Re: [PATCH v14 5/5] x86/boot/KASLR: Limit KASLR to extracting kernel in immovable memory
From: Chao Fan
Date: Mon Dec 17 2018 - 21:50:20 EST
On Mon, Dec 17, 2018 at 06:43:24PM +0100, Ingo Molnar wrote:
>
>* Chao Fan <fanc.fnst@xxxxxxxxxxxxxx> wrote:
>
>> KASLR randomly chooses some positions which may locate in movable
>> memory regions. It will break memory hotplug feature and make the
>> movable memory chosen by KASLR practically immovable.
>>
>> The solution is to limit KASLR to choose memory regions in immovable
>> node according to SRAT tables.
>> When CONFIG_EARLY_PARSE_RSDP is enabled, walk through SRAT to get the
>> information of immovable memory so that KASLR knows where should be
>> chosen for randomization.
>>
>> Rename process_mem_region() as __process_mem_region() and name new
>> function as process_mem_region().
>>
>> Signed-off-by: Chao Fan <fanc.fnst@xxxxxxxxxxxxxx>
>> ---
>> arch/x86/boot/compressed/kaslr.c | 75 +++++++++++++++++++++++++++-----
>> 1 file changed, 64 insertions(+), 11 deletions(-)
>
>Ok, I like this basic approach of automatically detecing memory areas we
>should not KASLR into - it's far better than earlier iterations.
Thanks,
>
>> +++ b/arch/x86/boot/compressed/kaslr.c
>> @@ -97,6 +97,11 @@ static bool memmap_too_large;
>> /* Store memory limit specified by "mem=nn[KMG]" or "memmap=nn[KMG]" */
>> static unsigned long long mem_limit = ULLONG_MAX;
>>
>> +#ifdef CONFIG_EARLY_SRAT_PARSE
>> +/* The immovable memory regions */
>> +extern struct mem_vector immovable_mem[MAX_NUMNODES*2];
>> +#endif
>
>What logic is the maximum size of this array based on?
>
Oh, sorry for that, I ever explained for that, I would add
more comments in next PATCH.
See arch/x86/mm/numa_internal.h:
struct numa_meminfo {
int nr_blks;
struct numa_memblk blk[NR_NODE_MEMBLKS];
};
In arch/x86/include/asm/numa.h:
#define NR_NODE_MEMBLKS (MAX_NUMNODES*2)
That means the memory in one node may be devided into 1 or 2 memory
regions(Also I saw that in the dmesg).
So think about how many regions we need to store the immovable memory.
The worst condition is:
1. There are MAX_NUMANODES nodes on this machine.
2. In SRAT table, every node is devided into 2 memory regions.
3. All of them are immovable.
So MAX_NUMNODES*2 is the biggest amount.
Thanks,
Chao Fan
>Thanks,
>
> Ingo
>
>