| Message ID | 20221018222503.90118-1-giulio.benetti@benettiengineering.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | [v2,1/2] ARM: mm: fix no-MMU ZERO_PAGE() implementation | expand |
Hello Arnd, Russell, All, is this patch ok or has it some changes to do? While instead [PATCH 2/2] has a NAK and can be dropped. Best regards
On Wed, Oct 19, 2022, at 00:25, Giulio Benetti wrote: > 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: > ``` > #define virt_to_pfn(0) \ > ((((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/#m1266ceb63ad140743174d6b3070364d3c9a5179b > > Signed-off-by: Giulio Benetti <giulio.benetti@benettiengineering.com> Reviewed-by: Arnd Bergmann <arnd@arndb.de>
On Fri, Nov 4, 2022, at 21:07, Giulio Benetti wrote: > Hello Arnd, Russell, All, > > is this patch ok or has it some changes to do? Looks ok to me, please add it to Russell's patch tracker at: https://www.arm.linux.org.uk/developer/patches/ The patch description could be improved a little ("Link:" tag for the URL, avoiding the wikitext markup, etc), but it's more important to actually get the bug fixed. Arnd
Hi Arnd, On 04/11/22 21:28, Arnd Bergmann wrote: > On Fri, Nov 4, 2022, at 21:07, Giulio Benetti wrote: >> Hello Arnd, Russell, All, >> >> is this patch ok or has it some changes to do? > > Looks ok to me, please add it to Russell's patch > tracker at: > https://www.arm.linux.org.uk/developer/patches/ I've submitted it with your Reviewed-by: https://www.arm.linux.org.uk/developer/patches/viewpatch.php?id=9266/1 > The patch description could be improved a little ("Link:" > tag for the URL, avoiding the wikitext markup, etc), Oh, I was not aware of "Link:" and avoiding wikitext markup. Next time I won't use it and I will use Link instead. I'll grep some other commit log to find out a correct way. > but > it's more important to actually get the bug fixed. Thank you for guiding me Best regards
diff --git a/arch/arm/include/asm/pgtable-nommu.h b/arch/arm/include/asm/pgtable-nommu.h index d16aba48fa0a..090011394477 100644 --- a/arch/arm/include/asm/pgtable-nommu.h +++ b/arch/arm/include/asm/pgtable-nommu.h @@ -44,12 +44,6 @@ typedef pte_t *pte_addr_t; -/* - * ZERO_PAGE is a global shared page that is always zero: used - * for zero-mapped memory areas etc.. - */ -#define ZERO_PAGE(vaddr) (virt_to_page(0)) - /* * Mark the prot value as uncacheable and unbufferable. */ diff --git a/arch/arm/include/asm/pgtable.h b/arch/arm/include/asm/pgtable.h index 78a532068fec..ef48a55e9af8 100644 --- a/arch/arm/include/asm/pgtable.h +++ b/arch/arm/include/asm/pgtable.h @@ -10,6 +10,15 @@ #include <linux/const.h> #include <asm/proc-fns.h> +#ifndef __ASSEMBLY__ +/* + * ZERO_PAGE is a global shared page that is always zero: used + * for zero-mapped memory areas etc.. + */ +extern struct page *empty_zero_page; +#define ZERO_PAGE(vaddr) (empty_zero_page) +#endif + #ifndef CONFIG_MMU #include <asm-generic/pgtable-nopud.h> @@ -139,13 +148,6 @@ extern pgprot_t phys_mem_access_prot(struct file *file, unsigned long pfn, */ #ifndef __ASSEMBLY__ -/* - * ZERO_PAGE is a global shared page that is always zero: used - * for zero-mapped memory areas etc.. - */ -extern struct page *empty_zero_page; -#define ZERO_PAGE(vaddr) (empty_zero_page) - extern pgd_t swapper_pg_dir[PTRS_PER_PGD]; diff --git a/arch/arm/mm/nommu.c b/arch/arm/mm/nommu.c index c42debaded95..c1494a4dee25 100644 --- a/arch/arm/mm/nommu.c +++ b/arch/arm/mm/nommu.c @@ -26,6 +26,13 @@ unsigned long vectors_base; +/* + * empty_zero_page is a special page that is used for + * zero-initialized data and COW. + */ +struct page *empty_zero_page; +EXPORT_SYMBOL(empty_zero_page); + #ifdef CONFIG_ARM_MPU struct mpu_rgn_info mpu_rgn_info; #endif @@ -148,9 +155,21 @@ void __init adjust_lowmem_bounds(void) */ void __init paging_init(const struct machine_desc *mdesc) { + void *zero_page; + early_trap_init((void *)vectors_base); mpu_setup(); + + /* allocate the zero page. */ + zero_page = memblock_alloc(PAGE_SIZE, PAGE_SIZE); + if (!zero_page) + panic("%s: Failed to allocate %lu bytes align=0x%lx\n", + __func__, PAGE_SIZE, PAGE_SIZE); + bootmem_init(); + + empty_zero_page = virt_to_page(zero_page); + flush_dcache_page(empty_zero_page); } /*
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: ``` #define virt_to_pfn(0) \ ((((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/#m1266ceb63ad140743174d6b3070364d3c9a5179b Signed-off-by: Giulio Benetti <giulio.benetti@benettiengineering.com> --- V1->V2: * improve commit log as suggested by Arnd Bergmann --- arch/arm/include/asm/pgtable-nommu.h | 6 ------ arch/arm/include/asm/pgtable.h | 16 +++++++++------- arch/arm/mm/nommu.c | 19 +++++++++++++++++++ 3 files changed, 28 insertions(+), 13 deletions(-)