lkml.org 
[lkml]   [2022]   [Sep]   [20]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
SubjectRe: [PATCH v8 2/3] Documentation: add a isolation strategy sysfs node for uacce
From
Date


On 2022/9/19 17:34, Greg KH wrote:
> On Mon, Sep 19, 2022 at 11:21:30AM +0800, yekai (A) wrote:
>>
>> On 2022/9/9 16:27, Greg KH wrote:
>>> On Fri, Sep 02, 2022 at 03:13:03AM +0000, Kai Ye wrote:
>>>> Update documentation describing sysfs node that could help to
>>>> configure isolation strategy for users in the user space. And
>>>> describing sysfs node that could read the device isolated state.
>>>>
>>>> Signed-off-by: Kai Ye <yekai13@huawei.com>
>>>> ---
>>>> Documentation/ABI/testing/sysfs-driver-uacce | 26 ++++++++++++++++++++
>>>> 1 file changed, 26 insertions(+)
>>>>
>>>> diff --git a/Documentation/ABI/testing/sysfs-driver-uacce b/Documentation/ABI/testing/sysfs-driver-uacce
>>>> index 08f2591138af..af5bc2f326d2 100644
>>>> --- a/Documentation/ABI/testing/sysfs-driver-uacce
>>>> +++ b/Documentation/ABI/testing/sysfs-driver-uacce
>>>> @@ -19,6 +19,32 @@ Contact: linux-accelerators@lists.ozlabs.org
>>>> Description: Available instances left of the device
>>>> Return -ENODEV if uacce_ops get_available_instances is not provided
>>>>
>>>> +What: /sys/class/uacce/<dev_name>/isolate_strategy
>>>> +Date: Sep 2022
>>>> +KernelVersion: 6.0
>>>> +Contact: linux-accelerators@lists.ozlabs.org
>>>> +Description: (RW) Configure the frequency size for the hardware error
>>>> + isolation strategy. This size is a configured integer value.
>>>> + The default is 0. The maximum value is 65535. This value is a
>>>> + threshold based on your driver strategies.
>>> I do not understand what the units are here.
>>>
>>> How is anyone supposed to know what they are?
>> This unit is the number of times. Number of occurrences in a period, also means threshold.
>> If the number of device pci AER error exceeds the threshold in a time window, the device is
>> isolated.
> Please document this very very well.
>
>>>> + For example, in the hisilicon accelerator engine, first we will
>>>> + time-stamp every slot AER error. Then check the AER error log
>>>> + when the device AER error occurred. if the device slot AER error
>>>> + count exceeds the preset the number of times in one hour, the
>>>> + isolated state will be set to true. So the device will be
>>>> + isolated. And the AER error log that exceed one hour will be
>>>> + cleared. Of course, different strategies can be defined in
>>>> + different drivers.
>>> So this file can contain values of different units depending on the
>>> different driver that creates it? How is anyone supposed to know what
>>> it is and what it should be?
>>>
>>> This feels very loose, please define this much better so that it can be
>>> understood and maintained properly.
>>>
>>> thanks,
>>>
>>> greg k-h
>>> .
>>>
>> Yes, We started out with the idea of not restricting the different drive, only specifying the input and output.
>> Because we think different drivers require different processing strategy.
> What different drivers? You only have one! And why do you need a
> framework for only one driver? You should only add that when you have
> multiple users to ensure you got the framework correct otherwise you do
> not know how it will be used.
>
> thanks,
>
> greg k-h
> .
>
ok . I will move isolation strategy from qm to uacce in the next version.

it can be understood and maintained properly.

thanks

Kai

\
 
 \ /
  Last update: 2022-09-21 05:06    [W:0.071 / U:0.008 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site