In message 1208530299-9817-1-git-send-email-sr@denx.de you wrote:
dcache_enable() was missing for 440 and the patch 017e9b7925f74878d0e9475388cca9bda5ef9482 ["allow ports to override bootelf "] behavior uses this function.
Note: Currently the cache handling functions like d/icache_disable/enable() are NOP's on 440. This may be changed in the future.
Sorry, but I don't think this is a good idea. Actually I tend to reject this patch.
--- a/cpu/ppc4xx/cache.S +++ b/cpu/ppc4xx/cache.S @@ -166,9 +166,11 @@ _GLOBAL(invalidate_dcache) #ifdef CONFIG_440
.globl dcache_disable
.globl dcache_enable .globl icache_disable .globl icache_enabledcache_disable: +dcache_enable: icache_disable: icache_enable: blr
I was not are that we have such code in U-Boot; I think it is plain unacceptable. Either we implement these functions correctly, i. e. such that they perform the actions their name suggests, or we do not define these functions at all.
But providing a function that promises to enable or disable the I resp. D cache, and which then actually does nothing, is *extremely* misleading. I see all kind of regressions and lost time because of people debug completely different parts of the code just because they don't know that these functions are dummies.
Please: let's not do this. May I please ask to clean this up?
Yes, I am aware that the current (new) implementation of do_bootelf_exec() needs to be fixed for this, too - and maybe some other places as well. But this is important enough to me.
Thanks.
Best regards,
Wolfgang Denk