[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-ID: <27c926bf-f889-4e54-83c7-28134ca84899@gmail.com>
Date: Thu, 4 Dec 2025 16:21:40 +0500
From: Ali Tariq <alitariq45892@...il.com>
To: Andy Shevchenko <andriy.shevchenko@...el.com>
Cc: broonie@...nel.org, linux-spi@...r.kernel.org,
linux-kernel@...r.kernel.org, khairul.anuar.romli@...era.com,
adrianhoyin.ng@...era.com, nirav.rabara@...era.com,
matthew.gerlach@...era.com
Subject: Re: [PATCH] spi: cadence-qspi: Fix runtime PM imbalance in probe
Thanks for the merge. Recently, there was a patch which removed the
redundant pm_runtime_mark_last_busy() call:
https://lore.kernel.org/linux-kernel/20251203181921.97171-1-akifejaz40@gmail.com/
The patch removes pm_runtime_mark_last_busy() from the probe function,
as pm_runtime_put_autosuspend() already handles this internally.
Best Regards,
Ali
On 12/3/25 1:50 AM, Andy Shevchenko wrote:
> On Sun, Nov 30, 2025 at 09:12:51AM +0000, Ali Tariq wrote:
>> The probe function incorrectly calls pm_runtime_put_autosuspend()
>> twice in succession at the end of successful probe, dropping two
>> runtime PM references while only one was acquired earlier with
>> pm_runtime_get_sync(). This causes a usage count underflow:
>>
>> cadence-qspi 13010000.spi: Runtime PM usage count underflow!
>>
>> Remove the first redundant pm_runtime_put_autosuspend() call to
>> balance the reference count.
>>
>> Tested on StarFive VisionFive 2 v1.2A board.
>
> ...
>
>> if (!(ddata && (ddata->quirks & CQSPI_DISABLE_RUNTIME_PM))) {
>> - pm_runtime_put_autosuspend(dev);
>> pm_runtime_mark_last_busy(dev);
>
> This one — pm_runtime_mark_last_busy() — also may be removed as it's integrated
> into the _put_auutosuspend().
>
>> pm_runtime_put_autosuspend(dev);
>> }
>
Powered by blists - more mailing lists