From: Kai-Heng Feng <kai.heng.feng@canonical.com>
To: Hans de Goede <hdegoede@redhat.com>
Cc: ktsai@capellamicro.com, jic23@kernel.org, lars@metafoo.de,
Wahaj <wahajaved@protonmail.com>,
linux-iio@vger.kernel.org, linux-kernel@vger.kernel.org
Subject: Re: [PATCH] iio: light: cm32181: Fix PM support on system with 2 I2C resources
Date: Thu, 19 Jan 2023 01:03:57 +0800 [thread overview]
Message-ID: <CAAd53p76cC5P+5-8Dbego6XCkRcvBaQqzf_8SikYEND+_o7vSw@mail.gmail.com> (raw)
In-Reply-To: <a9db708f-59a6-586a-8728-81622a93ef3e@redhat.com>
Hi Hans,
On Wed, Jan 18, 2023 at 6:52 PM Hans de Goede <hdegoede@redhat.com> wrote:
>
> Hi,
>
> On 1/18/23 06:15, Kai-Heng Feng wrote:
> > On Wed, Jan 18, 2023 at 11:29 AM Kai-Heng Feng
> > <kai.heng.feng@canonical.com> wrote:
> >>
> >> Hi Hans,
> >>
> >> On Wed, Jan 18, 2023 at 1:21 AM Hans de Goede <hdegoede@redhat.com> wrote:
> >>>
> >>> Hi,
> >>>
> >>> On 1/17/23 17:09, Kai-Heng Feng wrote:
> >>>> Commit c1e62062ff54 ("iio: light: cm32181: Handle CM3218 ACPI devices
> >>>> with 2 I2C resources") creates a second client for the actual I2C
> >>>> address, but the "struct device" passed to PM ops is the first client
> >>>> that can't talk to the sensor.
> >>>>
> >>>> That means the I2C transfers in both suspend and resume routines can
> >>>> fail and blocking the whole suspend process.
> >>>>
> >>>> Instead of using the first client for I2C transfer, store the cm32181
> >>>> private struct on both cases so the PM ops can get the correct I2C
> >>>> client to perfrom suspend and resume.
> >>>>
> >>>> Fixes: 68c1b3dd5c48 ("iio: light: cm32181: Add PM support")
> >>>> Tested-by: Wahaj <wahajaved@protonmail.com>
> >>>> Signed-off-by: Kai-Heng Feng <kai.heng.feng@canonical.com>
> >>>
> >>> Thank you for this fix. I had looking into this on my todo list,
> >>> since I have been seeing some bug reports about this too.
> >>>
> >>> One remark inline:
> >>>
> >>>> ---
> >>>> drivers/iio/light/cm32181.c | 11 +++++++----
> >>>> 1 file changed, 7 insertions(+), 4 deletions(-)
> >>>>
> >>>> diff --git a/drivers/iio/light/cm32181.c b/drivers/iio/light/cm32181.c
> >>>> index 001055d097509..0f319c891353c 100644
> >>>> --- a/drivers/iio/light/cm32181.c
> >>>> +++ b/drivers/iio/light/cm32181.c
> >>>> @@ -440,6 +440,8 @@ static int cm32181_probe(struct i2c_client *client)
> >>>> if (!indio_dev)
> >>>> return -ENOMEM;
> >>>>
> >>>> + i2c_set_clientdata(client, indio_dev);
> >>>> +
> >>>
> >>> Why move this up, the suspend/resume callbacks cannot run until
> >>> probe() completes, so no need for this change.
> >>
> >> The intention is to save indio_dev as drvdata in the first (i.e.
> >> original) i2c_client's dev.
> >>
> >>>
> >>>> /*
> >>>> * Some ACPI systems list 2 I2C resources for the CM3218 sensor, the
> >>>> * SMBus Alert Response Address (ARA, 0x0c) and the actual I2C address.
> >>>> @@ -458,9 +460,9 @@ static int cm32181_probe(struct i2c_client *client)
> >>>> client = i2c_acpi_new_device(dev, 1, &board_info);
> >>>> if (IS_ERR(client))
> >>>> return PTR_ERR(client);
> >>>> - }
> >>>>
> >>>> - i2c_set_clientdata(client, indio_dev);
> >>>> + i2c_set_clientdata(client, indio_dev);
> >>>> + }
> >>>
> >>> And moving it inside the if block here (instead of just dropping it)
> >>> is also weird. I guess you meant to just delete it since you moved it up.
> >>
> >> Doesn't i2c_acpi_new_device() creates a new i2c_client (and its dev embedded)?
> >>
> >> So the intention is to save indio_dev for the second (ARA case) i2c_client too.
> >>
> >>>
> >>>>
> >>>> cm32181 = iio_priv(indio_dev);
> >>>> cm32181->client = client;
> >>>
> >>> Also note that the ->client used in suspend/resume now is not set until
> >>> here, so moving the i2c_set_clientdata() up really does not do anything.
> >>>
> >>> I beleive it would be best to just these 2 hunks from the patch and
> >>> only keep the changes to the suspend/resume callbacks.
> >>
> >> Yes, it seems like those 2 hunks are not necessary. Let me send a new patch.
> >
> > if (ACPI_HANDLE(dev) && client->addr == SMBUS_ALERT_RESPONSE_ADDRESS) {
> > ...
> > client = i2c_acpi_new_device(dev, 1, &board_info);
> > ...
> > }
> > i2c_set_clientdata(client, indio_dev);
> >
> > It means the indio_dev is only assigned to the new i2c_client->dev's
> > drvdata, the original dev's drvdata remains NULL.
> > So we need to assign it before the original client gets replaced by
> > the new one, otherwise we can't get cm32181 in PM ops.
>
> You are right, my bad. The original code has a bug where it indeed was
> making the i2c_set_clientdata() call on the wrong client device.
>
> So the i2c_set_clientdata() call needs to be moved up.
>
> There is no need to also call i2c_set_clientdata() on the dummy
> i2c-client though. That one does not have a driver attached.
Sure, will update this in v2.
>
> The suspend/resume callbacks are made on the original client-dev,
> not on the one of the dummy-client (which is the one which we
> actually use to communicate).
>
> >> But I do wonder what happens for the removing case? Will the second
> >> i2c_client leak?
>
> Yes it does, good point. That should probably also be fixed, but
> that needs to be a different / second patch.
Agree, thanks for the input.
Kai-Heng
>
> Regards,
>
> Hans
>
>
>
> >>>> @@ -490,7 +492,8 @@ static int cm32181_probe(struct i2c_client *client)
> >>>>
> >>>> static int cm32181_suspend(struct device *dev)
> >>>> {
> >>>> - struct i2c_client *client = to_i2c_client(dev);
> >>>> + struct cm32181_chip *cm32181 = iio_priv(dev_get_drvdata(dev));
> >>>> + struct i2c_client *client = cm32181->client;
> >>>>
> >>>> return i2c_smbus_write_word_data(client, CM32181_REG_ADDR_CMD,
> >>>> CM32181_CMD_ALS_DISABLE);
> >>>> @@ -498,8 +501,8 @@ static int cm32181_suspend(struct device *dev)
> >>>>
> >>>> static int cm32181_resume(struct device *dev)
> >>>> {
> >>>> - struct i2c_client *client = to_i2c_client(dev);
> >>>> struct cm32181_chip *cm32181 = iio_priv(dev_get_drvdata(dev));
> >>>> + struct i2c_client *client = cm32181->client;
> >>>>
> >>>> return i2c_smbus_write_word_data(client, CM32181_REG_ADDR_CMD,
> >>>> cm32181->conf_regs[CM32181_REG_ADDR_CMD]);
> >>>
> >
>
prev parent reply other threads:[~2023-01-18 17:04 UTC|newest]
Thread overview: 7+ messages / expand[flat|nested] mbox.gz Atom feed top
2023-01-17 16:09 Kai-Heng Feng
2023-01-17 17:19 ` Hans de Goede
2023-01-17 21:02 ` Hans de Goede
2023-01-18 3:29 ` Kai-Heng Feng
2023-01-18 5:15 ` Kai-Heng Feng
2023-01-18 10:52 ` Hans de Goede
2023-01-18 17:03 ` Kai-Heng Feng [this message]
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=CAAd53p76cC5P+5-8Dbego6XCkRcvBaQqzf_8SikYEND+_o7vSw@mail.gmail.com \
--to=kai.heng.feng@canonical.com \
--cc=hdegoede@redhat.com \
--cc=jic23@kernel.org \
--cc=ktsai@capellamicro.com \
--cc=lars@metafoo.de \
--cc=linux-iio@vger.kernel.org \
--cc=linux-kernel@vger.kernel.org \
--cc=wahajaved@protonmail.com \
/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 a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox
all inboxes | Powered by JetHome®