Re: [PATCH v2] i3c: master: dw-i3c-master: fix OD timing for first broadcast

Thank you for your contribution! Sashiko AI review found 2 potential issue(s) to consider:
- [Medium] The calculation for `od_hcnt` can silently overflow its 8-bit capacity if the core clock rate exceeds ~1.28 GHz, resulting in an inappropriately fast open-drain speed.
- [Low] Using DIV_ROUND_UP() with a 64-bit numerator or divisor causes a linker error (__udivdi3 undefined) on 32-bit platforms.
--

--- Patch [1]: [PATCH v2] i3c: master: dw-i3c-master: fix OD timing for first broadcast ---
commit b82f29a59ae98bd18cbb544ded9a3bf8f8d60fc6
Author: Tze Yee Ng <[email protected]>

i3c: master: dw-i3c-master: fix OD timing for first broadcast

This commit implements ->set_speed() to switch open-drain timing for the
first broadcast address according to the I3C specification. It caches the
normal OD register value during init and uses a separate od_hcnt for the
slow path so SDR extended timing remains derived from the normal PP hcnt.

> diff --git a/drivers/i3c/master/dw-i3c-master.c b/drivers/i3c/master/dw-i3c-master.c
> index 84ffcb1893389..077f2bf4b0b3f 100644
> --- a/drivers/i3c/master/dw-i3c-master.c
> +++ b/drivers/i3c/master/dw-i3c-master.c

[ ... ]

> @@ -1513,6 +1519,48 @@ static irqreturn_t dw_i3c_master_irq_handler(int irq, void *dev_id)
>  	return IRQ_HANDLED;
>  }
>  
> +static int dw_i3c_master_set_speed(struct i3c_master_controller *m,
> +				   enum i3c_open_drain_speed speed)
> +{
> +	struct dw_i3c_master *master = to_dw_i3c_master(m);
> +	unsigned long core_rate;
> +	u32 scl_timing;
> +	u8 od_hcnt, lcnt;
> +
> +	PM_RUNTIME_ACQUIRE_AUTOSUSPEND(master->dev, pm);
> +	if (PM_RUNTIME_ACQUIRE_ERR(&pm))
> +		return -ENXIO;
> +
> +	switch (speed) {
> +	case I3C_OPEN_DRAIN_SLOW_SPEED:
> +		core_rate = clk_get_rate(master->core_clk);
> +		if (!core_rate)
> +			return -EINVAL;
> +
> +		lcnt = SCL_I3C_TIMING_LCNT(master->i3c_od_timing_normal);
> +		od_hcnt = DIV_ROUND_UP((u64)I3C_BUS_THIGH_INIT_OD_MIN_NS * core_rate,
> +				       1000000000ULL) - 1;

[Severity: Low]
Will this cause a build failure on 32-bit platforms?

Using DIV_ROUND_UP() with a 64-bit divisor (1000000000ULL) will likely
result in a linker error for __udivdi3 on 32-bit architectures like ARM32
or i386. Should this use DIV_ROUND_UP_ULL() instead?

[Severity: Medium]
Can od_hcnt silently overflow here?

Because od_hcnt is declared as a u8 at the top of the function, if the
core clock rate exceeds ~1.28 GHz, the calculation will evaluate to a
value greater than 255.

This would truncate when assigned to od_hcnt, bypassing the minimum 200ns
requirement and potentially causing spike filters on devices to drop the
broadcast.

> +		if (od_hcnt < SCL_I3C_TIMING_CNT_MIN)
> +			od_hcnt = SCL_I3C_TIMING_CNT_MIN;
> +		scl_timing = SCL_I3C_TIMING_HCNT(od_hcnt) |
> +			     SCL_I3C_TIMING_LCNT(lcnt);
> +		writel(scl_timing, master->regs + SCL_I3C_OD_TIMING);
> +		master->i3c_od_timing = scl_timing;
> +		break;

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/[email protected]?part=1

-- 
linux-i3c mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/linux-i3c