[PATCH 1/2] arm64: Fix STRICT_MM_TYPECHECKS errors from pgprot

[Ard Biesheuvel]

On 11 November 2015 at 06:51, Ard Biesheuvel <ard.biesheuvel at linaro.org> wrote:
> Hi Laura,
>
> On 11 November 2015 at 03:03, Laura Abbott <labbott at fedoraproject.org> wrote:
>>
>> Several accesses of pgprot values are incorrect when compiled with
>> STRICT_MM_TYPECHECKS. Use the appropriate pgprot_val/__pgprot wrappers
>> to access the structures appropriately.
>>
>
> I spotted 2 out of these, and Catalin has already queued fixes for
> them (see below)
>
>> Signed-off-by: Laura Abbott <labbott at fedoraproject.org>
>> ---
>> Found while working on the set_memory_* work
>> ---
>>  arch/arm64/mm/mmu.c | 6 +++---
>>  1 file changed, 3 insertions(+), 3 deletions(-)
>>
>> diff --git a/arch/arm64/mm/mmu.c b/arch/arm64/mm/mmu.c
>> index c2fa6b5..83a1162 100644
>> --- a/arch/arm64/mm/mmu.c
>> +++ b/arch/arm64/mm/mmu.c
>> @@ -146,7 +146,7 @@ static void alloc_init_pte(pmd_t *pmd, unsigned long addr,
>>                 if (((addr | next | phys) & ~CONT_MASK) == 0) {
>>                         /* a block of CONT_PTES  */
>>                         __populate_init_pte(pte, addr, next, phys,
>> -                                           prot | __pgprot(PTE_CONT));
>> +                                        __pgprot(pgprot_val(prot) | PTE_CONT));
>
> Got this one
>
>>                 } else {
>>                         /*
>>                          * If the range being split is already inside of a
>> @@ -475,7 +475,7 @@ void mark_rodata_ro(void)
>>  {
>>         create_mapping_late(__pa(_stext), (unsigned long)_stext,
>>                                 (unsigned long)_etext - (unsigned long)_stext,
>> -                               PAGE_KERNEL_EXEC | PTE_RDONLY);
>> +                               __pgprot(pgprot_val(PAGE_KERNEL_EXEC) | PTE_RDONLY));
>>
>
> This needs PAGE_KERNEL_RO (which was just introduced). The reason is
> that PAGE_KERNEL_EXEC has PTE_WRITE set as well, making the range
> writeable under the ARMv8.1 DBM feature, that manages the dirty bit in
> hardware (writing to a page with the PTE_RDONLY and PTE_WRITE bits
> both set will clear the PTE_RDONLY bit in that case)
>

...only you'd obviously need to clear the PTE_PXN bit (or introduce a
new PAGE_KERNEL_xx define?)

>>  }
>>  #endif
>> @@ -691,7 +691,7 @@ void __set_fixmap(enum fixed_addresses idx,
>>  void *__init fixmap_remap_fdt(phys_addr_t dt_phys)
>>  {
>>         const u64 dt_virt_base = __fix_to_virt(FIX_FDT);
>> -       pgprot_t prot = PAGE_KERNEL | PTE_RDONLY;
>> +       pgprot_t prot = __pgprot(pgprot_val(PAGE_KERNEL) | PTE_RDONLY);
>
> Got this one as well (using PAGE_KERNEL_RO)
>
>>         int size, offset;
>>         void *dt_virt;
>>
>> --
>> 2.5.0
>>