Hello Jean,
On 05/09/2019 14:57, Jean Delvare wrote:
> If this is of any value to you, I tried implementing it in i2c-i801 a
> few days ago. I can't really test it though as I don't have any device
> which triggers an alert on my system, but I am sharing it with you if
> you want to give it a try. You would still need to write the code in
> the slave driver.
>
> ---
> drivers/i2c/busses/i2c-i801.c | 77 +++++++++++++++++++++++++++++++++++++-----
> 1 file changed, 69 insertions(+), 8 deletions(-)
>
> --- linux-5.2.orig/drivers/i2c/busses/i2c-i801.c 2019-08-28 15:58:52.725828215 +0200
> +++ linux-5.2/drivers/i2c/busses/i2c-i801.c 2019-08-28 16:50:09.212696037 +0200
> @@ -196,6 +196,7 @@
>
> /* Host Notify Command register bits */
> #define SMBSLVCMD_HST_NTFY_INTREN BIT(0)
> +#define SMBSLVCMD_SMBALERT_DIS BIT(2)
>
> #define STATUS_ERROR_FLAGS (SMBHSTSTS_FAILED | SMBHSTSTS_BUS_ERR | \
> SMBHSTSTS_DEV_ERR)
> @@ -281,6 +282,10 @@ struct i801_priv {
> */
> bool acpi_reserved;
> struct mutex acpi_lock;
> +
> + /* SMBus alert */
> + struct i2c_smbus_alert_setup alert_data;
> + struct i2c_client *ara;
> };
>
> #define FEATURE_SMBUS_PEC BIT(0)
> @@ -289,6 +294,7 @@ struct i801_priv {
> #define FEATURE_I2C_BLOCK_READ BIT(3)
> #define FEATURE_IRQ BIT(4)
> #define FEATURE_HOST_NOTIFY BIT(5)
> +#define FEATURE_SMBUS_ALERT BIT(6)
> /* Not really a feature, but it's convenient to handle it as such */
> #define FEATURE_IDF BIT(15)
> #define FEATURE_TCO_SPT BIT(16)
> @@ -301,6 +307,7 @@ static const char *i801_feature_names[]
> "I2C block read",
> "Interrupt",
> "SMBus Host Notify",
> + "SMBus Alert",
> };
>
> static unsigned int disable_features;
> @@ -310,7 +317,8 @@ MODULE_PARM_DESC(disable_features, "Disa
> "\t\t 0x02 disable the block buffer\n"
> "\t\t 0x08 disable the I2C block read functionality\n"
> "\t\t 0x10 don't use interrupts\n"
> - "\t\t 0x20 disable SMBus Host Notify ");
> + "\t\t 0x20 disable SMBus Host Notify\n"
> + "\t\t 0x40 disable SMBus Alert ");
>
> /* Make sure the SMBus host is ready to start transmitting.
> Return 0 if it is, -EBUSY if it is not. */
> @@ -620,8 +628,24 @@ static irqreturn_t i801_host_notify_isr(
> return IRQ_HANDLED;
> }
>
> +static irqreturn_t i801_smbus_alert_isr(struct i801_priv *priv)
> +{
> + struct i2c_client *ara = priv->ara;
> +
> + if (ara) {
> + dev_dbg(&ara->dev, "SMBus alert received\n");
> + i2c_handle_smbus_alert(ara);
I've just tested this approach and must conclude that it doesn't work.
Seems that SMBALERT# in i801 is level-triggered (I wasn't able to
confirm or decline this statement using i801 documentation) and
once SMBALERT# is triggered, I experience the interrupt flood.
Clearing the interrupt request bit below has no effect at this point.
Unfortunately, current smbalert API doesn't provide any synchronous
interface to test another approach.
> + } else
> + dev_dbg(&priv->adapter.dev,
> + "SMBus alert received but no ARA client!\n");
> +
> + /* clear SMBus Alert bit and return */
> + outb_p(SMBHSTSTS_SMBALERT_STS, SMBHSTSTS(priv));
> + return IRQ_HANDLED;
> +}
> +
> /*
> - * There are three kinds of interrupts:
> + * There are four kinds of interrupts:
> *
> * 1) i801 signals transaction completion with one of these interrupts:
> * INTR - Success
Therefore, I'd like to ask you (and other I2C maintainers), not that
I spend a time for a solution which has no upstream perspectives, what
shall be the preferred way for i801 driver:
a) We disable SMBALERT# completely
or
b) I export smbus_alert(), so that we can
- catch the IRQ
- disable SMBALERT#
- synchronously smbus_alert()
- re-enable SMBALERT#
?
--
Best regards,
Alexander Sverdlin.
