On 11/09/11 20:23, Josh Poimboeuf wrote:
On Wed, 2011-11-09 at 12:03 +0530, Suzuki Poulose wrote:
On Tue, 08 Nov 2011 10:19:05 -0600
Josh Poimboeuf<jpoim...@linux.vnet.ibm.com> wrote:
On Tue, 2011-11-08 at 12:41 +0530, Suzuki Poulose wrote:
What I was suggesting is, instead of flushing the cache in
relocate(), lets do it like:
for e.g, on 440x, (in head_44x.S :)
#ifdef CONFIG_RELOCATABLE
...
bl relocate
#Flush the d-cache and invalidate the i-cache here
#endif
This would let the different platforms do the the cache
invalidation in their own way.
Btw, I didn't find an instruction to flush the entire d-cache in
PPC440 manual. We have instructions to flush only a block
corresponding to an address.
However, we have 'iccci' which would invalidate the entire i-cache
which, which I think is better than 80,000 i-cache invalidates.
In misc_32.S there are already some platform-independent cache
management functions. If we use those, then relocate() could simply
call them. Then the different platforms calling relocate() wouldn't
have to worry about flushing/invalidating caches.
For example, there's a clean_dcache_range() function. Given any range
twice the size of the d-cache, it should flush the entire d-cache.
But the only drawback is that it would require the caller to know the
size of the d-cache.
Instead, I think it would be preferable to create a new clean_dcache()
(or clean_dcache_all()?) function in misc_32.S, which could call
clean_dcache_range() with the appropriate args for flushing the entire
d-cache. relocate() could then call the platform-independent
clean_dcache().
How about using clean_dcache_range() to flush the range runtime
address range [ _stext, _end ] ? That would flush the entire
instructions.
Wouldn't that result in more cache flushing than the original solution?
For example, my kernel is 3.5MB. Assuming a 32 byte cache line size,
clean_dcache_range(_stext, _end) would result in about 115,000 dcbst's
(3.5MB / 32).
Oops ! You are right. We could go back to the clean_dcache_all() or the
initial approach that you suggested. (dcbst).
I am not sure how do we flush the entire dcache(only). Could you post a
patch which does the same ?
Another option is to, change the current mapping to 'Write Through' before
calling the relocate() and revert back to the original setting after relocate().
For i-cache invalidation there's already the (incorrectly named?)
flush_instruction_cache(). It uses the appropriate platform-specific
methods (e.g. iccci for 44x) to invalidate the entire i-cache.
Agreed. The only thing that worries me is the use of KERNELBASE in the
flush_instruction_cache() for CONFIG_4xx. Can we safely assume all 4xx
implementations ignore the arguments passed to iccci ?
Good question. I don't know the answer. :-)
That also may suggest a bigger can of worms. A grep of the powerpc code
shows many uses of KERNELBASE. For a relocatable kernel, nobody should
be relying on KERNELBASE except for the early relocation code. Are we
sure that all the other usages of KERNELBASE are "safe"?
I think we could simply replace the occurrences of KERNELBASE (after the
relocate())
with '_stext' which would give the virtual start address of the kernel.
Thanks
Suzuki
_______________________________________________
Linuxppc-dev mailing list
Linuxppc-dev@lists.ozlabs.org
https://lists.ozlabs.org/listinfo/linuxppc-dev
