Skip to contentSkip to conversation

i2c: i801: Fix block process call transactions

PATCH Applied Jean Delvare 3 participants 5 messages Feb 14, 2024 – Feb 15, 2024

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

PATCH 1/1 i2c: i801: Fix block process call transactions

According to the Intel datasheets, software must reset the block buffer index twice for block process call transactions: once before writing the outgoing data to the buffer, and once again before reading the incoming data from the buffer.

The driver is currently missing the second reset, causing the wrong portion of the block buffer to be read.

Fixes 315cd67c9453 ("i2c: i801: Add Block Write-Block Read Process Call support")

---
Piotr, does this change make your tests succeed?

Alexander, I don't suppose you still have access to the hardware on which you were using block process call transactions back in 2019, but maybe you remember having to do the same change to make it work?

 drivers/i2c/busses/i2c-i801.c |    4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)
drivers/i2c/busses/i2c-i801.c +2 -2
@@ -500,11 +500,10 @@ static int i801_block_transaction_by_blo
 	/* Set block buffer mode */
 	outb_p(inb_p(SMBAUXCTL(priv)) | SMBAUXCTL_E32B, SMBAUXCTL(priv));
 
-	inb_p(SMBHSTCNT(priv)); /* reset the data buffer index */
-
 	if (read_write == I2C_SMBUS_WRITE) {
 		len = data->block[0];
 		outb_p(len, SMBHSTDAT0(priv));
+		inb_p(SMBHSTCNT(priv));	/* reset the data buffer index */
 		for (i = 0; i < len; i++)
 			outb_p(data->block[i+1], SMBBLKDAT(priv));
 	}
@@ -522,6 +521,7 @@ static int i801_block_transaction_by_blo
 		}
 
 		data->block[0] = len;
+		inb_p(SMBHSTCNT(priv));	/* reset the data buffer index */
 		for (i = 0; i < len; i++)
 			data->block[i + 1] = inb_p(SMBBLKDAT(priv));
 	}
Signature · 2 lines
Jean Delvare
SUSE L3 Support
Reviewed-by on v1 1/1
#1 lore raw

Hi Jean,

thanks for looking into this!

Quoted 5 lines · Jean Delvare

According to the Intel datasheets, software must reset the block buffer index twice for block process call transactions: once before writing the outgoing data to the buffer, and once again before reading the incoming data from the buffer.

Quoted 7 lines

The driver is currently missing the second reset, causing the wrong portion of the block buffer to be read.

Fixes 315cd67c9453 ("i2c: i801: Add Block Write-Block Read Process Call support")

According to all the datasheets I've been able to find, the patch makes sense to me.

Quoted 6 lines

---
Piotr, does this change make your tests succeed?

Alexander, I don't suppose you still have access to the hardware on which you were using block process call transactions back in 2019, but maybe you remember having to do the same change to make it work?

Nope, I cannot explain how this went unnoticed... I don't think we had a downstream patch on top back than.

Quoted 27 lines

 drivers/i2c/busses/i2c-i801.c |    4 ++--

 1 file changed, 2 insertions(+), 2 deletions(-)
drivers/i2c/busses/i2c-i801.c +0 -0
@@ -500,11 +500,10 @@ static int i801_block_transaction_by_blo
  	/* Set block buffer mode */
  	outb_p(inb_p(SMBAUXCTL(priv)) | SMBAUXCTL_E32B, SMBAUXCTL(priv));

drivers/i2c/busses/i2c-i801.c +0 -2
-	inb_p(SMBHSTCNT(priv)); /* reset the data buffer index */
-

  if (read_write == I2C_SMBUS_WRITE) {
  len = data->block[0];
  outb_p(len, SMBHSTDAT0(priv));

drivers/i2c/busses/i2c-i801.c +1 -0
+		inb_p(SMBHSTCNT(priv));	/* reset the data buffer index */

  for (i = 0; i < len; i++) outb_p(data->block[i+1], SMBBLKDAT(priv)); }

drivers/i2c/busses/i2c-i801.c +0 -0
@@ -522,6 +521,7 @@ static int i801_block_transaction_by_blo
  		}


  data->block[0] = len;

drivers/i2c/busses/i2c-i801.c +1 -0
+		inb_p(SMBHSTCNT(priv));	/* reset the data buffer index */

  for (i = 0; i < len; i++)
  data->block[i + 1] = inb_p(SMBBLKDAT(priv)); }

Signature · 1 line
Alexander Sverdlin.
maintainer Applied
#2 lore raw
In reply to Jean Delvare

Hi Jean,

Quoted 5 lines · Jean Delvare

According to the Intel datasheets, software must reset the block buffer index twice for block process call transactions: once before writing the outgoing data to the buffer, and once again before reading the incoming data from the buffer.

Quoted 8 lines

The driver is currently missing the second reset, causing the wrong portion of the block buffer to be read.

Fixes 315cd67c9453 ("i2c: i801: Add Block Write-Block Read Process Call support")

---

applied to i2c/i2c-host-fixes.

b4 ty failed to send the thank you letter because there are some addresses that were not accepted ([email protected]?)

Anyway, it's applied :-)

Andi

maintainer
#3 lore raw
Quoted 15 lines · Andi Shyti
Quoted 5 lines · Jean Delvare

According to the Intel datasheets, software must reset the block buffer index twice for block process call transactions: once before writing the outgoing data to the buffer, and once again before reading the incoming data from the buffer.

Quoted 8 lines

The driver is currently missing the second reset, causing the wrong portion of the block buffer to be read.

Fixes 315cd67c9453 ("i2c: i801: Add Block Write-Block Read Process Call support")

---

Quoted 7 lines

applied to i2c/i2c-host-fixes.

b4 ty failed to send the thank you letter because there are some addresses that were not accepted ([email protected]?)

Anyway, it's applied :-)

BTW, thanks for the fast action on this bug report!

Andi

#4 lore raw
In reply to Andi Shyti

Hi Andi,

Quoted 3 lines · Andi Shyti

b4 ty failed to send the thank you letter because there are some addresses that were not accepted ([email protected]?)

Sorry for the mess, this seems to be a bug in my email client (Claws Mail 4.1.1) when the recipient names are quoted and/or include a comma. Ideally my SMTP server should have rejected the message as invalid, but it tried its best to still process it.

I'll report the issue to the Claws Mail developers.

Signature · 2 lines
Jean Delvare
SUSE L3 Support