Re: [PATCH v2] sound: rawmidi: Add framing mode

From: Takashi Sakamoto <o-takashi@sakamocchi.jp>
To: David Henningsson <coding@diwic.se>
Cc: tiwai@suse.de, alsa-devel@alsa-project.org
Subject: Re: [PATCH v2] sound: rawmidi: Add framing mode
Date: Wed, 24 Mar 2021 21:44:30 +0900	[thread overview]
Message-ID: <20210324124430.GA3711@workstation> (raw)
In-Reply-To: <20210324054253.34642-1-coding@diwic.se>

Hi,

On Wed, Mar 24, 2021 at 06:42:53AM +0100, David Henningsson wrote:
> diff --git a/include/uapi/sound/asound.h b/include/uapi/sound/asound.h
> index 535a7229e1d9..f33076755025 100644
> --- a/include/uapi/sound/asound.h
> +++ b/include/uapi/sound/asound.h
> @@ -736,12 +736,28 @@ struct snd_rawmidi_info {
>  	unsigned char reserved[64];	/* reserved for future use */
>  };
>  
> +enum {
> +	SNDRV_RAWMIDI_FRAMING_NONE = 0,
> +	SNDRV_RAWMIDI_FRAMING_TSTAMP_MONOTONIC_RAW,
> +	SNDRV_RAWMIDI_FRAMING_LAST = SNDRV_RAWMIDI_FRAMING_TSTAMP_MONOTONIC_RAW,
> +};

In C language specification, enumeration is for value of int storage. In
my opinion, int type should be used for the framing member, perhaps.

(I think you can easily understand my insistent since you're Rust
programmer.)

I note that in UAPI of Linux kernel, we have some macros to represent
system clocks; e.g. CLOCK_REALTIME, CLOCK_MONOTONIC:
https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/include/uapi/linux/time.h#n46

We can use the series of macro, instead of defining the specific
enumerations. However I have one concern that the 'None' value cannot be
zero in the case since CLOCK_REALTIME is zero. This is a bit inconvenient
since we need initializer function in both of kernel space and user
space...

For the idea to record system timestamp when drivers call helper
function to put MIDI message bytes into intermediate buffer in
hardware/software IRQ context, I have some concerns and I'll post
another message to thread, later.

Regards

Takashi Sakamoto