I am using the Quad SPI peripheral of the STM32F767ZIT6.
After I played around with it and watched the signals with an logic analyzer and oscilloscope, I noticed something strange.
After the dummy cycles phase the first nibble of the first data byte gets on the data lines with the rising clock edge.
But since I am in SPI mode 0, the data lines should have the right state some time before the rising edge.
Thus there is a high chance that the first data nibble gets received corrupted by the slave.
After looking into the errata it became more clear whats going on.
However, the workaround is not really helpful.
The errata recommends to use the alternate byte phase instead of the dummy phase.
But the alternate byte phase drives the data lines to an specific state instead of high impedance used in the dummy cycle phase.
So there is an chance of hardware damage in case the master switches from transmitting to receiving mode and for a moment both trying to drive the data lines to an specific state.
Is there a better workaround?
Is there an alternative STM32 device with fully working Quad SPI?
I explained the problem with more details on StackExchange but I got no solutions so far.