x86 vDSO: consolidate vdso32

This makes x86_64's ia32 emulation support share the sources used in the 32-bit kernel for the 32-bit vDSO and much of its setup code. The 32-bit vDSO mapping now behaves the same on x86_64 as on native 32-bit. The abi.syscall32 sysctl on x86_64 now takes the same values that vm.vdso_enabled takes on the 32-bit kernel. That is, 1 means a randomized vDSO location, 2 means the fixed old address. The CONFIG_COMPAT_VDSO option is now available to make this the default setting, the same meaning it has for the 32-bit kernel. (This does not affect the 64-bit vDSO.) The argument vdso32=[012] can be used on both 32-bit and 64-bit kernels to set this paramter at boot time. The vdso=[012] argument still does this same thing on the 32-bit kernel. Signed-off-by: Roland McGrath <roland@redhat.com> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Signed-off-by: Ingo Molnar <mingo@elte.hu> Signed-off-by: Thomas Gleixner <tglx@linutronix.de>

x86 vDSO: consolidate vdso32
This makes x86_64's ia32 emulation support share the sources used in the 32-bit kernel for the 32-bit vDSO and much of its setup code. The 32-bit vDSO mapping now behaves the same on x86_64 as on native 32-bit. The abi.syscall32 sysctl on x86_64 now takes the same values that vm.vdso_enabled takes on the 32-bit kernel. That is, 1 means a randomized vDSO location, 2 means the fixed old address. The CONFIG_COMPAT_VDSO option is now available to make this the default setting, the same meaning it has for the 32-bit kernel. (This does not affect the 64-bit vDSO.) The argument vdso32=[012] can be used on both 32-bit and 64-bit kernels to set this paramter at boot time. The vdso=[012] argument still does this same thing on the 32-bit kernel. Signed-off-by: Roland McGrath <roland@redhat.com> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Signed-off-by: Ingo Molnar <mingo@elte.hu> Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Roland McGrath · Ingo Molnar
1 parent 00f8b1bc0e
Showing 10 changed files with 146 additions and 60 deletions Side-by-side Diff
Documentation/kernel-parameters.txt
arch/x86/Kconfig
arch/x86/ia32/Makefile
arch/x86/ia32/ia32_binfmt.c
arch/x86/ia32/ia32_signal.c
arch/x86/vdso/Makefile
arch/x86/vdso/vdso32-setup.c
arch/x86/vdso/vdso32.S
arch/x86/xen/setup.c
include/asm-x86/elf.h
@@ -1988,6 +1988,11 @@
 			vdso=1: enable VDSO (default)
 			vdso=0: disable VDSO mapping
  
+	vdso32=		[X86-32,X86-64]
+			vdso32=2: enable compat VDSO (default with COMPAT_VDSO)
+			vdso32=1: enable 32-bit VDSO (default)
+			vdso32=0: disable 32-bit VDSO mapping
+
 	vector=		[IA-64,SMP]
 			vector=percpu: enable percpu vector domain
  
@@ -1191,9 +1191,9 @@
 config COMPAT_VDSO
 	bool "Compat VDSO support"
 	default y
-	depends on X86_32
+	depends on X86_32 || IA32_EMULATION
 	help
-	  Map the VDSO to the predictable old-style address too.
+	  Map the 32-bit VDSO to the predictable old-style address too.
 	---help---
 	  Say N here if you are running a sufficiently recent glibc
 	  version (2.3.3 or later), to remove the high-mapped
@@ -3,7 +3,7 @@
 #
  
 obj-$(CONFIG_IA32_EMULATION) := ia32entry.o sys_ia32.o ia32_signal.o tls32.o \
-	ia32_binfmt.o fpu32.o ptrace32.o syscall32.o syscall32_syscall.o
+	ia32_binfmt.o fpu32.o ptrace32.o
  
 sysv-$(CONFIG_SYSVIPC) := ipc32.o
 obj-$(CONFIG_IA32_EMULATION) += $(sysv-y)
@@ -26,7 +26,7 @@
 #include <asm/i387.h>
 #include <asm/uaccess.h>
 #include <asm/ia32.h>
-#include <asm/vsyscall32.h>
+#include <asm/vdso.h>
  
 #undef	ELF_ARCH
 #undef	ELF_CLASS
  
@@ -47,14 +47,13 @@
 #define AT_SYSINFO 32
 #define AT_SYSINFO_EHDR		33
  
-int sysctl_vsyscall32 = 1;
+extern int sysctl_vsyscall32;
  
 #undef ARCH_DLINFO
 #define ARCH_DLINFO do {  \
 	if (sysctl_vsyscall32) { \
-		current->mm->context.vdso = (void *)VSYSCALL32_BASE;	\
-		NEW_AUX_ENT(AT_SYSINFO, (u32)(u64)VSYSCALL32_VSYSCALL); \
-		NEW_AUX_ENT(AT_SYSINFO_EHDR, VSYSCALL32_BASE);    \
+		NEW_AUX_ENT(AT_SYSINFO, (u32)VDSO_ENTRY);		\
+		NEW_AUX_ENT(AT_SYSINFO_EHDR, (u32)VDSO_CURRENT_BASE);	\
 	}	\
 } while(0)
  
@@ -31,7 +31,7 @@
 #include <asm/sigcontext32.h>
 #include <asm/fpu32.h>
 #include <asm/proto.h>
-#include <asm/vsyscall32.h>
+#include <asm/vdso.h>
  
 #define DEBUG_SIG 0
  
  
@@ -465,13 +465,16 @@
 			goto give_sigsegv;
 	}
  
-	/* Return stub is in 32bit vsyscall page */
-	if (current->binfmt->hasvdso)
-		restorer = VSYSCALL32_SIGRETURN;
-	else
-		restorer = (void *)&frame->retcode;
-	if (ka->sa.sa_flags & SA_RESTORER)
+	if (ka->sa.sa_flags & SA_RESTORER) {
 		restorer = ka->sa.sa_restorer;
+	} else {
+		/* Return stub is in 32bit vsyscall page */
+		if (current->binfmt->hasvdso)
+			restorer = VDSO32_SYMBOL(current->mm->context.vdso,
+						 sigreturn);
+		else
+			restorer = (void *)&frame->retcode;
+	}
 	err |= __put_user(ptr_to_compat(restorer), &frame->pretcode);
  
 	/*
@@ -519,7 +522,7 @@
 {
 	struct rt_sigframe __user *frame;
 	struct exec_domain *ed = current_thread_info()->exec_domain;
-	void __user *restorer = VSYSCALL32_RTSIGRETURN;
+	void __user *restorer;
 	int err = 0;
  
 	/* __copy_to_user optimizes that into a single 8 byte store */
@@ -564,6 +567,9 @@
  
 	if (ka->sa.sa_flags & SA_RESTORER)
 		restorer = ka->sa.sa_restorer;
+	else
+		restorer = VDSO32_SYMBOL(current->mm->context.vdso,
+					 rt_sigreturn);
 	err |= __put_user(ptr_to_compat(restorer), &frame->pretcode);
  
 	/*
@@ -15,7 +15,7 @@
  
 # files to link into kernel
 obj-$(VDSO64-y)			+= vma.o vdso.o
-obj-$(CONFIG_X86_32)		+= vdso32.o vdso32-setup.o
+obj-$(VDSO32-y)			+= vdso32.o vdso32-setup.o
  
 vobjs := $(foreach F,$(vobjs-y),$(obj)/$F)
  
@@ -24,6 +24,7 @@
 #include <asm/elf.h>
 #include <asm/tlbflush.h>
 #include <asm/vdso.h>
+#include <asm/proto.h>
  
 enum {
 	VDSO_DISABLED = 0,
  
  
@@ -37,14 +38,24 @@
 #define VDSO_DEFAULT	VDSO_ENABLED
 #endif
  
+#ifdef CONFIG_X86_64
+#define vdso_enabled			sysctl_vsyscall32
+#define arch_setup_additional_pages	syscall32_setup_pages
+#endif
+
 /*
+ * This is the difference between the prelinked addresses in the vDSO images
+ * and the VDSO_HIGH_BASE address where CONFIG_COMPAT_VDSO places the vDSO
+ * in the user address space.
+ */
+#define VDSO_ADDR_ADJUST	(VDSO_HIGH_BASE - (unsigned long)VDSO32_PRELINK)
+
+/*
  * Should the kernel map a VDSO page into processes and pass its
  * address down to glibc upon exec()?
  */
 unsigned int __read_mostly vdso_enabled = VDSO_DEFAULT;
  
-EXPORT_SYMBOL_GPL(vdso_enabled);
-
 static int __init vdso_setup(char *s)
 {
 	vdso_enabled = simple_strtoul(s, NULL, 0);
  
  
@@ -52,10 +63,19 @@
 	return 1;
 }
  
-__setup("vdso=", vdso_setup);
+/*
+ * For consistency, the argument vdso32=[012] affects the 32-bit vDSO
+ * behavior on both 64-bit and 32-bit kernels.
+ * On 32-bit kernels, vdso=[012] means the same thing.
+ */
+__setup("vdso32=", vdso_setup);
  
-extern asmlinkage void ia32_sysenter_target(void);
+#ifdef CONFIG_X86_32
+__setup_param("vdso=", vdso32_setup, vdso_setup, 0);
  
+EXPORT_SYMBOL_GPL(vdso_enabled);
+#endif
+
 static __init void reloc_symtab(Elf32_Ehdr *ehdr,
 				unsigned offset, unsigned size)
 {
@@ -79,7 +99,7 @@
 		case STT_FUNC:
 		case STT_SECTION:
 		case STT_FILE:
-			sym->st_value += VDSO_HIGH_BASE;
+			sym->st_value += VDSO_ADDR_ADJUST;
 		}
 	}
 }
@@ -105,7 +125,7 @@
 		case DT_VERNEED:
 		case DT_ADDRRNGLO ... DT_ADDRRNGHI:
 			/* definitely pointers needing relocation */
-			dyn->d_un.d_ptr += VDSO_HIGH_BASE;
+			dyn->d_un.d_ptr += VDSO_ADDR_ADJUST;
 			break;
  
 		case DT_ENCODING ... OLD_DT_LOOS-1:
@@ -114,7 +134,7 @@
 			   they're even */
 			if (dyn->d_tag >= DT_ENCODING &&
 			    (dyn->d_tag & 1) == 0)
-				dyn->d_un.d_ptr += VDSO_HIGH_BASE;
+				dyn->d_un.d_ptr += VDSO_ADDR_ADJUST;
 			break;
  
 		case DT_VERDEFNUM:
  
  
@@ -143,15 +163,15 @@
 	int i;
  
 	BUG_ON(memcmp(ehdr->e_ident, ELFMAG, 4) != 0 ||
-	       !elf_check_arch(ehdr) ||
+	       !elf_check_arch_ia32(ehdr) ||
 	       ehdr->e_type != ET_DYN);
  
-	ehdr->e_entry += VDSO_HIGH_BASE;
+	ehdr->e_entry += VDSO_ADDR_ADJUST;
  
 	/* rebase phdrs */
 	phdr = (void *)ehdr + ehdr->e_phoff;
 	for (i = 0; i < ehdr->e_phnum; i++) {
-		phdr[i].p_vaddr += VDSO_HIGH_BASE;
+		phdr[i].p_vaddr += VDSO_ADDR_ADJUST;
  
 		/* relocate dynamic stuff */
 		if (phdr[i].p_type == PT_DYNAMIC)
@@ -164,7 +184,7 @@
 		if (!(shdr[i].sh_flags & SHF_ALLOC))
 			continue;
  
-		shdr[i].sh_addr += VDSO_HIGH_BASE;
+		shdr[i].sh_addr += VDSO_ADDR_ADJUST;
  
 		if (shdr[i].sh_type == SHT_SYMTAB ||
 		    shdr[i].sh_type == SHT_DYNSYM)
@@ -173,6 +193,45 @@
 	}
 }
  
+/*
+ * These symbols are defined by vdso32.S to mark the bounds
+ * of the ELF DSO images included therein.
+ */
+extern const char vdso32_default_start, vdso32_default_end;
+extern const char vdso32_sysenter_start, vdso32_sysenter_end;
+static struct page *vdso32_pages[1];
+
+#ifdef CONFIG_X86_64
+
+static int use_sysenter __read_mostly = -1;
+
+#define	vdso32_sysenter()	(use_sysenter > 0)
+
+/* May not be __init: called during resume */
+void syscall32_cpu_init(void)
+{
+	if (use_sysenter < 0)
+		use_sysenter = (boot_cpu_data.x86_vendor == X86_VENDOR_INTEL);
+
+	/* Load these always in case some future AMD CPU supports
+	   SYSENTER from compat mode too. */
+	checking_wrmsrl(MSR_IA32_SYSENTER_CS, (u64)__KERNEL_CS);
+	checking_wrmsrl(MSR_IA32_SYSENTER_ESP, 0ULL);
+	checking_wrmsrl(MSR_IA32_SYSENTER_EIP, (u64)ia32_sysenter_target);
+
+	wrmsrl(MSR_CSTAR, ia32_cstar_target);
+}
+
+#define compat_uses_vma		1
+
+static inline void map_compat_vdso(int map)
+{
+}
+
+#else  /* CONFIG_X86_32 */
+
+#define vdso32_sysenter()	(boot_cpu_has(X86_FEATURE_SEP))
+
 void enable_sep_cpu(void)
 {
 	int cpu = get_cpu();
@@ -210,13 +269,7 @@
 	return 0;
 }
  
-/*
- * These symbols are defined by vsyscall.o to mark the bounds
- * of the ELF DSO images included therein.
- */
-extern const char vsyscall_int80_start, vsyscall_int80_end;
-extern const char vsyscall_sysenter_start, vsyscall_sysenter_end;
-static struct page *syscall_pages[1];
+#define compat_uses_vma		0
  
 static void map_compat_vdso(int map)
 {
  
  
  
  
  
  
@@ -227,31 +280,35 @@
  
 	vdso_mapped = map;
  
-	__set_fixmap(FIX_VDSO, page_to_pfn(syscall_pages[0]) << PAGE_SHIFT,
+	__set_fixmap(FIX_VDSO, page_to_pfn(vdso32_pages[0]) << PAGE_SHIFT,
 		     map ? PAGE_READONLY_EXEC : PAGE_NONE);
  
 	/* flush stray tlbs */
 	flush_tlb_all();
 }
  
+#endif	/* CONFIG_X86_64 */
+
 int __init sysenter_setup(void)
 {
 	void *syscall_page = (void *)get_zeroed_page(GFP_ATOMIC);
 	const void *vsyscall;
 	size_t vsyscall_len;
  
-	syscall_pages[0] = virt_to_page(syscall_page);
+	vdso32_pages[0] = virt_to_page(syscall_page);
  
+#ifdef CONFIG_X86_32
 	gate_vma_init();
  
 	printk("Compat vDSO mapped to %08lx.\n", __fix_to_virt(FIX_VDSO));
+#endif
  
-	if (!boot_cpu_has(X86_FEATURE_SEP)) {
-		vsyscall = &vsyscall_int80_start;
-		vsyscall_len = &vsyscall_int80_end - &vsyscall_int80_start;
+	if (!vdso32_sysenter()) {
+		vsyscall = &vdso32_default_start;
+		vsyscall_len = &vdso32_default_end - &vdso32_default_start;
 	} else {
-		vsyscall = &vsyscall_sysenter_start;
-		vsyscall_len = &vsyscall_sysenter_end - &vsyscall_sysenter_start;
+		vsyscall = &vdso32_sysenter_start;
+		vsyscall_len = &vdso32_sysenter_end - &vdso32_sysenter_start;
 	}
  
 	memcpy(syscall_page, vsyscall, vsyscall_len);
  
@@ -284,7 +341,9 @@
 			ret = addr;
 			goto up_fail;
 		}
+	}
  
+	if (compat_uses_vma || !compat) {
 		/*
 		 * MAYWRITE to allow gdb to COW and set breakpoints
 		 *
@@ -298,7 +357,7 @@
 					      VM_READ|VM_EXEC|
 					      VM_MAYREAD|VM_MAYWRITE|VM_MAYEXEC|
 					      VM_ALWAYSDUMP,
-					      syscall_pages);
+					      vdso32_pages);
  
 		if (ret)
 			goto up_fail;
@@ -314,6 +373,12 @@
 	return ret;
 }
  
+#ifdef CONFIG_X86_64
+
+__initcall(sysenter_setup);
+
+#else  /* CONFIG_X86_32 */
+
 const char *arch_vma_name(struct vm_area_struct *vma)
 {
 	if (vma->vm_mm && vma->vm_start == (long)vma->vm_mm->context.vdso)
@@ -342,4 +407,6 @@
 {
 	return 0;
 }
+
+#endif	/* CONFIG_X86_64 */
@@ -2,15 +2,19 @@
  
 __INITDATA
  
-	.globl vsyscall_int80_start, vsyscall_int80_end
-vsyscall_int80_start:
+	.globl vdso32_default_start, vdso32_default_end
+vdso32_default_start:
+#ifdef CONFIG_X86_32
 	.incbin "arch/x86/vdso/vdso32-int80.so"
-vsyscall_int80_end:
+#else
+	.incbin "arch/x86/vdso/vdso32-syscall.so"
+#endif
+vdso32_default_end:
  
-	.globl vsyscall_sysenter_start, vsyscall_sysenter_end
-vsyscall_sysenter_start:
+	.globl vdso32_sysenter_start, vdso32_sysenter_end
+vdso32_sysenter_start:
 	.incbin "arch/x86/vdso/vdso32-sysenter.so"
-vsyscall_sysenter_end:
+vdso32_sysenter_end:
  
 __FINIT
@@ -62,8 +62,8 @@
  */
 static void fiddle_vdso(void)
 {
-	extern char vsyscall_int80_start;
-	u32 *mask = VDSO32_SYMBOL(&vsyscall_int80_start, NOTE_MASK);
+	extern const char vdso32_default_start;
+	u32 *mask = VDSO32_SYMBOL(&vdso32_default_start, NOTE_MASK);
 	*mask |= 1 << VDSO_NOTE_NONEGSEG_BIT;
 }
  
@@ -74,17 +74,19 @@
  
 #ifdef __KERNEL__
  
+/*
+ * This is used to ensure we don't load something for the wrong architecture.
+ */
+#define elf_check_arch_ia32(x) \
+	(((x)->e_machine == EM_386) || ((x)->e_machine == EM_486))
+
 #ifdef CONFIG_X86_32
 #include <asm/processor.h>
 #include <asm/system.h>		/* for savesegment */
 #include <asm/desc.h>
 #include <asm/vdso.h>
  
-/*
- * This is used to ensure we don't load something for the wrong architecture.
- */
-#define elf_check_arch(x) \
-	(((x)->e_machine == EM_386) || ((x)->e_machine == EM_486))
+#define elf_check_arch(x)	elf_check_arch_ia32(x)
  
 /* SVR4/i386 ABI (pages 3-31, 3-32) says that when the program starts %edx
    contains a pointer to a function which might be registered using `atexit'.
  
@@ -247,11 +249,7 @@
 #define ELF_CORE_XFPREG_TYPE NT_PRXFPREG
  
 #define VDSO_HIGH_BASE		(__fix_to_virt(FIX_VDSO))
-#define VDSO_CURRENT_BASE	((unsigned long)current->mm->context.vdso)
  
-#define VDSO_ENTRY \
-	((unsigned long) VDSO32_SYMBOL(VDSO_CURRENT_BASE, vsyscall))
-
 /* update AT_VECTOR_SIZE_ARCH if the number of NEW_AUX_ENT entries changes */
  
 #define ARCH_DLINFO \
@@ -262,6 +260,8 @@
  
 #else /* CONFIG_X86_32 */
  
+#define VDSO_HIGH_BASE		0xffffe000U /* CONFIG_COMPAT_VDSO address */
+
 /* 1GB for 64bit, 8MB for 32bit */
 #define STACK_RND_MASK (test_thread_flag(TIF_IA32) ? 0x7ff : 0x3fffff)
  
@@ -271,6 +271,11 @@
 } while (0)
  
 #endif /* !CONFIG_X86_32 */
+
+#define VDSO_CURRENT_BASE	((unsigned long)current->mm->context.vdso)
+
+#define VDSO_ENTRY \
+	((unsigned long) VDSO32_SYMBOL(VDSO_CURRENT_BASE, vsyscall))
  
 struct linux_binprm;
...	...	@@ -1988,6 +1988,11 @@
1988	1988	vdso=1: enable VDSO (default)
1989	1989	vdso=0: disable VDSO mapping
1990	1990
	1991	+ vdso32= [X86-32,X86-64]
	1992	+ vdso32=2: enable compat VDSO (default with COMPAT_VDSO)
	1993	+ vdso32=1: enable 32-bit VDSO (default)
	1994	+ vdso32=0: disable 32-bit VDSO mapping
	1995	+
1991	1996	vector= [IA-64,SMP]
1992	1997	vector=percpu: enable percpu vector domain
1993	1998
...	...	@@ -1191,9 +1191,9 @@
1191	1191	config COMPAT_VDSO
1192	1192	bool "Compat VDSO support"
1193	1193	default y
1194		- depends on X86_32
	1194	+ depends on X86_32 \|\| IA32_EMULATION
1195	1195	help
1196		- Map the VDSO to the predictable old-style address too.
	1196	+ Map the 32-bit VDSO to the predictable old-style address too.
1197	1197	---help---
1198	1198	Say N here if you are running a sufficiently recent glibc
1199	1199	version (2.3.3 or later), to remove the high-mapped
...	...	@@ -3,7 +3,7 @@
3	3	#
4	4
5	5	obj-$(CONFIG_IA32_EMULATION) := ia32entry.o sys_ia32.o ia32_signal.o tls32.o \
6		- ia32_binfmt.o fpu32.o ptrace32.o syscall32.o syscall32_syscall.o
	6	+ ia32_binfmt.o fpu32.o ptrace32.o
7	7
8	8	sysv-$(CONFIG_SYSVIPC) := ipc32.o
9	9	obj-$(CONFIG_IA32_EMULATION) += $(sysv-y)
...	...	@@ -26,7 +26,7 @@
26	26	#include <asm/i387.h>
27	27	#include <asm/uaccess.h>
28	28	#include <asm/ia32.h>
29		-#include <asm/vsyscall32.h>
	29	+#include <asm/vdso.h>
30	30
31	31	#undef ELF_ARCH
32	32	#undef ELF_CLASS
33	33
...	...	@@ -47,14 +47,13 @@
47	47	#define AT_SYSINFO 32
48	48	#define AT_SYSINFO_EHDR 33
49	49
50		-int sysctl_vsyscall32 = 1;
	50	+extern int sysctl_vsyscall32;
51	51
52	52	#undef ARCH_DLINFO
53	53	#define ARCH_DLINFO do { \
54	54	if (sysctl_vsyscall32) { \
55		- current->mm->context.vdso = (void *)VSYSCALL32_BASE; \
56		- NEW_AUX_ENT(AT_SYSINFO, (u32)(u64)VSYSCALL32_VSYSCALL); \
57		- NEW_AUX_ENT(AT_SYSINFO_EHDR, VSYSCALL32_BASE); \
	55	+ NEW_AUX_ENT(AT_SYSINFO, (u32)VDSO_ENTRY); \
	56	+ NEW_AUX_ENT(AT_SYSINFO_EHDR, (u32)VDSO_CURRENT_BASE); \
58	57	} \
59	58	} while(0)
60	59
...	...	@@ -31,7 +31,7 @@
31	31	#include <asm/sigcontext32.h>
32	32	#include <asm/fpu32.h>
33	33	#include <asm/proto.h>
34		-#include <asm/vsyscall32.h>
	34	+#include <asm/vdso.h>
35	35
36	36	#define DEBUG_SIG 0
37	37
38	38
...	...	@@ -465,13 +465,16 @@
465	465	goto give_sigsegv;
466	466	}
467	467
468		- /* Return stub is in 32bit vsyscall page */
469		- if (current->binfmt->hasvdso)
470		- restorer = VSYSCALL32_SIGRETURN;
471		- else
472		- restorer = (void *)&frame->retcode;
473		- if (ka->sa.sa_flags & SA_RESTORER)
	468	+ if (ka->sa.sa_flags & SA_RESTORER) {
474	469	restorer = ka->sa.sa_restorer;
	470	+ } else {
	471	+ /* Return stub is in 32bit vsyscall page */
	472	+ if (current->binfmt->hasvdso)
	473	+ restorer = VDSO32_SYMBOL(current->mm->context.vdso,
	474	+ sigreturn);
	475	+ else
	476	+ restorer = (void *)&frame->retcode;
	477	+ }
475	478	err \|= __put_user(ptr_to_compat(restorer), &frame->pretcode);
476	479
477	480	/*
...	...	@@ -519,7 +522,7 @@
519	522	{
520	523	struct rt_sigframe __user *frame;
521	524	struct exec_domain *ed = current_thread_info()->exec_domain;
522		- void __user *restorer = VSYSCALL32_RTSIGRETURN;
	525	+ void __user *restorer;
523	526	int err = 0;
524	527
525	528	/* __copy_to_user optimizes that into a single 8 byte store */
...	...	@@ -564,6 +567,9 @@
564	567
565	568	if (ka->sa.sa_flags & SA_RESTORER)
566	569	restorer = ka->sa.sa_restorer;
	570	+ else
	571	+ restorer = VDSO32_SYMBOL(current->mm->context.vdso,
	572	+ rt_sigreturn);
567	573	err \|= __put_user(ptr_to_compat(restorer), &frame->pretcode);
568	574
569	575	/*
...	...	@@ -15,7 +15,7 @@
15	15
16	16	# files to link into kernel
17	17	obj-$(VDSO64-y) += vma.o vdso.o
18		-obj-$(CONFIG_X86_32) += vdso32.o vdso32-setup.o
	18	+obj-$(VDSO32-y) += vdso32.o vdso32-setup.o
19	19
20	20	vobjs := $(foreach F,$(vobjs-y),$(obj)/$F)
21	21
...	...	@@ -24,6 +24,7 @@
24	24	#include <asm/elf.h>
25	25	#include <asm/tlbflush.h>
26	26	#include <asm/vdso.h>
	27	+#include <asm/proto.h>
27	28
28	29	enum {
29	30	VDSO_DISABLED = 0,
30	31
31	32
...	...	@@ -37,14 +38,24 @@
37	38	#define VDSO_DEFAULT VDSO_ENABLED
38	39	#endif
39	40
	41	+#ifdef CONFIG_X86_64
	42	+#define vdso_enabled sysctl_vsyscall32
	43	+#define arch_setup_additional_pages syscall32_setup_pages
	44	+#endif
	45	+
40	46	/*
	47	+ * This is the difference between the prelinked addresses in the vDSO images
	48	+ * and the VDSO_HIGH_BASE address where CONFIG_COMPAT_VDSO places the vDSO
	49	+ * in the user address space.
	50	+ */
	51	+#define VDSO_ADDR_ADJUST (VDSO_HIGH_BASE - (unsigned long)VDSO32_PRELINK)
	52	+
	53	+/*
41	54	* Should the kernel map a VDSO page into processes and pass its
42	55	* address down to glibc upon exec()?
43	56	*/
44	57	unsigned int __read_mostly vdso_enabled = VDSO_DEFAULT;
45	58
46		-EXPORT_SYMBOL_GPL(vdso_enabled);
47		-
48	59	static int __init vdso_setup(char *s)
49	60	{
50	61	vdso_enabled = simple_strtoul(s, NULL, 0);
51	62
52	63
...	...	@@ -52,10 +63,19 @@
52	63	return 1;
53	64	}
54	65
55		-__setup("vdso=", vdso_setup);
	66	+/*
	67	+ * For consistency, the argument vdso32=[012] affects the 32-bit vDSO
	68	+ * behavior on both 64-bit and 32-bit kernels.
	69	+ * On 32-bit kernels, vdso=[012] means the same thing.
	70	+ */
	71	+__setup("vdso32=", vdso_setup);
56	72
57		-extern asmlinkage void ia32_sysenter_target(void);
	73	+#ifdef CONFIG_X86_32
	74	+__setup_param("vdso=", vdso32_setup, vdso_setup, 0);
58	75
	76	+EXPORT_SYMBOL_GPL(vdso_enabled);
	77	+#endif
	78	+
59	79	static __init void reloc_symtab(Elf32_Ehdr *ehdr,
60	80	unsigned offset, unsigned size)
61	81	{
...	...	@@ -79,7 +99,7 @@
79	99	case STT_FUNC:
80	100	case STT_SECTION:
81	101	case STT_FILE:
82		- sym->st_value += VDSO_HIGH_BASE;
	102	+ sym->st_value += VDSO_ADDR_ADJUST;
83	103	}
84	104	}
85	105	}
...	...	@@ -105,7 +125,7 @@
105	125	case DT_VERNEED:
106	126	case DT_ADDRRNGLO ... DT_ADDRRNGHI:
107	127	/* definitely pointers needing relocation */
108		- dyn->d_un.d_ptr += VDSO_HIGH_BASE;
	128	+ dyn->d_un.d_ptr += VDSO_ADDR_ADJUST;
109	129	break;
110	130
111	131	case DT_ENCODING ... OLD_DT_LOOS-1:
...	...	@@ -114,7 +134,7 @@
114	134	they're even */
115	135	if (dyn->d_tag >= DT_ENCODING &&
116	136	(dyn->d_tag & 1) == 0)
117		- dyn->d_un.d_ptr += VDSO_HIGH_BASE;
	137	+ dyn->d_un.d_ptr += VDSO_ADDR_ADJUST;
118	138	break;
119	139
120	140	case DT_VERDEFNUM:
121	141
122	142
...	...	@@ -143,15 +163,15 @@
143	163	int i;
144	164
145	165	BUG_ON(memcmp(ehdr->e_ident, ELFMAG, 4) != 0 \|\|
146		- !elf_check_arch(ehdr) \|\|
	166	+ !elf_check_arch_ia32(ehdr) \|\|
147	167	ehdr->e_type != ET_DYN);
148	168
149		- ehdr->e_entry += VDSO_HIGH_BASE;
	169	+ ehdr->e_entry += VDSO_ADDR_ADJUST;
150	170
151	171	/* rebase phdrs */
152	172	phdr = (void *)ehdr + ehdr->e_phoff;
153	173	for (i = 0; i < ehdr->e_phnum; i++) {
154		- phdr[i].p_vaddr += VDSO_HIGH_BASE;
	174	+ phdr[i].p_vaddr += VDSO_ADDR_ADJUST;
155	175
156	176	/* relocate dynamic stuff */
157	177	if (phdr[i].p_type == PT_DYNAMIC)
...	...	@@ -164,7 +184,7 @@
164	184	if (!(shdr[i].sh_flags & SHF_ALLOC))
165	185	continue;
166	186
167		- shdr[i].sh_addr += VDSO_HIGH_BASE;
	187	+ shdr[i].sh_addr += VDSO_ADDR_ADJUST;
168	188
169	189	if (shdr[i].sh_type == SHT_SYMTAB \|\|
170	190	shdr[i].sh_type == SHT_DYNSYM)
...	...	@@ -173,6 +193,45 @@
173	193	}
174	194	}
175	195
	196	+/*
	197	+ * These symbols are defined by vdso32.S to mark the bounds
	198	+ * of the ELF DSO images included therein.
	199	+ */
	200	+extern const char vdso32_default_start, vdso32_default_end;
	201	+extern const char vdso32_sysenter_start, vdso32_sysenter_end;
	202	+static struct page *vdso32_pages[1];
	203	+
	204	+#ifdef CONFIG_X86_64
	205	+
	206	+static int use_sysenter __read_mostly = -1;
	207	+
	208	+#define vdso32_sysenter() (use_sysenter > 0)
	209	+
	210	+/* May not be __init: called during resume */
	211	+void syscall32_cpu_init(void)
	212	+{
	213	+ if (use_sysenter < 0)
	214	+ use_sysenter = (boot_cpu_data.x86_vendor == X86_VENDOR_INTEL);
	215	+
	216	+ /* Load these always in case some future AMD CPU supports
	217	+ SYSENTER from compat mode too. */
	218	+ checking_wrmsrl(MSR_IA32_SYSENTER_CS, (u64)__KERNEL_CS);
	219	+ checking_wrmsrl(MSR_IA32_SYSENTER_ESP, 0ULL);
	220	+ checking_wrmsrl(MSR_IA32_SYSENTER_EIP, (u64)ia32_sysenter_target);
	221	+
	222	+ wrmsrl(MSR_CSTAR, ia32_cstar_target);
	223	+}
	224	+
	225	+#define compat_uses_vma 1
	226	+
	227	+static inline void map_compat_vdso(int map)
	228	+{
	229	+}
	230	+
	231	+#else /* CONFIG_X86_32 */
	232	+
	233	+#define vdso32_sysenter() (boot_cpu_has(X86_FEATURE_SEP))
	234	+
176	235	void enable_sep_cpu(void)
177	236	{
178	237	int cpu = get_cpu();
...	...	@@ -210,13 +269,7 @@
210	269	return 0;
211	270	}
212	271
213		-/*
214		- * These symbols are defined by vsyscall.o to mark the bounds
215		- * of the ELF DSO images included therein.
216		- */
217		-extern const char vsyscall_int80_start, vsyscall_int80_end;
218		-extern const char vsyscall_sysenter_start, vsyscall_sysenter_end;
219		-static struct page *syscall_pages[1];
	272	+#define compat_uses_vma 0
220	273
221	274	static void map_compat_vdso(int map)
222	275	{
223	276
224	277
225	278
226	279
227	280
228	281
...	...	@@ -227,31 +280,35 @@
227	280
228	281	vdso_mapped = map;
229	282
230		- __set_fixmap(FIX_VDSO, page_to_pfn(syscall_pages[0]) << PAGE_SHIFT,
	283	+ __set_fixmap(FIX_VDSO, page_to_pfn(vdso32_pages[0]) << PAGE_SHIFT,
231	284	map ? PAGE_READONLY_EXEC : PAGE_NONE);
232	285
233	286	/* flush stray tlbs */
234	287	flush_tlb_all();
235	288	}
236	289
	290	+#endif /* CONFIG_X86_64 */
	291	+
237	292	int __init sysenter_setup(void)
238	293	{
239	294	void syscall_page = (void )get_zeroed_page(GFP_ATOMIC);
240	295	const void *vsyscall;
241	296	size_t vsyscall_len;
242	297
243		- syscall_pages[0] = virt_to_page(syscall_page);
	298	+ vdso32_pages[0] = virt_to_page(syscall_page);
244	299
	300	+#ifdef CONFIG_X86_32
245	301	gate_vma_init();
246	302
247	303	printk("Compat vDSO mapped to %08lx.\n", __fix_to_virt(FIX_VDSO));
	304	+#endif
248	305
249		- if (!boot_cpu_has(X86_FEATURE_SEP)) {
250		- vsyscall = &vsyscall_int80_start;
251		- vsyscall_len = &vsyscall_int80_end - &vsyscall_int80_start;
	306	+ if (!vdso32_sysenter()) {
	307	+ vsyscall = &vdso32_default_start;
	308	+ vsyscall_len = &vdso32_default_end - &vdso32_default_start;
252	309	} else {
253		- vsyscall = &vsyscall_sysenter_start;
254		- vsyscall_len = &vsyscall_sysenter_end - &vsyscall_sysenter_start;
	310	+ vsyscall = &vdso32_sysenter_start;
	311	+ vsyscall_len = &vdso32_sysenter_end - &vdso32_sysenter_start;
255	312	}
256	313
257	314	memcpy(syscall_page, vsyscall, vsyscall_len);
258	315
...	...	@@ -284,7 +341,9 @@
284	341	ret = addr;
285	342	goto up_fail;
286	343	}
	344	+ }
287	345
	346	+ if (compat_uses_vma \|\| !compat) {
288	347	/*
289	348	* MAYWRITE to allow gdb to COW and set breakpoints
290	349	*
...	...	@@ -298,7 +357,7 @@
298	357	VM_READ\|VM_EXEC\|
299	358	VM_MAYREAD\|VM_MAYWRITE\|VM_MAYEXEC\|
300	359	VM_ALWAYSDUMP,
301		- syscall_pages);
	360	+ vdso32_pages);
302	361
303	362	if (ret)
304	363	goto up_fail;
...	...	@@ -314,6 +373,12 @@
314	373	return ret;
315	374	}
316	375
	376	+#ifdef CONFIG_X86_64
	377	+
	378	+__initcall(sysenter_setup);
	379	+
	380	+#else /* CONFIG_X86_32 */
	381	+
317	382	const char arch_vma_name(struct vm_area_struct vma)
318	383	{
319	384	if (vma->vm_mm && vma->vm_start == (long)vma->vm_mm->context.vdso)
...	...	@@ -342,4 +407,6 @@
342	407	{
343	408	return 0;
344	409	}
	410	+
	411	+#endif /* CONFIG_X86_64 */
...	...	@@ -2,15 +2,19 @@
2	2
3	3	__INITDATA
4	4
5		- .globl vsyscall_int80_start, vsyscall_int80_end
6		-vsyscall_int80_start:
	5	+ .globl vdso32_default_start, vdso32_default_end
	6	+vdso32_default_start:
	7	+#ifdef CONFIG_X86_32
7	8	.incbin "arch/x86/vdso/vdso32-int80.so"
8		-vsyscall_int80_end:
	9	+#else
	10	+ .incbin "arch/x86/vdso/vdso32-syscall.so"
	11	+#endif
	12	+vdso32_default_end:
9	13
10		- .globl vsyscall_sysenter_start, vsyscall_sysenter_end
11		-vsyscall_sysenter_start:
	14	+ .globl vdso32_sysenter_start, vdso32_sysenter_end
	15	+vdso32_sysenter_start:
12	16	.incbin "arch/x86/vdso/vdso32-sysenter.so"
13		-vsyscall_sysenter_end:
	17	+vdso32_sysenter_end:
14	18
15	19	__FINIT
...	...	@@ -62,8 +62,8 @@
62	62	*/
63	63	static void fiddle_vdso(void)
64	64	{
65		- extern char vsyscall_int80_start;
66		- u32 *mask = VDSO32_SYMBOL(&vsyscall_int80_start, NOTE_MASK);
	65	+ extern const char vdso32_default_start;
	66	+ u32 *mask = VDSO32_SYMBOL(&vdso32_default_start, NOTE_MASK);
67	67	*mask \|= 1 << VDSO_NOTE_NONEGSEG_BIT;
68	68	}
69	69
...	...	@@ -74,17 +74,19 @@
74	74
75	75	#ifdef __KERNEL__
76	76
	77	+/*
	78	+ * This is used to ensure we don't load something for the wrong architecture.
	79	+ */
	80	+#define elf_check_arch_ia32(x) \
	81	+ (((x)->e_machine == EM_386) \|\| ((x)->e_machine == EM_486))
	82	+
77	83	#ifdef CONFIG_X86_32
78	84	#include <asm/processor.h>
79	85	#include <asm/system.h> /* for savesegment */
80	86	#include <asm/desc.h>
81	87	#include <asm/vdso.h>
82	88
83		-/*
84		- * This is used to ensure we don't load something for the wrong architecture.
85		- */
86		-#define elf_check_arch(x) \
87		- (((x)->e_machine == EM_386) \|\| ((x)->e_machine == EM_486))
	89	+#define elf_check_arch(x) elf_check_arch_ia32(x)
88	90
89	91	/* SVR4/i386 ABI (pages 3-31, 3-32) says that when the program starts %edx
90	92	contains a pointer to a function which might be registered using `atexit'.
91	93
...	...	@@ -247,11 +249,7 @@
247	249	#define ELF_CORE_XFPREG_TYPE NT_PRXFPREG
248	250
249	251	#define VDSO_HIGH_BASE (__fix_to_virt(FIX_VDSO))
250		-#define VDSO_CURRENT_BASE ((unsigned long)current->mm->context.vdso)
251	252
252		-#define VDSO_ENTRY \
253		- ((unsigned long) VDSO32_SYMBOL(VDSO_CURRENT_BASE, vsyscall))
254		-
255	253	/* update AT_VECTOR_SIZE_ARCH if the number of NEW_AUX_ENT entries changes */
256	254
257	255	#define ARCH_DLINFO \
...	...	@@ -262,6 +260,8 @@
262	260
263	261	#else /* CONFIG_X86_32 */
264	262
	263	+#define VDSO_HIGH_BASE 0xffffe000U /* CONFIG_COMPAT_VDSO address */
	264	+
265	265	/* 1GB for 64bit, 8MB for 32bit */
266	266	#define STACK_RND_MASK (test_thread_flag(TIF_IA32) ? 0x7ff : 0x3fffff)
267	267
...	...	@@ -271,6 +271,11 @@
271	271	} while (0)
272	272
273	273	#endif /* !CONFIG_X86_32 */
	274	+
	275	+#define VDSO_CURRENT_BASE ((unsigned long)current->mm->context.vdso)
	276	+
	277	+#define VDSO_ENTRY \
	278	+ ((unsigned long) VDSO32_SYMBOL(VDSO_CURRENT_BASE, vsyscall))
274	279
275	280	struct linux_binprm;
276	281