Linux Perf Users
 help / color / mirror / Atom feed
From: sashiko-bot@kernel.org
To: "Matt Turner" <mattst88@gmail.com>
Cc: linux-perf-users@vger.kernel.org
Subject: Re: [PATCH v3 5/7] alpha/perf: implement ProfileMe sampling
Date: Sun, 01 Nov 2026 20:12:26 +0000	[thread overview]
Message-ID: <sashiko-outbox-192656@kernel.org> (raw)
In-Reply-To: <20261101-alpha-profileme-v3-5-92d511b5a044@gmail.com>

Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:
- [Medium] alpha/perf: perf_get_regs_user() checks PMPC instead of interrupted PC
--

commit b49acea08699ec26b74c2c7c29ccc098435ea611
Author: Matt Turner <mattst88@gmail.com>

alpha/perf: implement ProfileMe sampling

This commit implements ProfileMe sampling mode on Alpha EV67 and later
processors, selectable via the profileme config attribute or precise_ip. It
attributes samples to the instruction address latched in PMPC, adjusts for
I_STAT instruction overcount, and works around PALcode counter overflow
restrictions when stopping or rearming counters.

> diff --git a/arch/alpha/kernel/perf_event.c b/arch/alpha/kernel/perf_event.c
> index ecec78c355cd3..f55ddab7478e0 100644
> --- a/arch/alpha/kernel/perf_event.c
> +++ b/arch/alpha/kernel/perf_event.c
[ ... ]
> @@ -835,6 +939,131 @@ void perf_event_print_debug(void)
[ ... ]
> +static void alpha_perf_event_profileme(struct perf_event *event,
> +				       struct hw_perf_event *hwc,
> +				       struct pt_regs *regs)
> +{
[ ... ]
> +	pm_regs = regs;
> +	pc = pmpc & EV67_PMPC_PC_MASK;
> +	if (pmpc & EV67_PMPC_PAL) {
> +		pc = regs->pc;
> +	} else if (pc < TASK_SIZE && !user_mode(regs)) {
> +		if (is_user_task(current)) {
> +			pc_only = true;
> +			pm_regs = this_cpu_ptr(&alpha_profileme_regs);
> +			memset(pm_regs, 0, sizeof(*pm_regs));
> +			/* PS bit 3 is the mode bit that user_mode() tests. */
> +			pm_regs->ps = 8;
> +		} else {
> +			pc = regs->pc;
> +		}
> +	} else if (pc >= TASK_SIZE && user_mode(regs)) {
> +		pm_regs = this_cpu_ptr(&alpha_profileme_regs);
> +		memset(pm_regs, 0, sizeof(*pm_regs));
> +	}
> +
> +	perf_sample_data_init(&data, 0, hwc->last_period);
> +	if (pc_only)
> +		alpha_profileme_drop_regs(event, &data, PERF_SAMPLE_REGS_USER |
> +							PERF_SAMPLE_REGS_INTR);
> +	/*
> +	 * perf_get_regs_user() has the same check, but it will find the PMPC
> +	 * in regs->pc rather than where the interrupt was taken.
> +	 */
> +	else if (!user_mode(regs) && in_entry_save_code(regs->pc))
> +		alpha_profileme_drop_regs(event, &data, PERF_SAMPLE_REGS_USER);
> +
> +	if (!alpha_perf_event_set_period(event, hwc, 0))
> +		return;
> +
> +	saved_pc = pm_regs->pc;
> +	saved_ps = pm_regs->ps;
> +	pm_regs->pc = pc;
> +	/* PS bit 3 is the mode bit that user_mode() tests. */
> +	if (pc >= TASK_SIZE)
> +		pm_regs->ps &= ~8UL;
> +
> +	perf_event_overflow(event, &data, pm_regs);

[Severity: Medium]
When a ProfileMe sample has a kernel PMPC inside the entry save range
(in_entry_save_code(pc) is true), but the performance counter interrupt itself
is taken after leaving that range—either later in kernel mode after SAVE_ALL
has populated task_pt_regs(current), or after returning to user mode where
user_mode(regs) is true—alpha_perf_event_profileme() does not call
alpha_profileme_drop_regs() because in_entry_save_code(regs->pc) is false.

It then sets pm_regs->pc = pc and clears the user mode bit in pm_regs->ps (or
uses the zeroed per-CPU alpha_profileme_regs when user_mode(regs) is true)
before calling perf_event_overflow().

During sample preparation, perf_event_overflow() -> __perf_event_overflow() ->
perf_prepare_sample() -> perf_sample_regs_user() sees !user_mode(pm_regs) and
calls perf_get_regs_user():

arch/alpha/kernel/perf_regs.c:perf_get_regs_user() {
	if (in_entry_save_code(regs->pc)) {
		regs_user->regs = NULL;
		regs_user->abi = PERF_SAMPLE_REGS_ABI_NONE;
		return;
	}

	regs_user->regs = task_pt_regs(current);
	regs_user->abi = perf_reg_abi(current);
}

Because perf_get_regs_user() checks in_entry_save_code(pm_regs->pc), which is
now the PMPC rather than the interrupted PC regs->pc, won't it falsely discard
valid PERF_SAMPLE_REGS_USER and PERF_SAMPLE_STACK_USER data whenever the
profiled instruction is in __entry_save_start..__entry_save_end even though
the interrupt was delivered after the user frame was completely saved (or
after returning to user mode)?

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20261101-alpha-profileme-v3-0-92d511b5a044@gmail.com?part=5

  reply	other threads:[~2026-11-01 20:12 UTC|newest]

Thread overview: 15+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2026-11-01 19:28 [PATCH v3 0/7] alpha/perf: add ProfileMe sampling Matt Turner
2026-11-01 19:28 ` [PATCH v3 1/7] alpha/perf: pass PERFMON_CMD_LOGGING_ALL to the logging options call Matt Turner
2026-11-01 19:31   ` sashiko-bot
2026-11-01 19:28 ` [PATCH v3 2/7] alpha/perf: add sysfs format and event attributes Matt Turner
2026-11-01 19:33   ` sashiko-bot
2026-11-01 19:28 ` [PATCH v3 3/7] alpha/perf: clear PERF_HES_UPTODATE when an event starts counting Matt Turner
2026-11-01 19:39   ` sashiko-bot
2026-11-01 19:28 ` [PATCH v3 4/7] alpha/perf: do not report user registers from the entry code Matt Turner
2026-11-01 19:48   ` sashiko-bot
2026-11-01 19:28 ` [PATCH v3 5/7] alpha/perf: implement ProfileMe sampling Matt Turner
2026-11-01 20:12   ` sashiko-bot [this message]
2026-11-01 19:28 ` [PATCH v3 6/7] alpha/perf: attach ProfileMe status as PERF_SAMPLE_RAW data Matt Turner
2026-11-01 20:17   ` sashiko-bot
2026-11-01 19:28 ` [PATCH v3 7/7] alpha/perf: optionally randomize the ProfileMe sampling interval Matt Turner
2026-11-01 20:25   ` sashiko-bot

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-192656@kernel.org \
    --to=sashiko-bot@kernel.org \
    --cc=linux-perf-users@vger.kernel.org \
    --cc=mattst88@gmail.com \
    --cc=sashiko-reviews@lists.linux.dev \
    /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