From 063b0a4207e43acbeff3d4b09f43e750e0212b48 Mon Sep 17 00:00:00 2001 From: Russell King Date: Fri, 31 Oct 2008 15:08:35 +0000 Subject: [ARM] copypage: provide our own copy_user_highpage() We used to override the copy_user_page() function. However, this is not only inefficient, it also causes additional complexity for highmem support, since we convert from a struct page to a kernel direct mapped address and back to a struct page again. Moreover, with highmem support, we end up pointlessly setting up kmap entries for pages which we're going to remap. So, push the kmapping down into the copypage implementation files where it's required. Signed-off-by: Russell King --- arch/arm/mm/copypage-v4mc.c | 21 ++++++++++++--------- 1 file changed, 12 insertions(+), 9 deletions(-) (limited to 'arch/arm/mm/copypage-v4mc.c') diff --git a/arch/arm/mm/copypage-v4mc.c b/arch/arm/mm/copypage-v4mc.c index 8d33e25..a7dc838 100644 --- a/arch/arm/mm/copypage-v4mc.c +++ b/arch/arm/mm/copypage-v4mc.c @@ -15,8 +15,8 @@ */ #include #include +#include -#include #include #include #include @@ -33,7 +33,7 @@ static DEFINE_SPINLOCK(minicache_lock); /* - * ARMv4 mini-dcache optimised copy_user_page + * ARMv4 mini-dcache optimised copy_user_highpage * * We flush the destination cache lines just before we write the data into the * corresponding address. Since the Dcache is read-allocate, this removes the @@ -42,7 +42,7 @@ static DEFINE_SPINLOCK(minicache_lock); * * Note: We rely on all ARMv4 processors implementing the "invalidate D line" * instruction. If your processor does not supply this, you have to write your - * own copy_user_page that does the right thing. + * own copy_user_highpage that does the right thing. */ static void __attribute__((naked)) mc_copy_user_page(void *from, void *to) @@ -68,21 +68,24 @@ mc_copy_user_page(void *from, void *to) : "r" (from), "r" (to), "I" (PAGE_SIZE / 64)); } -void v4_mc_copy_user_page(void *kto, const void *kfrom, unsigned long vaddr) +void v4_mc_copy_user_highpage(struct page *from, struct page *to, + unsigned long vaddr) { - struct page *page = virt_to_page(kfrom); + void *kto = kmap_atomic(to, KM_USER1); - if (test_and_clear_bit(PG_dcache_dirty, &page->flags)) - __flush_dcache_page(page_mapping(page), page); + if (test_and_clear_bit(PG_dcache_dirty, &from->flags)) + __flush_dcache_page(page_mapping(from), from); spin_lock(&minicache_lock); - set_pte_ext(TOP_PTE(0xffff8000), pfn_pte(__pa(kfrom) >> PAGE_SHIFT, minicache_pgprot), 0); + set_pte_ext(TOP_PTE(0xffff8000), pfn_pte(page_to_pfn(from), minicache_pgprot), 0); flush_tlb_kernel_page(0xffff8000); mc_copy_user_page((void *)0xffff8000, kto); spin_unlock(&minicache_lock); + + kunmap_atomic(kto, KM_USER1); } /* @@ -113,5 +116,5 @@ v4_mc_clear_user_page(void *kaddr, unsigned long vaddr) struct cpu_user_fns v4_mc_user_fns __initdata = { .cpu_clear_user_page = v4_mc_clear_user_page, - .cpu_copy_user_page = v4_mc_copy_user_page, + .cpu_copy_user_highpage = v4_mc_copy_user_highpage, }; -- cgit v1.1 From 303c6443659bc1dc911356f5de149f48ff1d97b8 Mon Sep 17 00:00:00 2001 From: Russell King Date: Fri, 31 Oct 2008 16:32:19 +0000 Subject: [ARM] clearpage: provide our own clear_user_highpage() For similar reasons as copy_user_page(), we want to avoid the additional kmap_atomic if it's unnecessary. Signed-off-by: Russell King --- arch/arm/mm/copypage-v4mc.c | 28 ++++++++++++++-------------- 1 file changed, 14 insertions(+), 14 deletions(-) (limited to 'arch/arm/mm/copypage-v4mc.c') diff --git a/arch/arm/mm/copypage-v4mc.c b/arch/arm/mm/copypage-v4mc.c index a7dc838..a5eae50 100644 --- a/arch/arm/mm/copypage-v4mc.c +++ b/arch/arm/mm/copypage-v4mc.c @@ -91,30 +91,30 @@ void v4_mc_copy_user_highpage(struct page *from, struct page *to, /* * ARMv4 optimised clear_user_page */ -void __attribute__((naked)) -v4_mc_clear_user_page(void *kaddr, unsigned long vaddr) +void v4_mc_clear_user_highpage(struct page *page, unsigned long vaddr) { - asm volatile( - "str lr, [sp, #-4]!\n\ + void *kaddr = kmap_atomic(page, KM_USER0); + asm volatile("\ mov r1, %0 @ 1\n\ mov r2, #0 @ 1\n\ mov r3, #0 @ 1\n\ mov ip, #0 @ 1\n\ mov lr, #0 @ 1\n\ -1: mcr p15, 0, r0, c7, c6, 1 @ 1 invalidate D line\n\ - stmia r0!, {r2, r3, ip, lr} @ 4\n\ - stmia r0!, {r2, r3, ip, lr} @ 4\n\ - mcr p15, 0, r0, c7, c6, 1 @ 1 invalidate D line\n\ - stmia r0!, {r2, r3, ip, lr} @ 4\n\ - stmia r0!, {r2, r3, ip, lr} @ 4\n\ +1: mcr p15, 0, %0, c7, c6, 1 @ 1 invalidate D line\n\ + stmia %0!, {r2, r3, ip, lr} @ 4\n\ + stmia %0!, {r2, r3, ip, lr} @ 4\n\ + mcr p15, 0, %0, c7, c6, 1 @ 1 invalidate D line\n\ + stmia %0!, {r2, r3, ip, lr} @ 4\n\ + stmia %0!, {r2, r3, ip, lr} @ 4\n\ subs r1, r1, #1 @ 1\n\ - bne 1b @ 1\n\ - ldr pc, [sp], #4" + bne 1b @ 1" : - : "I" (PAGE_SIZE / 64)); + : "r" (kaddr), "I" (PAGE_SIZE / 64) + : "r1", "r2", "r3", "ip", "lr"); + kunmap_atomic(kaddr, KM_USER0); } struct cpu_user_fns v4_mc_user_fns __initdata = { - .cpu_clear_user_page = v4_mc_clear_user_page, + .cpu_clear_user_highpage = v4_mc_clear_user_highpage, .cpu_copy_user_highpage = v4_mc_copy_user_highpage, }; -- cgit v1.1 From 43ae286b7d4d8c4983bc263ef2e3cccc10dedb2b Mon Sep 17 00:00:00 2001 From: Nicolas Pitre Date: Tue, 4 Nov 2008 02:42:27 -0500 Subject: [ARM] fix a couple clear_user_highpage assembly constraints In all cases the kaddr is assigned an input register even though it is modified in the assembly code. Let's assign a new variable to the modified value and mark those inline asm with volatile otherwise they get optimized away because the output variable is otherwise not used. Also fix a few conversion errors in copypage-feroceon.c and copypage-v4mc.c. Signed-off-by: Nicolas Pitre Signed-off-by: Russell King --- arch/arm/mm/copypage-v4mc.c | 8 ++++---- 1 file changed, 4 insertions(+), 4 deletions(-) (limited to 'arch/arm/mm/copypage-v4mc.c') diff --git a/arch/arm/mm/copypage-v4mc.c b/arch/arm/mm/copypage-v4mc.c index a5eae50..bdb5fd9 100644 --- a/arch/arm/mm/copypage-v4mc.c +++ b/arch/arm/mm/copypage-v4mc.c @@ -93,9 +93,9 @@ void v4_mc_copy_user_highpage(struct page *from, struct page *to, */ void v4_mc_clear_user_highpage(struct page *page, unsigned long vaddr) { - void *kaddr = kmap_atomic(page, KM_USER0); + void *ptr, *kaddr = kmap_atomic(page, KM_USER0); asm volatile("\ - mov r1, %0 @ 1\n\ + mov r1, %2 @ 1\n\ mov r2, #0 @ 1\n\ mov r3, #0 @ 1\n\ mov ip, #0 @ 1\n\ @@ -108,8 +108,8 @@ void v4_mc_clear_user_highpage(struct page *page, unsigned long vaddr) stmia %0!, {r2, r3, ip, lr} @ 4\n\ subs r1, r1, #1 @ 1\n\ bne 1b @ 1" - : - : "r" (kaddr), "I" (PAGE_SIZE / 64) + : "=r" (ptr) + : "0" (kaddr), "I" (PAGE_SIZE / 64) : "r1", "r2", "r3", "ip", "lr"); kunmap_atomic(kaddr, KM_USER0); } -- cgit v1.1