On 26/07/2023 05:38, Ryan Chen wrote:
>>> +
>>> +static int ast2600_i2c_remove(struct platform_device *pdev) {
>>> + struct ast2600_i2c_bus *i2c_bus = platform_get_drvdata(pdev);
>>> +
>>> + /* Disable everything. */
>>> + writel(0, i2c_bus->reg_base + AST2600_I2CC_FUN_CTRL);
>>> + writel(0, i2c_bus->reg_base + AST2600_I2CM_IER);
>>> +
>>> + i2c_del_adapter(&i2c_bus->adap);
>>
>>> I have doubts that you tested this. I think you have here double free/del of the adapter.
>> Sorry, i can't catch your point for double free the adapter.
>> It should use i2c_del_adapter in driver remove function.
>> All the driver doing this
>> https://github.com/torvalds/linux/blob/master/drivers/i2c/busses/i2c-n
>> pcm7xx.c#L2373
>> https://github.com/torvalds/linux/blob/master/drivers/i2c/busses/i2c-a
>> ltera.c#L473
>>
>> Do you mean it is not necessary?
>
>> Instead of giving you the fish, I think much more learning experience is to teach you how to fish. Please unbind your driver (echo the device name to proper unbind file in sysfs). The best if you build your kernel with KASAN.
>
>> Thanks, will do this test with unbind to understand your point.
> I do my the test with enable kernel config KASAN.
> I don't see any error dump. You can see the following. It use i2c-0 bind/unbind.
> Did I miss some test ??
>
> root@ast2600-default:/sys/bus/platform/drivers/i2c_ast2600# i2cdetect -l
> i2c-0 i2c 1e78a080.i2c-bus I2C adapter
> i2c-1 i2c 1e78a100.i2c-bus I2C adapter
> i2c-2 i2c 1e78a180.i2c-bus I2C adapter
> i2c-3 i2c 1e78a200.i2c-bus I2C adapter
> i2c-4 i2c 1e78a280.i2c-bus I2C adapter
> i2c-5 i2c 1e78a300.i2c-bus I2C adapter
> i2c-6 i2c 1e78a380.i2c-bus I2C adapter
> i2c-7 i2c 1e78a400.i2c-bus I2C adapter
> i2c-8 i2c 1e78a480.i2c-bus I2C adapter
> i2c-9 i2c 1e78a500.i2c-bus I2C adapter
> i2c-12 i2c 1e78a680.i2c-bus I2C adapter
> i2c-13 i2c 1e78a700.i2c-bus I2C adapter
> i2c-14 i2c 1e78a780.i2c-bus I2C adapter
> i2c-15 i2c 1e78a800.i2c-bus I2C adapter
> root@ast2600-default:/sys/bus/platform/drivers/i2c_ast2600# echo 1e78a080.i2c-bus > unbind
It should fail... if it doesn't, maybe you can tell why?
Best regards,
Krzysztof
