On Wed, May 19, 2021 at 02:49:32PM +0700, Quan Nguyen wrote:
> Slave i2c device on AST2500 received a lot of slave irq while it is
> busy processing the response. To handle this case, adds and exports
> aspeed_set_slave_busy() for controller to temporary stop slave irq
> while slave is handling the response, and re-enable them again when
> the response is ready.
>
> Signed-off-by: Quan Nguyen <quan@xxxxxxxxxxxxxxxxxxxxxx>
> ---
> v3:
> + First introduce in v3 [Quan]
>
> drivers/i2c/busses/i2c-aspeed.c | 20 ++++++++++++++++++++
> 1 file changed, 20 insertions(+)
>
> diff --git a/drivers/i2c/busses/i2c-aspeed.c b/drivers/i2c/busses/i2c-aspeed.c
> index b2e9c8f0ddf7..9926d04831a2 100644
> --- a/drivers/i2c/busses/i2c-aspeed.c
> +++ b/drivers/i2c/busses/i2c-aspeed.c
> @@ -944,6 +944,26 @@ static int aspeed_i2c_init(struct aspeed_i2c_bus *bus,
> return 0;
> }
>
> +#if IS_ENABLED(CONFIG_I2C_SLAVE)
> +void aspeed_set_slave_busy(struct i2c_adapter *adap, bool busy)
> +{
> + struct aspeed_i2c_bus *bus = i2c_get_adapdata(adap);
> + unsigned long current_mask, flags;
> +
> + spin_lock_irqsave(&bus->lock, flags);
This as far as I can see is still a recursive spinlock, and the spinlock
debugger seems to agree with me.
Graeme
> +
> + current_mask = readl(bus->base + ASPEED_I2C_INTR_CTRL_REG);
> + if (busy)
> + current_mask &= ~(ASPEED_I2CD_INTR_RX_DONE | ASPEED_I2CD_INTR_SLAVE_MATCH);
> + else
> + current_mask |= ASPEED_I2CD_INTR_RX_DONE | ASPEED_I2CD_INTR_SLAVE_MATCH;
> + writel(current_mask, bus->base + ASPEED_I2C_INTR_CTRL_REG);
> +
> + spin_unlock_irqrestore(&bus->lock, flags);
> +}
> +EXPORT_SYMBOL_GPL(aspeed_set_slave_busy);
> +#endif
> +
> static int aspeed_i2c_reset(struct aspeed_i2c_bus *bus)
> {
> struct platform_device *pdev = to_platform_device(bus->dev);
> --
> 2.28.0
>
