Re: [PATCH 04/19] LoongArch: Add common headers

On Sat, Aug 14, 2021 at 4:50 AM Huacai Chen [off-list ref] wrote:

On Fri, Aug 13, 2021 at 5:08 PM Arnd Bergmann [off-list ref] wrote:

quoted

After some thinking, I found that ARM64 is "define kernel VABITS
depends on page table layout", and MIPS (also LoongArch) is "define
page table layout depends on kernel VABITS". So you can see:

#ifdef CONFIG_VA_BITS_40
#ifdef CONFIG_PAGE_SIZE_4KB
#define PGD_ORDER               1
#define PUD_ORDER               aieeee_attempt_to_allocate_pud
#define PMD_ORDER               0
#define PTE_ORDER               0
#endif

I have no idea what aieeee_attempt_to_allocate_pud means, but
this part seems fine.

#ifdef CONFIG_PAGE_SIZE_16KB
#define PGD_ORDER               0
#define PUD_ORDER               aieeee_attempt_to_allocate_pud
#define PMD_ORDER               0
#define PTE_ORDER               0
#endif

This doesn't seem to make sense at all however: it looks like you have
three levels of 16KB page tables, so you get 47 bits, not 40. This
means you waste 99% of the available address space when you
run this kernel on a CPU that is able to access the entire space.

#ifdef CONFIG_PAGE_SIZE_64KB
#define PGD_ORDER               0
#define PUD_ORDER               aieeee_attempt_to_allocate_pud
#define PMD_ORDER               aieeee_attempt_to_allocate_pmd
#define PTE_ORDER               0
#endif
#endif

Similarly, here it seems you get 42 bits.

#ifdef CONFIG_VA_BITS_48
#ifdef CONFIG_PAGE_SIZE_4KB
#define PGD_ORDER               0
#define PUD_ORDER               0
#define PMD_ORDER               0
#define PTE_ORDER               0
#endif
#ifdef CONFIG_PAGE_SIZE_16KB
#define PGD_ORDER               1
#define PUD_ORDER               aieeee_attempt_to_allocate_pud
#define PMD_ORDER               0
#define PTE_ORDER               0
#endif

This again looks reasonable, though I don't see why you care about
having the extra pgd_order here, instead of just going with
47 bits.

#ifdef CONFIG_PAGE_SIZE_64KB
#define PGD_ORDER               0
#define PUD_ORDER               aieeee_attempt_to_allocate_pud
#define PMD_ORDER               0
#define PTE_ORDER               0
#endif
#endif

I suppose you can't ever have more than 48 bits? Otherwise this option
would give you 55 bits of address space.

Since 40 and 48 is the most popular VABITS of LoongArch hardware, and
LoongArch has a software-managed TLB, it seems "define page table
layout depends on kernel VABITS" is more natural for LoongArch.

How common are Loongarch64 CPUs that limit the virtual address space
to 40 bits instead of the full 48 bits? What is the purpose of limiting the
CPU this way?

       Arnd