Merged the tlb.c changes from the skas patch.

arch/um/kernel/skas/Makefile

@@ -4,7 +4,7 @@
 #
 
 obj-y = exec_kern.o exec_user.o mem.o mem_user.o mmu.o process.o \
-	process_kern.o syscall_kern.o syscall_user.o time.o trap_user.o \
+	process_kern.o syscall_kern.o syscall_user.o time.o tlb.o trap_user.o \
 	sys-$(SUBARCH)/
 
 USER_OBJS = $(filter %_user.o,$(obj-y)) process.o time.o

arch/um/kernel/skas/tlb.c

+/* 
+ * Copyright (C) 2002 Jeff Dike (jdike@karaya.com)
+ * Licensed under the GPL
+ */
+
+#include "linux/stddef.h"
+#include "linux/sched.h"
+#include "asm/page.h"
+#include "asm/pgtable.h"
+#include "asm/mmu.h"
+#include "user_util.h"
+#include "mem_user.h"
+#include "skas.h"
+#include "os.h"
+
+static void fix_range(struct mm_struct *mm, unsigned long start_addr,
+		      unsigned long end_addr, int force)
+{
+	pgd_t *npgd;
+	pmd_t *npmd;
+	pte_t *npte;
+	unsigned long addr;
+	int r, w, x, err, fd;
+
+	if(mm == NULL) return;
+	fd = mm->context.skas.mm_fd;
+	for(addr = start_addr; addr < end_addr;){
+		npgd = pgd_offset(mm, addr);
+		npmd = pmd_offset(npgd, addr);
+		if(pmd_present(*npmd)){
+			npte = pte_offset(npmd, addr);
+			r = pte_read(*npte);
+			w = pte_write(*npte);
+			x = pte_exec(*npte);
+			if(!pte_dirty(*npte)) w = 0;
+			if(!pte_young(*npte)){
+				r = 0;
+				w = 0;
+			}
+			if(force || pte_newpage(*npte)){
+				err = unmap(fd, (void *) addr, PAGE_SIZE);
+				if(err < 0)
+					panic("munmap failed, errno = %d\n",
+					      -err);
+				if(pte_present(*npte))
+					map(fd, addr, 
+					    pte_val(*npte) & PAGE_MASK,
+					    PAGE_SIZE, r, w, x);
+			}
+			else if(pte_newprot(*npte)){
+				protect(fd, addr, PAGE_SIZE, r, w, x, 1);
+			}
+			*npte = pte_mkuptodate(*npte);
+			addr += PAGE_SIZE;
+		}
+		else {
+			if(force || pmd_newpage(*npmd)){
+				err = unmap(fd, (void *) addr, PMD_SIZE);
+				if(err < 0)
+					panic("munmap failed, errno = %d\n",
+					      -err);
+				pmd_mkuptodate(*npmd);
+			}
+			addr += PMD_SIZE;
+		}
+	}
+}
+
+static void flush_kernel_vm_range(unsigned long start, unsigned long end)
+{
+	struct mm_struct *mm;
+	pgd_t *pgd;
+	pmd_t *pmd;
+	pte_t *pte;
+	unsigned long addr;
+	int updated = 0, err;
+
+	mm = &init_mm;
+	for(addr = start_vm; addr < end_vm;){
+		pgd = pgd_offset(mm, addr);
+		pmd = pmd_offset(pgd, addr);
+		if(pmd_present(*pmd)){
+			pte = pte_offset(pmd, addr);
+			if(!pte_present(*pte) || pte_newpage(*pte)){
+				updated = 1;
+				err = os_unmap_memory((void *) addr, 
+						      PAGE_SIZE);
+				if(err < 0)
+					panic("munmap failed, errno = %d\n",
+					      -err);
+				if(pte_present(*pte))
+					map_memory(addr, 
+						   pte_val(*pte) & PAGE_MASK,
+						   PAGE_SIZE, 1, 1, 1);
+			}
+			else if(pte_newprot(*pte)){
+				updated = 1;
+				protect_memory(addr, PAGE_SIZE, 1, 1, 1, 1);
+			}
+			addr += PAGE_SIZE;
+
+		}
+		else {
+			if(pmd_newpage(*pmd)){
+				updated = 1;
+				err = os_unmap_memory((void *) addr, PMD_SIZE);
+				if(err < 0)
+					panic("munmap failed, errno = %d\n",
+					      -err);
+			}
+			addr += PMD_SIZE;
+		}
+	}
+}
+
+void flush_tlb_kernel_vm_skas(void)
+{
+	flush_kernel_vm_range(start_vm, end_vm);
+}
+
+void __flush_tlb_one_skas(unsigned long addr)
+{
+	flush_kernel_vm_range(addr, addr + PAGE_SIZE);
+}
+
+void flush_tlb_range_skas(struct mm_struct *mm, unsigned long start, 
+		     unsigned long end)
+{
+	if(mm == NULL)
+		flush_kernel_vm_range(start, end);
+	else fix_range(mm, start, end, 0);
+}
+
+void flush_tlb_mm_skas(struct mm_struct *mm)
+{
+	if(mm == NULL)
+		flush_tlb_kernel_vm_skas();
+	else fix_range(mm, 0, host_task_size, 0);
+}
+
+void force_flush_all_skas(void)
+{
+	fix_range(current->mm, 0, host_task_size, 1);
+}
+
+/*
+ * Overrides for Emacs so that we follow Linus's tabbing style.
+ * Emacs will notice this stuff at the end of the file and automatically
+ * adjust the settings for this buffer only.  This must remain at the end
+ * of the file.
+ * ---------------------------------------------------------------------------
+ * Local variables:
+ * c-file-style: "linux"
+ * End:
+ */

arch/um/kernel/tlb.c

 /* 
- * Copyright (C) 2000 Jeff Dike (jdike@karaya.com)
+ * Copyright (C) 2000, 2001, 2002 Jeff Dike (jdike@karaya.com)
  * Licensed under the GPL
  */
 
-#include "linux/sched.h"
-#include "linux/slab.h"
-#include "linux/bootmem.h"
+#include "linux/mm.h"
+#include "asm/page.h"
 #include "asm/pgalloc.h"
-#include "asm-generic/tlb.h"
-#include "asm/pgtable.h"
-#include "asm/a.out.h"
-#include "asm/processor.h"
-#include "asm/mmu_context.h"
-#include "asm/uaccess.h"
-#include "asm/atomic.h"
-#include "mem_user.h"
-#include "user_util.h"
-#include "kern_util.h"
-#include "kern.h"
-#include "tlb.h"
-#include "os.h"
-
-static void fix_range(struct mm_struct *mm, unsigned long start_addr, 
-		      unsigned long end_addr, int force)
-{
-	pgd_t *npgd;
-	pmd_t *npmd;
-	pte_t *npte;
-	unsigned long addr;
-	int r, w, x, err;
-
-	if((current->thread.mode.tt.extern_pid != -1) && 
-	   (current->thread.mode.tt.extern_pid != os_getpid()))
-		panic("fix_range fixing wrong address space, current = 0x%p",
-		      current);
-	if(mm == NULL) return;
-	for(addr=start_addr;addr<end_addr;){
-		if(addr == TASK_SIZE){
-			/* Skip over kernel text, kernel data, and physical
-			 * memory, which don't have ptes, plus kernel virtual
-			 * memory, which is flushed separately, and remap
-			 * the process stack.  The only way to get here is
-			 * if (end_addr == STACK_TOP) > TASK_SIZE, which is
-			 * only true in the honeypot case.
-			 */
-			addr = STACK_TOP - ABOVE_KMEM;
-			continue;
-		}
-		npgd = pgd_offset(mm, addr);
-		npmd = pmd_offset(npgd, addr);
-		if(pmd_present(*npmd)){
-			npte = pte_offset_kernel(npmd, addr);
-			r = pte_read(*npte);
-			w = pte_write(*npte);
-			x = pte_exec(*npte);
-			if(!pte_dirty(*npte)) w = 0;
-			if(!pte_young(*npte)){
-				r = 0;
-				w = 0;
-			}
-			if(force || pte_newpage(*npte)){
-				err = os_unmap_memory((void *) addr, 
-						      PAGE_SIZE);
-				if(err < 0)
-					panic("munmap failed, errno = %d\n",
-					      -err);
-				if(pte_present(*npte))
-					map_memory(addr, 
-						   pte_val(*npte) & PAGE_MASK,
-						   PAGE_SIZE, r, w, x);
-			}
-			else if(pte_newprot(*npte)){
-				protect_memory(addr, PAGE_SIZE, r, w, x, 1);
-			}
-			*npte = pte_mkuptodate(*npte);
-			addr += PAGE_SIZE;
-		}
-		else {
-			if(force || pmd_newpage(*npmd)){
-				err = os_unmap_memory((void *) addr, PMD_SIZE);
-				if(err < 0)
-					panic("munmap failed, errno = %d\n",
-					      -err);
-				pmd_mkuptodate(*npmd);
-			}
-			addr += PMD_SIZE;
-		}
-	}
-}
-
-atomic_t vmchange_seq = ATOMIC_INIT(1);
-
-void flush_kernel_range(unsigned long start, unsigned long end, int update_seq)
-{
-	struct mm_struct *mm;
-	pgd_t *pgd;
-	pmd_t *pmd;
-	pte_t *pte;
-	unsigned long addr;
-	int updated = 0, err;
-
-	mm = &init_mm;
-	for(addr = start; addr < end;){
-		pgd = pgd_offset(mm, addr);
-		pmd = pmd_offset(pgd, addr);
-		if(pmd_present(*pmd)){
-			pte = pte_offset_kernel(pmd, addr);
-			if(!pte_present(*pte) || pte_newpage(*pte)){
-				updated = 1;
-				err = os_unmap_memory((void *) addr, 
-						      PAGE_SIZE);
-				if(err < 0)
-					panic("munmap failed, errno = %d\n",
-					      -err);
-				if(pte_present(*pte))
-					map_memory(addr, 
-						   pte_val(*pte) & PAGE_MASK,
-						   PAGE_SIZE, 1, 1, 1);
-			}
-			else if(pte_newprot(*pte)){
-				updated = 1;
-				protect_memory(addr, PAGE_SIZE, 1, 1, 1, 1);
-			}
-			addr += PAGE_SIZE;
-		}
-		else {
-			if(pmd_newpage(*pmd)){
-				updated = 1;
-				err = os_unmap_memory((void *) addr, PMD_SIZE);
-				if(err < 0)
-					panic("munmap failed, errno = %d\n",
-					      -err);
-			}
-			addr += PMD_SIZE;
-		}
-	}
-	if(updated && update_seq) atomic_inc(&vmchange_seq);
-}
+#include "choose-mode.h"
+#include "mode_kern.h"
 
 void flush_tlb_kernel_range(unsigned long start, unsigned long end)
 {
 	flush_kernel_range(start, end, 1);
 }
 
-static void protect_vm_page(unsigned long addr, int w, int must_succeed)
+void flush_tlb_page(struct vm_area_struct *vma, unsigned long address)
 {
-	int err;
-
-	err = protect_memory(addr, PAGE_SIZE, 1, w, 1, must_succeed);
-	if(err == 0) return;
-	else if((err == -EFAULT) || (err == -ENOMEM)){
-		flush_tlb_kernel_range(addr, addr + PAGE_SIZE);
-		protect_vm_page(addr, w, 1);
-	}
-	else panic("protect_vm_page : protect failed, errno = %d\n", err);
+	address &= PAGE_MASK;
+	flush_tlb_range(vma->vm_mm, address, address + PAGE_SIZE);
 }
 
-void mprotect_kernel_vm(int w)
+void flush_tlb_all(void)
 {
-	struct mm_struct *mm;
-	pgd_t *pgd;
-	pmd_t *pmd;
-	pte_t *pte;
-	unsigned long addr;
-	
-	mm = &init_mm;
-	for(addr = start_vm; addr < end_vm;){
-		pgd = pgd_offset(mm, addr);
-		pmd = pmd_offset(pgd, addr);
-		if(pmd_present(*pmd)){
-			pte = pte_offset_kernel(pmd, addr);
-			if(pte_present(*pte)) protect_vm_page(addr, w, 0);
-			addr += PAGE_SIZE;
-		}
-		else addr += PMD_SIZE;
-	}
+	flush_tlb_mm(current->mm);
 }
   
 void flush_tlb_kernel_vm(void)
 {
-	flush_tlb_kernel_range(start_vm, end_vm);
+	CHOOSE_MODE(flush_tlb_kernel_vm_tt(), flush_tlb_kernel_vm_skas());
 }
 
 void __flush_tlb_one(unsigned long addr)
 {
-	flush_tlb_kernel_range(addr, addr + PAGE_SIZE);
+	CHOOSE_MODE_PROC(__flush_tlb_one_tt, __flush_tlb_one_skas, addr);
 }
 
-void flush_tlb_range(struct vm_area_struct *vma, unsigned long start, 
+void flush_tlb_range(struct mm_struct *mm, unsigned long start, 
 		     unsigned long end)
 {
-	if(vma->vm_mm != current->mm) 
-		return;
-
-	/* Assumes that the range start ... end is entirely within
-	 * either process memory or kernel vm
-	 */
-	if((start >= start_vm) && (start < end_vm)) 
-		flush_kernel_range(start, end, 1);
-	else fix_range(vma->vm_mm, start, end, 0);
+	CHOOSE_MODE_PROC(flush_tlb_range_tt, flush_tlb_range_skas, mm, start, 
+			 end);
 }
 
 void flush_tlb_mm(struct mm_struct *mm)
 {
-	unsigned long seq;
-
-	if(mm != current->mm) 
-		return;
-
-	fix_range(mm, 0, STACK_TOP, 0);
-
-	seq = atomic_read(&vmchange_seq);
-	if(current->thread.mode.tt.vm_seq == seq) return;
-	current->thread.mode.tt.vm_seq = seq;
-	flush_kernel_range(start_vm, end_vm, 0);
-}
-
-void flush_tlb_page(struct vm_area_struct *vma, unsigned long address)
-{
-	address &= PAGE_MASK;
-	flush_tlb_range(vma, address, address + PAGE_SIZE);
-}
-
-void flush_tlb_all(void)
-{
-	flush_tlb_mm(current->mm);
+	CHOOSE_MODE_PROC(flush_tlb_mm_tt, flush_tlb_mm_skas, mm);
 }
 
 void force_flush_all(void)
 {
-	fix_range(current->mm, 0, STACK_TOP, 1);
-	flush_kernel_range(start_vm, end_vm, 0);
+	CHOOSE_MODE(force_flush_all_tt(), force_flush_all_skas());
 }
 
 pgd_t *pgd_offset_proc(struct mm_struct *mm, unsigned long address)

arch/um/kernel/tt/Makefile

@@ -4,7 +4,7 @@
 #
 
 obj-y = exec_kern.o exec_user.o gdb.o gdb_kern.o ksyms.o mem.o process_kern.o \
-	syscall_kern.o syscall_user.o time.o tracer.o trap_user.o \
+	syscall_kern.o syscall_user.o time.o tlb.o tracer.o trap_user.o \
 	uaccess_user.o sys-$(SUBARCH)/
 
 obj-$(CONFIG_PT_PROXY) += ptproxy/

arch/um/kernel/tt/tlb.c

+/* 
+ * Copyright (C) 2002 Jeff Dike (jdike@karaya.com)
+ * Licensed under the GPL
+ */
+
+#include "linux/stddef.h"
+#include "linux/kernel.h"
+#include "linux/sched.h"
+#include "asm/page.h"
+#include "asm/pgtable.h"
+#include "asm/uaccess.h"
+#include "user_util.h"
+#include "mem_user.h"
+#include "os.h"
+
+static void fix_range(struct mm_struct *mm, unsigned long start_addr, 
+		      unsigned long end_addr, int force)
+{
+	pgd_t *npgd;
+	pmd_t *npmd;
+	pte_t *npte;
+	unsigned long addr;
+	int r, w, x, err;
+
+	if((current->thread.mode.tt.extern_pid != -1) && 
+	   (current->thread.mode.tt.extern_pid != os_getpid()))
+		panic("fix_range fixing wrong address space, current = 0x%p",
+		      current);
+	if(mm == NULL) return;
+	for(addr=start_addr;addr<end_addr;){
+		if(addr == TASK_SIZE){
+			/* Skip over kernel text, kernel data, and physical
+			 * memory, which don't have ptes, plus kernel virtual
+			 * memory, which is flushed separately, and remap
+			 * the process stack.  The only way to get here is
+			 * if (end_addr == STACK_TOP) > TASK_SIZE, which is
+			 * only true in the honeypot case.
+			 */
+			addr = STACK_TOP - ABOVE_KMEM;
+			continue;
+		}
+		npgd = pgd_offset(mm, addr);
+		npmd = pmd_offset(npgd, addr);
+		if(pmd_present(*npmd)){
+			npte = pte_offset_kernel(npmd, addr);
+			r = pte_read(*npte);
+			w = pte_write(*npte);
+			x = pte_exec(*npte);
+			if(!pte_dirty(*npte)) w = 0;
+			if(!pte_young(*npte)){
+				r = 0;
+				w = 0;
+			}
+			if(force || pte_newpage(*npte)){
+				err = os_unmap_memory((void *) addr, 
+						      PAGE_SIZE);
+				if(err < 0)
+					panic("munmap failed, errno = %d\n",
+					      -err);
+				if(pte_present(*npte))
+					map_memory(addr, 
+						   pte_val(*npte) & PAGE_MASK,
+						   PAGE_SIZE, r, w, x);
+			}
+			else if(pte_newprot(*npte)){
+				protect_memory(addr, PAGE_SIZE, r, w, x, 1);
+			}
+			*npte = pte_mkuptodate(*npte);
+			addr += PAGE_SIZE;
+		}
+		else {
+			if(force || pmd_newpage(*npmd)){
+				err = os_unmap_memory((void *) addr, PMD_SIZE);
+				if(err < 0)
+					panic("munmap failed, errno = %d\n",
+					      -err);
+				pmd_mkuptodate(*npmd);
+			}
+			addr += PMD_SIZE;
+		}
+	}
+}
+
+atomic_t vmchange_seq = ATOMIC_INIT(1);
+
+void flush_kernel_range(unsigned long start, unsigned long end, int update_seq)
+{
+	struct mm_struct *mm;
+	pgd_t *pgd;
+	pmd_t *pmd;
+	pte_t *pte;
+	unsigned long addr;
+	int updated = 0, err;
+
+	mm = &init_mm;
+	for(addr = start; addr < end;){
+		pgd = pgd_offset(mm, addr);
+		pmd = pmd_offset(pgd, addr);
+		if(pmd_present(*pmd)){
+			pte = pte_offset_kernel(pmd, addr);
+			if(!pte_present(*pte) || pte_newpage(*pte)){
+				updated = 1;
+				err = os_unmap_memory((void *) addr, 
+						      PAGE_SIZE);
+				if(err < 0)
+					panic("munmap failed, errno = %d\n",
+					      -err);
+				if(pte_present(*pte))
+					map_memory(addr, 
+						   pte_val(*pte) & PAGE_MASK,
+						   PAGE_SIZE, 1, 1, 1);
+			}
+			else if(pte_newprot(*pte)){
+				updated = 1;
+				protect_memory(addr, PAGE_SIZE, 1, 1, 1, 1);
+			}
+			addr += PAGE_SIZE;
+		}
+		else {
+			if(pmd_newpage(*pmd)){
+				updated = 1;
+				err = os_unmap_memory((void *) addr, PMD_SIZE);
+				if(err < 0)
+					panic("munmap failed, errno = %d\n",
+					      -err);
+			}
+			addr += PMD_SIZE;
+		}
+	}
+	if(updated && update_seq) atomic_inc(&vmchange_seq);
+}
+
+static void protect_vm_page(unsigned long addr, int w, int must_succeed)
+{
+	int err;
+
+	err = protect_memory(addr, PAGE_SIZE, 1, w, 1, must_succeed);
+	if(err == 0) return;
+	else if((err == -EFAULT) || (err == -ENOMEM)){
+		flush_tlb_kernel_range(addr, addr + PAGE_SIZE);
+		protect_vm_page(addr, w, 1);
+	}
+	else panic("protect_vm_page : protect failed, errno = %d\n", err);
+}
+
+void mprotect_kernel_vm(int w)
+{
+	struct mm_struct *mm;
+	pgd_t *pgd;
+	pmd_t *pmd;
+	pte_t *pte;
+	unsigned long addr;
+	
+	mm = &init_mm;
+	for(addr = start_vm; addr < end_vm;){
+		pgd = pgd_offset(mm, addr);
+		pmd = pmd_offset(pgd, addr);
+		if(pmd_present(*pmd)){
+			pte = pte_offset_kernel(pmd, addr);
+			if(pte_present(*pte)) protect_vm_page(addr, w, 0);
+			addr += PAGE_SIZE;
+		}
+		else addr += PMD_SIZE;
+	}
+}
+
+void flush_tlb_kernel_vm_tt(void)
+{
+        flush_tlb_kernel_vm_range(start_vm, end_vm);
+}
+
+void __flush_tlb_one_tt(unsigned long addr)
+{
+        flush_tlb_kernel_vm_range(addr, addr + PAGE_SIZE);
+}
+  
+void flush_tlb_range_tt(struct mm_struct *mm, unsigned long start, 
+		     unsigned long end)
+{
+	if(mm != current->mm) return;
+
+	/* Assumes that the range start ... end is entirely within
+	 * either process memory or kernel vm
+	 */
+	if((start >= start_vm) && (start < end_vm)) 
+		flush_kernel_vm_range(start, end, 1);
+	else fix_range(mm, start, end, 0);
+}
+
+void flush_tlb_mm_tt(struct mm_struct *mm)
+{
+	unsigned long seq;
+
+	if(mm != current->mm) return;
+
+	fix_range(mm, 0, STACK_TOP, 0);
+
+	seq = atomic_read(&vmchange_seq);
+	if(current->thread.mode.tt.vm_seq == seq) return;
+	current->thread.mode.tt.vm_seq = seq;
+	flush_kernel_vm_range(start_vm, end_vm, 0);
+}
+
+void force_flush_all_tt(void)
+{
+	fix_range(current->mm, 0, STACK_TOP, 1);
+	flush_kernel_vm_range(start_vm, end_vm, 0);
+}
+
+/*
+ * Overrides for Emacs so that we follow Linus's tabbing style.
+ * Emacs will notice this stuff at the end of the file and automatically
+ * adjust the settings for this buffer only.  This must remain at the end
+ * of the file.
+ * ---------------------------------------------------------------------------
+ * Local variables:
+ * c-file-style: "linux"
+ * End:
+ */