[PATCH 0/2] Update vDSO examples for 32-bit systems

[PATCH 0/2] Update vDSO examples for 32-bit systems

[Andy Lutomirski]

We have useful code in the 32-bit x86 vDSO now; update the sample code.

Despite the late submission, this might be 3.16 material: it's just
documentation.  If I'd been more on top of this, I would have done it
for 3.15.

Andy Lutomirski (2):
  doc,vdso: Rename vdso_test.c to vdso_standalone_test_x86.c
  doc,vdso: Make vDSO examples more portable

 Documentation/vDSO/parse_vdso.c               |  59 ++++++------
 Documentation/vDSO/vdso_standalone_test_x86.c | 128 ++++++++++++++++++++++++++
 Documentation/vDSO/vdso_test.c                | 103 +++++----------------
 3 files changed, 181 insertions(+), 109 deletions(-)
 create mode 100644 Documentation/vDSO/vdso_standalone_test_x86.c

-- 
1.9.3

[PATCH 1/2] doc,vdso: Rename vdso_test.c to vdso_standalone_test_x86.c

[Andy Lutomirski]

This thing is hopelessly x86_64-specific: it's an example of how to
access the vDSO without any runtime support at all.

Signed-off-by: Andy Lutomirski <luto@amacapital.net>
---
 Documentation/vDSO/vdso_standalone_test_x86.c | 111 ++++++++++++++++++++++++++
 Documentation/vDSO/vdso_test.c                | 111 --------------------------
 2 files changed, 111 insertions(+), 111 deletions(-)
 create mode 100644 Documentation/vDSO/vdso_standalone_test_x86.c
 delete mode 100644 Documentation/vDSO/vdso_test.c

diff --git a/Documentation/vDSO/vdso_standalone_test_x86.c b/Documentation/vDSO/vdso_standalone_test_x86.c
new file mode 100644
index 0000000..1523f5d
--- /dev/null
+++ b/Documentation/vDSO/vdso_standalone_test_x86.c
@@ -0,0 +1,111 @@
+/*
+ * vdso_test.c: Sample code to test parse_vdso.c on x86_64
+ * Copyright (c) 2011 Andy Lutomirski
+ * Subject to the GNU General Public License, version 2
+ *
+ * You can amuse yourself by compiling with:
+ * gcc -std=gnu99 -nostdlib
+ *     -Os -fno-asynchronous-unwind-tables -flto
+ *      vdso_standalone_test_x86.c parse_vdso.c
+ * to generate a small binary with no dependencies at all.
+ */
+
+#include <sys/syscall.h>
+#include <sys/time.h>
+#include <unistd.h>
+#include <stdint.h>
+
+extern void *vdso_sym(const char *version, const char *name);
+extern void vdso_init_from_sysinfo_ehdr(uintptr_t base);
+extern void vdso_init_from_auxv(void *auxv);
+
+/* We need a libc functions... */
+int strcmp(const char *a, const char *b)
+{
+	/* This implementation is buggy: it never returns -1. */
+	while (*a || *b) {
+		if (*a != *b)
+			return 1;
+		if (*a == 0 || *b == 0)
+			return 1;
+		a++;
+		b++;
+	}
+
+	return 0;
+}
+
+/* ...and two syscalls.  This is x86_64-specific. */
+static inline long linux_write(int fd, const void *data, size_t len)
+{
+
+	long ret;
+	asm volatile ("syscall" : "=a" (ret) : "a" (__NR_write),
+		      "D" (fd), "S" (data), "d" (len) :
+		      "cc", "memory", "rcx",
+		      "r8", "r9", "r10", "r11" );
+	return ret;
+}
+
+static inline void linux_exit(int code)
+{
+	asm volatile ("syscall" : : "a" (__NR_exit), "D" (code));
+}
+
+void to_base10(char *lastdig, uint64_t n)
+{
+	while (n) {
+		*lastdig = (n % 10) + '0';
+		n /= 10;
+		lastdig--;
+	}
+}
+
+__attribute__((externally_visible)) void c_main(void **stack)
+{
+	/* Parse the stack */
+	long argc = (long)*stack;
+	stack += argc + 2;
+
+	/* Now we're pointing at the environment.  Skip it. */
+	while(*stack)
+		stack++;
+	stack++;
+
+	/* Now we're pointing at auxv.  Initialize the vDSO parser. */
+	vdso_init_from_auxv((void *)stack);
+
+	/* Find gettimeofday. */
+	typedef long (*gtod_t)(struct timeval *tv, struct timezone *tz);
+	gtod_t gtod = (gtod_t)vdso_sym("LINUX_2.6", "__vdso_gettimeofday");
+
+	if (!gtod)
+		linux_exit(1);
+
+	struct timeval tv;
+	long ret = gtod(&tv, 0);
+
+	if (ret == 0) {
+		char buf[] = "The time is                     .000000\n";
+		to_base10(buf + 31, tv.tv_sec);
+		to_base10(buf + 38, tv.tv_usec);
+		linux_write(1, buf, sizeof(buf) - 1);
+	} else {
+		linux_exit(ret);
+	}
+
+	linux_exit(0);
+}
+
+/*
+ * This is the real entry point.  It passes the initial stack into
+ * the C entry point.
+ */
+asm (
+	".text\n"
+	".global _start\n"
+        ".type _start,@function\n"
+        "_start:\n\t"
+        "mov %rsp,%rdi\n\t"
+        "jmp c_main"
+	);
diff --git a/Documentation/vDSO/vdso_test.c b/Documentation/vDSO/vdso_test.c
deleted file mode 100644
index fff6334..0000000
--- a/Documentation/vDSO/vdso_test.c
+++ /dev/null
@@ -1,111 +0,0 @@
-/*
- * vdso_test.c: Sample code to test parse_vdso.c on x86_64
- * Copyright (c) 2011 Andy Lutomirski
- * Subject to the GNU General Public License, version 2
- *
- * You can amuse yourself by compiling with:
- * gcc -std=gnu99 -nostdlib
- *     -Os -fno-asynchronous-unwind-tables -flto
- *      vdso_test.c parse_vdso.c -o vdso_test
- * to generate a small binary with no dependencies at all.
- */
-
-#include <sys/syscall.h>
-#include <sys/time.h>
-#include <unistd.h>
-#include <stdint.h>
-
-extern void *vdso_sym(const char *version, const char *name);
-extern void vdso_init_from_sysinfo_ehdr(uintptr_t base);
-extern void vdso_init_from_auxv(void *auxv);
-
-/* We need a libc functions... */
-int strcmp(const char *a, const char *b)
-{
-	/* This implementation is buggy: it never returns -1. */
-	while (*a || *b) {
-		if (*a != *b)
-			return 1;
-		if (*a == 0 || *b == 0)
-			return 1;
-		a++;
-		b++;
-	}
-
-	return 0;
-}
-
-/* ...and two syscalls.  This is x86_64-specific. */
-static inline long linux_write(int fd, const void *data, size_t len)
-{
-
-	long ret;
-	asm volatile ("syscall" : "=a" (ret) : "a" (__NR_write),
-		      "D" (fd), "S" (data), "d" (len) :
-		      "cc", "memory", "rcx",
-		      "r8", "r9", "r10", "r11" );
-	return ret;
-}
-
-static inline void linux_exit(int code)
-{
-	asm volatile ("syscall" : : "a" (__NR_exit), "D" (code));
-}
-
-void to_base10(char *lastdig, uint64_t n)
-{
-	while (n) {
-		*lastdig = (n % 10) + '0';
-		n /= 10;
-		lastdig--;
-	}
-}
-
-__attribute__((externally_visible)) void c_main(void **stack)
-{
-	/* Parse the stack */
-	long argc = (long)*stack;
-	stack += argc + 2;
-
-	/* Now we're pointing at the environment.  Skip it. */
-	while(*stack)
-		stack++;
-	stack++;
-
-	/* Now we're pointing at auxv.  Initialize the vDSO parser. */
-	vdso_init_from_auxv((void *)stack);
-
-	/* Find gettimeofday. */
-	typedef long (*gtod_t)(struct timeval *tv, struct timezone *tz);
-	gtod_t gtod = (gtod_t)vdso_sym("LINUX_2.6", "__vdso_gettimeofday");
-
-	if (!gtod)
-		linux_exit(1);
-
-	struct timeval tv;
-	long ret = gtod(&tv, 0);
-
-	if (ret == 0) {
-		char buf[] = "The time is                     .000000\n";
-		to_base10(buf + 31, tv.tv_sec);
-		to_base10(buf + 38, tv.tv_usec);
-		linux_write(1, buf, sizeof(buf) - 1);
-	} else {
-		linux_exit(ret);
-	}
-
-	linux_exit(0);
-}
-
-/*
- * This is the real entry point.  It passes the initial stack into
- * the C entry point.
- */
-asm (
-	".text\n"
-	".global _start\n"
-        ".type _start,@function\n"
-        "_start:\n\t"
-        "mov %rsp,%rdi\n\t"
-        "jmp c_main"
-	);
-- 
1.9.3

[PATCH 2/2] doc,vdso: Make vDSO examples more portable

[Andy Lutomirski]

This adds a new vdso_test.c that's written entirely in C.  It also
makes all of the vDSO examples work on 32-bit x86.

Cc: Stefani Seibold <stefani@seibold.net>
Signed-off-by: Andy Lutomirski <luto@amacapital.net>
---
 Documentation/vDSO/parse_vdso.c               | 59 +++++++++++++++------------
 Documentation/vDSO/vdso_standalone_test_x86.c | 45 +++++++++++++-------
 Documentation/vDSO/vdso_test.c                | 50 +++++++++++++++++++++++
 3 files changed, 113 insertions(+), 41 deletions(-)
 create mode 100644 Documentation/vDSO/vdso_test.c

diff --git a/Documentation/vDSO/parse_vdso.c b/Documentation/vDSO/parse_vdso.c
index 8587020..ca23f06 100644
--- a/Documentation/vDSO/parse_vdso.c
+++ b/Documentation/vDSO/parse_vdso.c
@@ -1,6 +1,6 @@
 /*
  * parse_vdso.c: Linux reference vDSO parser
- * Written by Andrew Lutomirski, 2011.
+ * Written by Andrew Lutomirski, 2011-2014.
  *
  * This code is meant to be linked in to various programs that run on Linux.
  * As such, it is available with as few restrictions as possible.  This file
@@ -11,7 +11,7 @@
  * it starts a program.  It works equally well in statically and dynamically
  * linked binaries.
  *
- * This code is tested on x86_64.  In principle it should work on any 64-bit
+ * This code is tested on x86.  In principle it should work on any
  * architecture that has a vDSO.
  */
 
@@ -20,6 +20,8 @@
 #include <string.h>
 #include <elf.h>
 
+#include <asm/bitsperlong.h>
+
 /*
  * To use this vDSO parser, first call one of the vdso_init_* functions.
  * If you've already parsed auxv, then pass the value of AT_SYSINFO_EHDR
@@ -45,11 +47,14 @@ extern void *vdso_sym(const char *version, const char *name);
 
 
 /* And here's the code. */
-
-#ifndef __x86_64__
-# error Not yet ported to non-x86_64 architectures
+#ifndef __BITS_PER_LONG
+#error We need __BITS_PER_LONG
 #endif
 
+#define ELF_BITS2(bits, x) Elf##bits##_##x
+#define ELF_BITS(bits, x) ELF_BITS2(bits, x)
+#define ELF(x) ELF_BITS(__BITS_PER_LONG, x)
+
 static struct vdso_info
 {
 	bool valid;
@@ -59,14 +64,14 @@ static struct vdso_info
 	uintptr_t load_offset;  /* load_addr - recorded vaddr */
 
 	/* Symbol table */
-	Elf64_Sym *symtab;
+	ELF(Sym) *symtab;
 	const char *symstrings;
-	Elf64_Word *bucket, *chain;
-	Elf64_Word nbucket, nchain;
+	ELF(Word) *bucket, *chain;
+	ELF(Word) nbucket, nchain;
 
 	/* Version table */
-	Elf64_Versym *versym;
-	Elf64_Verdef *verdef;
+	ELF(Versym) *versym;
+	ELF(Verdef) *verdef;
 } vdso_info;
 
 /* Straight from the ELF specification. */
@@ -92,9 +97,9 @@ void vdso_init_from_sysinfo_ehdr(uintptr_t base)
 
 	vdso_info.load_addr = base;
 
-	Elf64_Ehdr *hdr = (Elf64_Ehdr*)base;
-	Elf64_Phdr *pt = (Elf64_Phdr*)(vdso_info.load_addr + hdr->e_phoff);
-	Elf64_Dyn *dyn = 0;
+	ELF(Ehdr) *hdr = (ELF(Ehdr)*)base;
+	ELF(Phdr) *pt = (ELF(Phdr)*)(vdso_info.load_addr + hdr->e_phoff);
+	ELF(Dyn) *dyn = 0;
 
 	/*
 	 * We need two things from the segment table: the load offset
@@ -108,7 +113,7 @@ void vdso_init_from_sysinfo_ehdr(uintptr_t base)
 				+ (uintptr_t)pt[i].p_offset
 				- (uintptr_t)pt[i].p_vaddr;
 		} else if (pt[i].p_type == PT_DYNAMIC) {
-			dyn = (Elf64_Dyn*)(base + pt[i].p_offset);
+			dyn = (ELF(Dyn)*)(base + pt[i].p_offset);
 		}
 	}
 
@@ -118,7 +123,7 @@ void vdso_init_from_sysinfo_ehdr(uintptr_t base)
 	/*
 	 * Fish out the useful bits of the dynamic table.
 	 */
-	Elf64_Word *hash = 0;
+	ELF(Word) *hash = 0;
 	vdso_info.symstrings = 0;
 	vdso_info.symtab = 0;
 	vdso_info.versym = 0;
@@ -131,22 +136,22 @@ void vdso_init_from_sysinfo_ehdr(uintptr_t base)
 				 + vdso_info.load_offset);
 			break;
 		case DT_SYMTAB:
-			vdso_info.symtab = (Elf64_Sym *)
+			vdso_info.symtab = (ELF(Sym) *)
 				((uintptr_t)dyn[i].d_un.d_ptr
 				 + vdso_info.load_offset);
 			break;
 		case DT_HASH:
-			hash = (Elf64_Word *)
+			hash = (ELF(Word) *)
 				((uintptr_t)dyn[i].d_un.d_ptr
 				 + vdso_info.load_offset);
 			break;
 		case DT_VERSYM:
-			vdso_info.versym = (Elf64_Versym *)
+			vdso_info.versym = (ELF(Versym) *)
 				((uintptr_t)dyn[i].d_un.d_ptr
 				 + vdso_info.load_offset);
 			break;
 		case DT_VERDEF:
-			vdso_info.verdef = (Elf64_Verdef *)
+			vdso_info.verdef = (ELF(Verdef) *)
 				((uintptr_t)dyn[i].d_un.d_ptr
 				 + vdso_info.load_offset);
 			break;
@@ -168,8 +173,8 @@ void vdso_init_from_sysinfo_ehdr(uintptr_t base)
 	vdso_info.valid = true;
 }
 
-static bool vdso_match_version(Elf64_Versym ver,
-			       const char *name, Elf64_Word hash)
+static bool vdso_match_version(ELF(Versym) ver,
+			       const char *name, ELF(Word) hash)
 {
 	/*
 	 * This is a helper function to check if the version indexed by
@@ -188,7 +193,7 @@ static bool vdso_match_version(Elf64_Versym ver,
 
 	/* First step: find the version definition */
 	ver &= 0x7fff;  /* Apparently bit 15 means "hidden" */
-	Elf64_Verdef *def = vdso_info.verdef;
+	ELF(Verdef) *def = vdso_info.verdef;
 	while(true) {
 		if ((def->vd_flags & VER_FLG_BASE) == 0
 		    && (def->vd_ndx & 0x7fff) == ver)
@@ -197,11 +202,11 @@ static bool vdso_match_version(Elf64_Versym ver,
 		if (def->vd_next == 0)
 			return false;  /* No definition. */
 
-		def = (Elf64_Verdef *)((char *)def + def->vd_next);
+		def = (ELF(Verdef) *)((char *)def + def->vd_next);
 	}
 
 	/* Now figure out whether it matches. */
-	Elf64_Verdaux *aux = (Elf64_Verdaux*)((char *)def + def->vd_aux);
+	ELF(Verdaux) *aux = (ELF(Verdaux)*)((char *)def + def->vd_aux);
 	return def->vd_hash == hash
 		&& !strcmp(name, vdso_info.symstrings + aux->vda_name);
 }
@@ -213,10 +218,10 @@ void *vdso_sym(const char *version, const char *name)
 		return 0;
 
 	ver_hash = elf_hash(version);
-	Elf64_Word chain = vdso_info.bucket[elf_hash(name) % vdso_info.nbucket];
+	ELF(Word) chain = vdso_info.bucket[elf_hash(name) % vdso_info.nbucket];
 
 	for (; chain != STN_UNDEF; chain = vdso_info.chain[chain]) {
-		Elf64_Sym *sym = &vdso_info.symtab[chain];
+		ELF(Sym) *sym = &vdso_info.symtab[chain];
 
 		/* Check for a defined global or weak function w/ right name. */
 		if (ELF64_ST_TYPE(sym->st_info) != STT_FUNC)
@@ -243,7 +248,7 @@ void *vdso_sym(const char *version, const char *name)
 
 void vdso_init_from_auxv(void *auxv)
 {
-	Elf64_auxv_t *elf_auxv = auxv;
+	ELF(auxv_t) *elf_auxv = auxv;
 	for (int i = 0; elf_auxv[i].a_type != AT_NULL; i++)
 	{
 		if (elf_auxv[i].a_type == AT_SYSINFO_EHDR) {
diff --git a/Documentation/vDSO/vdso_standalone_test_x86.c b/Documentation/vDSO/vdso_standalone_test_x86.c
index 1523f5d..d462402 100644
--- a/Documentation/vDSO/vdso_standalone_test_x86.c
+++ b/Documentation/vDSO/vdso_standalone_test_x86.c
@@ -1,13 +1,14 @@
 /*
- * vdso_test.c: Sample code to test parse_vdso.c on x86_64
- * Copyright (c) 2011 Andy Lutomirski
+ * vdso_test.c: Sample code to test parse_vdso.c on x86
+ * Copyright (c) 2011-2014 Andy Lutomirski
  * Subject to the GNU General Public License, version 2
  *
  * You can amuse yourself by compiling with:
  * gcc -std=gnu99 -nostdlib
- *     -Os -fno-asynchronous-unwind-tables -flto
+ *     -Os -fno-asynchronous-unwind-tables -flto -lgcc_s
  *      vdso_standalone_test_x86.c parse_vdso.c
- * to generate a small binary with no dependencies at all.
+ * to generate a small binary.  On x86_64, you can omit -lgcc_s
+ * if you want the binary to be completely standalone.
  */
 
 #include <sys/syscall.h>
@@ -35,21 +36,31 @@ int strcmp(const char *a, const char *b)
 	return 0;
 }
 
-/* ...and two syscalls.  This is x86_64-specific. */
-static inline long linux_write(int fd, const void *data, size_t len)
+/* ...and two syscalls.  This is x86-specific. */
+static inline long x86_syscall3(long nr, long a0, long a1, long a2)
 {
-
 	long ret;
-	asm volatile ("syscall" : "=a" (ret) : "a" (__NR_write),
-		      "D" (fd), "S" (data), "d" (len) :
+#ifdef __x86_64__
+	asm volatile ("syscall" : "=a" (ret) : "a" (nr),
+		      "D" (a0), "S" (a1), "d" (a2) :
 		      "cc", "memory", "rcx",
 		      "r8", "r9", "r10", "r11" );
+#else
+	asm volatile ("int $0x80" : "=a" (ret) : "a" (nr),
+		      "b" (a0), "c" (a1), "d" (a2) :
+		      "cc", "memory" );
+#endif
 	return ret;
 }
 
+static inline long linux_write(int fd, const void *data, size_t len)
+{
+	return x86_syscall3(__NR_write, fd, (long)data, (long)len);
+}
+
 static inline void linux_exit(int code)
 {
-	asm volatile ("syscall" : : "a" (__NR_exit), "D" (code));
+	x86_syscall3(__NR_exit, code, 0, 0);
 }
 
 void to_base10(char *lastdig, uint64_t n)
@@ -104,8 +115,14 @@ __attribute__((externally_visible)) void c_main(void **stack)
 asm (
 	".text\n"
 	".global _start\n"
-        ".type _start,@function\n"
-        "_start:\n\t"
-        "mov %rsp,%rdi\n\t"
-        "jmp c_main"
+	".type _start,@function\n"
+	"_start:\n\t"
+#ifdef __x86_64__
+	"mov %rsp,%rdi\n\t"
+	"jmp c_main"
+#else
+	"push %esp\n\t"
+	"call c_main\n\t"
+	"int $3"
+#endif
 	);
diff --git a/Documentation/vDSO/vdso_test.c b/Documentation/vDSO/vdso_test.c
new file mode 100644
index 0000000..75b1da2
--- /dev/null
+++ b/Documentation/vDSO/vdso_test.c
@@ -0,0 +1,50 @@
+/*
+ * vdso_test.c: Sample code to test parse_vdso.c
+ * Copyright (c) 2014 Andy Lutomirski
+ * Subject to the GNU General Public License, version 2
+ *
+ * Compile with:
+ * gcc -std=gnu99 vdso_test.c parse_vdso.c
+ *
+ * Tested on x86, 32-bit and 64-bit.  It may work on other architectures, too.
+ */
+
+#include <stdint.h>
+#include <elf.h>
+#include <stdio.h>
+#include <sys/auxv.h>
+#include <sys/time.h>
+
+extern void *vdso_sym(const char *version, const char *name);
+extern void vdso_init_from_sysinfo_ehdr(uintptr_t base);
+extern void vdso_init_from_auxv(void *auxv);
+
+int main(int argc, char **argv)
+{
+	unsigned long sysinfo_ehdr = getauxval(AT_SYSINFO_EHDR);
+	if (!sysinfo_ehdr) {
+		printf("AT_SYSINFO_EHDR is not present!\n");
+		return 0;
+	}
+
+	vdso_init_from_sysinfo_ehdr(getauxval(AT_SYSINFO_EHDR));
+
+	/* Find gettimeofday. */
+	typedef long (*gtod_t)(struct timeval *tv, struct timezone *tz);
+	gtod_t gtod = (gtod_t)vdso_sym("LINUX_2.6", "__vdso_gettimeofday");
+
+	if (!gtod)
+		printf("Could not find __vdso_gettimeofday\n");
+
+	struct timeval tv;
+	long ret = gtod(&tv, 0);
+
+	if (ret == 0) {
+		printf("The time is %lld.%06lld\n",
+		       (long long)tv.tv_sec, (long long)tv.tv_usec);
+	} else {
+		printf("__vdso_gettimeofday failed\n");
+	}
+
+	return 0;
+}
-- 
1.9.3