From: Xi Pardee <xi.pardee@linux.intel.com>
To: "Tzung-Bi Shih" <tzungbi@kernel.org>,
"Rajneesh Bhardwaj" <irenic.rajneesh@gmail.com>,
"Hans de Goede" <hansg@kernel.org>,
"Ilpo Järvinen" <ilpo.jarvinen@linux.intel.com>
Cc: platform-driver-x86@vger.kernel.org,
linux-kernel@vger.kernel.org, tfiga@chromium.org
Subject: Re: [PATCH v3] platform/x86/intel/pmc: Disable ACPI PM timer during suspend on Panther Lake
Date: Mon, 16 Nov 2026 21:17:13 -0800 [thread overview]
Message-ID: <766daa32-2247-40c2-980f-2a2317733b96@linux.intel.com> (raw)
In-Reply-To: <20261112032040.930597-1-tzungbi@kernel.org>
Hi Tzung-Bi,
Thanks for the v3. Avoiding pmcdev->lock and the tick_freeze() atomic
context resolves the problem behind the revert in 5fa607880168, and the
save and restore logic looks correct.
Two requests:
1. Ordering vs. the acpi_pm clocksource.
• The reverted code toggled the timer from the clocksource's own
callbacks. These run inside timekeeping_suspend() after the last read
and inside timekeeping_resume() before the first read, so the counter
is never read while stopped.
• Here the timer is stopped in suspend_late and restarted in
resume_early . That is before timekeeping_forward_now() on suspend and
after tk_clock_read() on resume.
• If acpi_pm is the active clocksource, for example when TSC is
unstable, timekeeping can read a stopped counter, and ktime_get() stops
advancing from suspend_late on.
• Have you tested with clocksource=acpi_pm ? If TSC is always used on
PTL, please say so in the commit message.
2. Evidence for S0i2.2.
• Please share S0i2.1 and S0i2.2 residency with and without the patch.
• Include the platform, firmware and kernel version, and say whether
firmware enabled the timer and wdat_wdt was active.
Also, does WCL, which reuses the PTL ops, need the same fix?
Thanks,
Xi
On 11/11/2026 7:20 PM, Tzung-Bi Shih wrote:
> The ACPI PM timer is used to drive the ACPI Watchdog Action Table
> (wdat_wdt) hardware watchdog. When the ACPI PM timer is enabled by
> firmware on Panther Lake, leaving it running during s2idle prevents the
> SoC from entering its deepest S0i2.2 substate (staying in S0i2.1).
>
> Since the watchdog is stopped during system suspend, the ACPI PM timer
> does not need to remain active while suspended. Disable the timer via
> ACPI_TMR_CTL (PWRMBASE + 0x18fc) in ptl_suspend() and re-enable it in
> ptl_resume() only if it was active prior to suspend.
>
> Unlike commit e86c8186d03a ("platform/x86:intel/pmc: Enable the ACPI PM
> Timer to be turned off when suspended"), which was reverted in commit
> 5fa607880168 because it acquired pmcdev->lock from the acpi_pm
> clocksource suspend callback inside tick_freeze() atomic context, toggle
> the register in ptl_suspend() and ptl_resume() without taking
> pmcdev->lock.
>
> Signed-off-by: Tzung-Bi Shih <tzungbi@kernel.org>
> ---
> v3:
> - Rebase to next-20261111.
>
> v2: https://lore.kernel.org/all/20261016075416.3139759-1-tzungbi@kernel.org
> - Include <linux/bits.h> for BIT().
> - Rename ptl_set_acpi_pm_timer_on_suspend() -> ptl_set_acpi_pm_timer().
> - Update register values on separate lines.
>
> v1: https://lore.kernel.org/all/20261015035712.3604634-1-tzungbi@kernel.org
> ---
> drivers/platform/x86/intel/pmc/core.h | 2 ++
> drivers/platform/x86/intel/pmc/ptl.c | 40 ++++++++++++++++++++++++++-
> 2 files changed, 41 insertions(+), 1 deletion(-)
>
> diff --git a/drivers/platform/x86/intel/pmc/core.h b/drivers/platform/x86/intel/pmc/core.h
> index e55164e5807f..8aae69775ab3 100644
> --- a/drivers/platform/x86/intel/pmc/core.h
> +++ b/drivers/platform/x86/intel/pmc/core.h
> @@ -487,6 +487,7 @@ struct pmc {
> * @s0ix_counter: S0ix residency (step adjusted)
> * @suspend: Function to perform platform specific suspend
> * @resume: Function to perform platform specific resume
> + * @pm_timer_enabled: Whether the ACPI PM timer was enabled before suspend
> *
> * @pkgc_ltr_blocker_counters: Array of PKGC LTR blocker counters
> * @pkgc_ltr_blocker_offset: Offset to PKGC LTR blockers in telemetry region
> @@ -506,6 +507,7 @@ struct pmc_dev {
> u64 s0ix_counter;
> void (*suspend)(struct pmc_dev *pmcdev);
> int (*resume)(struct pmc_dev *pmcdev);
> + bool pm_timer_enabled;
>
> u64 *pkgc_res_cnt;
> u8 num_of_pkgc;
> diff --git a/drivers/platform/x86/intel/pmc/ptl.c b/drivers/platform/x86/intel/pmc/ptl.c
> index a7d5d4f77934..3cb063944007 100644
> --- a/drivers/platform/x86/intel/pmc/ptl.c
> +++ b/drivers/platform/x86/intel/pmc/ptl.c
> @@ -6,6 +6,8 @@
> * Copyright (c) 2025, Intel Corporation.
> */
>
> +#include <linux/bits.h>
> +#include <linux/io.h>
> #include <linux/pci.h>
>
> #include "core.h"
> @@ -548,6 +550,9 @@ static const u8 ptl_pmc_list[] = {PMC_IDX_MAIN};
> #define PTL_NPU_PCI_DEV 0xb03e
> #define PTL_IPU_PCI_DEV 0xb05d
>
> +#define ACPI_TMR_CTL_OFFSET 0x18fc
> +#define ACPI_TIM_DIS BIT(1)
> +
> /*
> * Set power state of select devices that do not have drivers to D3
> * so that they do not block Package C entry.
> @@ -558,9 +563,42 @@ static void ptl_d3_fixup(void)
> pmc_core_set_device_d3(PTL_NPU_PCI_DEV);
> }
>
> +static void ptl_set_acpi_pm_timer(struct pmc_dev *pmcdev, bool suspend)
> +{
> + struct pmc *pmc = pmcdev->pmcs[PMC_IDX_MAIN];
> + u8 val, old_val;
> +
> + if (!pmc || !pmc->regbase)
> + return;
> +
> + if (!suspend && !pmcdev->pm_timer_enabled)
> + return;
> +
> + val = readb(pmc->regbase + ACPI_TMR_CTL_OFFSET);
> + old_val = val;
> +
> + if (suspend) {
> + pmcdev->pm_timer_enabled = !(old_val & ACPI_TIM_DIS);
> + if (pmcdev->pm_timer_enabled)
> + val |= ACPI_TIM_DIS;
> + } else {
> + val &= ~ACPI_TIM_DIS;
> + }
> +
> + if (val != old_val)
> + writeb(val, pmc->regbase + ACPI_TMR_CTL_OFFSET);
> +}
> +
> +static void ptl_suspend(struct pmc_dev *pmcdev)
> +{
> + ptl_set_acpi_pm_timer(pmcdev, true);
> + cnl_suspend(pmcdev);
> +}
> +
> static int ptl_resume(struct pmc_dev *pmcdev)
> {
> ptl_d3_fixup();
> + ptl_set_acpi_pm_timer(pmcdev, false);
> return cnl_resume(pmcdev);
> }
>
> @@ -576,7 +614,7 @@ struct pmc_dev_info ptl_pmc_dev = {
> .regmap_list = ptl_pmc_info_list,
> .map = &ptl_pcdp_reg_map,
> .sub_req_show = &pmc_core_substate_blk_req_fops,
> - .suspend = cnl_suspend,
> + .suspend = ptl_suspend,
> .resume = ptl_resume,
> .init = ptl_core_init,
> .sub_req = pmc_core_pmt_get_blk_sub_req,
next prev parent reply other threads:[~2026-11-17 5:17 UTC|newest]
Thread overview: 3+ messages / expand[flat|nested] mbox.gz Atom feed top
2026-11-12 3:20 [PATCH v3] platform/x86/intel/pmc: Disable ACPI PM timer during suspend on Panther Lake Tzung-Bi Shih
2026-11-17 5:17 ` Xi Pardee [this message]
2026-11-18 7:15 ` Tzung-Bi Shih
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=766daa32-2247-40c2-980f-2a2317733b96@linux.intel.com \
--to=xi.pardee@linux.intel.com \
--cc=hansg@kernel.org \
--cc=ilpo.jarvinen@linux.intel.com \
--cc=irenic.rajneesh@gmail.com \
--cc=linux-kernel@vger.kernel.org \
--cc=platform-driver-x86@vger.kernel.org \
--cc=tfiga@chromium.org \
--cc=tzungbi@kernel.org \
/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 an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.