sh: Fix up and optimize the kmap_coherent() interface.

This fixes up the kmap_coherent/kunmap_coherent() interface for recent
changes both in the page fault path and the shared cache flushers, as
well as adding in some optimizations.

One of the key things to note here is that the TLB flush itself is
deferred until the unmap, and the call in to update_mmu_cache() itself
goes away, relying on the regular page fault path to handle the lazy
dcache writeback if necessary.
Signed-off-by: NPaul Mundt <lethal@linux-sh.org>

arch/sh/include/asm/cacheflush.h

@@ -85,7 +85,7 @@ extern void copy_from_user_page(struct vm_area_struct *vma,
 
 void kmap_coherent_init(void);
 void *kmap_coherent(struct page *page, unsigned long addr);
-void kunmap_coherent(void);
+void kunmap_coherent(void *kvaddr);
 
 #define PG_dcache_dirty	PG_arch_1
 

arch/sh/include/asm/pgtable_32.h

@@ -408,13 +408,19 @@ static inline pte_t pte_modify(pte_t pte, pgprot_t newprot)
 
 /* to find an entry in a page-table-directory. */
 #define pgd_index(address)	(((address) >> PGDIR_SHIFT) & (PTRS_PER_PGD-1))
-#define pgd_offset(mm, address)	((mm)->pgd+pgd_index(address))
+#define pgd_offset(mm, address)	((mm)->pgd + pgd_index(address))
+#define __pgd_offset(address)	pgd_index(address)
 
 /* to find an entry in a kernel page-table-directory */
 #define pgd_offset_k(address)	pgd_offset(&init_mm, address)
 
+#define __pud_offset(address)	(((address) >> PUD_SHIFT) & (PTRS_PER_PUD-1))
+#define __pmd_offset(address)	(((address) >> PMD_SHIFT) & (PTRS_PER_PMD-1))
+
 /* Find an entry in the third-level page table.. */
 #define pte_index(address)	((address >> PAGE_SHIFT) & (PTRS_PER_PTE - 1))
+#define __pte_offset(address)	pte_index(address)
+
 #define pte_offset_kernel(dir, address) \
 	((pte_t *) pmd_page_vaddr(*(dir)) + pte_index(address))
 #define pte_offset_map(dir, address)		pte_offset_kernel(dir, address)

arch/sh/include/asm/pgtable_64.h

@@ -60,6 +60,9 @@ static __inline__ void pmd_set(pmd_t *pmdp,pte_t *ptep)
 /* To find an entry in a kernel PGD. */
 #define pgd_offset_k(address) pgd_offset(&init_mm, address)
 
+#define __pud_offset(address)	(((address) >> PUD_SHIFT) & (PTRS_PER_PUD-1))
+#define __pmd_offset(address)	(((address) >> PMD_SHIFT) & (PTRS_PER_PMD-1))
+
 /*
  * PMD level access routines. Same notes as above.
  */
@@ -80,6 +83,8 @@ static __inline__ void pmd_set(pmd_t *pmdp,pte_t *ptep)
 #define pte_index(address) \
 		((address >> PAGE_SHIFT) & (PTRS_PER_PTE - 1))
 
+#define __pte_offset(address)	pte_index(address)
+
 #define pte_offset_kernel(dir, addr) \
 		((pte_t *) ((pmd_val(*(dir))) & PAGE_MASK) + pte_index((addr)))
 

arch/sh/mm/cache.c

@@ -51,7 +51,7 @@ void copy_to_user_page(struct vm_area_struct *vma, struct page *page,
 	    !test_bit(PG_dcache_dirty, &page->flags)) {
 		void *vto = kmap_coherent(page, vaddr) + (vaddr & ~PAGE_MASK);
 		memcpy(vto, src, len);
-		kunmap_coherent();
+		kunmap_coherent(vto);
 	} else {
 		memcpy(dst, src, len);
 		if (boot_cpu_data.dcache.n_aliases)
@@ -70,7 +70,7 @@ void copy_from_user_page(struct vm_area_struct *vma, struct page *page,
 	    !test_bit(PG_dcache_dirty, &page->flags)) {
 		void *vfrom = kmap_coherent(page, vaddr) + (vaddr & ~PAGE_MASK);
 		memcpy(dst, vfrom, len);
-		kunmap_coherent();
+		kunmap_coherent(vfrom);
 	} else {
 		memcpy(dst, src, len);
 		if (boot_cpu_data.dcache.n_aliases)
@@ -89,7 +89,7 @@ void copy_user_highpage(struct page *to, struct page *from,
 	    !test_bit(PG_dcache_dirty, &from->flags)) {
 		vfrom = kmap_coherent(from, vaddr);
 		copy_page(vto, vfrom);
-		kunmap_coherent();
+		kunmap_coherent(vfrom);
 	} else {
 		vfrom = kmap_atomic(from, KM_USER0);
 		copy_page(vto, vfrom);
@@ -150,7 +150,7 @@ void __flush_anon_page(struct page *page, unsigned long vmaddr)
 
 			kaddr = kmap_coherent(page, vmaddr);
 			__flush_wback_region((void *)kaddr, PAGE_SIZE);
-			kunmap_coherent();
+			kunmap_coherent(kaddr);
 		} else
 			__flush_wback_region((void *)addr, PAGE_SIZE);
 	}

arch/sh/mm/fault_32.c

@@ -82,8 +82,8 @@ static noinline int vmalloc_fault(unsigned long address)
 	pmd_t *pmd_k;
 	pte_t *pte_k;
 
-	/* Make sure we are in vmalloc area: */
-	if (!(address >= VMALLOC_START && address < VMALLOC_END))
+	/* Make sure we are in vmalloc/module/P3 area: */
+	if (!(address >= VMALLOC_START && address < P3_ADDR_MAX))
 		return -1;
 
 	/*

arch/sh/mm/init.c

@@ -106,27 +106,31 @@ void __init page_table_range_init(unsigned long start, unsigned long end,
 	pgd_t *pgd;
 	pud_t *pud;
 	pmd_t *pmd;
-	int pgd_idx;
+	pte_t *pte;
+	int i, j, k;
 	unsigned long vaddr;
 
-	vaddr = start & PMD_MASK;
-	end = (end + PMD_SIZE - 1) & PMD_MASK;
-	pgd_idx = pgd_index(vaddr);
-	pgd = pgd_base + pgd_idx;
-
-	for ( ; (pgd_idx < PTRS_PER_PGD) && (vaddr != end); pgd++, pgd_idx++) {
-		BUG_ON(pgd_none(*pgd));
-		pud = pud_offset(pgd, 0);
-		BUG_ON(pud_none(*pud));
-		pmd = pmd_offset(pud, 0);
-
-		if (!pmd_present(*pmd)) {
-			pte_t *pte_table;
-			pte_table = (pte_t *)alloc_bootmem_low_pages(PAGE_SIZE);
-			pmd_populate_kernel(&init_mm, pmd, pte_table);
+	vaddr = start;
+	i = __pgd_offset(vaddr);
+	j = __pud_offset(vaddr);
+	k = __pmd_offset(vaddr);
+	pgd = pgd_base + i;
+
+	for ( ; (i < PTRS_PER_PGD) && (vaddr != end); pgd++, i++) {
+		pud = (pud_t *)pgd;
+		for ( ; (j < PTRS_PER_PUD) && (vaddr != end); pud++, j++) {
+			pmd = (pmd_t *)pud;
+			for (; (k < PTRS_PER_PMD) && (vaddr != end); pmd++, k++) {
+				if (pmd_none(*pmd)) {
+					pte = (pte_t *) alloc_bootmem_low_pages(PAGE_SIZE);
+					pmd_populate_kernel(&init_mm, pmd, pte);
+					BUG_ON(pte != pte_offset_kernel(pmd, 0));
+				}
+				vaddr += PMD_SIZE;
+			}
+			k = 0;
 		}
-
-		vaddr += PMD_SIZE;
+		j = 0;
 	}
 }
 #endif	/* CONFIG_MMU */
@@ -137,7 +141,7 @@ void __init page_table_range_init(unsigned long start, unsigned long end,
 void __init paging_init(void)
 {
 	unsigned long max_zone_pfns[MAX_NR_ZONES];
-	unsigned long vaddr;
+	unsigned long vaddr, end;
 	int nid;
 
 	/* We don't need to map the kernel through the TLB, as
@@ -155,7 +159,8 @@ void __init paging_init(void)
 	 * pte's will be filled in by __set_fixmap().
 	 */
 	vaddr = __fix_to_virt(__end_of_fixed_addresses - 1) & PMD_MASK;
-	page_table_range_init(vaddr, 0, swapper_pg_dir);
+	end = (FIXADDR_TOP + PMD_SIZE - 1) & PMD_MASK;
+	page_table_range_init(vaddr, end, swapper_pg_dir);
 
 	kmap_coherent_init();
 

arch/sh/mm/kmap.c

@@ -24,9 +24,6 @@ void __init kmap_coherent_init(void)
 {
 	unsigned long vaddr;
 
-	if (!boot_cpu_data.dcache.n_aliases)
-		return;
-
 	/* cache the first coherent kmap pte */
 	vaddr = __fix_to_virt(FIX_CMAP_BEGIN);
 	kmap_coherent_pte = kmap_get_fixmap_pte(vaddr);
@@ -35,30 +32,31 @@ void __init kmap_coherent_init(void)
 void *kmap_coherent(struct page *page, unsigned long addr)
 {
 	enum fixed_addresses idx;
-	unsigned long vaddr, flags;
-	pte_t pte;
+	unsigned long vaddr;
 
 	BUG_ON(test_bit(PG_dcache_dirty, &page->flags));
 
-	inc_preempt_count();
-
-	idx = (addr & current_cpu_data.dcache.alias_mask) >> PAGE_SHIFT;
-	vaddr = __fix_to_virt(FIX_CMAP_END - idx);
-	pte = mk_pte(page, PAGE_KERNEL);
+	pagefault_disable();
 
-	local_irq_save(flags);
-	flush_tlb_one(get_asid(), vaddr);
-	local_irq_restore(flags);
+	idx = FIX_CMAP_END -
+		((addr & current_cpu_data.dcache.alias_mask) >> PAGE_SHIFT);
+	vaddr = __fix_to_virt(idx);
 
-	update_mmu_cache(NULL, vaddr, pte);
-
-	set_pte(kmap_coherent_pte - (FIX_CMAP_END - idx), pte);
+	BUG_ON(!pte_none(*(kmap_coherent_pte - idx)));
+	set_pte(kmap_coherent_pte - idx, mk_pte(page, PAGE_KERNEL));
 
 	return (void *)vaddr;
 }
 
-void kunmap_coherent(void)
+void kunmap_coherent(void *kvaddr)
 {
-	dec_preempt_count();
-	preempt_check_resched();
+	if (kvaddr >= (void *)FIXADDR_START) {
+		unsigned long vaddr = (unsigned long)kvaddr & PAGE_MASK;
+		enum fixed_addresses idx = __virt_to_fix(vaddr);
+
+		pte_clear(&init_mm, vaddr, kmap_coherent_pte - idx);
+		local_flush_tlb_one(get_asid(), vaddr);
+	}
+
+	pagefault_enable();
 }

arch/sh/mm/nommu.c

@@ -81,7 +81,7 @@ void *kmap_coherent(struct page *page, unsigned long addr)
 	return NULL;
 }
 
-void kunmap_coherent(void)
+void kunmap_coherent(void *kvaddr)
 {
 	BUG();
 }