Re: [tip:x86/urgent] x86, efi: Delete efi_ioremap() and fix CONFIG_X86_32 oops

On Fri, Mar 16, 2012 at 11:36 AM, Matt Fleming [off-list ref] wrote:

On Thu, 2012-03-15 at 10:54 -0700, Yinghai Lu wrote:

quoted

quoted

� � � � � � � �if (end <= max_low_pfn << PAGE_SHIFT)
� � � � � � � � � � � �max_low_pfn_mapped = last_pfn_mapped;

why max_low_pfn is used here?

The idea is that we only want to update max_low_pfn_mapped when we've
mapped a region at or below max_low_pfn. This maintains compatibility
with behaviour prior to this patch.

quoted

quoted

The max_low_pfn checks are only for CONFIG_X86_32 so that the behaviour
is the same as before this patch, i.e. we don't try to map above
max_low_pfn.

ok, to simplify the code, in setup.c you could move
#ifdef CONFIG_X86_64
� � � � if (max_pfn > max_low_pfn) {
� � � � � � � � /* can we preseve max_low_pfn ?*/
� � � � � � � � max_low_pfn = max_pfn;
� � � � }
#endif

before calling new init_memory_mapping()...

so you could remove the #idef. in init_memory_mapping.

Hmm.. if we do this then max_low_pfn_mapped will be set to max_pfn on
CONFIG_X86_64 by the time we've finished looping in
init_memory_mapping(). This is not how things work currently. Will that
cause a problem?

oh, before this patch could be applied, you need to make sure all
max_low_pfn_mapped reference
is removed.
current assumption is under max_low_pfn_mapped all mem is mapped.
now with this patch, that assumption is not right anymore.

Yinghai