| Message ID | 1491494594.2950.7.camel@linaro.org (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
Hi, On Thu, Apr 06, 2017 at 05:03:14PM +0100, Jon Medhurst (Tixy) wrote: > To cope with the variety in ARM architectures and configurations, the > pagetable attributes for kernel memory are generated at runtime to match > the system the kernel finds itself on. This calculated value is stored > in pgprot_kernel. > > However, when early fixmap support was added for ARM (commit > a5f4c561b3b1) the attributes used for mappings were hard coded because > pgprot_kernel is not set up early enough. Unfortunately, when fixmap is > used after early boot this means the memory being mapped can have > different attributes to existing mappings, potentially leading to > unpredictable behaviour. A specific problem also exists due to the hard > coded values not include the 'shareable' attribute which means on > systems where this matters (e.g. those with multiple CPU clusters) the > cache contents for a memory location can become inconsistent between > CPUs. > > To resolve these issues we change fixmap to use the same memory > attributes (from pgprot_kernel) that the rest of the kernel uses. To > enable this we need to refactor the initialisation code so > build_mem_type_table() is called early enough. Note, that relies on early > param parsing for memory type overrides passed via the kernel command > line, so we need to make sure this call is still after > parse_early_params(). Tested-by: afzal mohammed <afzal.mohd.ma@gmail.com> with an emphasis on no-MMU's Regards afzal
On Sat, 2017-04-08 at 22:21 +0530, afzal mohammed wrote: > Hi, > > On Thu, Apr 06, 2017 at 05:03:14PM +0100, Jon Medhurst (Tixy) wrote: > > To cope with the variety in ARM architectures and configurations, the > > pagetable attributes for kernel memory are generated at runtime to match > > the system the kernel finds itself on. This calculated value is stored > > in pgprot_kernel. > > > > However, when early fixmap support was added for ARM (commit > > a5f4c561b3b1) the attributes used for mappings were hard coded because > > pgprot_kernel is not set up early enough. Unfortunately, when fixmap is > > used after early boot this means the memory being mapped can have > > different attributes to existing mappings, potentially leading to > > unpredictable behaviour. A specific problem also exists due to the hard > > coded values not include the 'shareable' attribute which means on > > systems where this matters (e.g. those with multiple CPU clusters) the > > cache contents for a memory location can become inconsistent between > > CPUs. > > > > To resolve these issues we change fixmap to use the same memory > > attributes (from pgprot_kernel) that the rest of the kernel uses. To > > enable this we need to refactor the initialisation code so > > build_mem_type_table() is called early enough. Note, that relies on early > > param parsing for memory type overrides passed via the kernel command > > line, so we need to make sure this call is still after > > parse_early_params(). > > Tested-by: afzal mohammed <afzal.mohd.ma@gmail.com> > > with an emphasis on no-MMU's Thanks. I've now submitted this new version to the patch tracker... http://www.armlinux.org.uk/developer/patches/viewpatch.php?id=8667/3
diff --git a/arch/arm/include/asm/fixmap.h b/arch/arm/include/asm/fixmap.h index 5c17d2dec777..8f967d1373f6 100644 --- a/arch/arm/include/asm/fixmap.h +++ b/arch/arm/include/asm/fixmap.h @@ -41,7 +41,7 @@ static const enum fixed_addresses __end_of_fixed_addresses = #define FIXMAP_PAGE_COMMON (L_PTE_YOUNG | L_PTE_PRESENT | L_PTE_XN | L_PTE_DIRTY) -#define FIXMAP_PAGE_NORMAL (FIXMAP_PAGE_COMMON | L_PTE_MT_WRITEBACK) +#define FIXMAP_PAGE_NORMAL (pgprot_kernel | L_PTE_XN) #define FIXMAP_PAGE_RO (FIXMAP_PAGE_NORMAL | L_PTE_RDONLY) /* Used by set_fixmap_(io|nocache), both meant for mapping a device */ diff --git a/arch/arm/kernel/setup.c b/arch/arm/kernel/setup.c index f4e54503afa9..32e1a9513dc7 100644 --- a/arch/arm/kernel/setup.c +++ b/arch/arm/kernel/setup.c @@ -80,7 +80,7 @@ __setup("fpe=", fpe_setup); extern void init_default_cache_policy(unsigned long); extern void paging_init(const struct machine_desc *desc); -extern void early_paging_init(const struct machine_desc *); +extern void early_mm_init(const struct machine_desc *); extern void adjust_lowmem_bounds(void); extern enum reboot_mode reboot_mode; extern void setup_dma_zone(const struct machine_desc *desc); @@ -1088,7 +1088,7 @@ void __init setup_arch(char **cmdline_p) parse_early_param(); #ifdef CONFIG_MMU - early_paging_init(mdesc); + early_mm_init(mdesc); #endif setup_dma_zone(mdesc); xen_early_init(); diff --git a/arch/arm/mm/mmu.c b/arch/arm/mm/mmu.c index 4e016d7f37b3..347cca965783 100644 --- a/arch/arm/mm/mmu.c +++ b/arch/arm/mm/mmu.c @@ -414,6 +414,11 @@ void __set_fixmap(enum fixed_addresses idx, phys_addr_t phys, pgprot_t prot) FIXADDR_END); BUG_ON(idx >= __end_of_fixed_addresses); + /* we only support device mappings until pgprot_kernel has been set */ + if (WARN_ON(pgprot_val(prot) != pgprot_val(FIXMAP_PAGE_IO) && + pgprot_val(pgprot_kernel) == 0)) + return; + if (pgprot_val(prot)) set_pte_at(NULL, vaddr, pte, pfn_pte(phys >> PAGE_SHIFT, prot)); @@ -1492,7 +1497,7 @@ pgtables_remap lpae_pgtables_remap_asm; * early_paging_init() recreates boot time page table setup, allowing machines * to switch over to a high (>4G) address space on LPAE systems */ -void __init early_paging_init(const struct machine_desc *mdesc) +static void __init early_paging_init(const struct machine_desc *mdesc) { pgtables_remap *lpae_pgtables_remap; unsigned long pa_pgd; @@ -1560,7 +1565,7 @@ void __init early_paging_init(const struct machine_desc *mdesc) #else -void __init early_paging_init(const struct machine_desc *mdesc) +static void __init early_paging_init(const struct machine_desc *mdesc) { long long offset; @@ -1616,7 +1621,6 @@ void __init paging_init(const struct machine_desc *mdesc) { void *zero_page; - build_mem_type_table(); prepare_page_table(); map_lowmem(); memblock_set_current_limit(arm_lowmem_limit); @@ -1636,3 +1640,9 @@ void __init paging_init(const struct machine_desc *mdesc) empty_zero_page = virt_to_page(zero_page); __flush_dcache_page(NULL, empty_zero_page); } + +void __init early_mm_init(const struct machine_desc *mdesc) +{ + build_mem_type_table(); + early_paging_init(mdesc); +}
To cope with the variety in ARM architectures and configurations, the pagetable attributes for kernel memory are generated at runtime to match the system the kernel finds itself on. This calculated value is stored in pgprot_kernel. However, when early fixmap support was added for ARM (commit a5f4c561b3b1) the attributes used for mappings were hard coded because pgprot_kernel is not set up early enough. Unfortunately, when fixmap is used after early boot this means the memory being mapped can have different attributes to existing mappings, potentially leading to unpredictable behaviour. A specific problem also exists due to the hard coded values not include the 'shareable' attribute which means on systems where this matters (e.g. those with multiple CPU clusters) the cache contents for a memory location can become inconsistent between CPUs. To resolve these issues we change fixmap to use the same memory attributes (from pgprot_kernel) that the rest of the kernel uses. To enable this we need to refactor the initialisation code so build_mem_type_table() is called early enough. Note, that relies on early param parsing for memory type overrides passed via the kernel command line, so we need to make sure this call is still after parse_early_params(). Fixes: a5f4c561b3b1 ("ARM: 8415/1: early fixmap support for earlycon") Cc: <stable@vger.kernel.org> # v4.3+ Signed-off-by: Jon Medhurst <tixy@linaro.org> [ardb: keep early_fixmap_init() before param parsing, for earlycon] Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org> --- v4: - Drop declaration of early_mm_init() from arch/arm/include/asm/mmu.h and make early_paging_init() static. Ard's v3: - Unfortunately, I failed to spot an issue with earlycon when testing Tixy's v2. Earlycon depends on fixmap, but only for device mappings, which are not affected by this change. So instead, preserve the calls to early_fixmap_init() and early_ioremap_init() in their original locations, and only move build_mem_type_table() to the earliest possible moment in boot, which is right after early param parsing (which may set cachepolicy/nocache/nowb variables) arch/arm/include/asm/fixmap.h | 2 +- arch/arm/kernel/setup.c | 4 ++-- arch/arm/mm/mmu.c | 16 +++++++++++++--- 3 files changed, 16 insertions(+), 6 deletions(-)