i2c: designware: fix rx fifo depth tracking
authorRussell King <rmk+kernel@armlinux.org.uk>
Fri, 18 Nov 2016 19:40:10 +0000 (19:40 +0000)
committerWolfram Sang <wsa@the-dreams.de>
Thu, 24 Nov 2016 15:18:15 +0000 (16:18 +0100)
commit4d6d5f1d08d2138dc43b28966eb6200e3db2e623
treeab07faebf8963d5c2d20dd3a0520affa05f7ca0a
parent2bf413d56b7de72ab800a6edb009177e5669b929
i2c: designware: fix rx fifo depth tracking

When loading the TX fifo to receive bytes on the I2C bus, we incorrectly
count the number of bytes:

rx_limit = dev->rx_fifo_depth - dw_readl(dev, DW_IC_RXFLR);

while (buf_len > 0 && tx_limit > 0 && rx_limit > 0) {
if (rx_limit - dev->rx_outstanding <= 0)
break;
rx_limit--;
dev->rx_outstanding++;
}

DW_IC_RXFLR indicates how many bytes are available to be read in the
FIFO, dev->rx_fifo_depth is the FIFO size, and dev->rx_outstanding is
the number of bytes that we've requested to be read so far, but which
have not been read.

Firstly, increasing dev->rx_outstanding and decreasing rx_limit and then
comparing them results in each byte consuming "two" bytes in this
tracking, so this is obviously wrong.

Secondly, the number of bytes that _could_ be received into the FIFO at
any time is the number of bytes we have so far requested but not yet
read from the FIFO - in other words dev->rx_outstanding.

So, in order to request enough bytes to fill the RX FIFO, we need to
request dev->rx_fifo_depth - dev->rx_outstanding bytes.

Modifying the code thusly results in us reaching the maximum number of
bytes outstanding each time we queue more "receive" operations, provided
the transfer allows that to happen.

Signed-off-by: Russell King <rmk+kernel@armlinux.org.uk>
Reviewed-by: Mika Westerberg <mika.westerberg@linux.intel.com>
Acked-by: Jarkko Nikula <jarkko.nikula@linux.intel.com>
Signed-off-by: Wolfram Sang <wsa@the-dreams.de>
drivers/i2c/busses/i2c-designware-core.c