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

[Will Deacon]

On Mon, Jul 04, 2022 at 10:34:07PM +0800, guanghui.fgh wrote:
> Thanks.
> 
> 在 2022/7/4 22:23, Will Deacon 写道:
> > On Mon, Jul 04, 2022 at 10:11:27PM +0800, guanghui.fgh wrote:
> > > 在 2022/7/4 21:15, Will Deacon 写道:
> > > > On Mon, Jul 04, 2022 at 08:05:59PM +0800, guanghui.fgh wrote:
> > > > > > > 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.
> > > > 
> > > There is another reason.
> > > 
> > > When loading crashkernel finish, the do_kexec_load will use
> > > arch_kexec_protect_crashkres to invalid all the pagetable for crashkernel
> > > mem(protect crashkernel mem from access).
> > > 
> > > arch_kexec_protect_crashkres--->set_memory_valid--->...--->apply_to_pmd_range
> > > 
> > > In the apply_to_pmd_range, there is a judement： BUG_ON(pud_huge(*pud)). And
> > > if the crashkernel use block/section mapping, there will be some error.
> > > 
> > > Namely, it's need to use non block/section mapping for crashkernel mem
> > > before shringking.
> > 
> > Well, yes, but we can change arch_kexec_[un]protect_crashkres() not to do
> > that if we're leaving the thing mapped, no?
> > 
> I think we should use arch_kexec_[un]protect_crashkres for crashkernel mem.
> 
> Because when invalid crashkernel mem pagetable, there is no chance to rd/wr
> the crashkernel mem by mistake.
> 
> If we don't use arch_kexec_[un]protect_crashkres to invalid crashkernel mem
> pagetable, there maybe some write operations to these mem by mistake which
> may cause crashkernel boot error and vmcore saving error.

I don't really buy this line of reasoning. The entire main kernel is
writable, so why do we care about protecting the crashkernel so much? The
_code_ to launch the crash kernel is writable! If you care about preventing
writes to memory which should not be writable, then you should use
rodata=full.

Will