Re: [RFC v3 21/21] x86/boot/compressed/64: use 1GB pages for mappings

[Anthony Yznaga <anthony.yznaga@oracle.com>]

On 4/27/23 11:40 AM, H. Peter Anvin wrote:
> On April 26, 2023 5:08:57 PM PDT, Anthony Yznaga <anthony.yznaga@oracle.com> wrote:
>> pkram kaslr code can incur multiple page faults when it walks its
>> preserved ranges list called via mem_avoid_overlap().  The multiple
>> faults can easily end up using up the small number of pages available
>> to be allocated for page table pages.
>>
>> This patch hacks things so that mappings are 1GB which results in the need
>> for far fewer page table pages.  As is this breaks AMD SEV-ES which expects
>> the mappings to be 2M.  This could possibly be fixed by updating split
>> code to split 1GB page if the aren't any other issues with using 1GB
>> mappings.
>>
>> Signed-off-by: Anthony Yznaga <anthony.yznaga@oracle.com>
>> ---
>> arch/x86/boot/compressed/ident_map_64.c | 9 +++++----
>> 1 file changed, 5 insertions(+), 4 deletions(-)
>>
>> diff --git a/arch/x86/boot/compressed/ident_map_64.c b/arch/x86/boot/compressed/ident_map_64.c
>> index 321a5011042d..1e02cf6dda3c 100644
>> --- a/arch/x86/boot/compressed/ident_map_64.c
>> +++ b/arch/x86/boot/compressed/ident_map_64.c
>> @@ -95,8 +95,8 @@ void kernel_add_identity_map(unsigned long start, unsigned long end)
>> 	int ret;
>>
>> 	/* Align boundary to 2M. */
>> -	start = round_down(start, PMD_SIZE);
>> -	end = round_up(end, PMD_SIZE);
>> +	start = round_down(start, PUD_SIZE);
>> +	end = round_up(end, PUD_SIZE);
>> 	if (start >= end)
>> 		return;
>>
>> @@ -120,6 +120,7 @@ void initialize_identity_maps(void *rmode)
>> 	mapping_info.context = &pgt_data;
>> 	mapping_info.page_flag = __PAGE_KERNEL_LARGE_EXEC | sme_me_mask;
>> 	mapping_info.kernpg_flag = _KERNPG_TABLE;
>> +	mapping_info.direct_gbpages = true;
>>
>> 	/*
>> 	 * It should be impossible for this not to already be true,
>> @@ -365,8 +366,8 @@ void do_boot_page_fault(struct pt_regs *regs, unsigned long error_code)
>>
>> 	ghcb_fault = sev_es_check_ghcb_fault(address);
>>
>> -	address   &= PMD_MASK;
>> -	end        = address + PMD_SIZE;
>> +	address   &= PUD_MASK;
>> +	end        = address + PUD_SIZE;
>>
>> 	/*
>> 	 * Check for unexpected error codes. Unexpected are:
> Strong NAK: 1G pages are not supported by all 64-bit CPUs, *and* by your own admission breaks things ...
>
I strongly suspected that this was a no-go. Thank you for taking a 
looking and confirming it. I'll look into alternative solutions.


Anthony