[PATCH v4] arm64: mm: fix linear mem mapping access performance degradation

[Will Deacon]

On Mon, Jul 04, 2022 at 08:05:59PM +0800, guanghui.fgh wrote:
> 
> 
> 在 2022/7/4 19:14, Will Deacon 写道:
> > On Mon, Jul 04, 2022 at 06:58:20PM +0800, guanghui.fgh wrote:
> > > 
> > > 
> > > 在 2022/7/4 18:35, Will Deacon 写道:
> > > > On Sat, Jul 02, 2022 at 11:57:53PM +0800, Guanghui Feng wrote:
> > > > > The arm64 can build 2M/1G block/sectiion mapping. When using DMA/DMA32 zone
> > > > > (enable crashkernel, disable rodata full, disable kfence), the mem_map will
> > > > > use non block/section mapping(for crashkernel requires to shrink the region
> > > > > in page granularity). But it will degrade performance when doing larging
> > > > > continuous mem access in kernel(memcpy/memmove, etc).
> > > > 
> > > > Hmm. It seems a bit silly to me that we take special care to unmap the
> > > > crashkernel from the linear map even when can_set_direct_map() is false, as
> > > > we won't be protecting the main kernel at all!
> > > > 
> > > > Why don't we just leave the crashkernel mapped if !can_set_direct_map()
> > > > and then this problem just goes away?
> > > 
> > > This question had been asked lask week.
> > 
> > Sorry, I didn't spot that. Please could you link me to the conversation, as
> > I'm still unable to find it in my inbox?
> 
> Please access this link:
> https://lore.kernel.org/linux-arm-kernel/075b0a8e-cb7e-70f6-b45a-54cd31886794@linux.alibaba.com/T/

Sorry, but I read through the thread and I still can't find where the
possibility of leaving the crashkernel mapped was discussed.

> > > 1.Quoted messages from arch/arm64/mm/init.c
> > > 
> > > "Memory reservation for crash kernel either done early or deferred
> > > depending on DMA memory zones configs (ZONE_DMA) --
> > > 
> > > In absence of ZONE_DMA configs arm64_dma_phys_limit initialized
> > > here instead of max_zone_phys().  This lets early reservation of
> > > crash kernel memory which has a dependency on arm64_dma_phys_limit.
> > > Reserving memory early for crash kernel allows linear creation of block
> > > mappings (greater than page-granularity) for all the memory bank rangs.
> > > In this scheme a comparatively quicker boot is observed.
> > > 
> > > If ZONE_DMA configs are defined, crash kernel memory reservation
> > > is delayed until DMA zone memory range size initialization performed in
> > > zone_sizes_init().  The defer is necessary to steer clear of DMA zone
> > > memory range to avoid overlap allocation.
> > > 
> > > [[[
> > > So crash kernel memory boundaries are not known when mapping all bank memory
> > > ranges, which otherwise means not possible to exclude crash kernel range
> > > from creating block mappings so page-granularity mappings are created for
> > > the entire memory range.
> > > ]]]"
> > > 
> > > Namely, the init order: memblock init--->linear mem mapping(4k mapping for
> > > crashkernel, requirinig page-granularity changing))--->zone dma
> > > limit--->reserve crashkernel.
> > > So when enable ZONE DMA and using crashkernel, the mem mapping using 4k
> > > mapping.
> > 
> > Yes, I understand that is how things work today but I'm saying that we may
> > as well leave the crashkernel mapped (at block granularity) if
> > !can_set_direct_map() and then I think your patch becomes a lot simpler.
> 
> But Page-granularity mapppings are necessary for crash kernel memory range
> for shrinking its size via /sys/kernel/kexec_crash_size interfac(Quoted from
> arch/arm64/mm/init.c).
> So this patch split block/section mapping to 4k page-granularity mapping for
> crashkernel mem.

Why? I don't see why the mapping granularity is relevant at all if we
always leave the whole thing mapped.

Will