From: Riwen Lu <[email protected]>
Change the maximum idr ID to INT_MAX for i2c adapter.
This fix the following adding i2c adapter error while the idr with
adap->nr has been allocated by other i2c driver, and find a proper ID
for it.
[ 357.620685][ 3] [ T3655] ------------[ cut here ]------------
[ 357.626544][ 3] [ T3655] couldn't get idr
Signed-off-by: Riwen Lu <[email protected]>
---
drivers/i2c/i2c-core-base.c | 3 ++-
1 file changed, 2 insertions(+), 1 deletion(-)
diff --git a/drivers/i2c/i2c-core-base.c b/drivers/i2c/i2c-core-base.c
index 7539b0740351..c2982861057a 100644
--- a/drivers/i2c/i2c-core-base.c
+++ b/drivers/i2c/i2c-core-base.c
@@ -1535,11 +1535,12 @@ static int __i2c_add_numbered_adapter(struct i2c_adapter *adap)
int id;
mutex_lock(&core_lock);
- id = idr_alloc(&i2c_adapter_idr, adap, adap->nr, adap->nr + 1, GFP_KERNEL);
+ id = idr_alloc(&i2c_adapter_idr, adap, adap->nr, 0, GFP_KERNEL);
mutex_unlock(&core_lock);
if (WARN(id < 0, "couldn't get idr"))
return id == -ENOSPC ? -EBUSY : id;
+ adap->nr = id;
return i2c_register_adapter(adap);
}
--
2.25.1
On Fri, Dec 16, 2022 at 10:07:51AM +0800, Riwen Lu wrote:
> From: Riwen Lu <[email protected]>
>
> Change the maximum idr ID to INT_MAX for i2c adapter.
> This fix the following adding i2c adapter error while the idr with
> adap->nr has been allocated by other i2c driver, and find a proper ID
How did this happen that another i2c driver could get this? Were you
able to follow the code paths that lead to this situation?
> for it.
> [ 357.620685][ 3] [ T3655] ------------[ cut here ]------------
> [ 357.626544][ 3] [ T3655] couldn't get idr
>
> Signed-off-by: Riwen Lu <[email protected]>
> ---
> drivers/i2c/i2c-core-base.c | 3 ++-
> 1 file changed, 2 insertions(+), 1 deletion(-)
>
> diff --git a/drivers/i2c/i2c-core-base.c b/drivers/i2c/i2c-core-base.c
> index 7539b0740351..c2982861057a 100644
> --- a/drivers/i2c/i2c-core-base.c
> +++ b/drivers/i2c/i2c-core-base.c
> @@ -1535,11 +1535,12 @@ static int __i2c_add_numbered_adapter(struct i2c_adapter *adap)
> int id;
>
> mutex_lock(&core_lock);
> - id = idr_alloc(&i2c_adapter_idr, adap, adap->nr, adap->nr + 1, GFP_KERNEL);
> + id = idr_alloc(&i2c_adapter_idr, adap, adap->nr, 0, GFP_KERNEL);
> mutex_unlock(&core_lock);
> if (WARN(id < 0, "couldn't get idr"))
> return id == -ENOSPC ? -EBUSY : id;
>
> + adap->nr = id;
> return i2c_register_adapter(adap);
> }
>
> --
> 2.25.1
>
在 2023/1/9 23:13, Wolfram Sang 写道:
> On Fri, Dec 16, 2022 at 10:07:51AM +0800, Riwen Lu wrote:
>> From: Riwen Lu <[email protected]>
>>
>> Change the maximum idr ID to INT_MAX for i2c adapter.
>> This fix the following adding i2c adapter error while the idr with
>> adap->nr has been allocated by other i2c driver, and find a proper ID
>
> How did this happen that another i2c driver could get this? Were you
> able to follow the code paths that lead to this situation?
>
Yes, it's happened in Phytium(Arm64) platform with MWV207 GPU.
1. The MWV207 GPU driver has registered I2C adapter, and allocated idr
id 0-6 with i2c_adapter_idr.
# cd /sys/class/i2c-adapter/i2c-0/
# ls
i2c-0 i2c-1 i2c-2 i2c-3 i2c-4 i2c-5 i2c-6
# cd i2c-0/
# ls
delete_device device i2c-dev name new_device power subsystem uevent
# cat name
MWV207_I2C_0
2. The BIOS's ACPI DSDT has declared I2C device with _UID 0 as following:
Device (I2C0)
{
Name (_HID, "PHYT0038") // _HID: Hardware ID
Name (_UID, Zero) // _UID: Unique ID
Name (_CRS, ResourceTemplate ()
{
Memory32Fixed (ReadWrite,
0x28014000,
0x00001000,
)
Interrupt (ResourceConsumer, Level, ActiveHigh,
Exclusive, ,, )
{
0x0000007C,
}
})
...
}
The Phytium i2c driver would register the i2c adapter with idr id 0 by
calling function i2c_add_numbered_adapter(adap). Then, function
idr_alloc(&i2c_adapter_idr, adap, adap->nr, adap->nr + 1, GFP_KERNEL)；
allocate the idr failed because idr id adap->nr(It equal to _UID) has
been allocated by MWV207 GPU, and print the following error:
couldn't get idr
i2c-phytium-platform PHYT0038:00: fail to add adapter: -16
This patch can solve the problem.
>> for it.
>> [ 357.620685][ 3] [ T3655] ------------[ cut here ]------------
>> [ 357.626544][ 3] [ T3655] couldn't get idr
>>
>> Signed-off-by: Riwen Lu <[email protected]>
>> ---
>> drivers/i2c/i2c-core-base.c | 3 ++-
>> 1 file changed, 2 insertions(+), 1 deletion(-)
>>
>> diff --git a/drivers/i2c/i2c-core-base.c b/drivers/i2c/i2c-core-base.c
>> index 7539b0740351..c2982861057a 100644
>> --- a/drivers/i2c/i2c-core-base.c
>> +++ b/drivers/i2c/i2c-core-base.c
>> @@ -1535,11 +1535,12 @@ static int __i2c_add_numbered_adapter(struct i2c_adapter *adap)
>> int id;
>>
>> mutex_lock(&core_lock);
>> - id = idr_alloc(&i2c_adapter_idr, adap, adap->nr, adap->nr + 1, GFP_KERNEL);
>> + id = idr_alloc(&i2c_adapter_idr, adap, adap->nr, 0, GFP_KERNEL);
>> mutex_unlock(&core_lock);
>> if (WARN(id < 0, "couldn't get idr"))
>> return id == -ENOSPC ? -EBUSY : id;
>>
>> + adap->nr = id;
>> return i2c_register_adapter(adap);
>> }
>>
>> --
>> 2.25.1
>>
> The Phytium i2c driver would register the i2c adapter with idr id 0 by
> calling function i2c_add_numbered_adapter(adap). Then, function
> idr_alloc(&i2c_adapter_idr, adap, adap->nr, adap->nr + 1, GFP_KERNEL)；
> allocate the idr failed because idr id adap->nr(It equal to _UID) has been
> allocated by MWV207 GPU, and print the following error:
>
> couldn't get idr
> i2c-phytium-platform PHYT0038:00: fail to add adapter: -16
i2c-phytium-platform is not an upstream driver, right? Why does it not
simply use i2c_add_adapter() instead of i2c_add_numbered_adapter()?