Messages in this thread | | | Date | Tue, 28 Nov 2023 13:06:56 +0000 | From | "Russell King (Oracle)" <> | Subject | Re: [net-next PATCH v2] net: phy: aquantia: drop wrong endianness conversion for addr and CRC |
| |
On Tue, Nov 28, 2023 at 01:44:39PM +0100, Christian Marangi wrote: > Also the CRC returned from the mailbox CRC has to be converted with > le16_to_cpu since it's LE and won't match on BE system. Am I wrong?
I think you are. As I've said before, everything transferred over the MDIO bus is totally independent of the CPU endianness. Bit 0 in the registers on the PHY will appear in bit 0 of the CPU register, and bit 15 in the registers on the PHY will appear in bit 15 of the CPU register.
If this weren't the case, then if we access the BMCR register, and the BMCR contains 0x1140 (indicating AN is enabled, full duplex, 1000Mbps) then if these were CPU endian-dependent, we'd end up reading 0x4011, and that will break phylib _and_ user applications (these register definitions are exported to userspace.)
-- RMK's Patch system: https://www.armlinux.org.uk/developer/patches/ FTTP is here! 80Mbps down 10Mbps up. Decent connectivity at last!
| |