[ARM] Add section support to ioremap
Allow section mappings to be setup using ioremap() and torn down
with iounmap(). This requires additional support in the MM
context switch to ensure that mappings are properly synchronised
when mapped in.
Based an original implementation by Deepak Saxena, reworked and
ARMv6 support added by rmk.
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
diff --git a/arch/arm/mm/ioremap.c b/arch/arm/mm/ioremap.c
index 7691cfd..6aa13d5 100644
--- a/arch/arm/mm/ioremap.c
+++ b/arch/arm/mm/ioremap.c
@@ -27,7 +27,16 @@
#include <asm/cacheflush.h>
#include <asm/io.h>
+#include <asm/mmu_context.h>
+#include <asm/pgalloc.h>
#include <asm/tlbflush.h>
+#include <asm/sizes.h>
+
+/*
+ * Used by ioremap() and iounmap() code to mark section-mapped I/O regions
+ * in vm_struct->flags field.
+ */
+#define VM_ARM_SECTION_MAPPING 0x80000000
static inline void
remap_area_pte(pte_t * pte, unsigned long address, unsigned long size,
@@ -113,10 +122,120 @@
dir++;
} while (address && (address < end));
- flush_cache_vmap(start, end);
return err;
}
+
+void __check_kvm_seq(struct mm_struct *mm)
+{
+ unsigned int seq;
+
+ do {
+ seq = init_mm.context.kvm_seq;
+ memcpy(pgd_offset(mm, VMALLOC_START),
+ pgd_offset_k(VMALLOC_START),
+ sizeof(pgd_t) * (pgd_index(VMALLOC_END) -
+ pgd_index(VMALLOC_START)));
+ mm->context.kvm_seq = seq;
+ } while (seq != init_mm.context.kvm_seq);
+}
+
+#ifndef CONFIG_SMP
+/*
+ * Section support is unsafe on SMP - If you iounmap and ioremap a region,
+ * the other CPUs will not see this change until their next context switch.
+ * Meanwhile, (eg) if an interrupt comes in on one of those other CPUs
+ * which requires the new ioremap'd region to be referenced, the CPU will
+ * reference the _old_ region.
+ *
+ * Note that get_vm_area() allocates a guard 4K page, so we need to mask
+ * the size back to 1MB aligned or we will overflow in the loop below.
+ */
+static void unmap_area_sections(unsigned long virt, unsigned long size)
+{
+ unsigned long addr = virt, end = virt + (size & ~SZ_1M);
+ pgd_t *pgd;
+
+ flush_cache_vunmap(addr, end);
+ pgd = pgd_offset_k(addr);
+ do {
+ pmd_t pmd, *pmdp = pmd_offset(pgd, addr);
+
+ pmd = *pmdp;
+ if (!pmd_none(pmd)) {
+ /*
+ * Clear the PMD from the page table, and
+ * increment the kvm sequence so others
+ * notice this change.
+ *
+ * Note: this is still racy on SMP machines.
+ */
+ pmd_clear(pmdp);
+ init_mm.context.kvm_seq++;
+
+ /*
+ * Free the page table, if there was one.
+ */
+ if ((pmd_val(pmd) & PMD_TYPE_MASK) == PMD_TYPE_TABLE)
+ pte_free_kernel(pmd_page_kernel(pmd));
+ }
+
+ addr += PGDIR_SIZE;
+ pgd++;
+ } while (addr < end);
+
+ /*
+ * Ensure that the active_mm is up to date - we want to
+ * catch any use-after-iounmap cases.
+ */
+ if (current->active_mm->context.kvm_seq != init_mm.context.kvm_seq)
+ __check_kvm_seq(current->active_mm);
+
+ flush_tlb_kernel_range(virt, end);
+}
+
+static int
+remap_area_sections(unsigned long virt, unsigned long pfn,
+ unsigned long size, unsigned long flags)
+{
+ unsigned long prot, addr = virt, end = virt + size;
+ pgd_t *pgd;
+
+ /*
+ * Remove and free any PTE-based mapping, and
+ * sync the current kernel mapping.
+ */
+ unmap_area_sections(virt, size);
+
+ prot = PMD_TYPE_SECT | PMD_SECT_AP_WRITE | PMD_DOMAIN(DOMAIN_IO) |
+ (flags & (L_PTE_CACHEABLE | L_PTE_BUFFERABLE));
+
+ /*
+ * ARMv6 and above need XN set to prevent speculative prefetches
+ * hitting IO.
+ */
+ if (cpu_architecture() >= CPU_ARCH_ARMv6)
+ prot |= PMD_SECT_XN;
+
+ pgd = pgd_offset_k(addr);
+ do {
+ pmd_t *pmd = pmd_offset(pgd, addr);
+
+ pmd[0] = __pmd(__pfn_to_phys(pfn) | prot);
+ pfn += SZ_1M >> PAGE_SHIFT;
+ pmd[1] = __pmd(__pfn_to_phys(pfn) | prot);
+ pfn += SZ_1M >> PAGE_SHIFT;
+ flush_pmd_entry(pmd);
+
+ addr += PGDIR_SIZE;
+ pgd++;
+ } while (addr < end);
+
+ return 0;
+}
+#endif
+
+
/*
* Remap an arbitrary physical address space into the kernel virtual
* address space. Needed when the kernel wants to access high addresses
@@ -133,6 +252,7 @@
__ioremap_pfn(unsigned long pfn, unsigned long offset, size_t size,
unsigned long flags)
{
+ int err;
unsigned long addr;
struct vm_struct * area;
@@ -140,11 +260,22 @@
if (!area)
return NULL;
addr = (unsigned long)area->addr;
- if (remap_area_pages(addr, pfn, size, flags)) {
+
+#ifndef CONFIG_SMP
+ if (!((__pfn_to_phys(pfn) | size | addr) & ~PMD_MASK)) {
+ area->flags |= VM_ARM_SECTION_MAPPING;
+ err = remap_area_sections(addr, pfn, size, flags);
+ } else
+#endif
+ err = remap_area_pages(addr, pfn, size, flags);
+
+ if (err) {
vunmap((void *)addr);
return NULL;
}
- return (void __iomem *) (offset + (char *)addr);
+
+ flush_cache_vmap(addr, addr + size);
+ return (void __iomem *) (offset + addr);
}
EXPORT_SYMBOL(__ioremap_pfn);
@@ -173,6 +304,34 @@
void __iounmap(void __iomem *addr)
{
- vunmap((void *)(PAGE_MASK & (unsigned long)addr));
+ struct vm_struct **p, *tmp;
+ unsigned int section_mapping = 0;
+
+ addr = (void __iomem *)(PAGE_MASK & (unsigned long)addr);
+
+ /*
+ * If this is a section based mapping we need to handle it
+ * specially as the VM subysystem does not know how to handle
+ * such a beast. We need the lock here b/c we need to clear
+ * all the mappings before the area can be reclaimed
+ * by someone else.
+ */
+ write_lock(&vmlist_lock);
+ for (p = &vmlist ; (tmp = *p) ; p = &tmp->next) {
+ if((tmp->flags & VM_IOREMAP) && (tmp->addr == addr)) {
+ if (tmp->flags & VM_ARM_SECTION_MAPPING) {
+ *p = tmp->next;
+ unmap_area_sections((unsigned long)tmp->addr,
+ tmp->size);
+ kfree(tmp);
+ section_mapping = 1;
+ }
+ break;
+ }
+ }
+ write_unlock(&vmlist_lock);
+
+ if (!section_mapping)
+ vunmap(addr);
}
EXPORT_SYMBOL(__iounmap);
diff --git a/include/asm-arm/memory.h b/include/asm-arm/memory.h
index 94f973b7..176a4fb 100644
--- a/include/asm-arm/memory.h
+++ b/include/asm-arm/memory.h
@@ -68,6 +68,11 @@
*/
#define XIP_VIRT_ADDR(physaddr) (MODULE_START + ((physaddr) & 0x000fffff))
+/*
+ * Allow 2MB-aligned ioremap pages
+ */
+#define IOREMAP_MAX_ORDER 21
+
#else /* CONFIG_MMU */
/*
diff --git a/include/asm-arm/mmu.h b/include/asm-arm/mmu.h
index 23dde52..fe2a23b 100644
--- a/include/asm-arm/mmu.h
+++ b/include/asm-arm/mmu.h
@@ -7,6 +7,7 @@
#if __LINUX_ARM_ARCH__ >= 6
unsigned int id;
#endif
+ unsigned int kvm_seq;
} mm_context_t;
#if __LINUX_ARM_ARCH__ >= 6
diff --git a/include/asm-arm/mmu_context.h b/include/asm-arm/mmu_context.h
index 9fadb01..d1a65b1 100644
--- a/include/asm-arm/mmu_context.h
+++ b/include/asm-arm/mmu_context.h
@@ -17,6 +17,8 @@
#include <asm/cacheflush.h>
#include <asm/proc-fns.h>
+void __check_kvm_seq(struct mm_struct *mm);
+
#if __LINUX_ARM_ARCH__ >= 6
/*
@@ -45,13 +47,21 @@
{
if (unlikely((mm->context.id ^ cpu_last_asid) >> ASID_BITS))
__new_context(mm);
+
+ if (unlikely(mm->context.kvm_seq != init_mm.context.kvm_seq))
+ __check_kvm_seq(mm);
}
#define init_new_context(tsk,mm) (__init_new_context(tsk,mm),0)
#else
-#define check_context(mm) do { } while (0)
+static inline void check_context(struct mm_struct *mm)
+{
+ if (unlikely(mm->context.kvm_seq != init_mm.context.kvm_seq))
+ __check_kvm_seq(mm);
+}
+
#define init_new_context(tsk,mm) 0
#endif
