On 3/15/2013 6:56 PM, Manjunathappa, Prakash wrote: > Hi Sekhar, > > On Thu, Feb 28, 2013 at 16:09:47, Nori, Sekhar wrote: >> Prakash, >> >> On 2/19/2013 2:02 PM, Manjunathappa, Prakash wrote: >>> DT kernel with latest of denx SPL U-boot boots with garbled UART >>> logs. This is because in U-boot UART2 gets sourced by PLL0_SYSCLK2 >>> configured for 150MHz. But later in kernel UART2 gets mapped to >>> PLL1_SYSCLK2 and is configured for 132MHz not for 150MHz. >>> >>> PLL1 is configured for 264MHz to support mDDR on the EVM. That is >>> memory controller driving mDDR can be configured for 150MHz and >>> mDDR it self can operate at 132MHz. >>> >>> So override UART1 and UART2 DT node clock-frequency property with >>> rate available on da850 EVM. >>> >>> Signed-off-by: Manjunathappa, Prakash <[email protected]> >> >> How about dropping the clock-frequency attribute altogether? of_serial.c >> seems to be falling back on clk apis if frequency is not passed and that >> should make the kernel work with all versions of U-Boot. >> > > Yes it can be dropped by having duplicate clock node for UART having only
You meant an additional clk_lookup entry? That's different from a new 'struct clk' node. > dev_id. This is required because DT kernel does the clk_get with only dev_id > and non-DT kernel only with con_id. In DT kernel clk_get is done from generic > code of_platform_serial_setup:of_serial.c and in non-DT kernel is done from > platform code, davinci_serial_setup_clk:arch/arm/mach-davinci/serial.c > > Please let me know your opinion on this. When only one clock is present, con_id of NULL is more appropriate. How about fixing arch/arm/mach-davinci/serial.c and existing clk_lookup entries to use this? Thanks, Sekhar _______________________________________________ devicetree-discuss mailing list [email protected] https://lists.ozlabs.org/listinfo/devicetree-discuss
