Re: [PATCH 1/3] mfd: cros_ec: Add Chrome OS EC userspace device interface

[Olof Johansson]

On Wed, Nov 19, 2014 at 10:37 AM, Javier Martinez Canillas
<javier.martinez@xxxxxxxxxxxxxxx> wrote:
> Hello Alan,
>
> Thanks a lot for your feedback.
>
> On 11/18/2014 06:00 PM, One Thousand Gnomes wrote:
>>> +#ifdef CONFIG_COMPAT
>>> +struct compat_cros_ec_command {
>>> +    uint32_t version;
>>> +    uint32_t command;
>>> +    compat_uptr_t outdata;
>>> +    uint32_t outsize;
>>> +    compat_uptr_t indata;
>>> +    uint32_t insize;
>>> +    uint32_t result;
>>> +};
>>> +
>>> +struct compat_cros_ec_readmem {
>>> +    uint32_t offset;
>>> +    uint32_t bytes;
>>> +    compat_uptr_t buffer;
>>> +};
>>>
>>
>> This is a new API - arrange them to be 64bit safe and properly padded,
>> there is no excuse for needing compat crap except for legacy interfaces
>> you can't fix.
>>
>
> Is true that this is a new API for mainline but there is a lot of ChromeOS
> installations that depends on this API which means that just replacing the
> kernel with a mainline one there, will break existing user-space programs.

I think we can deal with that, at least if we pick new ioctl numbers
so we can tell from the userspace tool which interface is in use
during transition.

> But I understand that since those binaries were using a non-ustream kernel
> it is expected that the kernel API could be changed.
>
> I think it would be great to keep existing binaries working but if changing
> the API is required, then I can certainly do that when doing a re-spin.

I think there's some value in that, but i'm also somewhat embarrassed
to have missed this aspect when doing internal review, and do agree
with Alan. :) And we have only a few tools that use this interface so
we should be able to cope with it.


-Olof
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/