The driver was clearing the hold bit in the control register before
writing to the address register which resulted in a stop condition
being generated rather than a repeated start.

This issue was only observed when a system was running much
slower than a normal processor would execute.  The IP data sheet
mentions a ordering of writing to the address register before
clearing the hold.

Fixes: df8eb569 ("i2c: Add driver for Cadence I2C controller")
Signed-off-by: John Linn <john.linn@xilinx.com>
Signed-off-by: Paresh Chaudhary <paresh.chaudhary@rockwellcollins.com>
Signed-off-by: Matthew Weber <matthew.weber@rockwellcollins.com>
Signed-off-by: Wolfram Sang <wsa@the-dreams.de>