Currently kernel direct mappings are created for all pfns between [ 0 to max_low_pfn ) and [ 4GB to max_pfn ). When we introduce memory holes, we end up mapping memory ranges that are not backed by physical DRAM. This is fine for lower memory addresses which can be marked as UC by fixed/variable range MTRRs, however we run in to trouble with high addresses.
The following patchset creates direct mappings only for E820_RAM regions between 0 ~ max_low_pfn and 4GB ~ max_pfn. And leaves non-E820_RAM and memory holes unmapped. This revision of the patchset attempts to resolve comments and concerns from the following threads: https://lkml.org/lkml/2012/8/11/95 and https://lkml.org/lkml/2011/12/16/486 Jacob Shin (5): x86: Only direct map addresses that are marked as E820_RAM x86: find_early_table_space based on memory ranges that are being mapped x86: Keep track of direct mapped pfn ranges x86: Fixup code testing if a pfn is direct mapped x86: Move enabling of PSE and PGE out of init_memory_mapping arch/x86/include/asm/page_types.h | 9 +++ arch/x86/kernel/amd_gart_64.c | 4 +- arch/x86/kernel/cpu/amd.c | 6 +- arch/x86/kernel/setup.c | 118 ++++++++++++++++++++++++++++++++----- arch/x86/mm/init.c | 72 +++++++++++----------- arch/x86/mm/init_64.c | 3 +- arch/x86/platform/efi/efi.c | 8 +-- arch/x86/platform/efi/efi_64.c | 2 + 8 files changed, 157 insertions(+), 65 deletions(-) -- 1.7.9.5 -- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
