sh: Initial vsyscall page support.

This implements initial support for the vsyscall page on SH. At the moment we leave it configurable due to having nommu to support from the same code base. We hook it up for the signal trampoline return at present, with more to be added later, once uClibc catches up. Signed-off-by: Paul Mundt <lethal@linux-sh.org>

sh: Initial vsyscall page support.
This implements initial support for the vsyscall page on SH. At the moment we leave it configurable due to having nommu to support from the same code base. We hook it up for the signal trampoline return at present, with more to be added later, once uClibc catches up. Signed-off-by: Paul Mundt <lethal@linux-sh.org>
Paul Mundt
1 parent 8c12b5dc13
Showing 19 changed files with 473 additions and 17 deletions Side-by-side Diff
arch/sh/kernel/Makefile
arch/sh/kernel/process.c
arch/sh/kernel/signal.c
arch/sh/kernel/vsyscall/Makefile
arch/sh/kernel/vsyscall/vsyscall-note.S
arch/sh/kernel/vsyscall/vsyscall-sigreturn.S
arch/sh/kernel/vsyscall/vsyscall-syscall.S
arch/sh/kernel/vsyscall/vsyscall-trapa.S
arch/sh/kernel/vsyscall/vsyscall.c
arch/sh/kernel/vsyscall/vsyscall.lds.S
arch/sh/mm/Kconfig
arch/sh/mm/init.c
arch/sh/mm/tlb-flush.c
include/asm-sh/auxvec.h
include/asm-sh/elf.h
include/asm-sh/mmu.h
include/asm-sh/mmu_context.h
include/asm-sh/page.h
include/asm-sh/processor.h
@@ -9,6 +9,7 @@
 	io.o io_generic.o sh_ksyms.o syscalls.o
  
 obj-y				+= cpu/ timers/
+obj-$(CONFIG_VSYSCALL)		+= vsyscall/
  
 obj-$(CONFIG_SMP)		+= smp.o
 obj-$(CONFIG_CF_ENABLER)	+= cf-enabler.o
@@ -355,7 +355,7 @@
 	else if (next->thread.ubc_pc && next->mm) {
 		int asid = 0;
 #ifdef CONFIG_MMU
-		asid |= next->mm->context & MMU_CONTEXT_ASID_MASK;
+		asid |= next->mm->context.id & MMU_CONTEXT_ASID_MASK;
 #endif
 		ubc_set_tracing(asid, next->thread.ubc_pc);
 	} else {
@@ -8,7 +8,6 @@
  *  SuperH version:  Copyright (C) 1999, 2000  Niibe Yutaka & Kaz Kojima
  *
  */
-
 #include <linux/sched.h>
 #include <linux/mm.h>
 #include <linux/smp.h>
@@ -21,6 +20,7 @@
 #include <linux/unistd.h>
 #include <linux/stddef.h>
 #include <linux/tty.h>
+#include <linux/elf.h>
 #include <linux/personality.h>
 #include <linux/binfmts.h>
  
@@ -29,8 +29,6 @@
 #include <asm/pgtable.h>
 #include <asm/cacheflush.h>
  
-#undef DEBUG
-
 #define _BLOCKABLE (~(sigmask(SIGKILL) | sigmask(SIGSTOP)))
  
 /*
@@ -312,6 +310,11 @@
 	return (void __user *)((sp - frame_size) & -8ul);
 }
  
+/* These symbols are defined with the addresses in the vsyscall page.
+   See vsyscall-trapa.S.  */
+extern void __user __kernel_sigreturn;
+extern void __user __kernel_rt_sigreturn;
+
 static int setup_frame(int sig, struct k_sigaction *ka,
 			sigset_t *set, struct pt_regs *regs)
 {
@@ -340,6 +343,10 @@
 	   already in userspace.  */
 	if (ka->sa.sa_flags & SA_RESTORER) {
 		regs->pr = (unsigned long) ka->sa.sa_restorer;
+#ifdef CONFIG_VSYSCALL
+	} else if (likely(current->mm->context.vdso)) {
+		regs->pr = VDSO_SYM(&__kernel_sigreturn);
+#endif
 	} else {
 		/* Generate return code (system call to sigreturn) */
 		err |= __put_user(MOVW(7), &frame->retcode[0]);
@@ -416,6 +423,10 @@
 	   already in userspace.  */
 	if (ka->sa.sa_flags & SA_RESTORER) {
 		regs->pr = (unsigned long) ka->sa.sa_restorer;
+#ifdef CONFIG_VSYSCALL
+	} else if (likely(current->mm->context.vdso)) {
+		regs->pr = VDSO_SYM(&__kernel_rt_sigreturn);
+#endif
 	} else {
 		/* Generate return code (system call to rt_sigreturn) */
 		err |= __put_user(MOVW(7), &frame->retcode[0]);
+obj-y += vsyscall.o vsyscall-syscall.o
+
+$(obj)/vsyscall-syscall.o: \
+	$(foreach F,trapa,$(obj)/vsyscall-$F.so)
+
+# Teach kbuild about targets
+targets += $(foreach F,trapa,vsyscall-$F.o vsyscall-$F.so)
+targets += vsyscall-note.o vsyscall.lds
+
+# The DSO images are built using a special linker script
+quiet_cmd_syscall = SYSCALL $@
+      cmd_syscall = $(CC) -nostdlib $(SYSCFLAGS_$(@F)) \
+			   -Wl,-T,$(filter-out FORCE,$^) -o $@
+
+export CPPFLAGS_vsyscall.lds += -P -C -Ush
+
+vsyscall-flags = -shared -s -Wl,-soname=linux-gate.so.1 \
+		$(call ld-option, -Wl$(comma)--hash-style=sysv)
+
+SYSCFLAGS_vsyscall-trapa.so	= $(vsyscall-flags)
+
+$(obj)/vsyscall-trapa.so: \
+$(obj)/vsyscall-%.so: $(src)/vsyscall.lds $(obj)/vsyscall-%.o FORCE
+	$(call if_changed,syscall)
+
+# We also create a special relocatable object that should mirror the symbol
+# table and layout of the linked DSO.  With ld -R we can then refer to
+# these symbols in the kernel code rather than hand-coded addresses.
+extra-y += vsyscall-syms.o
+$(obj)/built-in.o: $(obj)/vsyscall-syms.o
+$(obj)/built-in.o: ld_flags += -R $(obj)/vsyscall-syms.o
+
+SYSCFLAGS_vsyscall-syms.o = -r
+$(obj)/vsyscall-syms.o: $(src)/vsyscall.lds \
+			$(obj)/vsyscall-trapa.o $(obj)/vsyscall-note.o FORCE
+	$(call if_changed,syscall)
+/*
+ * This supplies .note.* sections to go into the PT_NOTE inside the vDSO text.
+ * Here we can supply some information useful to userland.
+ */
+
+#include <linux/uts.h>
+#include <linux/version.h>
+
+#define ASM_ELF_NOTE_BEGIN(name, flags, vendor, type)			      \
+	.section name, flags;						      \
+	.balign 4;							      \
+	.long 1f - 0f;		/* name length */			      \
+	.long 3f - 2f;		/* data length */			      \
+	.long type;		/* note type */				      \
+0:	.asciz vendor;		/* vendor name */			      \
+1:	.balign 4;							      \
+2:
+
+#define ASM_ELF_NOTE_END						      \
+3:	.balign 4;		/* pad out section */			      \
+	.previous
+
+	ASM_ELF_NOTE_BEGIN(".note.kernel-version", "a", UTS_SYSNAME, 0)
+	.long LINUX_VERSION_CODE
+	ASM_ELF_NOTE_END
+#include <asm/unistd.h>
+
+	.text
+	.balign 32
+	.globl __kernel_sigreturn
+	.type __kernel_sigreturn,@function
+__kernel_sigreturn:
+.LSTART_sigreturn:
+	mov.w	1f, r3
+	trapa	#0x10
+	or	r0, r0
+	or	r0, r0
+	or	r0, r0
+	or	r0, r0
+	or	r0, r0
+
+1:	.short	__NR_sigreturn
+.LEND_sigreturn:
+	.size __kernel_sigreturn,.-.LSTART_sigreturn
+
+	.balign 32
+	.globl __kernel_rt_sigreturn
+	.type __kernel_rt_sigreturn,@function
+__kernel_rt_sigreturn:
+.LSTART_rt_sigreturn:
+	mov.w	1f, r3
+	trapa	#0x10
+	or	r0, r0
+	or	r0, r0
+	or	r0, r0
+	or	r0, r0
+	or	r0, r0
+
+1:	.short	__NR_rt_sigreturn
+.LEND_rt_sigreturn:
+	.size __kernel_rt_sigreturn,.-.LSTART_rt_sigreturn
+
+	.section .eh_frame,"a",@progbits
+	.previous
+#include <linux/init.h>
+
+__INITDATA
+
+	.globl vsyscall_trapa_start, vsyscall_trapa_end
+vsyscall_trapa_start:
+	.incbin "arch/sh/kernel/vsyscall/vsyscall-trapa.so"
+vsyscall_trapa_end:
+
+__FINIT
+	.text
+	.globl __kernel_vsyscall
+	.type __kernel_vsyscall,@function
+__kernel_vsyscall:
+.LSTART_vsyscall:
+	/* XXX: We'll have to do something here once we opt to use the vDSO
+	 * page for something other than the signal trampoline.. as well as
+	 * fill out .eh_frame -- PFM. */
+.LEND_vsyscall:
+	.size __kernel_vsyscall,.-.LSTART_vsyscall
+	.previous
+
+	.section .eh_frame,"a",@progbits
+.LCIE:
+	.ualong	.LCIE_end - .LCIE_start
+.LCIE_start:
+	.ualong	0		/* CIE ID */
+	.byte	0x1		/* Version number */
+	.string	"zRS"		/* NUL-terminated augmentation string */
+	.uleb128 0x1		/* Code alignment factor */
+	.sleb128 -4		/* Data alignment factor */
+	.byte	0x11		/* Return address register column */
+				/* Augmentation length and data (none) */
+	.byte	0xc		/* DW_CFA_def_cfa */
+	.uleb128 0xf		/* r15 */
+	.uleb128 0x0		/* offset 0 */
+
+	.align 2
+.LCIE_end:
+
+	.ualong	.LFDE_end-.LFDE_start	/* Length FDE */
+.LFDE_start:
+	.ualong	.LCIE			/* CIE pointer */
+	.ualong	.LSTART_vsyscall-.	/* start address */
+	.ualong	.LEND_vsyscall-.LSTART_vsyscall
+	.uleb128 0
+	.align 2
+.LFDE_end:
+	.previous
+
+/* Get the common code for the sigreturn entry points */
+#include "vsyscall-sigreturn.S"
+/*
+ * arch/sh/kernel/vsyscall.c
+ *
+ *  Copyright (C) 2006 Paul Mundt
+ *
+ * vDSO randomization
+ * Copyright(C) 2005-2006, Red Hat, Inc., Ingo Molnar
+ *
+ * This file is subject to the terms and conditions of the GNU General Public
+ * License.  See the file "COPYING" in the main directory of this archive
+ * for more details.
+ */
+#include <linux/mm.h>
+#include <linux/slab.h>
+#include <linux/kernel.h>
+#include <linux/init.h>
+#include <linux/gfp.h>
+#include <linux/module.h>
+#include <linux/elf.h>
+
+/*
+ * Should the kernel map a VDSO page into processes and pass its
+ * address down to glibc upon exec()?
+ */
+unsigned int __read_mostly vdso_enabled = 1;
+EXPORT_SYMBOL_GPL(vdso_enabled);
+
+static int __init vdso_setup(char *s)
+{
+	vdso_enabled = simple_strtoul(s, NULL, 0);
+	return 1;
+}
+__setup("vdso=", vdso_setup);
+
+/*
+ * These symbols are defined by vsyscall.o to mark the bounds
+ * of the ELF DSO images included therein.
+ */
+extern const char vsyscall_trapa_start, vsyscall_trapa_end;
+static void *syscall_page;
+
+int __init vsyscall_init(void)
+{
+	syscall_page = (void *)get_zeroed_page(GFP_ATOMIC);
+
+	/*
+	 * XXX: Map this page to a fixmap entry if we get around
+	 * to adding the page to ELF core dumps
+	 */
+
+	memcpy(syscall_page,
+	       &vsyscall_trapa_start,
+	       &vsyscall_trapa_end - &vsyscall_trapa_start);
+
+	return 0;
+}
+
+static struct page *syscall_vma_nopage(struct vm_area_struct *vma,
+				       unsigned long address, int *type)
+{
+	unsigned long offset = address - vma->vm_start;
+	struct page *page;
+
+	if (address < vma->vm_start || address > vma->vm_end)
+		return NOPAGE_SIGBUS;
+
+	page = virt_to_page(syscall_page + offset);
+
+	get_page(page);
+
+	return page;
+}
+
+/* Prevent VMA merging */
+static void syscall_vma_close(struct vm_area_struct *vma)
+{
+}
+
+static struct vm_operations_struct syscall_vm_ops = {
+	.nopage	= syscall_vma_nopage,
+	.close	= syscall_vma_close,
+};
+
+/* Setup a VMA at program startup for the vsyscall page */
+int arch_setup_additional_pages(struct linux_binprm *bprm,
+				int executable_stack)
+{
+	struct vm_area_struct *vma;
+	struct mm_struct *mm = current->mm;
+	unsigned long addr;
+	int ret;
+
+	down_write(&mm->mmap_sem);
+	addr = get_unmapped_area(NULL, 0, PAGE_SIZE, 0, 0);
+	if (IS_ERR_VALUE(addr)) {
+		ret = addr;
+		goto up_fail;
+	}
+
+	vma = kmem_cache_zalloc(vm_area_cachep, SLAB_KERNEL);
+	if (!vma) {
+		ret = -ENOMEM;
+		goto up_fail;
+	}
+
+	vma->vm_start = addr;
+	vma->vm_end = addr + PAGE_SIZE;
+	/* MAYWRITE to allow gdb to COW and set breakpoints */
+	vma->vm_flags = VM_READ|VM_EXEC|VM_MAYREAD|VM_MAYEXEC|VM_MAYWRITE;
+	vma->vm_flags |= mm->def_flags;
+	vma->vm_page_prot = protection_map[vma->vm_flags & 7];
+	vma->vm_ops = &syscall_vm_ops;
+	vma->vm_mm = mm;
+
+	ret = insert_vm_struct(mm, vma);
+	if (unlikely(ret)) {
+		kmem_cache_free(vm_area_cachep, vma);
+		goto up_fail;
+	}
+
+	current->mm->context.vdso = (void *)addr;
+
+	mm->total_vm++;
+up_fail:
+	up_write(&mm->mmap_sem);
+	return ret;
+}
+
+const char *arch_vma_name(struct vm_area_struct *vma)
+{
+	if (vma->vm_mm && vma->vm_start == (long)vma->vm_mm->context.vdso)
+		return "[vdso]";
+
+	return NULL;
+}
+
+struct vm_area_struct *get_gate_vma(struct task_struct *task)
+{
+	return NULL;
+}
+
+int in_gate_area(struct task_struct *task, unsigned long address)
+{
+	return 0;
+}
+
+int in_gate_area_no_task(unsigned long address)
+{
+	return 0;
+}
+/*
+ * Linker script for vsyscall DSO.  The vsyscall page is an ELF shared
+ * object prelinked to its virtual address, and with only one read-only
+ * segment (that fits in one page).  This script controls its layout.
+ */
+#include <asm/asm-offsets.h>
+
+#ifdef CONFIG_CPU_LITTLE_ENDIAN
+OUTPUT_FORMAT("elf32-sh-linux", "elf32-sh-linux", "elf32-sh-linux")
+#else
+OUTPUT_FORMAT("elf32-shbig-linux", "elf32-shbig-linux", "elf32-shbig-linux")
+#endif
+OUTPUT_ARCH(sh)
+
+/* The ELF entry point can be used to set the AT_SYSINFO value.  */
+ENTRY(__kernel_vsyscall);
+
+SECTIONS
+{
+  . = SIZEOF_HEADERS;
+
+  .hash           : { *(.hash) }		:text
+  .gnu.hash       : { *(.gnu.hash) }
+  .dynsym         : { *(.dynsym) }
+  .dynstr         : { *(.dynstr) }
+  .gnu.version    : { *(.gnu.version) }
+  .gnu.version_d  : { *(.gnu.version_d) }
+  .gnu.version_r  : { *(.gnu.version_r) }
+
+  /* This linker script is used both with -r and with -shared.
+     For the layouts to match, we need to skip more than enough
+     space for the dynamic symbol table et al.  If this amount
+     is insufficient, ld -shared will barf.  Just increase it here.  */
+  . = 0x400;
+
+  .text           : { *(.text) }		:text =0x90909090
+  .note		  : { *(.note.*) }		:text :note
+  .eh_frame_hdr   : { *(.eh_frame_hdr) }	:text :eh_frame_hdr
+  .eh_frame       : { KEEP (*(.eh_frame)) }	:text
+  .dynamic        : { *(.dynamic) }		:text :dynamic
+  .useless        : {
+  	*(.got.plt) *(.got)
+	*(.data .data.* .gnu.linkonce.d.*)
+	*(.dynbss)
+	*(.bss .bss.* .gnu.linkonce.b.*)
+  }						:text
+}
+
+/*
+ * We must supply the ELF program headers explicitly to get just one
+ * PT_LOAD segment, and set the flags explicitly to make segments read-only.
+ */
+PHDRS
+{
+  text PT_LOAD FILEHDR PHDRS FLAGS(5); /* PF_R|PF_X */
+  dynamic PT_DYNAMIC FLAGS(4); /* PF_R */
+  note PT_NOTE FLAGS(4); /* PF_R */
+  eh_frame_hdr 0x6474e550; /* PT_GNU_EH_FRAME, but ld doesn't match the name */
+}
+
+/*
+ * This controls what symbols we export from the DSO.
+ */
+VERSION
+{
+  LINUX_2.6 {
+    global:
+    	__kernel_vsyscall;
+    	__kernel_sigreturn;
+    	__kernel_rt_sigreturn;
+
+    local: *;
+  };
+}
@@ -223,6 +223,19 @@
 	  32-bits through the SH-4A PMB. If this is not set, legacy
 	  29-bit physical addressing will be used.
  
+config VSYSCALL
+	bool "Support vsyscall page"
+	depends on MMU
+	default y
+	help
+	  This will enable support for the kernel mapping a vDSO page
+	  in process space, and subsequently handing down the entry point
+	  to the libc through the ELF auxiliary vector.
+
+	  From the kernel side this is used for the signal trampoline.
+	  For systems with an MMU that can afford to give up a page,
+	  (the default value) say Y.
+
 choice
 	prompt "HugeTLB page size"
 	depends on HUGETLB_PAGE && CPU_SH4 && MMU
@@ -286,6 +286,9 @@
 		initsize >> 10);
  
 	p3_cache_init();
+
+	/* Initialize the vDSO */
+	vsyscall_init();
 }
  
 void free_initmem(void)
@@ -14,12 +14,12 @@
  
 void flush_tlb_page(struct vm_area_struct *vma, unsigned long page)
 {
-	if (vma->vm_mm && vma->vm_mm->context != NO_CONTEXT) {
+	if (vma->vm_mm && vma->vm_mm->context.id != NO_CONTEXT) {
 		unsigned long flags;
 		unsigned long asid;
 		unsigned long saved_asid = MMU_NO_ASID;
  
-		asid = vma->vm_mm->context & MMU_CONTEXT_ASID_MASK;
+		asid = vma->vm_mm->context.id & MMU_CONTEXT_ASID_MASK;
 		page &= PAGE_MASK;
  
 		local_irq_save(flags);
  
  
  
@@ -39,20 +39,21 @@
 {
 	struct mm_struct *mm = vma->vm_mm;
  
-	if (mm->context != NO_CONTEXT) {
+	if (mm->context.id != NO_CONTEXT) {
 		unsigned long flags;
 		int size;
  
 		local_irq_save(flags);
 		size = (end - start + (PAGE_SIZE - 1)) >> PAGE_SHIFT;
 		if (size > (MMU_NTLB_ENTRIES/4)) { /* Too many TLB to flush */
-			mm->context = NO_CONTEXT;
+			mm->context.id = NO_CONTEXT;
 			if (mm == current->mm)
 				activate_context(mm);
 		} else {
-			unsigned long asid = mm->context&MMU_CONTEXT_ASID_MASK;
+			unsigned long asid;
 			unsigned long saved_asid = MMU_NO_ASID;
  
+			asid = mm->context.id & MMU_CONTEXT_ASID_MASK;
 			start &= PAGE_MASK;
 			end += (PAGE_SIZE - 1);
 			end &= PAGE_MASK;
  
@@ -81,9 +82,10 @@
 	if (size > (MMU_NTLB_ENTRIES/4)) { /* Too many TLB to flush */
 		flush_tlb_all();
 	} else {
-		unsigned long asid = init_mm.context&MMU_CONTEXT_ASID_MASK;
+		unsigned long asid;
 		unsigned long saved_asid = get_asid();
  
+		asid = init_mm.context.id & MMU_CONTEXT_ASID_MASK;
 		start &= PAGE_MASK;
 		end += (PAGE_SIZE - 1);
 		end &= PAGE_MASK;
  
@@ -101,11 +103,11 @@
 {
 	/* Invalidate all TLB of this process. */
 	/* Instead of invalidating each TLB, we get new MMU context. */
-	if (mm->context != NO_CONTEXT) {
+	if (mm->context.id != NO_CONTEXT) {
 		unsigned long flags;
  
 		local_irq_save(flags);
-		mm->context = NO_CONTEXT;
+		mm->context.id = NO_CONTEXT;
 		if (mm == current->mm)
 			activate_context(mm);
 		local_irq_restore(flags);
 #ifndef __ASM_SH_AUXVEC_H
 #define __ASM_SH_AUXVEC_H
  
+/*
+ * Architecture-neutral AT_ values in 0-17, leave some room
+ * for more of them.
+ */
+
+#ifdef CONFIG_VSYSCALL
+/*
+ * Only define this in the vsyscall case, the entry point to
+ * the vsyscall page gets placed here. The kernel will attempt
+ * to build a gate VMA we don't care about otherwise..
+ */
+#define AT_SYSINFO_EHDR		33
+#endif
+
 #endif /* __ASM_SH_AUXVEC_H */
@@ -121,5 +121,25 @@
 #define ELF_CORE_COPY_FPREGS(tsk, elf_fpregs) dump_task_fpu(tsk, elf_fpregs)
 #endif
  
+#ifdef CONFIG_VSYSCALL
+/* vDSO has arch_setup_additional_pages */
+#define ARCH_HAS_SETUP_ADDITIONAL_PAGES
+struct linux_binprm;
+extern int arch_setup_additional_pages(struct linux_binprm *bprm,
+				       int executable_stack);
+
+extern unsigned int vdso_enabled;
+extern void __kernel_vsyscall;
+
+#define VDSO_BASE		((unsigned long)current->mm->context.vdso)
+#define VDSO_SYM(x)		(VDSO_BASE + (unsigned long)(x))
+
+#define ARCH_DLINFO						\
+do {								\
+	if (vdso_enabled)					\
+		NEW_AUX_ENT(AT_SYSINFO_EHDR, VDSO_BASE);	\
+} while (0)
+#endif /* CONFIG_VSYSCALL */
+
 #endif /* __ASM_SH_ELF_H */
@@ -11,7 +11,12 @@
 #else
  
 /* Default "unsigned long" context */
-typedef unsigned long mm_context_t;
+typedef unsigned long mm_context_id_t;
+
+typedef struct {
+	mm_context_id_t id;
+	void *vdso;
+} mm_context_t;
  
 #endif /* CONFIG_MMU */
  
@@ -49,7 +49,7 @@
 	unsigned long mc = mmu_context_cache;
  
 	/* Check if we have old version of context. */
-	if (((mm->context ^ mc) & MMU_CONTEXT_VERSION_MASK) == 0)
+	if (((mm->context.id ^ mc) & MMU_CONTEXT_VERSION_MASK) == 0)
 		/* It's up to date, do nothing */
 		return;
  
@@ -68,7 +68,7 @@
 		if (!mc)
 			mmu_context_cache = mc = MMU_CONTEXT_FIRST_VERSION;
 	}
-	mm->context = mc;
+	mm->context.id = mc;
 }
  
 /*
@@ -78,7 +78,7 @@
 static __inline__ int init_new_context(struct task_struct *tsk,
 				       struct mm_struct *mm)
 {
-	mm->context = NO_CONTEXT;
+	mm->context.id = NO_CONTEXT;
  
 	return 0;
 }
@@ -123,7 +123,7 @@
 static __inline__ void activate_context(struct mm_struct *mm)
 {
 	get_mmu_context(mm);
-	set_asid(mm->context & MMU_CONTEXT_ASID_MASK);
+	set_asid(mm->context.id & MMU_CONTEXT_ASID_MASK);
 }
  
 /* MMU_TTB can be used for optimizing the fault handling.
@@ -117,6 +117,11 @@
 #include <asm-generic/memory_model.h>
 #include <asm-generic/page.h>
  
+/* vDSO support */
+#ifdef CONFIG_VSYSCALL
+#define __HAVE_ARCH_GATE_AREA
+#endif
+
 #endif /* __KERNEL__ */
 #endif /* __ASM_SH_PAGE_H */
@@ -276,6 +276,12 @@
 #define prefetchw(x)	prefetch(x)
 #endif
  
+#ifdef CONFIG_VSYSCALL
+extern int vsyscall_init(void);
+#else
+#define vsyscall_init() do { } while (0)
+#endif
+
 #endif /* __KERNEL__ */
 #endif /* __ASM_SH_PROCESSOR_H */
...	...	@@ -9,6 +9,7 @@
9	9	io.o io_generic.o sh_ksyms.o syscalls.o
10	10
11	11	obj-y += cpu/ timers/
	12	+obj-$(CONFIG_VSYSCALL) += vsyscall/
12	13
13	14	obj-$(CONFIG_SMP) += smp.o
14	15	obj-$(CONFIG_CF_ENABLER) += cf-enabler.o
...	...	@@ -355,7 +355,7 @@
355	355	else if (next->thread.ubc_pc && next->mm) {
356	356	int asid = 0;
357	357	#ifdef CONFIG_MMU
358		- asid \|= next->mm->context & MMU_CONTEXT_ASID_MASK;
	358	+ asid \|= next->mm->context.id & MMU_CONTEXT_ASID_MASK;
359	359	#endif
360	360	ubc_set_tracing(asid, next->thread.ubc_pc);
361	361	} else {
...	...	@@ -8,7 +8,6 @@
8	8	* SuperH version: Copyright (C) 1999, 2000 Niibe Yutaka & Kaz Kojima
9	9	*
10	10	*/
11		-
12	11	#include <linux/sched.h>
13	12	#include <linux/mm.h>
14	13	#include <linux/smp.h>
...	...	@@ -21,6 +20,7 @@
21	20	#include <linux/unistd.h>
22	21	#include <linux/stddef.h>
23	22	#include <linux/tty.h>
	23	+#include <linux/elf.h>
24	24	#include <linux/personality.h>
25	25	#include <linux/binfmts.h>
26	26
...	...	@@ -29,8 +29,6 @@
29	29	#include <asm/pgtable.h>
30	30	#include <asm/cacheflush.h>
31	31
32		-#undef DEBUG
33		-
34	32	#define _BLOCKABLE (~(sigmask(SIGKILL) \| sigmask(SIGSTOP)))
35	33
36	34	/*
...	...	@@ -312,6 +310,11 @@
312	310	return (void __user *)((sp - frame_size) & -8ul);
313	311	}
314	312
	313	+/* These symbols are defined with the addresses in the vsyscall page.
	314	+ See vsyscall-trapa.S. */
	315	+extern void __user __kernel_sigreturn;
	316	+extern void __user __kernel_rt_sigreturn;
	317	+
315	318	static int setup_frame(int sig, struct k_sigaction *ka,
316	319	sigset_t set, struct pt_regs regs)
317	320	{
...	...	@@ -340,6 +343,10 @@
340	343	already in userspace. */
341	344	if (ka->sa.sa_flags & SA_RESTORER) {
342	345	regs->pr = (unsigned long) ka->sa.sa_restorer;
	346	+#ifdef CONFIG_VSYSCALL
	347	+ } else if (likely(current->mm->context.vdso)) {
	348	+ regs->pr = VDSO_SYM(&__kernel_sigreturn);
	349	+#endif
343	350	} else {
344	351	/* Generate return code (system call to sigreturn) */
345	352	err \|= __put_user(MOVW(7), &frame->retcode[0]);
...	...	@@ -416,6 +423,10 @@
416	423	already in userspace. */
417	424	if (ka->sa.sa_flags & SA_RESTORER) {
418	425	regs->pr = (unsigned long) ka->sa.sa_restorer;
	426	+#ifdef CONFIG_VSYSCALL
	427	+ } else if (likely(current->mm->context.vdso)) {
	428	+ regs->pr = VDSO_SYM(&__kernel_rt_sigreturn);
	429	+#endif
419	430	} else {
420	431	/* Generate return code (system call to rt_sigreturn) */
421	432	err \|= __put_user(MOVW(7), &frame->retcode[0]);
	1	+obj-y += vsyscall.o vsyscall-syscall.o
	2	+
	3	+$(obj)/vsyscall-syscall.o: \
	4	+ $(foreach F,trapa,$(obj)/vsyscall-$F.so)
	5	+
	6	+# Teach kbuild about targets
	7	+targets += $(foreach F,trapa,vsyscall-$F.o vsyscall-$F.so)
	8	+targets += vsyscall-note.o vsyscall.lds
	9	+
	10	+# The DSO images are built using a special linker script
	11	+quiet_cmd_syscall = SYSCALL $@
	12	+ cmd_syscall = $(CC) -nostdlib $(SYSCFLAGS_$(@F)) \
	13	+ -Wl,-T,$(filter-out FORCE,$^) -o $@
	14	+
	15	+export CPPFLAGS_vsyscall.lds += -P -C -Ush
	16	+
	17	+vsyscall-flags = -shared -s -Wl,-soname=linux-gate.so.1 \
	18	+ $(call ld-option, -Wl$(comma)--hash-style=sysv)
	19	+
	20	+SYSCFLAGS_vsyscall-trapa.so = $(vsyscall-flags)
	21	+
	22	+$(obj)/vsyscall-trapa.so: \
	23	+$(obj)/vsyscall-%.so: $(src)/vsyscall.lds $(obj)/vsyscall-%.o FORCE
	24	+ $(call if_changed,syscall)
	25	+
	26	+# We also create a special relocatable object that should mirror the symbol
	27	+# table and layout of the linked DSO. With ld -R we can then refer to
	28	+# these symbols in the kernel code rather than hand-coded addresses.
	29	+extra-y += vsyscall-syms.o
	30	+$(obj)/built-in.o: $(obj)/vsyscall-syms.o
	31	+$(obj)/built-in.o: ld_flags += -R $(obj)/vsyscall-syms.o
	32	+
	33	+SYSCFLAGS_vsyscall-syms.o = -r
	34	+$(obj)/vsyscall-syms.o: $(src)/vsyscall.lds \
	35	+ $(obj)/vsyscall-trapa.o $(obj)/vsyscall-note.o FORCE
	36	+ $(call if_changed,syscall)
	1	+/*
	2	+ * This supplies .note.* sections to go into the PT_NOTE inside the vDSO text.
	3	+ * Here we can supply some information useful to userland.
	4	+ */
	5	+
	6	+#include <linux/uts.h>
	7	+#include <linux/version.h>
	8	+
	9	+#define ASM_ELF_NOTE_BEGIN(name, flags, vendor, type) \
	10	+ .section name, flags; \
	11	+ .balign 4; \
	12	+ .long 1f - 0f; /* name length */ \
	13	+ .long 3f - 2f; /* data length */ \
	14	+ .long type; /* note type */ \
	15	+0: .asciz vendor; /* vendor name */ \
	16	+1: .balign 4; \
	17	+2:
	18	+
	19	+#define ASM_ELF_NOTE_END \
	20	+3: .balign 4; /* pad out section */ \
	21	+ .previous
	22	+
	23	+ ASM_ELF_NOTE_BEGIN(".note.kernel-version", "a", UTS_SYSNAME, 0)
	24	+ .long LINUX_VERSION_CODE
	25	+ ASM_ELF_NOTE_END
	1	+#include <asm/unistd.h>
	2	+
	3	+ .text
	4	+ .balign 32
	5	+ .globl __kernel_sigreturn
	6	+ .type __kernel_sigreturn,@function
	7	+__kernel_sigreturn:
	8	+.LSTART_sigreturn:
	9	+ mov.w 1f, r3
	10	+ trapa #0x10
	11	+ or r0, r0
	12	+ or r0, r0
	13	+ or r0, r0
	14	+ or r0, r0
	15	+ or r0, r0
	16	+
	17	+1: .short __NR_sigreturn
	18	+.LEND_sigreturn:
	19	+ .size __kernel_sigreturn,.-.LSTART_sigreturn
	20	+
	21	+ .balign 32
	22	+ .globl __kernel_rt_sigreturn
	23	+ .type __kernel_rt_sigreturn,@function
	24	+__kernel_rt_sigreturn:
	25	+.LSTART_rt_sigreturn:
	26	+ mov.w 1f, r3
	27	+ trapa #0x10
	28	+ or r0, r0
	29	+ or r0, r0
	30	+ or r0, r0
	31	+ or r0, r0
	32	+ or r0, r0
	33	+
	34	+1: .short __NR_rt_sigreturn
	35	+.LEND_rt_sigreturn:
	36	+ .size __kernel_rt_sigreturn,.-.LSTART_rt_sigreturn
	37	+
	38	+ .section .eh_frame,"a",@progbits
	39	+ .previous
	1	+#include <linux/init.h>
	2	+
	3	+__INITDATA
	4	+
	5	+ .globl vsyscall_trapa_start, vsyscall_trapa_end
	6	+vsyscall_trapa_start:
	7	+ .incbin "arch/sh/kernel/vsyscall/vsyscall-trapa.so"
	8	+vsyscall_trapa_end:
	9	+
	10	+__FINIT
	1	+ .text
	2	+ .globl __kernel_vsyscall
	3	+ .type __kernel_vsyscall,@function
	4	+__kernel_vsyscall:
	5	+.LSTART_vsyscall:
	6	+ /* XXX: We'll have to do something here once we opt to use the vDSO
	7	+ * page for something other than the signal trampoline.. as well as
	8	+ * fill out .eh_frame -- PFM. */
	9	+.LEND_vsyscall:
	10	+ .size __kernel_vsyscall,.-.LSTART_vsyscall
	11	+ .previous
	12	+
	13	+ .section .eh_frame,"a",@progbits
	14	+.LCIE:
	15	+ .ualong .LCIE_end - .LCIE_start
	16	+.LCIE_start:
	17	+ .ualong 0 /* CIE ID */
	18	+ .byte 0x1 /* Version number */
	19	+ .string "zRS" /* NUL-terminated augmentation string */
	20	+ .uleb128 0x1 /* Code alignment factor */
	21	+ .sleb128 -4 /* Data alignment factor */
	22	+ .byte 0x11 /* Return address register column */
	23	+ /* Augmentation length and data (none) */
	24	+ .byte 0xc /* DW_CFA_def_cfa */
	25	+ .uleb128 0xf /* r15 */
	26	+ .uleb128 0x0 /* offset 0 */
	27	+
	28	+ .align 2
	29	+.LCIE_end:
	30	+
	31	+ .ualong .LFDE_end-.LFDE_start /* Length FDE */
	32	+.LFDE_start:
	33	+ .ualong .LCIE /* CIE pointer */
	34	+ .ualong .LSTART_vsyscall-. /* start address */
	35	+ .ualong .LEND_vsyscall-.LSTART_vsyscall
	36	+ .uleb128 0
	37	+ .align 2
	38	+.LFDE_end:
	39	+ .previous
	40	+
	41	+/* Get the common code for the sigreturn entry points */
	42	+#include "vsyscall-sigreturn.S"
	1	+/*
	2	+ * arch/sh/kernel/vsyscall.c
	3	+ *
	4	+ * Copyright (C) 2006 Paul Mundt
	5	+ *
	6	+ * vDSO randomization
	7	+ * Copyright(C) 2005-2006, Red Hat, Inc., Ingo Molnar
	8	+ *
	9	+ * This file is subject to the terms and conditions of the GNU General Public
	10	+ * License. See the file "COPYING" in the main directory of this archive
	11	+ * for more details.
	12	+ */
	13	+#include <linux/mm.h>
	14	+#include <linux/slab.h>
	15	+#include <linux/kernel.h>
	16	+#include <linux/init.h>
	17	+#include <linux/gfp.h>
	18	+#include <linux/module.h>
	19	+#include <linux/elf.h>
	20	+
	21	+/*
	22	+ * Should the kernel map a VDSO page into processes and pass its
	23	+ * address down to glibc upon exec()?
	24	+ */
	25	+unsigned int __read_mostly vdso_enabled = 1;
	26	+EXPORT_SYMBOL_GPL(vdso_enabled);
	27	+
	28	+static int __init vdso_setup(char *s)
	29	+{
	30	+ vdso_enabled = simple_strtoul(s, NULL, 0);
	31	+ return 1;
	32	+}
	33	+__setup("vdso=", vdso_setup);
	34	+
	35	+/*
	36	+ * These symbols are defined by vsyscall.o to mark the bounds
	37	+ * of the ELF DSO images included therein.
	38	+ */
	39	+extern const char vsyscall_trapa_start, vsyscall_trapa_end;
	40	+static void *syscall_page;
	41	+
	42	+int __init vsyscall_init(void)
	43	+{
	44	+ syscall_page = (void *)get_zeroed_page(GFP_ATOMIC);
	45	+
	46	+ /*
	47	+ * XXX: Map this page to a fixmap entry if we get around
	48	+ * to adding the page to ELF core dumps
	49	+ */
	50	+
	51	+ memcpy(syscall_page,
	52	+ &vsyscall_trapa_start,
	53	+ &vsyscall_trapa_end - &vsyscall_trapa_start);
	54	+
	55	+ return 0;
	56	+}
	57	+
	58	+static struct page syscall_vma_nopage(struct vm_area_struct vma,
	59	+ unsigned long address, int *type)
	60	+{
	61	+ unsigned long offset = address - vma->vm_start;
	62	+ struct page *page;
	63	+
	64	+ if (address < vma->vm_start \|\| address > vma->vm_end)
	65	+ return NOPAGE_SIGBUS;
	66	+
	67	+ page = virt_to_page(syscall_page + offset);
	68	+
	69	+ get_page(page);
	70	+
	71	+ return page;
	72	+}
	73	+
	74	+/* Prevent VMA merging */
	75	+static void syscall_vma_close(struct vm_area_struct *vma)
	76	+{
	77	+}
	78	+
	79	+static struct vm_operations_struct syscall_vm_ops = {
	80	+ .nopage = syscall_vma_nopage,
	81	+ .close = syscall_vma_close,
	82	+};
	83	+
	84	+/* Setup a VMA at program startup for the vsyscall page */
	85	+int arch_setup_additional_pages(struct linux_binprm *bprm,
	86	+ int executable_stack)
	87	+{
	88	+ struct vm_area_struct *vma;
	89	+ struct mm_struct *mm = current->mm;
	90	+ unsigned long addr;
	91	+ int ret;
	92	+
	93	+ down_write(&mm->mmap_sem);
	94	+ addr = get_unmapped_area(NULL, 0, PAGE_SIZE, 0, 0);
	95	+ if (IS_ERR_VALUE(addr)) {
	96	+ ret = addr;
	97	+ goto up_fail;
	98	+ }
	99	+
	100	+ vma = kmem_cache_zalloc(vm_area_cachep, SLAB_KERNEL);
	101	+ if (!vma) {
	102	+ ret = -ENOMEM;
	103	+ goto up_fail;
	104	+ }
	105	+
	106	+ vma->vm_start = addr;
	107	+ vma->vm_end = addr + PAGE_SIZE;
	108	+ /* MAYWRITE to allow gdb to COW and set breakpoints */
	109	+ vma->vm_flags = VM_READ\|VM_EXEC\|VM_MAYREAD\|VM_MAYEXEC\|VM_MAYWRITE;
	110	+ vma->vm_flags \|= mm->def_flags;
	111	+ vma->vm_page_prot = protection_map[vma->vm_flags & 7];
	112	+ vma->vm_ops = &syscall_vm_ops;
	113	+ vma->vm_mm = mm;
	114	+
	115	+ ret = insert_vm_struct(mm, vma);
	116	+ if (unlikely(ret)) {
	117	+ kmem_cache_free(vm_area_cachep, vma);
	118	+ goto up_fail;
	119	+ }
	120	+
	121	+ current->mm->context.vdso = (void *)addr;
	122	+
	123	+ mm->total_vm++;
	124	+up_fail:
	125	+ up_write(&mm->mmap_sem);
	126	+ return ret;
	127	+}
	128	+
	129	+const char arch_vma_name(struct vm_area_struct vma)
	130	+{
	131	+ if (vma->vm_mm && vma->vm_start == (long)vma->vm_mm->context.vdso)
	132	+ return "[vdso]";
	133	+
	134	+ return NULL;
	135	+}
	136	+
	137	+struct vm_area_struct get_gate_vma(struct task_struct task)
	138	+{
	139	+ return NULL;
	140	+}
	141	+
	142	+int in_gate_area(struct task_struct *task, unsigned long address)
	143	+{
	144	+ return 0;
	145	+}
	146	+
	147	+int in_gate_area_no_task(unsigned long address)
	148	+{
	149	+ return 0;
	150	+}
	1	+/*
	2	+ * Linker script for vsyscall DSO. The vsyscall page is an ELF shared
	3	+ * object prelinked to its virtual address, and with only one read-only
	4	+ * segment (that fits in one page). This script controls its layout.
	5	+ */
	6	+#include <asm/asm-offsets.h>
	7	+
	8	+#ifdef CONFIG_CPU_LITTLE_ENDIAN
	9	+OUTPUT_FORMAT("elf32-sh-linux", "elf32-sh-linux", "elf32-sh-linux")
	10	+#else
	11	+OUTPUT_FORMAT("elf32-shbig-linux", "elf32-shbig-linux", "elf32-shbig-linux")
	12	+#endif
	13	+OUTPUT_ARCH(sh)
	14	+
	15	+/* The ELF entry point can be used to set the AT_SYSINFO value. */
	16	+ENTRY(__kernel_vsyscall);
	17	+
	18	+SECTIONS
	19	+{
	20	+ . = SIZEOF_HEADERS;
	21	+
	22	+ .hash : { *(.hash) } :text
	23	+ .gnu.hash : { *(.gnu.hash) }
	24	+ .dynsym : { *(.dynsym) }
	25	+ .dynstr : { *(.dynstr) }
	26	+ .gnu.version : { *(.gnu.version) }
	27	+ .gnu.version_d : { *(.gnu.version_d) }
	28	+ .gnu.version_r : { *(.gnu.version_r) }
	29	+
	30	+ /* This linker script is used both with -r and with -shared.
	31	+ For the layouts to match, we need to skip more than enough
	32	+ space for the dynamic symbol table et al. If this amount
	33	+ is insufficient, ld -shared will barf. Just increase it here. */
	34	+ . = 0x400;
	35	+
	36	+ .text : { *(.text) } :text =0x90909090
	37	+ .note : { (.note.) } :text :note
	38	+ .eh_frame_hdr : { *(.eh_frame_hdr) } :text :eh_frame_hdr
	39	+ .eh_frame : { KEEP (*(.eh_frame)) } :text
	40	+ .dynamic : { *(.dynamic) } :text :dynamic
	41	+ .useless : {
	42	+ (.got.plt) (.got)
	43	+ (.data .data. .gnu.linkonce.d.*)
	44	+ *(.dynbss)
	45	+ (.bss .bss. .gnu.linkonce.b.*)
	46	+ } :text
	47	+}
	48	+
	49	+/*
	50	+ * We must supply the ELF program headers explicitly to get just one
	51	+ * PT_LOAD segment, and set the flags explicitly to make segments read-only.
	52	+ */
	53	+PHDRS
	54	+{
	55	+ text PT_LOAD FILEHDR PHDRS FLAGS(5); /* PF_R\|PF_X */
	56	+ dynamic PT_DYNAMIC FLAGS(4); /* PF_R */
	57	+ note PT_NOTE FLAGS(4); /* PF_R */
	58	+ eh_frame_hdr 0x6474e550; /* PT_GNU_EH_FRAME, but ld doesn't match the name */
	59	+}
	60	+
	61	+/*
	62	+ * This controls what symbols we export from the DSO.
	63	+ */
	64	+VERSION
	65	+{
	66	+ LINUX_2.6 {
	67	+ global:
	68	+ __kernel_vsyscall;
	69	+ __kernel_sigreturn;
	70	+ __kernel_rt_sigreturn;
	71	+
	72	+ local: *;
	73	+ };
	74	+}