset;
+
+	/* CFA alignment 16 bytes */
+	if (cfa & 0x15)
+		return -EINVAL;
+
+	/* Get the Return Address (RA) */
+	switch (frame.ra.rule) {
+	case UNWIND_RULE_RETAIN:
+		ra = regs->regs[30];
+		source = KUNWIND_SOURCE_REGS_LR;
+		break;
+
+	/*
+	 * UNWIND_RULE_CFA_OFFSET doesn't make sense for RA.
+	 * The return address cannot legitimately be a stack address.
+	 */
+	case UNWIND_RULE_CFA_OFFSET_DEREF:
+		ra = cfa + frame.ra.offset;
+		break;
+	/*
+	 * UNWIND_RULE_REG_OFFSET and UNWIND_RULE_REG_OFFSET_DEREF not
+	 * implemented -- flexible FDEs are not currently generated by assembler
+	 * for arm64.
+	 */
+	default:
+		WARN_ON_ONCE(1);
+		return -EINVAL;
+	}
+
+	/* Get the Frame Pointer (FP) */
+	switch (frame.fp.rule) {
+	case UNWIND_RULE_RETAIN:
+		fp = state->common.fp;
+		break;
+	/*
+	 * UNWIND_RULE_CFA_OFFSET is currently not used for FP
+	 * (e.g. SFrame cannot represent this rule).
+	 */
+	case UNWIND_RULE_CFA_OFFSET_DEREF:
+		fp = cfa + frame.fp.offset;
+		break;
+	/*
+	 * UNWIND_RULE_REG_OFFSET and UNWIND_RULE_REG_OFFSET_DEREF not
+	 * implemented -- flexible FDEs are not currently generated by assembler
+	 * for arm64.
+	 */
+	default:
+		WARN_ON_ONCE(1);
+		return -EINVAL;
+	}
+
+	/*
+	 * Consume RA and FP from the stack. The frame record puts FP at a lower
+	 * address than RA, so we always read FP first.
+	 */
+	if (frame.fp.rule & UNWIND_RULE_DEREF &&
+	    !get_word(&state->common, &fp))
+		return -EINVAL;
+
+	if (frame.ra.rule & UNWIND_RULE_DEREF &&
+	    get_consume_word(&state->common, &ra))
+		return -EINVAL;
+
+	state->common.pc = ra;
+	state->common.fp = fp;
+
+	state->source = source;
+
+	return 0;
+}
+
+#else /* !CONFIG_HAVE_UNWIND_KERNEL_SFRAME */
+
+static __always_inline int
+unwind_next_frame_sframe(struct kunwind_state *state) { return -EINVAL; }
+
+#endif /* !CONFIG_HAVE_UNWIND_KERNEL_SFRAME*/
+
 /*
  * Unwind from one frame record (A) to the next frame record (B).
  *
@@ -259,10 +424,20 @@ kunwind_next(struct kunwind_state *state)
 	state->flags.all = 0;
 
 	switch (state->source) {
+	case KUNWIND_SOURCE_REGS_PC:
+		err = kunwind_next_regs_sframe(state);
+
+		if (err && err != -ENOENT) {
+			/* Fallback to FP based unwinder */
+			err = kunwind_next_frame_record(state);
+			state->unreliable = true;
+		}
+		state->regs = NULL;
+		break;
 	case KUNWIND_SOURCE_FRAME:
 	case KUNWIND_SOURCE_CALLER:
 	case KUNWIND_SOURCE_TASK:
-	case KUNWIND_SOURCE_REGS_PC:
+	case KUNWIND_SOURCE_REGS_LR:
 		err = kunwind_next_frame_record(state);
 		break;
 	default:
@@ -390,34 +565,40 @@ noinline noinstr void arch_stack_walk(stack_trace_consume_fn consume_entry,
 	kunwind_stack_walk(arch_kunwind_consume_entry, &data, task, regs);
 }
 
+struct kunwind_reliable_consume_entry_data {
+	stack_trace_consume_fn consume_entry;
+	void *cookie;
+	bool unreliable;
+};
+
 static __always_inline bool
-arch_reliable_kunwind_consume_entry(const struct kunwind_state *state, void *cookie)
+arch_kunwind_reliable_consume_entry(const struct kunwind_state *state, void *cookie)
 {
-	/*
-	 * At an exception boundary we can reliably consume the saved PC. We do
-	 * not know whether the LR was live when the exception was taken, and
-	 * so we cannot perform the next unwind step reliably.
-	 *
-	 * All that matters is whether the *entire* unwind is reliable, so give
-	 * up as soon as we hit an exception boundary.
-	 */
-	if (state->source == KUNWIND_SOURCE_REGS_PC)
-		return false;
+	struct kunwind_reliable_consume_entry_data *data = cookie;
 
-	return arch_kunwind_consume_entry(state, cookie);
+	if (state->unreliable) {
+		data->unreliable = true;
+		return false;
+	}
+	return data->consume_entry(data->cookie, state->common.pc);
 }
 
-noinline noinstr int arch_stack_walk_reliable(stack_trace_consume_fn consume_entry,
-					      void *cookie,
-					      struct task_struct *task)
+noinline notrace int arch_stack_walk_reliable(
+				stack_trace_consume_fn consume_entry,
+				void *cookie, struct task_struct *task)
 {
-	struct kunwind_consume_entry_data data = {
+	struct kunwind_reliable_consume_entry_data data = {
 		.consume_entry = consume_entry,
 		.cookie = cookie,
+		.unreliable = false,
 	};
 
-	return kunwind_stack_walk(arch_reliable_kunwind_consume_entry, &data,
-				  task, NULL);
+	kunwind_stack_walk(arch_kunwind_reliable_consume_entry, &data, task, NULL);
+
+	if (data.unreliable)
+		return -EINVAL;
+
+	return 0;
 }
 
 struct bpf_unwind_consume_entry_data {
@@ -452,6 +633,7 @@ static const char *state_source_string(const struct kunwind_state *state)
 	case KUNWIND_SOURCE_CALLER:	return "C";
 	case KUNWIND_SOURCE_TASK:	return "T";
 	case KUNWIND_SOURCE_REGS_PC:	return "P";
+	case KUNWIND_SOURCE_REGS_LR:	return "L";
 	default:			return "U";
 	}
 }
-- 
2.54.0.563.g4f69b47b94-goog


      parent reply	other threads:[~2026-05-19  6:50 UTC|newest]

Thread overview: 11+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2026-05-19  6:49 [PATCH v6 0/9] unwind, arm64: add sframe unwinder for kernel Dylan Hatch
2026-05-19  6:49 ` [PATCH v6 1/9] sframe: Allow kernelspace sframe sections Dylan Hatch
2026-05-19  6:49 ` [PATCH v6 2/9] arm64, unwind: build kernel with sframe V3 info Dylan Hatch
2026-05-19  6:49 ` [PATCH v6 3/9] arm64: entry: add unwind info for call_on_irq_stack() Dylan Hatch
2026-05-19  6:49 ` [PATCH v6 4/9] arm64, crypto/lib: Annotate leaf functions with CFI info Dylan Hatch
2026-05-19  6:49 ` [PATCH v6 5/9] sframe: Provide PC lookup for vmlinux .sframe section Dylan Hatch
2026-05-19  6:49 ` [PATCH v6 6/9] arm64/module, sframe: Add sframe support for modules Dylan Hatch
2026-05-19  6:49 ` [PATCH v6 7/9] sframe: Introduce in-kernel SFRAME_VALIDATION Dylan Hatch
2026-05-19  6:49 ` [PATCH v6 8/9] sframe: Initialize debug info for kernel sections Dylan Hatch
2026-05-19 11:20   ` Jens Remus
2026-05-19  6:49 ` Dylan Hatch [this message]

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20260519064950.493949-10-dylanbhatch@google.com \
    --to=dylanbhatch@google.com \
    --cc=catalin.marinas@arm.com \
    --cc=davem@davemloft.net \
    --cc=herbert@gondor.apana.org.au \
    --cc=ibhagatgnu@gmail.com \
    --cc=jikos@kernel.org \
    --cc=joe.lawrence@redhat.com \
    --cc=jpoimboe@kernel.org \
    --cc=jremus@linux.ibm.com \
    --cc=linux-arm-kernel@lists.infradead.org \
    --cc=linux-kernel@vger.kernel.org \
    --cc=linux-toolchains@vger.kernel.org \
    --cc=live-patching@vger.kernel.org \
    --cc=mark.rutland@arm.com \
    --cc=peterz@infradead.org \
    --cc=ptsm@linux.microsoft.com \
    --cc=puranjay@kernel.org \
    --cc=rdunlap@infradead.org \
    --cc=roman.gushchin@linux.dev \
    --cc=rostedt@goodmis.org \
    --cc=smostafa@google.com \
    --cc=song@kernel.org \
    --cc=will@kernel.org \
    --cc=wnliu@google.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox