Re: [PATCH v1] rust: serdev: use Delta<Jiffy> for timeouts

[FUJITA Tomonori <tomo@flapping.org>]

On Thu, 01 Oct 2026 01:39:51 +0100
"Gary Guo" <gary@garyguo.net> wrote:

> On Thu Oct 1, 2026 at 1:29 AM BST, FUJITA Tomonori wrote:
>> On Wed, 30 Sep 2026 14:52:50 +0100
>> "Gary Guo" <gary@garyguo.net> wrote:
>>
>>> I think the API should ideally use `Option<Delta<Jiffies>>` for this case, and
>>> use `None` to represent indefinite wait.
>>
>> For read_poll_timeout(), where a timeout of 0 means "never time out"
>> in C, we did not take Option for the timeout. Alice's comment [1]:
>>
>> | Another thing is the `timeout_delta` option. I would just have
>> | written it as two methods, one that takes a timeout and one that
>> | doesn't. That way, callers that don't need a timeout do not need to
>> | handle timeout errors. (Do we have any users without a timeout? If
>> | not, maybe just remove the Option.)
>>
>> serdev is in the same situation. Do we want an API that uses None for
>> "never" here?
> 
> I don't like the "using 0" to mean forever, because 0 has a more sensible
> meaning -- which is just do a poll and don't wait.

Fully agreed.

> This one doesn't even return a result at all, so `Option` sounds reasonable.

That's true for wait_until_sent(), but write_all() returns a result
and it can be ETIMEDOUT.

> Alternatively, just use `MAX_SCHEDULE_TIMEOUT` (i.e.
> `Delta::from_jiffies(isize::MAX)`). We can perhaps add a `Delta::MAX` for this.

I don't think Delta::<Jiffy>::MAX is a good idea. Whether the maximum
value means an indefinite wait depends on the function. For example,
it doesn't for Queue::enqueue_delayed().

Another option is to take Delta<Jiffy> and round 0 up to 1 jiffy. A
driver that needs an indefinite wait passes
Delta::from_jiffies(MAX_SCHEDULE_TIMEOUT) explicitly.

Markus, what do you think?