Do you mean 0xFFE address? In vendor driver such 'addresses' are used
for delay between writes (50 ms for RTL8812A).

The issue is in the assignment, not the write/read part. rtwn_rf_prog
wants a * uint8_t list, whereas the register size from Linux is a
uint32_t (but can cleanly fit into a uint16_t and might just be the
default register size on Linux). How do I reconcile those two? I hope I
was clearer there.

The first column are the Linux registers in question:

Thank you for your continued assistance.

On 10/01/2017 06:30 AM, Andriy Voskoboinyk wrote:

RF registers are using indirect addressing; you should use
rtwn_rf_write() / rtwn_rf_read() instead.

I am working on porting over a Linux Realtek driver to FreeBSD. I ran
into a register-size issue.

FreeBSD's PCI-write function is defined as follows:
rtwn_pci_write_4(struct rtwn_softc *sc, uint16_t addr)

Notice that the second parameter is of type uint16_t.

During initialization, the rtwn driver uses the rtwn_rf_prog structure
to write a pre-defined list of data to a pre-defined list of registers.
The structure to hold both lists is defined here:

Notice how the second parameter 'reg' is a uint8_t.

The rtwn_pci_write_4's addr is uint16_t, the rtwn_rf_prog's addr is a
uint8_t. How would I reconcile this type mismatch? Additionally, the
Linux version of this block of code has all register values as a
uint32_t. It is not a matter of a cast, because some values definitely
use more than 1 byte (ie 0xFFE).

Suggestions on how to reconcile and resolve this issue?

Farhan Khan
_______________________________________________ mailing list
To unsubscribe, send any mail to
_______________________________________________ mailing list
To unsubscribe, send any mail to ""

Reply via email to