linux-kernel - Re: [PATCH 3/7] i3c: master: Mark last_busy on IBI when runtime PM is allowed

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <66d63d13-8e49-4101-acab-a6a5203826ca@intel.com>
Date: Fri, 30 Jan 2026 09:48:07 +0200
From: Adrian Hunter <adrian.hunter@...el.com>
To: Frank Li <Frank.li@....com>
CC: <alexandre.belloni@...tlin.com>, <linux-i3c@...ts.infradead.org>,
	<linux-kernel@...r.kernel.org>, <linux-pm@...r.kernel.org>
Subject: Re: [PATCH 3/7] i3c: master: Mark last_busy on IBI when runtime PM is
 allowed

On 29/01/2026 22:55, Frank Li wrote:
> On Thu, Jan 29, 2026 at 10:42:32PM +0200, Adrian Hunter wrote:
>> On 29/01/2026 21:56, Frank Li wrote:
>>> On Thu, Jan 29, 2026 at 08:18:37PM +0200, Adrian Hunter wrote:
>>>> When an IBI can be received after the controller is
>>>> pm_runtime_put_autosuspend()'ed, the interrupt may occur just before the
>>>> device is auto‑suspended.  In such cases, the runtime PM core may not see
>>>> any recent activity and may suspend the device earlier than intended.
>>>>
>>>> Mark the controller as last busy whenever an IBI is queued (when
>>>> rpm_ibi_allowed is set) so that the auto-suspend delay correctly reflects
>>>> recent bus activity and avoids premature suspension.
>>>
>>> look like this can't resolve problem. pm_runtime_mark_last_busy() just
>>> change dev->power.last_busy. If suspend before it, nothing happen.
>>
>> It should be effective.
>>
>> rpm_suspend() recalculates the autosuspend expiry time based on
>> last_busy (see pm_runtime_autosuspend_expiration()) and restarts
>> the timer is it is in the future.
>>
>>>
>>> irq use thread irq, in irq thread call pm_runtime_resume() if needs.
>>>
>>> And this function call by irq handle, just put to work queue, what's impact
>>> if do nothing here?
>>
>> Just premature runtime suspension inconsistent with autosuspend_delay.
> 
> 
>    CPU 0            CPU 1
> 1. rpm_suspend()    2. pm_runtime_mark_last_busy(master->rpm_dev)
> 
> if 2 happen before 1, it can extend suspend. 2 happen after 1, it should
> do nothing.

2 happening after 1 is a separate issue.  It will never happen
in the wakeup case because the wakeup does a runtime resume:

	pm_runtime_put_autosuspend()
	IBI -> pm_runtime_mark_last_busy()
	another IBI -> pm_runtime_mark_last_busy() and so on
	<autosuspend_delay finally elapses>
	rpm_suspend() -> device suspended, PME activated
	IBI START -> PME -> pm_request_resume()
	IBI is delivered after controller runtime resumes

> 
> Frank
>>
>>>
>>> Frank
>>>
>>>>
>>>> Signed-off-by: Adrian Hunter <adrian.hunter@...el.com>
>>>> ---
>>>>  drivers/i3c/master.c | 5 +++++
>>>>  1 file changed, 5 insertions(+)
>>>>
>>>> diff --git a/drivers/i3c/master.c b/drivers/i3c/master.c
>>>> index bcc493dc9d04..dcc07ebc50a2 100644
>>>> --- a/drivers/i3c/master.c
>>>> +++ b/drivers/i3c/master.c
>>>> @@ -2721,9 +2721,14 @@ static void i3c_master_unregister_i3c_devs(struct i3c_master_controller *master)
>>>>   */
>>>>  void i3c_master_queue_ibi(struct i3c_dev_desc *dev, struct i3c_ibi_slot *slot)
>>>>  {
>>>> +	struct i3c_master_controller *master = i3c_dev_get_master(dev);
>>>> +
>>>>  	if (!dev->ibi || !slot)
>>>>  		return;
>>>>
>>>> +	if (master->rpm_ibi_allowed)
>>>> +		pm_runtime_mark_last_busy(master->rpm_dev);
>>>> +
>>>>  	atomic_inc(&dev->ibi->pending_ibis);
>>>>  	queue_work(dev->ibi->wq, &slot->work);
>>>>  }
>>>> --
>>>> 2.51.0
>>>>
>>