Skip to contentSkip to conversation

i2c: i801: Improve status polling

PATCH Applied Heiner Kallweit 3 participants 3 messages May 25, 2021 – Jun 20, 2021

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

PATCH 1/1 i2c: i801: Improve status polling

Polling uses the same timeout as irq mode: 400 * 500us = 200ms = HZ / 5. So let's use the adapter->timeout value also for polling. This has the advantage that userspace can control the timeout value for polling as well. In addition change the code to make it better readable. Last but not least remove the timeout debug messages. Calls to both functions are followed by a call to i801_check_post() that will print an error message in case of timeout.

---
 drivers/i2c/busses/i2c-i801.c | 36 +++++++++++++----------------------
 1 file changed, 13 insertions(+), 23 deletions(-)
drivers/i2c/busses/i2c-i801.c +13 -23
index f6d7866f1..2c6e84108 100644
@@ -156,9 +156,6 @@
 #define SMBAUXCTL_CRC		BIT(0)
 #define SMBAUXCTL_E32B		BIT(1)
 
-/* Other settings */
-#define MAX_RETRIES		400
-
 /* I801 command constants */
 #define I801_QUICK		0x00
 #define I801_BYTE		0x04
@@ -447,42 +444,35 @@ static int i801_check_post(struct i801_priv *priv, int status)
 /* Wait for BUSY being cleared and either INTR or an error flag being set */
 static int i801_wait_intr(struct i801_priv *priv)
 {
-	int timeout = 0;
-	int status;
+	unsigned long timeout = jiffies + priv->adapter.timeout;
+	int status, busy;
 
-	/* We will always wait for a fraction of a second! */
 	do {
 		usleep_range(250, 500);
 		status = inb_p(SMBHSTSTS(priv));
-	} while (((status & SMBHSTSTS_HOST_BUSY) ||
-		  !(status & (STATUS_ERROR_FLAGS | SMBHSTSTS_INTR))) &&
-		 (timeout++ < MAX_RETRIES));
+		busy = status & SMBHSTSTS_HOST_BUSY;
+		status &= STATUS_ERROR_FLAGS | SMBHSTSTS_INTR;
+		if (!busy && status)
+			return status;
+	} while (time_is_after_eq_jiffies(timeout));
 
-	if (timeout > MAX_RETRIES) {
-		dev_dbg(&priv->pci_dev->dev, "INTR Timeout!\n");
-		return -ETIMEDOUT;
-	}
-	return status & (STATUS_ERROR_FLAGS | SMBHSTSTS_INTR);
+	return -ETIMEDOUT;
 }
 
 /* Wait for either BYTE_DONE or an error flag being set */
 static int i801_wait_byte_done(struct i801_priv *priv)
 {
-	int timeout = 0;
+	unsigned long timeout = jiffies + priv->adapter.timeout;
 	int status;
 
-	/* We will always wait for a fraction of a second! */
 	do {
 		usleep_range(250, 500);
 		status = inb_p(SMBHSTSTS(priv));
-	} while (!(status & (STATUS_ERROR_FLAGS | SMBHSTSTS_BYTE_DONE)) &&
-		 (timeout++ < MAX_RETRIES));
+		if (status & (STATUS_ERROR_FLAGS | SMBHSTSTS_BYTE_DONE))
+			return status & STATUS_ERROR_FLAGS;
+	} while (time_is_after_eq_jiffies(timeout));
 
-	if (timeout > MAX_RETRIES) {
-		dev_dbg(&priv->pci_dev->dev, "BYTE_DONE Timeout!\n");
-		return -ETIMEDOUT;
-	}
-	return status & STATUS_ERROR_FLAGS;
+	return -ETIMEDOUT;
 }
 
 static int i801_transaction(struct i801_priv *priv, int xact)
Signature · 1 line
2.31.1
Reviewed-by Tested-by on v1 1/1
#1 lore raw

Hi Heiner,

Quoted 7 lines · Heiner Kallweit

Polling uses the same timeout as irq mode: 400 * 500us = 200ms = HZ / 5. So let's use the adapter->timeout value also for polling. This has the advantage that userspace can control the timeout value for polling as well. In addition change the code to make it better readable. Last but not least remove the timeout debug messages. Calls to both functions are followed by a call to i801_check_post() that will print an error message in case of timeout.

I think that the intent was to differentiate between regular polling and BYTE_DONE polling. But I agree it's not really important, as the caller probably already knows which type of transaction failed (simple or block).

Quoted 5 lines
---
 drivers/i2c/busses/i2c-i801.c | 36 +++++++++++++----------------------
 1 file changed, 13 insertions(+), 23 deletions(-)

(...)

Signature · 2 lines
Jean Delvare
SUSE L3 Support
Applied
#2 lore raw
In reply to Heiner Kallweit
Quoted 9 lines · Heiner Kallweit

Polling uses the same timeout as irq mode: 400 * 500us = 200ms = HZ / 5. So let's use the adapter->timeout value also for polling. This has the advantage that userspace can control the timeout value for polling as well. In addition change the code to make it better readable. Last but not least remove the timeout debug messages. Calls to both functions are followed by a call to i801_check_post() that will print an error message in case of timeout.

Applied to for-next, thanks!