Re: [REGRESSION] Re: [PATCH v3] PM: suspend: Make pm_test delay interruptible by wakeup events

[luriwen <luriwen@kylinos.cn>]

在 2025/12/24 19:44, Rafael J. Wysocki 写道:
> On Tue, Dec 23, 2025 at 10:09 PM Sebastian Reichel
> <sebastian.reichel@collabora.com> wrote:
>> Hi,
>>
>> On Fri, Nov 14, 2025 at 05:11:16PM +0100, Rafael J. Wysocki wrote:
>>> On Thu, Nov 13, 2025 at 2:26 AM Riwen Lu <luriwen@kylinos.cn> wrote:
>>>> Modify the suspend_test() function to make the test delay can be
>>>> interrupted by wakeup events.
>>>>
>>>> This improves the responsiveness of the system during suspend testing
>>>> when wakeup events occur, allowing the suspend process to proceed
>>>> without waiting for the full test delay to complete when wakeup events
>>>> are detected.
>>>>
>>>> Additionally, using msleep() instead of mdelay() avoids potential soft
>>>> lockup "CPU stuck" issues when long test delays are configured.
>>>>
>>>> Co-developed-by: xiongxin <xiongxin@kylinos.cn>
>>>> Signed-off-by: xiongxin <xiongxin@kylinos.cn>
>>>> Signed-off-by: Riwen Lu <luriwen@kylinos.cn>
>>>> ---
>>>>   kernel/power/suspend.c | 6 +++++-
>>>>   1 file changed, 5 insertions(+), 1 deletion(-)
>>>>
>>>> diff --git a/kernel/power/suspend.c b/kernel/power/suspend.c
>>>> index b4ca17c2fecf..1c2f777da367 100644
>>>> --- a/kernel/power/suspend.c
>>>> +++ b/kernel/power/suspend.c
>>>> @@ -344,10 +344,14 @@ MODULE_PARM_DESC(pm_test_delay,
>>>>   static int suspend_test(int level)
>>>>   {
>>>>   #ifdef CONFIG_PM_DEBUG
>>>> +       int i;
>>>> +
>>>>          if (pm_test_level == level) {
>>>>                  pr_info("suspend debug: Waiting for %d second(s).\n",
>>>>                                  pm_test_delay);
>>>> -               mdelay(pm_test_delay * 1000);
>>>> +               for (i = 0; i < pm_test_delay && !pm_wakeup_pending(); i++)
>>>> +                       msleep(1000);
>>>> +
>>>>                  return 1;
>>>>          }
>>>>   #endif /* !CONFIG_PM_DEBUG */
>>>> --
>>> Applied as 6.19 material, thanks!
>> This is now upstream as a10ad1b10402 ("PM: suspend: Make pm_test delay interruptible by wakeup events").
>> When doing the following PM debugging test on ROCK 4D on v6.19-rc2
>>
>> echo N > /sys/module/printk/parameters/console_suspend
>> echo 1 > /sys/power/pm_print_times
>> echo 1 > /sys/power/pm_debug_messages
>> echo core > /sys/power/pm_test
>> echo mem > /sys/power/state
>>
>> I see the following error triggered reliably, which did not happen on 6.18
>> and no longer appears after reverting this patch:
>>
>> [   49.647656] ------------[ cut here ]------------
>> [   49.647656] WARNING: kernel/time/timekeeping.c:821 at ktime_get+0xb8/0xd8, CPU#0: swapper/0/0
>> [   49.647656] Modules linked in: sha256 cfg80211 binfmt_misc fuse ipv6 snd_soc_hdmi_codec rk805_pwrkey rockchip_saradc dwmac_rk stmmac_platform rtc_hym8563 snd_soc_es8328_i2c phy_rockchip_usbdp rockchipdrm stmmac snd_soc_es8328 dw_hdmi_qp industrialio_triggered_buffer analogix_dp typec phy_rockchip_samsung_hdptx kfifo_buf dw_dp rockchip_thermal dw_mipi_dsi spi_rockchip_sfc phy_rockchip_naneng_combphy pcs_xpcs panfrost drm_shmem_helper gpu_sched snd_soc_rockchip_sai snd_soc_simple_card drm_dp_aux_bus dw_hdmi snd_soc_simple_card_utils rfkill_gpio rfkill snd_soc_core drm_display_helper snd_compress cec drm_client_lib drm_dma_helper display_connector snd_pcm_dmaengine drm_kms_helper snd_pcm drm snd_timer snd backlight soundcore adc_keys
>> [   49.647656] CPU: 0 UID: 0 PID: 0 Comm: swapper/0 Not tainted 6.19.0-rc2-g7211b2cf9c08 #1 PREEMPT
>> [   49.647656] Hardware name: Radxa ROCK 4D (DT)
>> [   49.647656] pstate: 800000c5 (Nzcv daIF -PAN -UAO -TCO -DIT -SSBS BTYPE=--)
>> [   49.647656] pc : ktime_get+0xb8/0xd8
>> [   49.647656] lr : tick_nohz_idle_enter+0x50/0x90
>> [   49.647656] sp : ffffc1a17b573d50
>> [   49.647656] x29: ffffc1a17b573d50 x28: 00000000fdf57ec0 x27: ffffc1a17b57b260
>> [   49.647656] x26: ffffc1a17b57e000 x25: 0000000000000000 x24: 00000000000000ea
>> [   49.647656] x23: ffffc1a17a470000 x22: ffffc1a17b57ad00 x21: ffff3e6282a25000
>> [   49.647656] x20: ffff0003fdead5d8 x19: ffffc1a17b4885d8 x18: 000000000000000a
>> [   49.647656] x17: ffff3e6282a25000 x16: ffff800080000000 x15: 00700ea231d1d404
>> [   49.647656] x14: 0000000000000396 x13: 0000000000000001 x12: 0000000000000001
>> [   49.647656] x11: 00000000000000c0 x10: 0000000000000aa0 x9 : ffffc1a17b573d00
>> [   49.647656] x8 : ffffc1a17b586680 x7 : 0000000000000000 x6 : 0000000000000000
>> [   49.647656] x5 : 0000000000000004 x4 : ffffc1a17b57eb68 x3 : 0000000000000000
>> [   49.647656] x2 : ffffc1a17b48ec40 x1 : 0000000000000000 x0 : 0000000000000001
>> [   49.647656] Call trace:
>> [   49.647656]  ktime_get+0xb8/0xd8 (P)
>> [   49.647656]  tick_nohz_idle_enter+0x50/0x90
>> [   49.647656]  do_idle+0x38/0x260
>> [   49.647656]  cpu_startup_entry+0x38/0x40
>> [   49.647656]  rest_init+0xd8/0xe0
>> [   49.647656]  console_on_rootfs+0x0/0x6c
>> [   49.647656]  __primary_switched+0x88/0x90
>> [   49.647656] ---[ end trace 0000000000000000 ]---
> msleep() is not suitable for the "core" level testing.
>
> Change msleep() in suspend_test() back to mdelay() and the backtrace
> should go away.
>
> I'll send a patch for that later this week unless someone beats me to it.

Maybe usingmdelay() only forTEST_CORE while keepingmsleep() for other 
test levels ?