ARM: 9266/1: mm: fix no-MMU ZERO_PAGE() implementation
authorGiulio Benetti <giulio.benetti@benettiengineering.com>
Fri, 4 Nov 2022 20:46:18 +0000 (21:46 +0100)
committerRussell King (Oracle) <rmk+kernel@armlinux.org.uk>
Mon, 7 Nov 2022 14:16:44 +0000 (14:16 +0000)
commit340a982825f76f1cff0daa605970fe47321b5ee7
treecdf55efdf516d84ac2fdee2bb6b1b9a7b2834916
parent612695bccfdbd52004551308a55bae410e7cd22f
ARM: 9266/1: mm: fix no-MMU ZERO_PAGE() implementation

Actually in no-MMU SoCs(i.e. i.MXRT) ZERO_PAGE(vaddr) expands to
```
virt_to_page(0)
```
that in order expands to:
```
pfn_to_page(virt_to_pfn(0))
```
and then virt_to_pfn(0) to:
```
        ((((unsigned long)(0) - PAGE_OFFSET) >> PAGE_SHIFT) +
         PHYS_PFN_OFFSET)
```
where PAGE_OFFSET and PHYS_PFN_OFFSET are the DRAM offset(0x80000000) and
PAGE_SHIFT is 12. This way we obtain 16MB(0x01000000) summed to the base of
DRAM(0x80000000).
When ZERO_PAGE(0) is then used, for example in bio_add_page(), the page
gets an address that is out of DRAM bounds.
So instead of using fake virtual page 0 let's allocate a dedicated
zero_page during paging_init() and assign it to a global 'struct page *
empty_zero_page' the same way mmu.c does and it's the same approach used
in m68k with commit dc068f462179 as discussed here[0]. Then let's move
ZERO_PAGE() definition to the top of pgtable.h to be in common between
mmu.c and nommu.c.

[0]: https://lore.kernel.org/linux-m68k/2a462b23-5b8e-bbf4-ec7d-778434a3b9d7@google.com/T/#m1266ceb63
ad140743174d6b3070364d3c9a5179b

Signed-off-by: Giulio Benetti <giulio.benetti@benettiengineering.com>
Reviewed-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Russell King (Oracle) <rmk+kernel@armlinux.org.uk>
arch/arm/include/asm/pgtable-nommu.h
arch/arm/include/asm/pgtable.h
arch/arm/mm/nommu.c