[RFC PATCH 2/2] arm64: Implement vmalloc based thread_info allocator

[Jungseok Lee]

On May 26, 2015, at 11:52 AM, yalin wang wrote:
> 2015-05-25 0:02 GMT+08:00 Jungseok Lee <jungseoklee85 at gmail.com>:
>> Fork-routine sometimes fails to get a physically contiguous region for
>> thread_info on 4KB page system although free memory is enough. That is,
>> a physically contiguous region, which is currently 16KB, is not available
>> since system memory is fragmented.
>> 
>> This patch tries to solve the problem as allocating thread_info memory
>> from vmalloc space, not 1:1 mapping one. The downside is one additional
>> page allocation in case of vmalloc. However, vmalloc space is large enough,
>> around 240GB, under a combination of 39-bit VA and 4KB page. Thus, it is
>> not a big tradeoff for fork-routine service.
>> 
>> Suggested-by: Sungjinn Chung <barami97 at gmail.com>
>> Signed-off-by: Jungseok Lee <jungseoklee85 at gmail.com>
>> Cc: Catalin Marinas <catalin.marinas at arm.com>
>> Cc: Will Deacon <will.deacon at arm.com>
>> Cc: linux-kernel at vger.kernel.org
>> Cc: linux-mm at kvack.org
>> ---
>> arch/arm64/Kconfig                   | 12 ++++++++++++
>> arch/arm64/include/asm/thread_info.h |  9 +++++++++
>> arch/arm64/kernel/process.c          |  7 +++++++
>> 3 files changed, 28 insertions(+)
>> 
>> diff --git a/arch/arm64/Kconfig b/arch/arm64/Kconfig
>> index 99930cf..93c236a 100644
>> --- a/arch/arm64/Kconfig
>> +++ b/arch/arm64/Kconfig
>> @@ -536,6 +536,18 @@ config ARCH_SELECT_MEMORY_MODEL
>> config HAVE_ARCH_PFN_VALID
>>        def_bool ARCH_HAS_HOLES_MEMORYMODEL || !SPARSEMEM
>> 
>> +config ARCH_THREAD_INFO_ALLOCATOR
>> +       bool "Enable vmalloc based thread_info allocator (EXPERIMENTAL)"
>> +       depends on ARM64_4K_PAGES
>> +       default n
>> +       help
>> +         This feature enables vmalloc based thread_info allocator. It
>> +         prevents fork-routine from begin failed to obtain physically
>> +         contiguour region due to memory fragmentation on low system
>> +         memory platforms.
>> +
>> +         If unsure, say N
>> +
>> config HW_PERF_EVENTS
>>        bool "Enable hardware performance counter support for perf events"
>>        depends on PERF_EVENTS
>> diff --git a/arch/arm64/include/asm/thread_info.h b/arch/arm64/include/asm/thread_info.h
>> index dcd06d1..e753e59 100644
>> --- a/arch/arm64/include/asm/thread_info.h
>> +++ b/arch/arm64/include/asm/thread_info.h
>> @@ -61,6 +61,15 @@ struct thread_info {
>> #define init_thread_info       (init_thread_union.thread_info)
>> #define init_stack             (init_thread_union.stack)
>> 
>> +#ifdef CONFIG_ARCH_THREAD_INFO_ALLOCATOR
>> +#define alloc_thread_info_node(tsk, node)                              \
>> +({                                                                     \
>> +       __vmalloc_node_range(THREAD_SIZE, THREAD_SIZE, VMALLOC_START,   \
>> +                       VMALLOC_END, GFP_KERNEL, PAGE_KERNEL, 0,        \
>> +                       NUMA_NO_NODE, __builtin_return_address(0));     \
>> +})
> why not add __GFP_HIGHMEM, if you decided to use vmalloc() alloc stack pages?

I do not add the flag since there is no high memory on a current ARM64 kernel. 

It would be helpful to review include/linux/gfp.h and the following code snippet
from arch/arm64/kernel/module.c.

void *module_alloc(unsigned long size)
{
	return __vmalloc_node_range(size, 1, MODULES_VADDR, MODULES_END,
				    GFP_KERNEL, PAGE_KERNEL_EXEC, 0,
				    NUMA_NO_NODE, __builtin_return_address(0));
}

Best Regards
Jungseok Lee