Hi Andreas, On Fri, Apr 02, 2010 at 14:28:15, Andreas Auer wrote: > Steve Chen schrieb: > > On Wed, Mar 31, 2010 at 2:44 AM, Andreas Auer <[email protected]> wrote: > >> Hello, > >> > >> I'm using a customized board similar to the EVM with a DM6446 cpu. The main > >> difference is that I'm using a Micrel KSZ8893M ethernet switch. The > >> ethernet > >> switch has an I2C config interface and the standard mdio interface. > >> Therefore, I'm using the DSA driver subsystem and wrote a small chip driver > >> (I took the KSZ8893 driver for the Blackfin processor as reference). > >> Now, my problem is that the ethernet mac of the Davinci cpu is initialized > >> (probed) later then the DSA driver. And therefore the DSA driver cannot > >> find > >> a valid netdevice. > >> Does anyone know how to change the initialization order of the devices?? > >> Changing the order of the register_device_driver function calls has > >> basically no influence on the init sequence. > >> > >> Thanks, > >> Andreas > > > > Hello, > > > > I encounter the same issue with DA830/OMAP-l137 EVM. Instead of > > probing PHY to get speed and duplexity, the driver just set the speed > > and duplexity based on platform data. May want to take a look at the > > OMAP-l137. > > Thanks for your answer. I got another solution. The davinci_emac driver > calls the init function very late because of > late_initcall(davinci_emac_init); > > I changed this line: > module_init(davinci_emac_init);
This change was made so EMAC can get initialized after the SPI subsystem (so mac address could be fetched from SPI flash). Seems like there are one too many initialization order dependencies around EMAC driver. Is it possible to get your DSA driver to be a late_initcall too? Within the same initcall level the order is determined by the order In which the makefile lists the .o file. Thanks, Sekhar _______________________________________________ Davinci-linux-open-source mailing list [email protected] http://linux.davincidsp.com/mailman/listinfo/davinci-linux-open-source
