Virtualization - Oct 2021 - [PATCH 1/2] i2c: virtio: disable timeout handling

On Wed, Oct 20, 2021 at 12:20:13PM +0800, Jie Deng
wrote:
> 
> On 2021/10/20 2:14, Wolfram Sang wrote:
> > > I think it is set to HZ currently, though I haven't tried big
> > > transfers but I still get into some issues with Qemu based stuff.
> > > Maybe we can bump it up to few seconds :)
> > If you use adapter->timeout, this can even be set at runtime using
a
> > ioctl. So, it can adapt to use cases. Of course, the driver should
> > initialize it to a sane default if the automatic default (HZ) is not
> > suitable.
> 
> 
> I think a big value may solve most cases. but the driver never know how big
> is enough by static configuration.
> 
> Can we make this value to be configurable, just let the other side provide
> this value ?
If an ioctl can change it, that would mean it is configurable, right?

On 2021/10/20 13:36, Greg KH wrote:
> On Wed, Oct 20, 2021 at 12:20:13PM +0800, Jie Deng wrote:
>> On 2021/10/20 2:14, Wolfram Sang wrote:
>>>> I think it is set to HZ currently, though I haven't tried
big
>>>> transfers but I still get into some issues with Qemu based
stuff.
>>>> Maybe we can bump it up to few seconds :)
>>> If you use adapter->timeout, this can even be set at runtime
using a
>>> ioctl. So, it can adapt to use cases. Of course, the driver should
>>> initialize it to a sane default if the automatic default (HZ) is
not
>>> suitable.
>>
>> I think a big value may solve most cases. but the driver never know how
big
>> is enough by static configuration.
>>
>> Can we make this value to be configurable, just let the other side
provide
>> this value ?
> If an ioctl can change it, that would mean it is configurable, right?

Yes, but we need to know what's the best value to be configured for a 
specific "other side".

I think the "other side" should be more aware of what value is 
reasonable to be used.