Live Patching
 help / color / mirror / Atom feed
From: sashiko-bot@kernel.org
To: "Dylan Hatch" <dylanbhatch@google.com>,
	"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: "Prasanna Kumar T S M" <ptsm@linux.microsoft.com>,
	"Puranjay Mohan" <puranjay@kernel.org>,
	"Song Liu" <song@kernel.org>,
	joe.lawrence@redhat.com, live-patching@vger.kernel.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: Re: [PATCH v9 11/11] unwind: arm64: Use sframe to unwind interrupt frames
Date: Tue, 17 Nov 2026 03:19:21 +0000	[thread overview]
Message-ID: <sashiko-outbox-218134@kernel.org> (raw)
In-Reply-To: <20261117020648.323464-12-dylanbhatch@google.com>

Thank you for your contribution! Sashiko AI review found 2 potential
issues to consider:
- [Medium] arm64: stacktrace: kunwind_next() clears state->regs after
  fallback
- [Low] unwind: arm64: consume FP stack slot in
  kunwind_next_regs_sframe()
--

commit e44539400678f653df77eaa983169650892957a4
Author: Dylan Hatch <dylanbhatch@google.com>

unwind: arm64: Use sframe to unwind interrupt frames

This commit adds kunwind_next_regs_sframe() to unwind across exception
boundaries using SFrame information, falling back to frame-pointer
unwinding and marking the stacktrace unreliable on failure.

> diff --git a/arch/arm64/kernel/stacktrace.c b/arch/arm64/kernel/stacktrace.c
> index 3ebcf8c53fb04..e876af54a4055 100644
> --- a/arch/arm64/kernel/stacktrace.c
> +++ b/arch/arm64/kernel/stacktrace.c

[ ... ]

> @@ -244,6 +266,163 @@ kunwind_next_frame_record(struct kunwind_state *state)

[ ... ]

> +	/*
> +	 * Consume RA and FP from the stack. The frame record puts FP at a lower
> +	 * address than RA, so we read FP first.
> +	 */
> +	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;
> +		if (!get_word(&state->common, &fp))
> +			return -EINVAL;
> +		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;
> +	}
> +
> +	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;
> +		if (get_consume_word(&state->common, &ra))
> +			return -EINVAL;
> +		break;

[Severity: Low]
Does UNWIND_RULE_CFA_OFFSET_DEREF for frame.fp.rule miss calling
unwind_consume_stack() when frame.ra.rule is UNWIND_RULE_RETAIN (or
frame.fp.offset >= frame.ra.offset)?

get_word() reads the saved FP from the stack without consuming it, relying on
get_consume_word() for frame.ra to advance state->common.stack.low.  If RA is
retained in LR, the FP slot is never consumed, leaving state->common.stack
unadvanced (or failing to transition stacks if CFA + frame.fp.offset lies on
the previous stack) while state->unreliable stays false.  Note that the stack
should only be consumed after both frame.fp and frame.ra checks succeed so
state->common.stack remains intact if kunwind_next_regs_sframe() fails and
falls back to kunwind_next_frame_record().

[ ... ]

> @@ -259,10 +438,23 @@ kunwind_next(struct kunwind_state *state)
>  	state->flags.all = 0;
>  
>  	switch (state->source) {
> +	case KUNWIND_SOURCE_REGS_PC:
> +		if (state->regs)
> +			err = kunwind_next_regs_sframe(state);
> +		else
> +			err = -EINVAL;
> +
> +		if (err) {
> +			/* Fallback to FP based unwinder */
> +			err = kunwind_next_frame_record(state);
> +			state->unreliable = true;
> +		}
> +		state->regs = NULL;
> +		break;

[Severity: Medium]
Should state->regs be cleared before calling the fallback
kunwind_next_frame_record(state)?

When kunwind_next() falls back to kunwind_next_frame_record() (for example on
a synthetic pt_regs or an exception in entry.S without SFrame data) and
state->common.fp points directly to an on-stack FRAME_META_TYPE_PT_REGS
record, kunwind_next_frame_record() calls kunwind_next_regs_pc(), which sets
state->regs = regs and state->source = KUNWIND_SOURCE_REGS_PC for that next
exception frame.  Executing state->regs = NULL afterwards overwrites that new
pointer, preventing SFrame unwinding on the next exception boundary.

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20261117020648.323464-1-dylanbhatch@google.com?part=11

      reply	other threads:[~2026-11-17  3:19 UTC|newest]

Thread overview: 14+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2026-11-17  2:06 [PATCH v9 00/11] unwind, arm64: add sframe unwinder for kernel Dylan Hatch
2026-11-17  2:06 ` [PATCH v9 01/11] arm64, unwind: build kernel with sframe V3 info Dylan Hatch
2026-11-17  2:06 ` [PATCH v9 02/11] arm64/sframe: Read vmlinux .sframe header Dylan Hatch
2026-11-17  2:06 ` [PATCH v9 03/11] sframe: Add support for reading vmlinux .sframe contents Dylan Hatch
2026-11-17  2:06 ` [PATCH v9 04/11] sframe: Separate reading of FRE from reading of FRE data words Dylan Hatch
2026-11-17  2:06 ` [PATCH v9 05/11] module, sframe: Add sframe support for modules Dylan Hatch
2026-11-17  2:34   ` sashiko-bot
2026-11-17  2:06 ` [PATCH v9 06/11] arm64/sframe: Validate IP addresses Dylan Hatch
2026-11-17  2:06 ` [PATCH v9 07/11] sframe: Add debug helpers with object name Dylan Hatch
2026-11-17  2:06 ` [PATCH v9 08/11] sframe: Add .sframe validation option Dylan Hatch
2026-11-17  2:06 ` [PATCH v9 09/11] arm64: entry: add unwind info for call_on_irq_stack() Dylan Hatch
2026-11-17  2:06 ` [PATCH v9 10/11] arm64, crypto/lib: Annotate leaf functions with CFI info Dylan Hatch
2026-11-17  2:06 ` [PATCH v9 11/11] unwind: arm64: Use sframe to unwind interrupt frames Dylan Hatch
2026-11-17  3:19   ` sashiko-bot [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=sashiko-outbox-218134@kernel.org \
    --to=sashiko-bot@kernel.org \
    --cc=catalin.marinas@arm.com \
    --cc=davem@davemloft.net \
    --cc=dylanbhatch@google.com \
    --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=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=sashiko-reviews@lists.linux.dev \
    --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