Re: [PATCH v5 1/1] i2c: mediatek: add runtime PM operations and bus regulator control

["Zoie Lin (林禹妡)" <Zoie.Lin@mediatek.com>]

Hi Andi,
Hi Daniel,
 
1. The i2c_transfer function uses a completion mechanism
for interrupt handling, so i2c_transfer cannot be used in
interrupt context.

2. From the crash log, it was found that rt5190a_probe -->
irqreturn_t rt5190a_irq_handler (in interrupt context) -->
regmap_raw_read --> i2c_transfer, which indicates that i2c_transfer is
being used in interrupt context. This usage is incorrect.
 
Based on the reason above, it appears that the issue is related to
the rt5190a calling context rather than this patch.
 
Could you please review and consider to merge this patch again?
 
Thanks,
Zoie


On Fri, 2025-03-21 at 01:34 +0100, Andi Shyti wrote:
> 
> External email : Please do not click links or open attachments until
> you have verified the sender or the content.
> 
> 
> Hi Daniel,
> 
> On Thu, Mar 20, 2025 at 10:08:42PM +0000, Daniel Golle wrote:
> > On Wed, Mar 19, 2025 at 12:34:53AM +0100, Andi Shyti wrote:
> > > On Fri, Mar 14, 2025 at 10:53:43PM +0800, Zoie Lin wrote:
> > > > Introduce support for runtime PM operations in
> > > > the I2C driver, enabling runtime suspend and resume
> > > > functionality.
> > > > 
> > > > Although in most platforms, the bus power of i2c is always
> > > > on, some platforms disable the i2c bus power in order to meet
> > > > low power request.
> > > > 
> > > > This implementation includes bus regulator control to
> > > > facilitate
> > > > proper handling of the bus power based on platform
> > > > requirements.
> > > > 
> > > > Signed-off-by: Zoie Lin <zoie.lin@mediatek.com>
> > > 
> > > merged to i2c/i2c-host.
> > 
> > the change causes a crash during boot on MT7988 which typically
> > uses
> > one of its I2C busses to connect a Richtek RT5190A PMIC.
> 
> Thanks a lot for testing it. I removed the commit.
> 
> Andi