Re: [PATCH] mfd: cros_ec_dev: Add missing mfd_remove_devices() call in remove
From: Enric Balletbo i Serra
Date: Tue Dec 11 2018 - 05:00:55 EST
Hi Lee,
On 11/12/18 7:12, Lee Jones wrote:
> On Mon, 10 Dec 2018, Enric Balletbo i Serra wrote:
>
>> The driver adds different MFD child devices via mfd_add_devices() and
>> hence it is required to call mfd_remove_devices() to remove MFD child
>> devices.
>>
>> Fixes: 5e0115581bbc ("cros_ec: Move cros_ec_dev module to drivers/mfd")
>> Cc: stable@xxxxxxxxxxxxxxx
>> Signed-off-by: Enric Balletbo i Serra <enric.balletbo@xxxxxxxxxxxxx>
>> ---
>> Hi Lee,
>>
>> I saw that you send a mfd-fixes pull request this morning, so sorry in
>> advance for sending this too late. This was broken since the driver
>> moved from platform/chrome to mfd (and probably before that), so
>> it's an old problem. Note that I plan to send a patch series that depends
>> on this to apply cleanly. If the patch is fine with you and there is any
>> possibility to go in this version that will be good, if not, let me know
>> if you prefer queue this in your for-next branch or if you prefer I
>> include the patch on the series I plan to send on top of it to not mess
>> things.
>
> It wouldn't have made the v4.20-rcs anyway. Even if you did send it
> earlier. I only send fixes to that -rcs which fix issues introduced
> during the current release cycle.
>
Ok.
> If memory serves, doesn't this driver now (or will in the very near
> future) use devm_* for device creation? That would make this patch
> either incorrect (should be devm_mfd_remove_devices() if really
> required) or moot?
>
I think you have in mind this patch [1], right? Note that in this patch we're
using devm_* for cros_ec driver, _not_ cros_ec_dev driver which is different.
For the cros_ec_dev driver we need to take care with device managed allocations
as explained in the last fix merged [2]. For the cros_ec_dev I was trying to no
mix device managed allocations with non-device managed allocations, and as we
can't use devm_kzalloc in this case, I thought that was worth use the
mfd_add/mfd_remove functions like is now. Makes sense? What are your thoughts here?
Thanks,
Enric
[1] https://lkml.org/lkml/2018/11/27/881
[2]
https://git.kernel.org/pub/scm/linux/kernel/git/lee/mfd.git/commit/?h=for-mfd-fixes&id=48a2ca0ee3994df53da230c7079a18a70ec914f9
>> drivers/mfd/cros_ec_dev.c | 1 +
>> 1 file changed, 1 insertion(+)
>>
>> diff --git a/drivers/mfd/cros_ec_dev.c b/drivers/mfd/cros_ec_dev.c
>> index b99a194ce5a4..2d0fee488c5a 100644
>> --- a/drivers/mfd/cros_ec_dev.c
>> +++ b/drivers/mfd/cros_ec_dev.c
>> @@ -499,6 +499,7 @@ static int ec_device_remove(struct platform_device *pdev)
>>
>> cros_ec_debugfs_remove(ec);
>>
>> + mfd_remove_devices(ec->dev);
>> cdev_del(&ec->cdev);
>> device_unregister(&ec->class_dev);
>> return 0;
>