Re: [PATCH] stm: class: Add MIPI OST protocol support

[Alexander Shishkin]

Mao Jinlong <quic_jinlmao@xxxxxxxxxxx> writes:

> Add MIPI OST protocol support for stm to format the traces.

Missing an explanation of what OST is, what it's used for, how it is
different from the SyS-T and others.

> Framework copied from drivers/hwtracing/stm.p-sys-t.c as of

You mean stm/p_sys-t.c. Also, it's not a framework, it's a driver.

> commit d69d5e83110f ("stm class: Add MIPI SyS-T protocol
> support").

Why is this significant?

> diff --git a/drivers/hwtracing/stm/p_ost.c b/drivers/hwtracing/stm/p_ost.c
> new file mode 100644
> index 000000000000..2ca1a3fda57f
> --- /dev/null
> +++ b/drivers/hwtracing/stm/p_ost.c
> @@ -0,0 +1,95 @@
> +// SPDX-License-Identifier: GPL-2.0-only
> +/*
> + * Copied from drivers/hwtracing/stm.p-sys-t.c as of commit d69d5e83110f
> + * ("stm class: Add MIPI SyS-T protocol support").

Same as in the commit message.

[...]

> +#define OST_TOKEN_STARTSIMPLE		(0x10)
> +#define OST_VERSION_MIPI1		(0x10 << 8)
> +#define OST_ENTITY_FTRACE		(0x01 << 16)
> +#define OST_CONTROL_PROTOCOL		(0x0 << 24)

These could use an explanation.

> +#define DATA_HEADER (OST_TOKEN_STARTSIMPLE | OST_VERSION_MIPI1 | \
> +			OST_ENTITY_FTRACE | OST_CONTROL_PROTOCOL)

Does this mean that everything is ftrace? Because it's not.

> +
> +#define STM_MAKE_VERSION(ma, mi)	((ma << 8) | mi)
> +#define STM_HEADER_MAGIC		(0x5953)
> +
> +static ssize_t notrace ost_write(struct stm_data *data,
> +		struct stm_output *output, unsigned int chan,
> +		const char *buf, size_t count)
> +{
> +	unsigned int c = output->channel + chan;
> +	unsigned int m = output->master;
> +	const unsigned char nil = 0;
> +	u32 header = DATA_HEADER;
> +	u8 trc_hdr[24];
> +	ssize_t sz;
> +
> +	/*
> +	 * STP framing rules for OST frames:
> +	 *   * the first packet of the OST frame is marked;
> +	 *   * the last packet is a FLAG.

Which in your case is also timestamped.

> +	 */
> +	/* Message layout: HEADER / DATA / TAIL */
> +	/* HEADER */
> +
> +	sz = data->packet(data, m, c, STP_PACKET_DATA, STP_PACKET_MARKED,
> +			  4, (u8 *)&header);

The /* HEADER */ comment applies to the above line, so it should
probably be directly before it.

> +	if (sz <= 0)
> +		return sz;
> +	*(uint16_t *)(trc_hdr) = STM_MAKE_VERSION(0, 3);
> +	*(uint16_t *)(trc_hdr + 2) = STM_HEADER_MAGIC;
> +	*(uint32_t *)(trc_hdr + 4) = raw_smp_processor_id();
> +	*(uint64_t *)(trc_hdr + 8) = sched_clock();

Why sched_clock()? It should, among other things, be called with
interrupts disabled, which is not the case here.

> +	*(uint64_t *)(trc_hdr + 16) = task_tgid_nr(get_current());

Is there a reason why trc_hdr is not a struct?

Thanks,
--
Alex