On 2020-01-01 17:55, Laurent Pinchart wrote:
> Hi Wolfram,
>
> Thank you for the patch.
>
> On Tue, Dec 31, 2019 at 05:13:58PM +0100, Wolfram Sang wrote:
>> Some devices are able to reprogram their I2C address at runtime. This
>> can prevent address collisions when one is able to activate and
>> reprogram these devices one by one. For that to work, they need to be
>> assigned an unused address. This new functions allows drivers to request
>> for such an address. It assumes all non-occupied addresses are free. It
>> will then send a message to such a free address to make sure there is
>> really nothing listening there.
>>
>> Signed-off-by: Wolfram Sang <wsa+renesas@xxxxxxxxxxxxxxxxxxxx>
>> ---
>> drivers/i2c/i2c-core-base.c | 22 ++++++++++++++++++++++
>> include/linux/i2c.h | 2 ++
>> 2 files changed, 24 insertions(+)
>>
>> diff --git a/drivers/i2c/i2c-core-base.c b/drivers/i2c/i2c-core-base.c
>> index 51bd953ddfb2..5a010e7e698f 100644
>> --- a/drivers/i2c/i2c-core-base.c
>> +++ b/drivers/i2c/i2c-core-base.c
>> @@ -2241,6 +2241,28 @@ static int i2c_detect(struct i2c_adapter *adapter, struct i2c_driver *driver)
>> return err;
>> }
>>
>
> Missing kerneldoc, but you already know about this.
>
>> +struct i2c_client *i2c_new_alias_device(struct i2c_adapter *adap)
>> +{
>> + struct i2c_client *alias = ERR_PTR(-EBUSY);
>> + int ret;
>> + u16 addr;
>> +
>> + i2c_lock_bus(adap, I2C_LOCK_SEGMENT);
>> +
>> + for (addr = 0x08; addr < 0x78; addr++) {
>> + ret = i2c_scan_for_client(adap, addr, i2c_unlocked_read_byte_probe);
>> + if (ret == -ENODEV) {
>> + alias = i2c_new_dummy_device(adap, addr);
>> + dev_dbg(&adap->dev, "Found alias: 0x%x\n", addr);
>> + break;
>> + }
>> + }
>
> This looks quite inefficient, especially if the beginning of the range
> is populated with devices. Furthermore, I think there's a high risk of
> false negatives, as acquiring a free address and reprogramming the
> client to make use of it are separate operations. Another call to
> i2c_new_alias_device() could occur in-between. There's also the issue
> that I2C hasn't been designed for scanning, so some devices may not
> appreciate this.
>
> What happened to the idea of reporting busy address ranges in the
> firmware (DT, ACPI, ...) ?
Another argument against probing (perhaps weak, but still) is that the
probed address might already be programmed in (one of) the reprogrammable
chip(s) making it respond to the probe and thus preventing the use of a
perfectly good alias address.
In /extreme/ situations this /might/ prevent finding a needed alias at
all.
Cheers,
Peter
