21 files changed, 3729 insertions, 0 deletions
diff --git a/x86/.gitignore b/x86/.gitignore
new file mode 100644
index 0000000..15034fe
--- /dev/null
+++ b/x86/.gitignore
@@ -0,0 +1,2 @@
+*_32
+*_64
diff --git a/x86/Makefile b/x86/Makefile
new file mode 100644
index 0000000..eabcff4
--- /dev/null
+++ b/x86/Makefile
@@ -0,0 +1,65 @@
+all:
+
+include ../lib.mk
+
+.PHONY: all all_32 all_64 warn_32bit_failure clean
+
+TARGETS_C_BOTHBITS := single_step_syscall sysret_ss_attrs ldt_gdt syscall_nt ptrace_syscall
+TARGETS_C_32BIT_ONLY := entry_from_vm86 syscall_arg_fault sigreturn test_syscall_vdso unwind_vdso \
+			test_FCMOV test_FCOMI test_FISTTP
+
+TARGETS_C_32BIT_ALL := $(TARGETS_C_BOTHBITS) $(TARGETS_C_32BIT_ONLY)
+BINARIES_32 := $(TARGETS_C_32BIT_ALL:%=%_32)
+BINARIES_64 := $(TARGETS_C_BOTHBITS:%=%_64)
+
+CFLAGS := -O2 -g -std=gnu99 -pthread -Wall
+
+UNAME_M := $(shell uname -m)
+CAN_BUILD_I386 := $(shell ./check_cc.sh $(CC) trivial_32bit_program.c -m32)
+CAN_BUILD_X86_64 := $(shell ./check_cc.sh $(CC) trivial_64bit_program.c)
+
+ifeq ($(CAN_BUILD_I386),1)
+all: all_32
+TEST_PROGS += $(BINARIES_32)
+endif
+
+ifeq ($(CAN_BUILD_X86_64),1)
+all: all_64
+TEST_PROGS += $(BINARIES_64)
+endif
+
+all_32: $(BINARIES_32)
+
+all_64: $(BINARIES_64)
+
+clean:
+	$(RM) $(BINARIES_32) $(BINARIES_64)
+
+$(TARGETS_C_32BIT_ALL:%=%_32): %_32: %.c
+	$(CC) -m32 -o $@ $(CFLAGS) $(EXTRA_CFLAGS) $^ -lrt -ldl -lm
+
+$(TARGETS_C_BOTHBITS:%=%_64): %_64: %.c
+	$(CC) -m64 -o $@ $(CFLAGS) $(EXTRA_CFLAGS) $^ -lrt -ldl
+
+# x86_64 users should be encouraged to install 32-bit libraries
+ifeq ($(CAN_BUILD_I386)$(CAN_BUILD_X86_64),01)
+all: warn_32bit_failure
+
+warn_32bit_failure:
+	@echo "Warning: you seem to have a broken 32-bit build" 2>&1; 	\
+	echo "environment.  This will reduce test coverage of 64-bit" 2>&1; \
+	echo "kernels.  If you are using a Debian-like distribution," 2>&1; \
+	echo "try:"; 2>&1; \
+	echo "";							\
+	echo "  apt-get install gcc-multilib libc6-i386 libc6-dev-i386"; \
+	echo "";							\
+	echo "If you are using a Fedora-like distribution, try:";	\
+	echo "";							\
+	echo "  yum install glibc-devel.*i686";				\
+	exit 0;
+endif
+
+# Some tests have additional dependencies.
+sysret_ss_attrs_64: thunks.S
+ptrace_syscall_32: raw_syscall_helper_32.S
+test_syscall_vdso_32: thunks_32.S
diff --git a/x86/check_cc.sh b/x86/check_cc.sh
new file mode 100755
index 0000000..172d329
--- /dev/null
+++ b/x86/check_cc.sh
@@ -0,0 +1,16 @@
+#!/bin/sh
+# check_cc.sh - Helper to test userspace compilation support
+# Copyright (c) 2015 Andrew Lutomirski
+# GPL v2
+
+CC="$1"
+TESTPROG="$2"
+shift 2
+
+if "$CC" -o /dev/null "$TESTPROG" -O0 "$@" 2>/dev/null; then
+    echo 1
+else
+    echo 0
+fi
+
+exit 0
diff --git a/x86/entry_from_vm86.c b/x86/entry_from_vm86.c
new file mode 100644
index 0000000..d075ea0
--- /dev/null
+++ b/x86/entry_from_vm86.c
@@ -0,0 +1,238 @@
+/*
+ * entry_from_vm86.c - tests kernel entries from vm86 mode
+ * Copyright (c) 2014-2015 Andrew Lutomirski
+ *
+ * This exercises a few paths that need to special-case vm86 mode.
+ *
+ * GPL v2.
+ */
+
+#define _GNU_SOURCE
+
+#include <assert.h>
+#include <stdlib.h>
+#include <sys/syscall.h>
+#include <sys/signal.h>
+#include <sys/ucontext.h>
+#include <unistd.h>
+#include <stdio.h>
+#include <string.h>
+#include <inttypes.h>
+#include <sys/mman.h>
+#include <err.h>
+#include <stddef.h>
+#include <stdbool.h>
+#include <errno.h>
+#include <sys/vm86.h>
+
+static unsigned long load_addr = 0x10000;
+static int nerrs = 0;
+
+static void sethandler(int sig, void (*handler)(int, siginfo_t *, void *),
+		       int flags)
+{
+	struct sigaction sa;
+	memset(&sa, 0, sizeof(sa));
+	sa.sa_sigaction = handler;
+	sa.sa_flags = SA_SIGINFO | flags;
+	sigemptyset(&sa.sa_mask);
+	if (sigaction(sig, &sa, 0))
+		err(1, "sigaction");
+}
+
+static void clearhandler(int sig)
+{
+	struct sigaction sa;
+	memset(&sa, 0, sizeof(sa));
+	sa.sa_handler = SIG_DFL;
+	sigemptyset(&sa.sa_mask);
+	if (sigaction(sig, &sa, 0))
+		err(1, "sigaction");
+}
+
+static sig_atomic_t got_signal;
+
+static void sighandler(int sig, siginfo_t *info, void *ctx_void)
+{
+	ucontext_t *ctx = (ucontext_t*)ctx_void;
+
+	if (ctx->uc_mcontext.gregs[REG_EFL] & X86_EFLAGS_VM ||
+	    (ctx->uc_mcontext.gregs[REG_CS] & 3) != 3) {
+		printf("[FAIL]\tSignal frame should not reflect vm86 mode\n");
+		nerrs++;
+	}
+
+	const char *signame;
+	if (sig == SIGSEGV)
+		signame = "SIGSEGV";
+	else if (sig == SIGILL)
+		signame = "SIGILL";
+	else
+		signame = "unexpected signal";
+
+	printf("[INFO]\t%s: FLAGS = 0x%lx, CS = 0x%hx\n", signame,
+	       (unsigned long)ctx->uc_mcontext.gregs[REG_EFL],
+	       (unsigned short)ctx->uc_mcontext.gregs[REG_CS]);
+
+	got_signal = 1;
+}
+
+asm (
+	".pushsection .rodata\n\t"
+	".type vmcode_bound, @object\n\t"
+	"vmcode:\n\t"
+	"vmcode_bound:\n\t"
+	".code16\n\t"
+	"bound %ax, (2048)\n\t"
+	"int3\n\t"
+	"vmcode_sysenter:\n\t"
+	"sysenter\n\t"
+	"vmcode_syscall:\n\t"
+	"syscall\n\t"
+	"vmcode_sti:\n\t"
+	"sti\n\t"
+	"vmcode_int3:\n\t"
+	"int3\n\t"
+	"vmcode_int80:\n\t"
+	"int $0x80\n\t"
+	".size vmcode, . - vmcode\n\t"
+	"end_vmcode:\n\t"
+	".code32\n\t"
+	".popsection"
+	);
+
+extern unsigned char vmcode[], end_vmcode[];
+extern unsigned char vmcode_bound[], vmcode_sysenter[], vmcode_syscall[],
+	vmcode_sti[], vmcode_int3[], vmcode_int80[];
+
+/* Returns false if the test was skipped. */
+static bool do_test(struct vm86plus_struct *v86, unsigned long eip,
+		    unsigned int rettype, unsigned int retarg,
+		    const char *text)
+{
+	long ret;
+
+	printf("[RUN]\t%s from vm86 mode\n", text);
+	v86->regs.eip = eip;
+	ret = vm86(VM86_ENTER, v86);
+
+	if (ret == -1 && (errno == ENOSYS || errno == EPERM)) {
+		printf("[SKIP]\tvm86 %s\n",
+		       errno == ENOSYS ? "not supported" : "not allowed");
+		return false;
+	}
+
+	if (VM86_TYPE(ret) == VM86_INTx) {
+		char trapname[32];
+		int trapno = VM86_ARG(ret);
+		if (trapno == 13)
+			strcpy(trapname, "GP");
+		else if (trapno == 5)
+			strcpy(trapname, "BR");
+		else if (trapno == 14)
+			strcpy(trapname, "PF");
+		else
+			sprintf(trapname, "%d", trapno);
+
+		printf("[INFO]\tExited vm86 mode due to #%s\n", trapname);
+	} else if (VM86_TYPE(ret) == VM86_UNKNOWN) {
+		printf("[INFO]\tExited vm86 mode due to unhandled GP fault\n");
+	} else if (VM86_TYPE(ret) == VM86_TRAP) {
+		printf("[INFO]\tExited vm86 mode due to a trap (arg=%ld)\n",
+		       VM86_ARG(ret));
+	} else if (VM86_TYPE(ret) == VM86_SIGNAL) {
+		printf("[INFO]\tExited vm86 mode due to a signal\n");
+	} else if (VM86_TYPE(ret) == VM86_STI) {
+		printf("[INFO]\tExited vm86 mode due to STI\n");
+	} else {
+		printf("[INFO]\tExited vm86 mode due to type %ld, arg %ld\n",
+		       VM86_TYPE(ret), VM86_ARG(ret));
+	}
+
+	if (rettype == -1 ||
+	    (VM86_TYPE(ret) == rettype && VM86_ARG(ret) == retarg)) {
+		printf("[OK]\tReturned correctly\n");
+	} else {
+		printf("[FAIL]\tIncorrect return reason\n");
+		nerrs++;
+	}
+
+	return true;
+}
+
+int main(void)
+{
+	struct vm86plus_struct v86;
+	unsigned char *addr = mmap((void *)load_addr, 4096,
+				   PROT_READ | PROT_WRITE | PROT_EXEC,
+				   MAP_ANONYMOUS | MAP_PRIVATE, -1,0);
+	if (addr != (unsigned char *)load_addr)
+		err(1, "mmap");
+
+	memcpy(addr, vmcode, end_vmcode - vmcode);
+	addr[2048] = 2;
+	addr[2050] = 3;
+
+	memset(&v86, 0, sizeof(v86));
+
+	v86.regs.cs = load_addr / 16;
+	v86.regs.ss = load_addr / 16;
+	v86.regs.ds = load_addr / 16;
+	v86.regs.es = load_addr / 16;
+
+	assert((v86.regs.cs & 3) == 0);	/* Looks like RPL = 0 */
+
+	/* #BR -- should deliver SIG??? */
+	do_test(&v86, vmcode_bound - vmcode, VM86_INTx, 5, "#BR");
+
+	/*
+	 * SYSENTER -- should cause #GP or #UD depending on CPU.
+	 * Expected return type -1 means that we shouldn't validate
+	 * the vm86 return value.  This will avoid problems on non-SEP
+	 * CPUs.
+	 */
+	sethandler(SIGILL, sighandler, 0);
+	do_test(&v86, vmcode_sysenter - vmcode, -1, 0, "SYSENTER");
+	clearhandler(SIGILL);
+
+	/*
+	 * SYSCALL would be a disaster in VM86 mode.  Fortunately,
+	 * there is no kernel that both enables SYSCALL and sets
+	 * EFER.SCE, so it's #UD on all systems.  But vm86 is
+	 * buggy (or has a "feature"), so the SIGILL will actually
+	 * be delivered.
+	 */
+	sethandler(SIGILL, sighandler, 0);
+	do_test(&v86, vmcode_syscall - vmcode, VM86_SIGNAL, 0, "SYSCALL");
+	clearhandler(SIGILL);
+
+	/* STI with VIP set */
+	v86.regs.eflags |= X86_EFLAGS_VIP;
+	v86.regs.eflags &= ~X86_EFLAGS_IF;
+	do_test(&v86, vmcode_sti - vmcode, VM86_STI, 0, "STI with VIP set");
+
+	/* INT3 -- should cause #BP */
+	do_test(&v86, vmcode_int3 - vmcode, VM86_TRAP, 3, "INT3");
+
+	/* INT80 -- should exit with "INTx 0x80" */
+	v86.regs.eax = (unsigned int)-1;
+	do_test(&v86, vmcode_int80 - vmcode, VM86_INTx, 0x80, "int80");
+
+	/* Execute a null pointer */
+	v86.regs.cs = 0;
+	v86.regs.ss = 0;
+	sethandler(SIGSEGV, sighandler, 0);
+	got_signal = 0;
+	if (do_test(&v86, 0, VM86_SIGNAL, 0, "Execute null pointer") &&
+	    !got_signal) {
+		printf("[FAIL]\tDid not receive SIGSEGV\n");
+		nerrs++;
+	}
+	clearhandler(SIGSEGV);
+
+	/* Make sure nothing explodes if we fork. */
+	if (fork() > 0)
+		return 0;
+
+	return (nerrs == 0 ? 0 : 1);
+}
diff --git a/x86/ldt_gdt.c b/x86/ldt_gdt.c
new file mode 100644
index 0000000..31a3035
--- /dev/null
+++ b/x86/ldt_gdt.c
@@ -0,0 +1,576 @@
+/*
+ * ldt_gdt.c - Test cases for LDT and GDT access
+ * Copyright (c) 2015 Andrew Lutomirski
+ */
+
+#define _GNU_SOURCE
+#include <err.h>
+#include <stdio.h>
+#include <stdint.h>
+#include <signal.h>
+#include <setjmp.h>
+#include <stdlib.h>
+#include <string.h>
+#include <errno.h>
+#include <unistd.h>
+#include <sys/syscall.h>
+#include <asm/ldt.h>
+#include <sys/types.h>
+#include <sys/wait.h>
+#include <stdbool.h>
+#include <pthread.h>
+#include <sched.h>
+#include <linux/futex.h>
+
+#define AR_ACCESSED		(1<<8)
+
+#define AR_TYPE_RODATA		(0 * (1<<9))
+#define AR_TYPE_RWDATA		(1 * (1<<9))
+#define AR_TYPE_RODATA_EXPDOWN	(2 * (1<<9))
+#define AR_TYPE_RWDATA_EXPDOWN	(3 * (1<<9))
+#define AR_TYPE_XOCODE		(4 * (1<<9))
+#define AR_TYPE_XRCODE		(5 * (1<<9))
+#define AR_TYPE_XOCODE_CONF	(6 * (1<<9))
+#define AR_TYPE_XRCODE_CONF	(7 * (1<<9))
+
+#define AR_DPL3			(3 * (1<<13))
+
+#define AR_S			(1 << 12)
+#define AR_P			(1 << 15)
+#define AR_AVL			(1 << 20)
+#define AR_L			(1 << 21)
+#define AR_DB			(1 << 22)
+#define AR_G			(1 << 23)
+
+static int nerrs;
+
+static void check_invalid_segment(uint16_t index, int ldt)
+{
+	uint32_t has_limit = 0, has_ar = 0, limit, ar;
+	uint32_t selector = (index << 3) | (ldt << 2) | 3;
+
+	asm ("lsl %[selector], %[limit]\n\t"
+	     "jnz 1f\n\t"
+	     "movl $1, %[has_limit]\n\t"
+	     "1:"
+	     : [limit] "=r" (limit), [has_limit] "+rm" (has_limit)
+	     : [selector] "r" (selector));
+	asm ("larl %[selector], %[ar]\n\t"
+	     "jnz 1f\n\t"
+	     "movl $1, %[has_ar]\n\t"
+	     "1:"
+	     : [ar] "=r" (ar), [has_ar] "+rm" (has_ar)
+	     : [selector] "r" (selector));
+
+	if (has_limit || has_ar) {
+		printf("[FAIL]\t%s entry %hu is valid but should be invalid\n",
+		       (ldt ? "LDT" : "GDT"), index);
+		nerrs++;
+	} else {
+		printf("[OK]\t%s entry %hu is invalid\n",
+		       (ldt ? "LDT" : "GDT"), index);
+	}
+}
+
+static void check_valid_segment(uint16_t index, int ldt,
+				uint32_t expected_ar, uint32_t expected_limit,
+				bool verbose)
+{
+	uint32_t has_limit = 0, has_ar = 0, limit, ar;
+	uint32_t selector = (index << 3) | (ldt << 2) | 3;
+
+	asm ("lsl %[selector], %[limit]\n\t"
+	     "jnz 1f\n\t"
+	     "movl $1, %[has_limit]\n\t"
+	     "1:"
+	     : [limit] "=r" (limit), [has_limit] "+rm" (has_limit)
+	     : [selector] "r" (selector));
+	asm ("larl %[selector], %[ar]\n\t"
+	     "jnz 1f\n\t"
+	     "movl $1, %[has_ar]\n\t"
+	     "1:"
+	     : [ar] "=r" (ar), [has_ar] "+rm" (has_ar)
+	     : [selector] "r" (selector));
+
+	if (!has_limit || !has_ar) {
+		printf("[FAIL]\t%s entry %hu is invalid but should be valid\n",
+		       (ldt ? "LDT" : "GDT"), index);
+		nerrs++;
+		return;
+	}
+
+	if (ar != expected_ar) {
+		printf("[FAIL]\t%s entry %hu has AR 0x%08X but expected 0x%08X\n",
+		       (ldt ? "LDT" : "GDT"), index, ar, expected_ar);
+		nerrs++;
+	} else if (limit != expected_limit) {
+		printf("[FAIL]\t%s entry %hu has limit 0x%08X but expected 0x%08X\n",
+		       (ldt ? "LDT" : "GDT"), index, limit, expected_limit);
+		nerrs++;
+	} else if (verbose) {
+		printf("[OK]\t%s entry %hu has AR 0x%08X and limit 0x%08X\n",
+		       (ldt ? "LDT" : "GDT"), index, ar, limit);
+	}
+}
+
+static bool install_valid_mode(const struct user_desc *desc, uint32_t ar,
+			       bool oldmode)
+{
+	int ret = syscall(SYS_modify_ldt, oldmode ? 1 : 0x11,
+			  desc, sizeof(*desc));
+	if (ret < -1)
+		errno = -ret;
+	if (ret == 0) {
+		uint32_t limit = desc->limit;
+		if (desc->limit_in_pages)
+			limit = (limit << 12) + 4095;
+		check_valid_segment(desc->entry_number, 1, ar, limit, true);
+		return true;
+	} else if (errno == ENOSYS) {
+		printf("[OK]\tmodify_ldt returned -ENOSYS\n");
+		return false;
+	} else {
+		if (desc->seg_32bit) {
+			printf("[FAIL]\tUnexpected modify_ldt failure %d\n",
+			       errno);
+			nerrs++;
+			return false;
+		} else {
+			printf("[OK]\tmodify_ldt rejected 16 bit segment\n");
+			return false;
+		}
+	}
+}
+
+static bool install_valid(const struct user_desc *desc, uint32_t ar)
+{
+	return install_valid_mode(desc, ar, false);
+}
+
+static void install_invalid(const struct user_desc *desc, bool oldmode)
+{
+	int ret = syscall(SYS_modify_ldt, oldmode ? 1 : 0x11,
+			  desc, sizeof(*desc));
+	if (ret < -1)
+		errno = -ret;
+	if (ret == 0) {
+		check_invalid_segment(desc->entry_number, 1);
+	} else if (errno == ENOSYS) {
+		printf("[OK]\tmodify_ldt returned -ENOSYS\n");
+	} else {
+		if (desc->seg_32bit) {
+			printf("[FAIL]\tUnexpected modify_ldt failure %d\n",
+			       errno);
+			nerrs++;
+		} else {
+			printf("[OK]\tmodify_ldt rejected 16 bit segment\n");
+		}
+	}
+}
+
+static int safe_modify_ldt(int func, struct user_desc *ptr,
+			   unsigned long bytecount)
+{
+	int ret = syscall(SYS_modify_ldt, 0x11, ptr, bytecount);
+	if (ret < -1)
+		errno = -ret;
+	return ret;
+}
+
+static void fail_install(struct user_desc *desc)
+{
+	if (safe_modify_ldt(0x11, desc, sizeof(*desc)) == 0) {
+		printf("[FAIL]\tmodify_ldt accepted a bad descriptor\n");
+		nerrs++;
+	} else if (errno == ENOSYS) {
+		printf("[OK]\tmodify_ldt returned -ENOSYS\n");
+	} else {
+		printf("[OK]\tmodify_ldt failure %d\n", errno);
+	}
+}
+
+static void do_simple_tests(void)
+{
+	struct user_desc desc = {
+		.entry_number    = 0,
+		.base_addr       = 0,
+		.limit           = 10,
+		.seg_32bit       = 1,
+		.contents        = 2, /* Code, not conforming */
+		.read_exec_only  = 0,
+		.limit_in_pages  = 0,
+		.seg_not_present = 0,
+		.useable         = 0
+	};
+	install_valid(&desc, AR_DPL3 | AR_TYPE_XRCODE | AR_S | AR_P | AR_DB);
+
+	desc.limit_in_pages = 1;
+	install_valid(&desc, AR_DPL3 | AR_TYPE_XRCODE |
+		      AR_S | AR_P | AR_DB | AR_G);
+
+	check_invalid_segment(1, 1);
+
+	desc.entry_number = 2;
+	install_valid(&desc, AR_DPL3 | AR_TYPE_XRCODE |
+		      AR_S | AR_P | AR_DB | AR_G);
+
+	check_invalid_segment(1, 1);
+
+	desc.base_addr = 0xf0000000;
+	install_valid(&desc, AR_DPL3 | AR_TYPE_XRCODE |
+		      AR_S | AR_P | AR_DB | AR_G);
+
+	desc.useable = 1;
+	install_valid(&desc, AR_DPL3 | AR_TYPE_XRCODE |
+		      AR_S | AR_P | AR_DB | AR_G | AR_AVL);
+
+	desc.seg_not_present = 1;
+	install_valid(&desc, AR_DPL3 | AR_TYPE_XRCODE |
+		      AR_S | AR_DB | AR_G | AR_AVL);
+
+	desc.seg_32bit = 0;
+	install_valid(&desc, AR_DPL3 | AR_TYPE_XRCODE |
+		      AR_S | AR_G | AR_AVL);
+
+	desc.seg_32bit = 1;
+	desc.contents = 0;
+	install_valid(&desc, AR_DPL3 | AR_TYPE_RWDATA |
+		      AR_S | AR_DB | AR_G | AR_AVL);
+
+	desc.read_exec_only = 1;
+	install_valid(&desc, AR_DPL3 | AR_TYPE_RODATA |
+		      AR_S | AR_DB | AR_G | AR_AVL);
+
+	desc.contents = 1;
+	install_valid(&desc, AR_DPL3 | AR_TYPE_RODATA_EXPDOWN |
+		      AR_S | AR_DB | AR_G | AR_AVL);
+
+	desc.read_exec_only = 0;
+	desc.limit_in_pages = 0;
+	install_valid(&desc, AR_DPL3 | AR_TYPE_RWDATA_EXPDOWN |
+		      AR_S | AR_DB | AR_AVL);
+
+	desc.contents = 3;
+	install_valid(&desc, AR_DPL3 | AR_TYPE_XRCODE_CONF |
+		      AR_S | AR_DB | AR_AVL);
+
+	desc.read_exec_only = 1;
+	install_valid(&desc, AR_DPL3 | AR_TYPE_XOCODE_CONF |
+		      AR_S | AR_DB | AR_AVL);
+
+	desc.read_exec_only = 0;
+	desc.contents = 2;
+	install_valid(&desc, AR_DPL3 | AR_TYPE_XRCODE |
+		      AR_S | AR_DB | AR_AVL);
+
+	desc.read_exec_only = 1;
+
+#ifdef __x86_64__
+	desc.lm = 1;
+	install_valid(&desc, AR_DPL3 | AR_TYPE_XOCODE |
+		      AR_S | AR_DB | AR_AVL);
+	desc.lm = 0;
+#endif
+
+	bool entry1_okay = install_valid(&desc, AR_DPL3 | AR_TYPE_XOCODE |
+					 AR_S | AR_DB | AR_AVL);
+
+	if (entry1_okay) {
+		printf("[RUN]\tTest fork\n");
+		pid_t child = fork();
+		if (child == 0) {
+			nerrs = 0;
+			check_valid_segment(desc.entry_number, 1,
+					    AR_DPL3 | AR_TYPE_XOCODE |
+					    AR_S | AR_DB | AR_AVL, desc.limit,
+					    true);
+			check_invalid_segment(1, 1);
+			exit(nerrs ? 1 : 0);
+		} else {
+			int status;
+			if (waitpid(child, &status, 0) != child ||
+			    !WIFEXITED(status)) {
+				printf("[FAIL]\tChild died\n");
+				nerrs++;
+			} else if (WEXITSTATUS(status) != 0) {
+				printf("[FAIL]\tChild failed\n");
+				nerrs++;
+			} else {
+				printf("[OK]\tChild succeeded\n");
+			}
+		}
+
+		printf("[RUN]\tTest size\n");
+		int i;
+		for (i = 0; i < 8192; i++) {
+			desc.entry_number = i;
+			desc.limit = i;
+			if (safe_modify_ldt(0x11, &desc, sizeof(desc)) != 0) {
+				printf("[FAIL]\tFailed to install entry %d\n", i);
+				nerrs++;
+				break;
+			}
+		}
+		for (int j = 0; j < i; j++) {
+			check_valid_segment(j, 1, AR_DPL3 | AR_TYPE_XOCODE |
+					    AR_S | AR_DB | AR_AVL, j, false);
+		}
+		printf("[DONE]\tSize test\n");
+	} else {
+		printf("[SKIP]\tSkipping fork and size tests because we have no LDT\n");
+	}
+
+	/* Test entry_number too high. */
+	desc.entry_number = 8192;
+	fail_install(&desc);
+
+	/* Test deletion and actions mistakeable for deletion. */
+	memset(&desc, 0, sizeof(desc));
+	install_valid(&desc, AR_DPL3 | AR_TYPE_RWDATA | AR_S | AR_P);
+
+	desc.seg_not_present = 1;
+	install_valid(&desc, AR_DPL3 | AR_TYPE_RWDATA | AR_S);
+
+	desc.seg_not_present = 0;
+	desc.read_exec_only = 1;
+	install_valid(&desc, AR_DPL3 | AR_TYPE_RODATA | AR_S | AR_P);
+
+	desc.read_exec_only = 0;
+	desc.seg_not_present = 1;
+	install_valid(&desc, AR_DPL3 | AR_TYPE_RWDATA | AR_S);
+
+	desc.read_exec_only = 1;
+	desc.limit = 1;
+	install_valid(&desc, AR_DPL3 | AR_TYPE_RODATA | AR_S);
+
+	desc.limit = 0;
+	desc.base_addr = 1;
+	install_valid(&desc, AR_DPL3 | AR_TYPE_RODATA | AR_S);
+
+	desc.base_addr = 0;
+	install_invalid(&desc, false);
+
+	desc.seg_not_present = 0;
+	desc.read_exec_only = 0;
+	desc.seg_32bit = 1;
+	install_valid(&desc, AR_DPL3 | AR_TYPE_RWDATA | AR_S | AR_P | AR_DB);
+	install_invalid(&desc, true);
+}
+
+/*
+ * 0: thread is idle
+ * 1: thread armed
+ * 2: thread should clear LDT entry 0
+ * 3: thread should exit
+ */
+static volatile unsigned int ftx;
+
+static void *threadproc(void *ctx)
+{
+	cpu_set_t cpuset;
+	CPU_ZERO(&cpuset);
+	CPU_SET(1, &cpuset);
+	if (sched_setaffinity(0, sizeof(cpuset), &cpuset) != 0)
+		err(1, "sched_setaffinity to CPU 1");	/* should never fail */
+
+	while (1) {
+		syscall(SYS_futex, &ftx, FUTEX_WAIT, 0, NULL, NULL, 0);
+		while (ftx != 2) {
+			if (ftx >= 3)
+				return NULL;
+		}
+
+		/* clear LDT entry 0 */
+		const struct user_desc desc = {};
+		if (syscall(SYS_modify_ldt, 1, &desc, sizeof(desc)) != 0)
+			err(1, "modify_ldt");
+
+		/* If ftx == 2, set it to zero.  If ftx == 100, quit. */
+		unsigned int x = -2;
+		asm volatile ("lock xaddl %[x], %[ftx]" :
+			      [x] "+r" (x), [ftx] "+m" (ftx));
+		if (x != 2)
+			return NULL;
+	}
+}
+
+static void sethandler(int sig, void (*handler)(int, siginfo_t *, void *),
+		       int flags)
+{
+	struct sigaction sa;
+	memset(&sa, 0, sizeof(sa));
+	sa.sa_sigaction = handler;
+	sa.sa_flags = SA_SIGINFO | flags;
+	sigemptyset(&sa.sa_mask);
+	if (sigaction(sig, &sa, 0))
+		err(1, "sigaction");
+
+}
+
+static jmp_buf jmpbuf;
+
+static void sigsegv(int sig, siginfo_t *info, void *ctx_void)
+{
+	siglongjmp(jmpbuf, 1);
+}
+
+static void do_multicpu_tests(void)
+{
+	cpu_set_t cpuset;
+	pthread_t thread;
+	int failures = 0, iters = 5, i;
+	unsigned short orig_ss;
+
+	CPU_ZERO(&cpuset);
+	CPU_SET(1, &cpuset);
+	if (sched_setaffinity(0, sizeof(cpuset), &cpuset) != 0) {
+		printf("[SKIP]\tCannot set affinity to CPU 1\n");
+		return;
+	}
+
+	CPU_ZERO(&cpuset);
+	CPU_SET(0, &cpuset);
+	if (sched_setaffinity(0, sizeof(cpuset), &cpuset) != 0) {
+		printf("[SKIP]\tCannot set affinity to CPU 0\n");
+		return;
+	}
+
+	sethandler(SIGSEGV, sigsegv, 0);
+#ifdef __i386__
+	/* True 32-bit kernels send SIGILL instead of SIGSEGV on IRET faults. */
+	sethandler(SIGILL, sigsegv, 0);
+#endif
+
+	printf("[RUN]\tCross-CPU LDT invalidation\n");
+
+	if (pthread_create(&thread, 0, threadproc, 0) != 0)
+		err(1, "pthread_create");
+
+	asm volatile ("mov %%ss, %0" : "=rm" (orig_ss));
+
+	for (i = 0; i < 5; i++) {
+		if (sigsetjmp(jmpbuf, 1) != 0)
+			continue;
+
+		/* Make sure the thread is ready after the last test. */
+		while (ftx != 0)
+			;
+
+		struct user_desc desc = {
+			.entry_number    = 0,
+			.base_addr       = 0,
+			.limit           = 0xfffff,
+			.seg_32bit       = 1,
+			.contents        = 0, /* Data */
+			.read_exec_only  = 0,
+			.limit_in_pages  = 1,
+			.seg_not_present = 0,
+			.useable         = 0
+		};
+
+		if (safe_modify_ldt(0x11, &desc, sizeof(desc)) != 0) {
+			if (errno != ENOSYS)
+				err(1, "modify_ldt");
+			printf("[SKIP]\tmodify_ldt unavailable\n");
+			break;
+		}
+
+		/* Arm the thread. */
+		ftx = 1;
+		syscall(SYS_futex, &ftx, FUTEX_WAKE, 0, NULL, NULL, 0);
+
+		asm volatile ("mov %0, %%ss" : : "r" (0x7));
+
+		/* Go! */
+		ftx = 2;
+
+		while (ftx != 0)
+			;
+
+		/*
+		 * On success, modify_ldt will segfault us synchronously,
+		 * and we'll escape via siglongjmp.
+		 */
+
+		failures++;
+		asm volatile ("mov %0, %%ss" : : "rm" (orig_ss));
+	};
+
+	ftx = 100;  /* Kill the thread. */
+	syscall(SYS_futex, &ftx, FUTEX_WAKE, 0, NULL, NULL, 0);
+
+	if (pthread_join(thread, NULL) != 0)
+		err(1, "pthread_join");
+
+	if (failures) {
+		printf("[FAIL]\t%d of %d iterations failed\n", failures, iters);
+		nerrs++;
+	} else {
+		printf("[OK]\tAll %d iterations succeeded\n", iters);
+	}
+}
+
+static int finish_exec_test(void)
+{
+	/*
+	 * In a sensible world, this would be check_invalid_segment(0, 1);
+	 * For better or for worse, though, the LDT is inherited across exec.
+	 * We can probably change this safely, but for now we test it.
+	 */
+	check_valid_segment(0, 1,
+			    AR_DPL3 | AR_TYPE_XRCODE | AR_S | AR_P | AR_DB,
+			    42, true);
+
+	return nerrs ? 1 : 0;
+}
+
+static void do_exec_test(void)
+{
+	printf("[RUN]\tTest exec\n");
+
+	struct user_desc desc = {
+		.entry_number    = 0,
+		.base_addr       = 0,
+		.limit           = 42,
+		.seg_32bit       = 1,
+		.contents        = 2, /* Code, not conforming */
+		.read_exec_only  = 0,
+		.limit_in_pages  = 0,
+		.seg_not_present = 0,
+		.useable         = 0
+	};
+	install_valid(&desc, AR_DPL3 | AR_TYPE_XRCODE | AR_S | AR_P | AR_DB);
+
+	pid_t child = fork();
+	if (child == 0) {
+		execl("/proc/self/exe", "ldt_gdt_test_exec", NULL);
+		printf("[FAIL]\tCould not exec self\n");
+		exit(1);	/* exec failed */
+	} else {
+		int status;
+		if (waitpid(child, &status, 0) != child ||
+		    !WIFEXITED(status)) {
+			printf("[FAIL]\tChild died\n");
+			nerrs++;
+		} else if (WEXITSTATUS(status) != 0) {
+			printf("[FAIL]\tChild failed\n");
+			nerrs++;
+		} else {
+			printf("[OK]\tChild succeeded\n");
+		}
+	}
+}
+
+int main(int argc, char **argv)
+{
+	if (argc == 1 && !strcmp(argv[0], "ldt_gdt_test_exec"))
+		return finish_exec_test();
+
+	do_simple_tests();
+
+	do_multicpu_tests();
+
+	do_exec_test();
+
+	return nerrs ? 1 : 0;
+}
diff --git a/x86/ptrace_syscall.c b/x86/ptrace_syscall.c
new file mode 100644
index 0000000..5105b49
--- /dev/null
+++ b/x86/ptrace_syscall.c
@@ -0,0 +1,294 @@
+#define _GNU_SOURCE
+
+#include <sys/ptrace.h>
+#include <sys/types.h>
+#include <sys/wait.h>
+#include <sys/syscall.h>
+#include <sys/user.h>
+#include <unistd.h>
+#include <errno.h>
+#include <stddef.h>
+#include <stdio.h>
+#include <err.h>
+#include <string.h>
+#include <asm/ptrace-abi.h>
+#include <sys/auxv.h>
+
+/* Bitness-agnostic defines for user_regs_struct fields. */
+#ifdef __x86_64__
+# define user_syscall_nr	orig_rax
+# define user_arg0		rdi
+# define user_arg1		rsi
+# define user_arg2		rdx
+# define user_arg3		r10
+# define user_arg4		r8
+# define user_arg5		r9
+# define user_ip		rip
+# define user_ax		rax
+#else
+# define user_syscall_nr	orig_eax
+# define user_arg0		ebx
+# define user_arg1		ecx
+# define user_arg2		edx
+# define user_arg3		esi
+# define user_arg4		edi
+# define user_arg5		ebp
+# define user_ip		eip
+# define user_ax		eax
+#endif
+
+static int nerrs = 0;
+
+struct syscall_args32 {
+	uint32_t nr, arg0, arg1, arg2, arg3, arg4, arg5;
+};
+
+#ifdef __i386__
+extern void sys32_helper(struct syscall_args32 *, void *);
+extern void int80_and_ret(void);
+#endif
+
+/*
+ * Helper to invoke int80 with controlled regs and capture the final regs.
+ */
+static void do_full_int80(struct syscall_args32 *args)
+{
+#ifdef __x86_64__
+	register unsigned long bp asm("bp") = args->arg5;
+	asm volatile ("int $0x80"
+		      : "+a" (args->nr),
+			"+b" (args->arg0), "+c" (args->arg1), "+d" (args->arg2),
+			"+S" (args->arg3), "+D" (args->arg4), "+r" (bp));
+	args->arg5 = bp;
+#else
+	sys32_helper(args, int80_and_ret);
+#endif
+}
+
+#ifdef __i386__
+static void (*vsyscall32)(void);
+
+/*
+ * Nasty helper to invoke AT_SYSINFO (i.e. __kernel_vsyscall) with
+ * controlled regs and capture the final regs.  This is so nasty that it
+ * crashes my copy of gdb :)
+ */
+static void do_full_vsyscall32(struct syscall_args32 *args)
+{
+	sys32_helper(args, vsyscall32);
+}
+#endif
+
+static siginfo_t wait_trap(pid_t chld)
+{
+	siginfo_t si;
+	if (waitid(P_PID, chld, &si, WEXITED|WSTOPPED) != 0)
+		err(1, "waitid");
+	if (si.si_pid != chld)
+		errx(1, "got unexpected pid in event\n");
+	if (si.si_code != CLD_TRAPPED)
+		errx(1, "got unexpected event type %d\n", si.si_code);
+	return si;
+}
+
+static void sethandler(int sig, void (*handler)(int, siginfo_t *, void *),
+		       int flags)
+{
+	struct sigaction sa;
+	memset(&sa, 0, sizeof(sa));
+	sa.sa_sigaction = handler;
+	sa.sa_flags = SA_SIGINFO | flags;
+	sigemptyset(&sa.sa_mask);
+	if (sigaction(sig, &sa, 0))
+		err(1, "sigaction");
+}
+
+static void clearhandler(int sig)
+{
+	struct sigaction sa;
+	memset(&sa, 0, sizeof(sa));
+	sa.sa_handler = SIG_DFL;
+	sigemptyset(&sa.sa_mask);
+	if (sigaction(sig, &sa, 0))
+		err(1, "sigaction");
+}
+
+#ifdef __x86_64__
+# define REG_BP REG_RBP
+#else
+# define REG_BP REG_EBP
+#endif
+
+static void empty_handler(int sig, siginfo_t *si, void *ctx_void)
+{
+}
+
+static void test_sys32_regs(void (*do_syscall)(struct syscall_args32 *))
+{
+	struct syscall_args32 args = {
+		.nr = 224,	/* gettid */
+		.arg0 = 10, .arg1 = 11, .arg2 = 12,
+		.arg3 = 13, .arg4 = 14, .arg5 = 15,
+	};
+
+	do_syscall(&args);
+
+	if (args.nr != getpid() ||
+	    args.arg0 != 10 || args.arg1 != 11 || args.arg2 != 12 ||
+	    args.arg3 != 13 || args.arg4 != 14 || args.arg5 != 15) {
+		printf("[FAIL]\tgetpid() failed to preseve regs\n");
+		nerrs++;
+	} else {
+		printf("[OK]\tgetpid() preserves regs\n");
+	}
+
+	sethandler(SIGUSR1, empty_handler, 0);
+
+	args.nr = 37;	/* kill */
+	args.arg0 = getpid();
+	args.arg1 = SIGUSR1;
+	do_syscall(&args);
+	if (args.nr != 0 ||
+	    args.arg0 != getpid() || args.arg1 != SIGUSR1 || args.arg2 != 12 ||
+	    args.arg3 != 13 || args.arg4 != 14 || args.arg5 != 15) {
+		printf("[FAIL]\tkill(getpid(), SIGUSR1) failed to preseve regs\n");
+		nerrs++;
+	} else {
+		printf("[OK]\tkill(getpid(), SIGUSR1) preserves regs\n");
+	}
+	clearhandler(SIGUSR1);
+}
+
+static void test_ptrace_syscall_restart(void)
+{
+	printf("[RUN]\tptrace-induced syscall restart\n");
+	pid_t chld = fork();
+	if (chld < 0)
+		err(1, "fork");
+
+	if (chld == 0) {
+		if (ptrace(PTRACE_TRACEME, 0, 0, 0) != 0)
+			err(1, "PTRACE_TRACEME");
+
+		printf("\tChild will make one syscall\n");
+		raise(SIGSTOP);
+
+		syscall(SYS_gettid, 10, 11, 12, 13, 14, 15);
+		_exit(0);
+	}
+
+	int status;
+
+	/* Wait for SIGSTOP. */
+	if (waitpid(chld, &status, 0) != chld || !WIFSTOPPED(status))
+		err(1, "waitpid");
+
+	struct user_regs_struct regs;
+
+	printf("[RUN]\tSYSEMU\n");
+	if (ptrace(PTRACE_SYSEMU, chld, 0, 0) != 0)
+		err(1, "PTRACE_SYSCALL");
+	wait_trap(chld);
+
+	if (ptrace(PTRACE_GETREGS, chld, 0, &regs) != 0)
+		err(1, "PTRACE_GETREGS");
+
+	if (regs.user_syscall_nr != SYS_gettid ||
+	    regs.user_arg0 != 10 || regs.user_arg1 != 11 ||
+	    regs.user_arg2 != 12 || regs.user_arg3 != 13 ||
+	    regs.user_arg4 != 14 || regs.user_arg5 != 15) {
+		printf("[FAIL]\tInitial args are wrong (nr=%lu, args=%lu %lu %lu %lu %lu %lu)\n", (unsigned long)regs.user_syscall_nr, (unsigned long)regs.user_arg0, (unsigned long)regs.user_arg1, (unsigned long)regs.user_arg2, (unsigned long)regs.user_arg3, (unsigned long)regs.user_arg4, (unsigned long)regs.user_arg5);
+		nerrs++;
+	} else {
+		printf("[OK]\tInitial nr and args are correct\n");
+	}
+
+	printf("[RUN]\tRestart the syscall (ip = 0x%lx)\n",
+	       (unsigned long)regs.user_ip);
+
+	/*
+	 * This does exactly what it appears to do if syscall is int80 or
+	 * SYSCALL64.  For SYSCALL32 or SYSENTER, though, this is highly
+	 * magical.  It needs to work so that ptrace and syscall restart
+	 * work as expected.
+	 */
+	regs.user_ax = regs.user_syscall_nr;
+	regs.user_ip -= 2;
+	if (ptrace(PTRACE_SETREGS, chld, 0, &regs) != 0)
+		err(1, "PTRACE_SETREGS");
+
+	if (ptrace(PTRACE_SYSEMU, chld, 0, 0) != 0)
+		err(1, "PTRACE_SYSCALL");
+	wait_trap(chld);
+
+	if (ptrace(PTRACE_GETREGS, chld, 0, &regs) != 0)
+		err(1, "PTRACE_GETREGS");
+
+	if (regs.user_syscall_nr != SYS_gettid ||
+	    regs.user_arg0 != 10 || regs.user_arg1 != 11 ||
+	    regs.user_arg2 != 12 || regs.user_arg3 != 13 ||
+	    regs.user_arg4 != 14 || regs.user_arg5 != 15) {
+		printf("[FAIL]\tRestart nr or args are wrong (nr=%lu, args=%lu %lu %lu %lu %lu %lu)\n", (unsigned long)regs.user_syscall_nr, (unsigned long)regs.user_arg0, (unsigned long)regs.user_arg1, (unsigned long)regs.user_arg2, (unsigned long)regs.user_arg3, (unsigned long)regs.user_arg4, (unsigned long)regs.user_arg5);
+		nerrs++;
+	} else {
+		printf("[OK]\tRestarted nr and args are correct\n");
+	}
+
+	printf("[RUN]\tChange nr and args and restart the syscall (ip = 0x%lx)\n",
+	       (unsigned long)regs.user_ip);
+
+	regs.user_ax = SYS_getpid;
+	regs.user_arg0 = 20;
+	regs.user_arg1 = 21;
+	regs.user_arg2 = 22;
+	regs.user_arg3 = 23;
+	regs.user_arg4 = 24;
+	regs.user_arg5 = 25;
+	regs.user_ip -= 2;
+
+	if (ptrace(PTRACE_SETREGS, chld, 0, &regs) != 0)
+		err(1, "PTRACE_SETREGS");
+
+	if (ptrace(PTRACE_SYSEMU, chld, 0, 0) != 0)
+		err(1, "PTRACE_SYSCALL");
+	wait_trap(chld);
+
+	if (ptrace(PTRACE_GETREGS, chld, 0, &regs) != 0)
+		err(1, "PTRACE_GETREGS");
+
+	if (regs.user_syscall_nr != SYS_getpid ||
+	    regs.user_arg0 != 20 || regs.user_arg1 != 21 || regs.user_arg2 != 22 ||
+	    regs.user_arg3 != 23 || regs.user_arg4 != 24 || regs.user_arg5 != 25) {
+		printf("[FAIL]\tRestart nr or args are wrong (nr=%lu, args=%lu %lu %lu %lu %lu %lu)\n", (unsigned long)regs.user_syscall_nr, (unsigned long)regs.user_arg0, (unsigned long)regs.user_arg1, (unsigned long)regs.user_arg2, (unsigned long)regs.user_arg3, (unsigned long)regs.user_arg4, (unsigned long)regs.user_arg5);
+		nerrs++;
+	} else {
+		printf("[OK]\tReplacement nr and args are correct\n");
+	}
+
+	if (ptrace(PTRACE_CONT, chld, 0, 0) != 0)
+		err(1, "PTRACE_CONT");
+	if (waitpid(chld, &status, 0) != chld)
+		err(1, "waitpid");
+	if (!WIFEXITED(status) || WEXITSTATUS(status) != 0) {
+		printf("[FAIL]\tChild failed\n");
+		nerrs++;
+	} else {
+		printf("[OK]\tChild exited cleanly\n");
+	}
+}
+
+int main()
+{
+	printf("[RUN]\tCheck int80 return regs\n");
+	test_sys32_regs(do_full_int80);
+
+#if defined(__i386__) && (!defined(__GLIBC__) || __GLIBC__ > 2 || __GLIBC_MINOR__ >= 16)
+	vsyscall32 = (void *)getauxval(AT_SYSINFO);
+	printf("[RUN]\tCheck AT_SYSINFO return regs\n");
+	test_sys32_regs(do_full_vsyscall32);
+#endif
+
+	test_ptrace_syscall_restart();
+
+	return 0;
+}
diff --git a/x86/raw_syscall_helper_32.S b/x86/raw_syscall_helper_32.S
new file mode 100644
index 0000000..534e71e
--- /dev/null
+++ b/x86/raw_syscall_helper_32.S
@@ -0,0 +1,46 @@
+.global sys32_helper
+sys32_helper:
+	/* Args: syscall_args_32*, function pointer */
+	pushl	%ebp
+	pushl	%ebx
+	pushl	%esi
+	pushl	%edi
+	movl	5*4(%esp), %eax	/* pointer to args struct */
+
+	movl	1*4(%eax), %ebx
+	movl	2*4(%eax), %ecx
+	movl	3*4(%eax), %edx
+	movl	4*4(%eax), %esi
+	movl	5*4(%eax), %edi
+	movl	6*4(%eax), %ebp
+	movl	0*4(%eax), %eax
+
+	call	*(6*4)(%esp)	/* Do the syscall */
+
+	/* Now we need to recover without losing any reg values */
+	pushl	%eax
+	movl	6*4(%esp), %eax
+	popl	0*4(%eax)
+	movl	%ebx, 1*4(%eax)
+	movl	%ecx, 2*4(%eax)
+	movl	%edx, 3*4(%eax)
+	movl	%esi, 4*4(%eax)
+	movl	%edi, 5*4(%eax)
+	movl	%ebp, 6*4(%eax)
+
+	popl	%edi
+	popl	%esi
+	popl	%ebx
+	popl	%ebp
+	ret
+
+	.type sys32_helper, @function
+	.size sys32_helper, .-sys32_helper
+
+.global int80_and_ret
+int80_and_ret:
+	int	$0x80
+	ret
+
+	.type int80_and_ret, @function
+	.size int80_and_ret, .-int80_and_ret
diff --git a/x86/sigreturn.c b/x86/sigreturn.c
new file mode 100644
index 0000000..b5aa1ba
--- /dev/null
+++ b/x86/sigreturn.c
@@ -0,0 +1,684 @@
+/*
+ * sigreturn.c - tests for x86 sigreturn(2) and exit-to-userspace
+ * Copyright (c) 2014-2015 Andrew Lutomirski
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms and conditions of the GNU General Public License,
+ * version 2, as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope it will be useful, but
+ * WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * This is a series of tests that exercises the sigreturn(2) syscall and
+ * the IRET / SYSRET paths in the kernel.
+ *
+ * For now, this focuses on the effects of unusual CS and SS values,
+ * and it has a bunch of tests to make sure that ESP/RSP is restored
+ * properly.
+ *
+ * The basic idea behind these tests is to raise(SIGUSR1) to create a
+ * sigcontext frame, plug in the values to be tested, and then return,
+ * which implicitly invokes sigreturn(2) and programs the user context
+ * as desired.
+ *
+ * For tests for which we expect sigreturn and the subsequent return to
+ * user mode to succeed, we return to a short trampoline that generates
+ * SIGTRAP so that the meat of the tests can be ordinary C code in a
+ * SIGTRAP handler.
+ *
+ * The inner workings of each test is documented below.
+ *
+ * Do not run on outdated, unpatched kernels at risk of nasty crashes.
+ */
+
+#define _GNU_SOURCE
+
+#include <sys/time.h>
+#include <time.h>
+#include <stdlib.h>
+#include <sys/syscall.h>
+#include <unistd.h>
+#include <stdio.h>
+#include <string.h>
+#include <inttypes.h>
+#include <sys/mman.h>
+#include <sys/signal.h>
+#include <sys/ucontext.h>
+#include <asm/ldt.h>
+#include <err.h>
+#include <setjmp.h>
+#include <stddef.h>
+#include <stdbool.h>
+#include <sys/ptrace.h>
+#include <sys/user.h>
+
+/*
+ * In principle, this test can run on Linux emulation layers (e.g.
+ * Illumos "LX branded zones").  Solaris-based kernels reserve LDT
+ * entries 0-5 for their own internal purposes, so start our LDT
+ * allocations above that reservation.  (The tests don't pass on LX
+ * branded zones, but at least this lets them run.)
+ */
+#define LDT_OFFSET 6
+
+/* An aligned stack accessible through some of our segments. */
+static unsigned char stack16[65536] __attribute__((aligned(4096)));
+
+/*
+ * An aligned int3 instruction used as a trampoline.  Some of the tests
+ * want to fish out their ss values, so this trampoline copies ss to eax
+ * before the int3.
+ */
+asm (".pushsection .text\n\t"
+     ".type int3, @function\n\t"
+     ".align 4096\n\t"
+     "int3:\n\t"
+     "mov %ss,%eax\n\t"
+     "int3\n\t"
+     ".size int3, . - int3\n\t"
+     ".align 4096, 0xcc\n\t"
+     ".popsection");
+extern char int3[4096];
+
+/*
+ * At startup, we prepapre:
+ *
+ * - ldt_nonexistent_sel: An LDT entry that doesn't exist (all-zero
+ *   descriptor or out of bounds).
+ * - code16_sel: A 16-bit LDT code segment pointing to int3.
+ * - data16_sel: A 16-bit LDT data segment pointing to stack16.
+ * - npcode32_sel: A 32-bit not-present LDT code segment pointing to int3.
+ * - npdata32_sel: A 32-bit not-present LDT data segment pointing to stack16.
+ * - gdt_data16_idx: A 16-bit GDT data segment pointing to stack16.
+ * - gdt_npdata32_idx: A 32-bit not-present GDT data segment pointing to
+ *   stack16.
+ *
+ * For no particularly good reason, xyz_sel is a selector value with the
+ * RPL and LDT bits filled in, whereas xyz_idx is just an index into the
+ * descriptor table.  These variables will be zero if their respective
+ * segments could not be allocated.
+ */
+static unsigned short ldt_nonexistent_sel;
+static unsigned short code16_sel, data16_sel, npcode32_sel, npdata32_sel;
+
+static unsigned short gdt_data16_idx, gdt_npdata32_idx;
+
+static unsigned short GDT3(int idx)
+{
+	return (idx << 3) | 3;
+}
+
+static unsigned short LDT3(int idx)
+{
+	return (idx << 3) | 7;
+}
+
+/* Our sigaltstack scratch space. */
+static char altstack_data[SIGSTKSZ];
+
+static void sethandler(int sig, void (*handler)(int, siginfo_t *, void *),
+		       int flags)
+{
+	struct sigaction sa;
+	memset(&sa, 0, sizeof(sa));
+	sa.sa_sigaction = handler;
+	sa.sa_flags = SA_SIGINFO | flags;
+	sigemptyset(&sa.sa_mask);
+	if (sigaction(sig, &sa, 0))
+		err(1, "sigaction");
+}
+
+static void clearhandler(int sig)
+{
+	struct sigaction sa;
+	memset(&sa, 0, sizeof(sa));
+	sa.sa_handler = SIG_DFL;
+	sigemptyset(&sa.sa_mask);
+	if (sigaction(sig, &sa, 0))
+		err(1, "sigaction");
+}
+
+static void add_ldt(const struct user_desc *desc, unsigned short *var,
+		    const char *name)
+{
+	if (syscall(SYS_modify_ldt, 1, desc, sizeof(*desc)) == 0) {
+		*var = LDT3(desc->entry_number);
+	} else {
+		printf("[NOTE]\tFailed to create %s segment\n", name);
+		*var = 0;
+	}
+}
+
+static void setup_ldt(void)
+{
+	if ((unsigned long)stack16 > (1ULL << 32) - sizeof(stack16))
+		errx(1, "stack16 is too high\n");
+	if ((unsigned long)int3 > (1ULL << 32) - sizeof(int3))
+		errx(1, "int3 is too high\n");
+
+	ldt_nonexistent_sel = LDT3(LDT_OFFSET + 2);
+
+	const struct user_desc code16_desc = {
+		.entry_number    = LDT_OFFSET + 0,
+		.base_addr       = (unsigned long)int3,
+		.limit           = 4095,
+		.seg_32bit       = 0,
+		.contents        = 2, /* Code, not conforming */
+		.read_exec_only  = 0,
+		.limit_in_pages  = 0,
+		.seg_not_present = 0,
+		.useable         = 0
+	};
+	add_ldt(&code16_desc, &code16_sel, "code16");
+
+	const struct user_desc data16_desc = {
+		.entry_number    = LDT_OFFSET + 1,
+		.base_addr       = (unsigned long)stack16,
+		.limit           = 0xffff,
+		.seg_32bit       = 0,
+		.contents        = 0, /* Data, grow-up */
+		.read_exec_only  = 0,
+		.limit_in_pages  = 0,
+		.seg_not_present = 0,
+		.useable         = 0
+	};
+	add_ldt(&data16_desc, &data16_sel, "data16");
+
+	const struct user_desc npcode32_desc = {
+		.entry_number    = LDT_OFFSET + 3,
+		.base_addr       = (unsigned long)int3,
+		.limit           = 4095,
+		.seg_32bit       = 1,
+		.contents        = 2, /* Code, not conforming */
+		.read_exec_only  = 0,
+		.limit_in_pages  = 0,
+		.seg_not_present = 1,
+		.useable         = 0
+	};
+	add_ldt(&npcode32_desc, &npcode32_sel, "npcode32");
+
+	const struct user_desc npdata32_desc = {
+		.entry_number    = LDT_OFFSET + 4,
+		.base_addr       = (unsigned long)stack16,
+		.limit           = 0xffff,
+		.seg_32bit       = 1,
+		.contents        = 0, /* Data, grow-up */
+		.read_exec_only  = 0,
+		.limit_in_pages  = 0,
+		.seg_not_present = 1,
+		.useable         = 0
+	};
+	add_ldt(&npdata32_desc, &npdata32_sel, "npdata32");
+
+	struct user_desc gdt_data16_desc = {
+		.entry_number    = -1,
+		.base_addr       = (unsigned long)stack16,
+		.limit           = 0xffff,
+		.seg_32bit       = 0,
+		.contents        = 0, /* Data, grow-up */
+		.read_exec_only  = 0,
+		.limit_in_pages  = 0,
+		.seg_not_present = 0,
+		.useable         = 0
+	};
+
+	if (syscall(SYS_set_thread_area, &gdt_data16_desc) == 0) {
+		/*
+		 * This probably indicates vulnerability to CVE-2014-8133.
+		 * Merely getting here isn't definitive, though, and we'll
+		 * diagnose the problem for real later on.
+		 */
+		printf("[WARN]\tset_thread_area allocated data16 at index %d\n",
+		       gdt_data16_desc.entry_number);
+		gdt_data16_idx = gdt_data16_desc.entry_number;
+	} else {
+		printf("[OK]\tset_thread_area refused 16-bit data\n");
+	}
+
+	struct user_desc gdt_npdata32_desc = {
+		.entry_number    = -1,
+		.base_addr       = (unsigned long)stack16,
+		.limit           = 0xffff,
+		.seg_32bit       = 1,
+		.contents        = 0, /* Data, grow-up */
+		.read_exec_only  = 0,
+		.limit_in_pages  = 0,
+		.seg_not_present = 1,
+		.useable         = 0
+	};
+
+	if (syscall(SYS_set_thread_area, &gdt_npdata32_desc) == 0) {
+		/*
+		 * As a hardening measure, newer kernels don't allow this.
+		 */
+		printf("[WARN]\tset_thread_area allocated npdata32 at index %d\n",
+		       gdt_npdata32_desc.entry_number);
+		gdt_npdata32_idx = gdt_npdata32_desc.entry_number;
+	} else {
+		printf("[OK]\tset_thread_area refused 16-bit data\n");
+	}
+}
+
+/* State used by our signal handlers. */
+static gregset_t initial_regs, requested_regs, resulting_regs;
+
+/* Instructions for the SIGUSR1 handler. */
+static volatile unsigned short sig_cs, sig_ss;
+static volatile sig_atomic_t sig_trapped, sig_err, sig_trapno;
+
+/* Abstractions for some 32-bit vs 64-bit differences. */
+#ifdef __x86_64__
+# define REG_IP REG_RIP
+# define REG_SP REG_RSP
+# define REG_AX REG_RAX
+
+struct selectors {
+	unsigned short cs, gs, fs, ss;
+};
+
+static unsigned short *ssptr(ucontext_t *ctx)
+{
+	struct selectors *sels = (void *)&ctx->uc_mcontext.gregs[REG_CSGSFS];
+	return &sels->ss;
+}
+
+static unsigned short *csptr(ucontext_t *ctx)
+{
+	struct selectors *sels = (void *)&ctx->uc_mcontext.gregs[REG_CSGSFS];
+	return &sels->cs;
+}
+#else
+# define REG_IP REG_EIP
+# define REG_SP REG_ESP
+# define REG_AX REG_EAX
+
+static greg_t *ssptr(ucontext_t *ctx)
+{
+	return &ctx->uc_mcontext.gregs[REG_SS];
+}
+
+static greg_t *csptr(ucontext_t *ctx)
+{
+	return &ctx->uc_mcontext.gregs[REG_CS];
+}
+#endif
+
+/* Number of errors in the current test case. */
+static volatile sig_atomic_t nerrs;
+
+/*
+ * SIGUSR1 handler.  Sets CS and SS as requested and points IP to the
+ * int3 trampoline.  Sets SP to a large known value so that we can see
+ * whether the value round-trips back to user mode correctly.
+ */
+static void sigusr1(int sig, siginfo_t *info, void *ctx_void)
+{
+	ucontext_t *ctx = (ucontext_t*)ctx_void;
+
+	memcpy(&initial_regs, &ctx->uc_mcontext.gregs, sizeof(gregset_t));
+
+	*csptr(ctx) = sig_cs;
+	*ssptr(ctx) = sig_ss;
+
+	ctx->uc_mcontext.gregs[REG_IP] =
+		sig_cs == code16_sel ? 0 : (unsigned long)&int3;
+	ctx->uc_mcontext.gregs[REG_SP] = (unsigned long)0x8badf00d5aadc0deULL;
+	ctx->uc_mcontext.gregs[REG_AX] = 0;
+
+	memcpy(&requested_regs, &ctx->uc_mcontext.gregs, sizeof(gregset_t));
+	requested_regs[REG_AX] = *ssptr(ctx);	/* The asm code does this. */
+
+	return;
+}
+
+/*
+ * Called after a successful sigreturn.  Restores our state so that
+ * the original raise(SIGUSR1) returns.
+ */
+static void sigtrap(int sig, siginfo_t *info, void *ctx_void)
+{
+	ucontext_t *ctx = (ucontext_t*)ctx_void;
+
+	sig_err = ctx->uc_mcontext.gregs[REG_ERR];
+	sig_trapno = ctx->uc_mcontext.gregs[REG_TRAPNO];
+
+	unsigned short ss;
+	asm ("mov %%ss,%0" : "=r" (ss));
+
+	greg_t asm_ss = ctx->uc_mcontext.gregs[REG_AX];
+	if (asm_ss != sig_ss && sig == SIGTRAP) {
+		/* Sanity check failure. */
+		printf("[FAIL]\tSIGTRAP: ss = %hx, frame ss = %hx, ax = %llx\n",
+		       ss, *ssptr(ctx), (unsigned long long)asm_ss);
+		nerrs++;
+	}
+
+	memcpy(&resulting_regs, &ctx->uc_mcontext.gregs, sizeof(gregset_t));
+	memcpy(&ctx->uc_mcontext.gregs, &initial_regs, sizeof(gregset_t));
+
+	sig_trapped = sig;
+}
+
+/*
+ * Checks a given selector for its code bitness or returns -1 if it's not
+ * a usable code segment selector.
+ */
+int cs_bitness(unsigned short cs)
+{
+	uint32_t valid = 0, ar;
+	asm ("lar %[cs], %[ar]\n\t"
+	     "jnz 1f\n\t"
+	     "mov $1, %[valid]\n\t"
+	     "1:"
+	     : [ar] "=r" (ar), [valid] "+rm" (valid)
+	     : [cs] "r" (cs));
+
+	if (!valid)
+		return -1;
+
+	bool db = (ar & (1 << 22));
+	bool l = (ar & (1 << 21));
+
+	if (!(ar & (1<<11)))
+	    return -1;	/* Not code. */
+
+	if (l && !db)
+		return 64;
+	else if (!l && db)
+		return 32;
+	else if (!l && !db)
+		return 16;
+	else
+		return -1;	/* Unknown bitness. */
+}
+
+/* Finds a usable code segment of the requested bitness. */
+int find_cs(int bitness)
+{
+	unsigned short my_cs;
+
+	asm ("mov %%cs,%0" :  "=r" (my_cs));
+
+	if (cs_bitness(my_cs) == bitness)
+		return my_cs;
+	if (cs_bitness(my_cs + (2 << 3)) == bitness)
+		return my_cs + (2 << 3);
+	if (my_cs > (2<<3) && cs_bitness(my_cs - (2 << 3)) == bitness)
+	    return my_cs - (2 << 3);
+	if (cs_bitness(code16_sel) == bitness)
+		return code16_sel;
+
+	printf("[WARN]\tCould not find %d-bit CS\n", bitness);
+	return -1;
+}
+
+static int test_valid_sigreturn(int cs_bits, bool use_16bit_ss, int force_ss)
+{
+	int cs = find_cs(cs_bits);
+	if (cs == -1) {
+		printf("[SKIP]\tCode segment unavailable for %d-bit CS, %d-bit SS\n",
+		       cs_bits, use_16bit_ss ? 16 : 32);
+		return 0;
+	}
+
+	if (force_ss != -1) {
+		sig_ss = force_ss;
+	} else {
+		if (use_16bit_ss) {
+			if (!data16_sel) {
+				printf("[SKIP]\tData segment unavailable for %d-bit CS, 16-bit SS\n",
+				       cs_bits);
+				return 0;
+			}
+			sig_ss = data16_sel;
+		} else {
+			asm volatile ("mov %%ss,%0" : "=r" (sig_ss));
+		}
+	}
+
+	sig_cs = cs;
+
+	printf("[RUN]\tValid sigreturn: %d-bit CS (%hx), %d-bit SS (%hx%s)\n",
+	       cs_bits, sig_cs, use_16bit_ss ? 16 : 32, sig_ss,
+	       (sig_ss & 4) ? "" : ", GDT");
+
+	raise(SIGUSR1);
+
+	nerrs = 0;
+
+	/*
+	 * Check that each register had an acceptable value when the
+	 * int3 trampoline was invoked.
+	 */
+	for (int i = 0; i < NGREG; i++) {
+		greg_t req = requested_regs[i], res = resulting_regs[i];
+		if (i == REG_TRAPNO || i == REG_IP)
+			continue;	/* don't care */
+		if (i == REG_SP) {
+			printf("\tSP: %llx -> %llx\n", (unsigned long long)req,
+			       (unsigned long long)res);
+
+			/*
+			 * In many circumstances, the high 32 bits of rsp
+			 * are zeroed.  For example, we could be a real
+			 * 32-bit program, or we could hit any of a number
+			 * of poorly-documented IRET or segmented ESP
+			 * oddities.  If this happens, it's okay.
+			 */
+			if (res == (req & 0xFFFFFFFF))
+				continue;  /* OK; not expected to work */
+		}
+
+		bool ignore_reg = false;
+#if __i386__
+		if (i == REG_UESP)
+			ignore_reg = true;
+#else
+		if (i == REG_CSGSFS) {
+			struct selectors *req_sels =
+				(void *)&requested_regs[REG_CSGSFS];
+			struct selectors *res_sels =
+				(void *)&resulting_regs[REG_CSGSFS];
+			if (req_sels->cs != res_sels->cs) {
+				printf("[FAIL]\tCS mismatch: requested 0x%hx; got 0x%hx\n",
+				       req_sels->cs, res_sels->cs);
+				nerrs++;
+			}
+
+			if (req_sels->ss != res_sels->ss) {
+				printf("[FAIL]\tSS mismatch: requested 0x%hx; got 0x%hx\n",
+				       req_sels->ss, res_sels->ss);
+				nerrs++;
+			}
+
+			continue;
+		}
+#endif
+
+		/* Sanity check on the kernel */
+		if (i == REG_AX && requested_regs[i] != resulting_regs[i]) {
+			printf("[FAIL]\tAX (saved SP) mismatch: requested 0x%llx; got 0x%llx\n",
+			       (unsigned long long)requested_regs[i],
+			       (unsigned long long)resulting_regs[i]);
+			nerrs++;
+			continue;
+		}
+
+		if (requested_regs[i] != resulting_regs[i] && !ignore_reg) {
+			/*
+			 * SP is particularly interesting here.  The
+			 * usual cause of failures is that we hit the
+			 * nasty IRET case of returning to a 16-bit SS,
+			 * in which case bits 16:31 of the *kernel*
+			 * stack pointer persist in ESP.
+			 */
+			printf("[FAIL]\tReg %d mismatch: requested 0x%llx; got 0x%llx\n",
+			       i, (unsigned long long)requested_regs[i],
+			       (unsigned long long)resulting_regs[i]);
+			nerrs++;
+		}
+	}
+
+	if (nerrs == 0)
+		printf("[OK]\tall registers okay\n");
+
+	return nerrs;
+}
+
+static int test_bad_iret(int cs_bits, unsigned short ss, int force_cs)
+{
+	int cs = force_cs == -1 ? find_cs(cs_bits) : force_cs;
+	if (cs == -1)
+		return 0;
+
+	sig_cs = cs;
+	sig_ss = ss;
+
+	printf("[RUN]\t%d-bit CS (%hx), bogus SS (%hx)\n",
+	       cs_bits, sig_cs, sig_ss);
+
+	sig_trapped = 0;
+	raise(SIGUSR1);
+	if (sig_trapped) {
+		char errdesc[32] = "";
+		if (sig_err) {
+			const char *src = (sig_err & 1) ? " EXT" : "";
+			const char *table;
+			if ((sig_err & 0x6) == 0x0)
+				table = "GDT";
+			else if ((sig_err & 0x6) == 0x4)
+				table = "LDT";
+			else if ((sig_err & 0x6) == 0x2)
+				table = "IDT";
+			else
+				table = "???";
+
+			sprintf(errdesc, "%s%s index %d, ",
+				table, src, sig_err >> 3);
+		}
+
+		char trapname[32];
+		if (sig_trapno == 13)
+			strcpy(trapname, "GP");
+		else if (sig_trapno == 11)
+			strcpy(trapname, "NP");
+		else if (sig_trapno == 12)
+			strcpy(trapname, "SS");
+		else if (sig_trapno == 32)
+			strcpy(trapname, "IRET");  /* X86_TRAP_IRET */
+		else
+			sprintf(trapname, "%d", sig_trapno);
+
+		printf("[OK]\tGot #%s(0x%lx) (i.e. %s%s)\n",
+		       trapname, (unsigned long)sig_err,
+		       errdesc, strsignal(sig_trapped));
+		return 0;
+	} else {
+		printf("[FAIL]\tDid not get SIGSEGV\n");
+		return 1;
+	}
+}
+
+int main()
+{
+	int total_nerrs = 0;
+	unsigned short my_cs, my_ss;
+
+	asm volatile ("mov %%cs,%0" : "=r" (my_cs));
+	asm volatile ("mov %%ss,%0" : "=r" (my_ss));
+	setup_ldt();
+
+	stack_t stack = {
+		.ss_sp = altstack_data,
+		.ss_size = SIGSTKSZ,
+	};
+	if (sigaltstack(&stack, NULL) != 0)
+		err(1, "sigaltstack");
+
+	sethandler(SIGUSR1, sigusr1, 0);
+	sethandler(SIGTRAP, sigtrap, SA_ONSTACK);
+
+	/* Easy cases: return to a 32-bit SS in each possible CS bitness. */
+	total_nerrs += test_valid_sigreturn(64, false, -1);
+	total_nerrs += test_valid_sigreturn(32, false, -1);
+	total_nerrs += test_valid_sigreturn(16, false, -1);
+
+	/*
+	 * Test easy espfix cases: return to a 16-bit LDT SS in each possible
+	 * CS bitness.  NB: with a long mode CS, the SS bitness is irrelevant.
+	 *
+	 * This catches the original missing-espfix-on-64-bit-kernels issue
+	 * as well as CVE-2014-8134.
+	 */
+	total_nerrs += test_valid_sigreturn(64, true, -1);
+	total_nerrs += test_valid_sigreturn(32, true, -1);
+	total_nerrs += test_valid_sigreturn(16, true, -1);
+
+	if (gdt_data16_idx) {
+		/*
+		 * For performance reasons, Linux skips espfix if SS points
+		 * to the GDT.  If we were able to allocate a 16-bit SS in
+		 * the GDT, see if it leaks parts of the kernel stack pointer.
+		 *
+		 * This tests for CVE-2014-8133.
+		 */
+		total_nerrs += test_valid_sigreturn(64, true,
+						    GDT3(gdt_data16_idx));
+		total_nerrs += test_valid_sigreturn(32, true,
+						    GDT3(gdt_data16_idx));
+		total_nerrs += test_valid_sigreturn(16, true,
+						    GDT3(gdt_data16_idx));
+	}
+
+	/*
+	 * We're done testing valid sigreturn cases.  Now we test states
+	 * for which sigreturn itself will succeed but the subsequent
+	 * entry to user mode will fail.
+	 *
+	 * Depending on the failure mode and the kernel bitness, these
+	 * entry failures can generate SIGSEGV, SIGBUS, or SIGILL.
+	 */
+	clearhandler(SIGTRAP);
+	sethandler(SIGSEGV, sigtrap, SA_ONSTACK);
+	sethandler(SIGBUS, sigtrap, SA_ONSTACK);
+	sethandler(SIGILL, sigtrap, SA_ONSTACK);  /* 32-bit kernels do this */
+
+	/* Easy failures: invalid SS, resulting in #GP(0) */
+	test_bad_iret(64, ldt_nonexistent_sel, -1);
+	test_bad_iret(32, ldt_nonexistent_sel, -1);
+	test_bad_iret(16, ldt_nonexistent_sel, -1);
+
+	/* These fail because SS isn't a data segment, resulting in #GP(SS) */
+	test_bad_iret(64, my_cs, -1);
+	test_bad_iret(32, my_cs, -1);
+	test_bad_iret(16, my_cs, -1);
+
+	/* Try to return to a not-present code segment, triggering #NP(SS). */
+	test_bad_iret(32, my_ss, npcode32_sel);
+
+	/*
+	 * Try to return to a not-present but otherwise valid data segment.
+	 * This will cause IRET to fail with #SS on the espfix stack.  This
+	 * exercises CVE-2014-9322.
+	 *
+	 * Note that, if espfix is enabled, 64-bit Linux will lose track
+	 * of the actual cause of failure and report #GP(0) instead.
+	 * This would be very difficult for Linux to avoid, because
+	 * espfix64 causes IRET failures to be promoted to #DF, so the
+	 * original exception frame is never pushed onto the stack.
+	 */
+	test_bad_iret(32, npdata32_sel, -1);
+
+	/*
+	 * Try to return to a not-present but otherwise valid data
+	 * segment without invoking espfix.  Newer kernels don't allow
+	 * this to happen in the first place.  On older kernels, though,
+	 * this can trigger CVE-2014-9322.
+	 */
+	if (gdt_npdata32_idx)
+		test_bad_iret(32, GDT3(gdt_npdata32_idx), -1);
+
+	return total_nerrs ? 1 : 0;
+}
diff --git a/x86/single_step_syscall.c b/x86/single_step_syscall.c
new file mode 100644
index 0000000..50c2635
--- /dev/null
+++ b/x86/single_step_syscall.c
@@ -0,0 +1,181 @@
+/*
+ * single_step_syscall.c - single-steps various x86 syscalls
+ * Copyright (c) 2014-2015 Andrew Lutomirski
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms and conditions of the GNU General Public License,
+ * version 2, as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope it will be useful, but
+ * WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * This is a very simple series of tests that makes system calls with
+ * the TF flag set.  This exercises some nasty kernel code in the
+ * SYSENTER case: SYSENTER does not clear TF, so SYSENTER with TF set
+ * immediately issues #DB from CPL 0.  This requires special handling in
+ * the kernel.
+ */
+
+#define _GNU_SOURCE
+
+#include <sys/time.h>
+#include <time.h>
+#include <stdlib.h>
+#include <sys/syscall.h>
+#include <unistd.h>
+#include <stdio.h>
+#include <string.h>
+#include <inttypes.h>
+#include <sys/mman.h>
+#include <sys/signal.h>
+#include <sys/ucontext.h>
+#include <asm/ldt.h>
+#include <err.h>
+#include <setjmp.h>
+#include <stddef.h>
+#include <stdbool.h>
+#include <sys/ptrace.h>
+#include <sys/user.h>
+
+static void sethandler(int sig, void (*handler)(int, siginfo_t *, void *),
+		       int flags)
+{
+	struct sigaction sa;
+	memset(&sa, 0, sizeof(sa));
+	sa.sa_sigaction = handler;
+	sa.sa_flags = SA_SIGINFO | flags;
+	sigemptyset(&sa.sa_mask);
+	if (sigaction(sig, &sa, 0))
+		err(1, "sigaction");
+}
+
+static volatile sig_atomic_t sig_traps;
+
+#ifdef __x86_64__
+# define REG_IP REG_RIP
+# define WIDTH "q"
+#else
+# define REG_IP REG_EIP
+# define WIDTH "l"
+#endif
+
+static unsigned long get_eflags(void)
+{
+	unsigned long eflags;
+	asm volatile ("pushf" WIDTH "\n\tpop" WIDTH " %0" : "=rm" (eflags));
+	return eflags;
+}
+
+static void set_eflags(unsigned long eflags)
+{
+	asm volatile ("push" WIDTH " %0\n\tpopf" WIDTH
+		      : : "rm" (eflags) : "flags");
+}
+
+#define X86_EFLAGS_TF (1UL << 8)
+
+static void sigtrap(int sig, siginfo_t *info, void *ctx_void)
+{
+	ucontext_t *ctx = (ucontext_t*)ctx_void;
+
+	if (get_eflags() & X86_EFLAGS_TF) {
+		set_eflags(get_eflags() & ~X86_EFLAGS_TF);
+		printf("[WARN]\tSIGTRAP handler had TF set\n");
+		_exit(1);
+	}
+
+	sig_traps++;
+
+	if (sig_traps == 10000 || sig_traps == 10001) {
+		printf("[WARN]\tHit %d SIGTRAPs with si_addr 0x%lx, ip 0x%lx\n",
+		       (int)sig_traps,
+		       (unsigned long)info->si_addr,
+		       (unsigned long)ctx->uc_mcontext.gregs[REG_IP]);
+	}
+}
+
+static void check_result(void)
+{
+	unsigned long new_eflags = get_eflags();
+	set_eflags(new_eflags & ~X86_EFLAGS_TF);
+
+	if (!sig_traps) {
+		printf("[FAIL]\tNo SIGTRAP\n");
+		exit(1);
+	}
+
+	if (!(new_eflags & X86_EFLAGS_TF)) {
+		printf("[FAIL]\tTF was cleared\n");
+		exit(1);
+	}
+
+	printf("[OK]\tSurvived with TF set and %d traps\n", (int)sig_traps);
+	sig_traps = 0;
+}
+
+int main()
+{
+	int tmp;
+
+	sethandler(SIGTRAP, sigtrap, 0);
+
+	printf("[RUN]\tSet TF and check nop\n");
+	set_eflags(get_eflags() | X86_EFLAGS_TF);
+	asm volatile ("nop");
+	check_result();
+
+#ifdef __x86_64__
+	printf("[RUN]\tSet TF and check syscall-less opportunistic sysret\n");
+	set_eflags(get_eflags() | X86_EFLAGS_TF);
+	extern unsigned char post_nop[];
+	asm volatile ("pushf" WIDTH "\n\t"
+		      "pop" WIDTH " %%r11\n\t"
+		      "nop\n\t"
+		      "post_nop:"
+		      : : "c" (post_nop) : "r11");
+	check_result();
+#endif
+
+	printf("[RUN]\tSet TF and check int80\n");
+	set_eflags(get_eflags() | X86_EFLAGS_TF);
+	asm volatile ("int $0x80" : "=a" (tmp) : "a" (SYS_getpid));
+	check_result();
+
+	/*
+	 * This test is particularly interesting if fast syscalls use
+	 * SYSENTER: it triggers a nasty design flaw in SYSENTER.
+	 * Specifically, SYSENTER does not clear TF, so either SYSENTER
+	 * or the next instruction traps at CPL0.  (Of course, Intel
+	 * mostly forgot to document exactly what happens here.)  So we
+	 * get a CPL0 fault with usergs (on 64-bit kernels) and possibly
+	 * no stack.  The only sane way the kernel can possibly handle
+	 * it is to clear TF on return from the #DB handler, but this
+	 * happens way too early to set TF in the saved pt_regs, so the
+	 * kernel has to do something clever to avoid losing track of
+	 * the TF bit.
+	 *
+	 * Needless to say, we've had bugs in this area.
+	 */
+	syscall(SYS_getpid);  /* Force symbol binding without TF set. */
+	printf("[RUN]\tSet TF and check a fast syscall\n");
+	set_eflags(get_eflags() | X86_EFLAGS_TF);
+	syscall(SYS_getpid);
+	check_result();
+
+	/* Now make sure that another fast syscall doesn't set TF again. */
+	printf("[RUN]\tFast syscall with TF cleared\n");
+	fflush(stdout);  /* Force a syscall */
+	if (get_eflags() & X86_EFLAGS_TF) {
+		printf("[FAIL]\tTF is now set\n");
+		exit(1);
+	}
+	if (sig_traps) {
+		printf("[FAIL]\tGot SIGTRAP\n");
+		exit(1);
+	}
+	printf("[OK]\tNothing unexpected happened\n");
+
+	return 0;
+}
diff --git a/x86/syscall_arg_fault.c b/x86/syscall_arg_fault.c
new file mode 100644
index 0000000..7db4fc9
--- /dev/null
+++ b/x86/syscall_arg_fault.c
@@ -0,0 +1,130 @@
+/*
+ * syscall_arg_fault.c - tests faults 32-bit fast syscall stack args
+ * Copyright (c) 2015 Andrew Lutomirski
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms and conditions of the GNU General Public License,
+ * version 2, as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope it will be useful, but
+ * WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ */
+
+#define _GNU_SOURCE
+
+#include <stdlib.h>
+#include <stdio.h>
+#include <string.h>
+#include <sys/signal.h>
+#include <sys/ucontext.h>
+#include <err.h>
+#include <setjmp.h>
+#include <errno.h>
+
+/* Our sigaltstack scratch space. */
+static unsigned char altstack_data[SIGSTKSZ];
+
+static void sethandler(int sig, void (*handler)(int, siginfo_t *, void *),
+		       int flags)
+{
+	struct sigaction sa;
+	memset(&sa, 0, sizeof(sa));
+	sa.sa_sigaction = handler;
+	sa.sa_flags = SA_SIGINFO | flags;
+	sigemptyset(&sa.sa_mask);
+	if (sigaction(sig, &sa, 0))
+		err(1, "sigaction");
+}
+
+static volatile sig_atomic_t sig_traps;
+static sigjmp_buf jmpbuf;
+
+static volatile sig_atomic_t n_errs;
+
+static void sigsegv(int sig, siginfo_t *info, void *ctx_void)
+{
+	ucontext_t *ctx = (ucontext_t*)ctx_void;
+
+	if (ctx->uc_mcontext.gregs[REG_EAX] != -EFAULT) {
+		printf("[FAIL]\tAX had the wrong value: 0x%x\n",
+		       ctx->uc_mcontext.gregs[REG_EAX]);
+		n_errs++;
+	} else {
+		printf("[OK]\tSeems okay\n");
+	}
+
+	siglongjmp(jmpbuf, 1);
+}
+
+static void sigill(int sig, siginfo_t *info, void *ctx_void)
+{
+	printf("[SKIP]\tIllegal instruction\n");
+	siglongjmp(jmpbuf, 1);
+}
+
+int main()
+{
+	stack_t stack = {
+		.ss_sp = altstack_data,
+		.ss_size = SIGSTKSZ,
+	};
+	if (sigaltstack(&stack, NULL) != 0)
+		err(1, "sigaltstack");
+
+	sethandler(SIGSEGV, sigsegv, SA_ONSTACK);
+	sethandler(SIGILL, sigill, SA_ONSTACK);
+
+	/*
+	 * Exercise another nasty special case.  The 32-bit SYSCALL
+	 * and SYSENTER instructions (even in compat mode) each
+	 * clobber one register.  A Linux system call has a syscall
+	 * number and six arguments, and the user stack pointer
+	 * needs to live in some register on return.  That means
+	 * that we need eight registers, but SYSCALL and SYSENTER
+	 * only preserve seven registers.  As a result, one argument
+	 * ends up on the stack.  The stack is user memory, which
+	 * means that the kernel can fail to read it.
+	 *
+	 * The 32-bit fast system calls don't have a defined ABI:
+	 * we're supposed to invoke them through the vDSO.  So we'll
+	 * fudge it: we set all regs to invalid pointer values and
+	 * invoke the entry instruction.  The return will fail no
+	 * matter what, and we completely lose our program state,
+	 * but we can fix it up with a signal handler.
+	 */
+
+	printf("[RUN]\tSYSENTER with invalid state\n");
+	if (sigsetjmp(jmpbuf, 1) == 0) {
+		asm volatile (
+			"movl $-1, %%eax\n\t"
+			"movl $-1, %%ebx\n\t"
+			"movl $-1, %%ecx\n\t"
+			"movl $-1, %%edx\n\t"
+			"movl $-1, %%esi\n\t"
+			"movl $-1, %%edi\n\t"
+			"movl $-1, %%ebp\n\t"
+			"movl $-1, %%esp\n\t"
+			"sysenter"
+			: : : "memory", "flags");
+	}
+
+	printf("[RUN]\tSYSCALL with invalid state\n");
+	if (sigsetjmp(jmpbuf, 1) == 0) {
+		asm volatile (
+			"movl $-1, %%eax\n\t"
+			"movl $-1, %%ebx\n\t"
+			"movl $-1, %%ecx\n\t"
+			"movl $-1, %%edx\n\t"
+			"movl $-1, %%esi\n\t"
+			"movl $-1, %%edi\n\t"
+			"movl $-1, %%ebp\n\t"
+			"movl $-1, %%esp\n\t"
+			"syscall\n\t"
+			"pushl $0"	/* make sure we segfault cleanly */
+			: : : "memory", "flags");
+	}
+
+	return 0;
+}
diff --git a/x86/syscall_nt.c b/x86/syscall_nt.c
new file mode 100644
index 0000000..60c06af
--- /dev/null
+++ b/x86/syscall_nt.c
@@ -0,0 +1,54 @@
+/*
+ * syscall_nt.c - checks syscalls with NT set
+ * Copyright (c) 2014-2015 Andrew Lutomirski
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms and conditions of the GNU General Public License,
+ * version 2, as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope it will be useful, but
+ * WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * Some obscure user-space code requires the ability to make system calls
+ * with FLAGS.NT set.  Make sure it works.
+ */
+
+#include <stdio.h>
+#include <unistd.h>
+#include <sys/syscall.h>
+#include <asm/processor-flags.h>
+
+#ifdef __x86_64__
+# define WIDTH "q"
+#else
+# define WIDTH "l"
+#endif
+
+static unsigned long get_eflags(void)
+{
+	unsigned long eflags;
+	asm volatile ("pushf" WIDTH "\n\tpop" WIDTH " %0" : "=rm" (eflags));
+	return eflags;
+}
+
+static void set_eflags(unsigned long eflags)
+{
+	asm volatile ("push" WIDTH " %0\n\tpopf" WIDTH
+		      : : "rm" (eflags) : "flags");
+}
+
+int main()
+{
+	printf("[RUN]\tSet NT and issue a syscall\n");
+	set_eflags(get_eflags() | X86_EFLAGS_NT);
+	syscall(SYS_getpid);
+	if (get_eflags() & X86_EFLAGS_NT) {
+		printf("[OK]\tThe syscall worked and NT is still set\n");
+		return 0;
+	} else {
+		printf("[FAIL]\tThe syscall worked but NT was cleared\n");
+		return 1;
+	}
+}
diff --git a/x86/sysret_ss_attrs.c b/x86/sysret_ss_attrs.c
new file mode 100644
index 0000000..ce42d5a
--- /dev/null
+++ b/x86/sysret_ss_attrs.c
@@ -0,0 +1,112 @@
+/*
+ * sysret_ss_attrs.c - test that syscalls return valid hidden SS attributes
+ * Copyright (c) 2015 Andrew Lutomirski
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms and conditions of the GNU General Public License,
+ * version 2, as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope it will be useful, but
+ * WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * On AMD CPUs, SYSRET can return with a valid SS descriptor with with
+ * the hidden attributes set to an unusable state.  Make sure the kernel
+ * doesn't let this happen.
+ */
+
+#define _GNU_SOURCE
+
+#include <stdlib.h>
+#include <unistd.h>
+#include <stdio.h>
+#include <string.h>
+#include <sys/mman.h>
+#include <err.h>
+#include <stddef.h>
+#include <stdbool.h>
+#include <pthread.h>
+
+static void *threadproc(void *ctx)
+{
+	/*
+	 * Do our best to cause sleeps on this CPU to exit the kernel and
+	 * re-enter with SS = 0.
+	 */
+	while (true)
+		;
+
+	return NULL;
+}
+
+#ifdef __x86_64__
+extern unsigned long call32_from_64(void *stack, void (*function)(void));
+
+asm (".pushsection .text\n\t"
+     ".code32\n\t"
+     "test_ss:\n\t"
+     "pushl $0\n\t"
+     "popl %eax\n\t"
+     "ret\n\t"
+     ".code64");
+extern void test_ss(void);
+#endif
+
+int main()
+{
+	/*
+	 * Start a busy-looping thread on the same CPU we're on.
+	 * For simplicity, just stick everything to CPU 0.  This will
+	 * fail in some containers, but that's probably okay.
+	 */
+	cpu_set_t cpuset;
+	CPU_ZERO(&cpuset);
+	CPU_SET(0, &cpuset);
+	if (sched_setaffinity(0, sizeof(cpuset), &cpuset) != 0)
+		printf("[WARN]\tsched_setaffinity failed\n");
+
+	pthread_t thread;
+	if (pthread_create(&thread, 0, threadproc, 0) != 0)
+		err(1, "pthread_create");
+
+#ifdef __x86_64__
+	unsigned char *stack32 = mmap(NULL, 4096, PROT_READ | PROT_WRITE,
+				      MAP_32BIT | MAP_ANONYMOUS | MAP_PRIVATE,
+				      -1, 0);
+	if (stack32 == MAP_FAILED)
+		err(1, "mmap");
+#endif
+
+	printf("[RUN]\tSyscalls followed by SS validation\n");
+
+	for (int i = 0; i < 1000; i++) {
+		/*
+		 * Go to sleep and return using sysret (if we're 64-bit
+		 * or we're 32-bit on AMD on a 64-bit kernel).  On AMD CPUs,
+		 * SYSRET doesn't fix up the cached SS descriptor, so the
+		 * kernel needs some kind of workaround to make sure that we
+		 * end the system call with a valid stack segment.  This
+		 * can be a confusing failure because the SS *selector*
+		 * is the same regardless.
+		 */
+		usleep(2);
+
+#ifdef __x86_64__
+		/*
+		 * On 32-bit, just doing a syscall through glibc is enough
+		 * to cause a crash if our cached SS descriptor is invalid.
+		 * On 64-bit, it's not, so try extra hard.
+		 */
+		call32_from_64(stack32 + 4088, test_ss);
+#endif
+	}
+
+	printf("[OK]\tWe survived\n");
+
+#ifdef __x86_64__
+	munmap(stack32, 4096);
+#endif
+
+	return 0;
+}
diff --git a/x86/test_FCMOV.c b/x86/test_FCMOV.c
new file mode 100644
index 0000000..4adcca0
--- /dev/null
+++ b/x86/test_FCMOV.c
@@ -0,0 +1,93 @@
+#undef _GNU_SOURCE
+#define _GNU_SOURCE 1
+#undef __USE_GNU
+#define __USE_GNU 1
+#include <unistd.h>
+#include <stdlib.h>
+#include <string.h>
+#include <stdio.h>
+#include <signal.h>
+#include <sys/types.h>
+#include <sys/select.h>
+#include <sys/time.h>
+#include <sys/wait.h>
+
+#define TEST(insn) \
+long double __attribute__((noinline)) insn(long flags) \
+{						\
+	long double out;			\
+	asm ("\n"				\
+	"	push	%1""\n"			\
+	"	popf""\n"			\
+	"	fldpi""\n"			\
+	"	fld1""\n"			\
+	"	" #insn " %%st(1), %%st" "\n"	\
+	"	ffree	%%st(1)" "\n"		\
+	: "=t" (out)				\
+	: "r" (flags)				\
+	);					\
+	return out;				\
+}
+
+TEST(fcmovb)
+TEST(fcmove)
+TEST(fcmovbe)
+TEST(fcmovu)
+TEST(fcmovnb)
+TEST(fcmovne)
+TEST(fcmovnbe)
+TEST(fcmovnu)
+
+enum {
+	CF = 1 << 0,
+	PF = 1 << 2,
+	ZF = 1 << 6,
+};
+
+void sighandler(int sig)
+{
+	printf("[FAIL]\tGot signal %d, exiting\n", sig);
+	exit(1);
+}
+
+int main(int argc, char **argv, char **envp)
+{
+	int err = 0;
+
+	/* SIGILL triggers on 32-bit kernels w/o fcomi emulation
+	 * when run with "no387 nofxsr". Other signals are caught
+	 * just in case.
+	 */
+	signal(SIGILL, sighandler);
+	signal(SIGFPE, sighandler);
+	signal(SIGSEGV, sighandler);
+
+	printf("[RUN]\tTesting fcmovCC instructions\n");
+	/* If fcmovCC() returns 1.0, the move wasn't done */
+	err |= !(fcmovb(0)   == 1.0); err |= !(fcmovnb(0)  != 1.0);
+	err |= !(fcmove(0)   == 1.0); err |= !(fcmovne(0)  != 1.0);
+	err |= !(fcmovbe(0)  == 1.0); err |= !(fcmovnbe(0) != 1.0);
+	err |= !(fcmovu(0)   == 1.0); err |= !(fcmovnu(0)  != 1.0);
+
+	err |= !(fcmovb(CF)  != 1.0); err |= !(fcmovnb(CF)  == 1.0);
+	err |= !(fcmove(CF)  == 1.0); err |= !(fcmovne(CF)  != 1.0);
+	err |= !(fcmovbe(CF) != 1.0); err |= !(fcmovnbe(CF) == 1.0);
+	err |= !(fcmovu(CF)  == 1.0); err |= !(fcmovnu(CF)  != 1.0);
+
+	err |= !(fcmovb(ZF)  == 1.0); err |= !(fcmovnb(ZF)  != 1.0);
+	err |= !(fcmove(ZF)  != 1.0); err |= !(fcmovne(ZF)  == 1.0);
+	err |= !(fcmovbe(ZF) != 1.0); err |= !(fcmovnbe(ZF) == 1.0);
+	err |= !(fcmovu(ZF)  == 1.0); err |= !(fcmovnu(ZF)  != 1.0);
+
+	err |= !(fcmovb(PF)  == 1.0); err |= !(fcmovnb(PF)  != 1.0);
+	err |= !(fcmove(PF)  == 1.0); err |= !(fcmovne(PF)  != 1.0);
+	err |= !(fcmovbe(PF) == 1.0); err |= !(fcmovnbe(PF) != 1.0);
+	err |= !(fcmovu(PF)  != 1.0); err |= !(fcmovnu(PF)  == 1.0);
+
+        if (!err)
+                printf("[OK]\tfcmovCC\n");
+	else
+		printf("[FAIL]\tfcmovCC errors: %d\n", err);
+
+	return err;
+}
diff --git a/x86/test_FCOMI.c b/x86/test_FCOMI.c
new file mode 100644
index 0000000..db4933e
--- /dev/null
+++ b/x86/test_FCOMI.c
@@ -0,0 +1,331 @@
+#undef _GNU_SOURCE
+#define _GNU_SOURCE 1
+#undef __USE_GNU
+#define __USE_GNU 1
+#include <unistd.h>
+#include <stdlib.h>
+#include <string.h>
+#include <stdio.h>
+#include <signal.h>
+#include <sys/types.h>
+#include <sys/select.h>
+#include <sys/time.h>
+#include <sys/wait.h>
+#include <fenv.h>
+
+enum {
+	CF = 1 << 0,
+	PF = 1 << 2,
+	ZF = 1 << 6,
+	ARITH = CF | PF | ZF,
+};
+
+long res_fcomi_pi_1;
+long res_fcomi_1_pi;
+long res_fcomi_1_1;
+long res_fcomi_nan_1;
+/* sNaN is s|111 1111 1|1xx xxxx xxxx xxxx xxxx xxxx */
+/* qNaN is s|111 1111 1|0xx xxxx xxxx xxxx xxxx xxxx (some x must be nonzero) */
+int snan = 0x7fc11111;
+int qnan = 0x7f811111;
+unsigned short snan1[5];
+/* sNaN80 is s|111 1111 1111 1111 |10xx xx...xx (some x must be nonzero) */
+unsigned short snan80[5] = { 0x1111, 0x1111, 0x1111, 0x8111, 0x7fff };
+
+int test(long flags)
+{
+	feclearexcept(FE_DIVBYZERO|FE_INEXACT|FE_INVALID|FE_OVERFLOW|FE_UNDERFLOW);
+
+	asm ("\n"
+
+	"	push	%0""\n"
+	"	popf""\n"
+	"	fld1""\n"
+	"	fldpi""\n"
+	"	fcomi	%%st(1), %%st" "\n"
+	"	ffree	%%st(0)" "\n"
+	"	ffree	%%st(1)" "\n"
+	"	pushf""\n"
+	"	pop	res_fcomi_1_pi""\n"
+
+	"	push	%0""\n"
+	"	popf""\n"
+	"	fldpi""\n"
+	"	fld1""\n"
+	"	fcomi	%%st(1), %%st" "\n"
+	"	ffree	%%st(0)" "\n"
+	"	ffree	%%st(1)" "\n"
+	"	pushf""\n"
+	"	pop	res_fcomi_pi_1""\n"
+
+	"	push	%0""\n"
+	"	popf""\n"
+	"	fld1""\n"
+	"	fld1""\n"
+	"	fcomi	%%st(1), %%st" "\n"
+	"	ffree	%%st(0)" "\n"
+	"	ffree	%%st(1)" "\n"
+	"	pushf""\n"
+	"	pop	res_fcomi_1_1""\n"
+	:
+	: "r" (flags)
+	);
+	if ((res_fcomi_1_pi & ARITH) != (0)) {
+		printf("[BAD]\tfcomi_1_pi with flags:%lx\n", flags);
+		return 1;
+	}
+	if ((res_fcomi_pi_1 & ARITH) != (CF)) {
+		printf("[BAD]\tfcomi_pi_1 with flags:%lx->%lx\n", flags, res_fcomi_pi_1 & ARITH);
+		return 1;
+	}
+	if ((res_fcomi_1_1 & ARITH) != (ZF)) {
+		printf("[BAD]\tfcomi_1_1 with flags:%lx\n", flags);
+		return 1;
+	}
+	if (fetestexcept(FE_INVALID) != 0) {
+		printf("[BAD]\tFE_INVALID is set in %s\n", __func__);
+		return 1;
+	}
+	return 0;
+}
+
+int test_qnan(long flags)
+{
+	feclearexcept(FE_DIVBYZERO|FE_INEXACT|FE_INVALID|FE_OVERFLOW|FE_UNDERFLOW);
+
+	asm ("\n"
+	"	push	%0""\n"
+	"	popf""\n"
+	"	flds	qnan""\n"
+	"	fld1""\n"
+	"	fnclex""\n"		// fld of a qnan raised FE_INVALID, clear it
+	"	fcomi	%%st(1), %%st" "\n"
+	"	ffree	%%st(0)" "\n"
+	"	ffree	%%st(1)" "\n"
+	"	pushf""\n"
+	"	pop	res_fcomi_nan_1""\n"
+	:
+	: "r" (flags)
+	);
+	if ((res_fcomi_nan_1 & ARITH) != (ZF|CF|PF)) {
+		printf("[BAD]\tfcomi_qnan_1 with flags:%lx\n", flags);
+		return 1;
+	}
+	if (fetestexcept(FE_INVALID) != FE_INVALID) {
+		printf("[BAD]\tFE_INVALID is not set in %s\n", __func__);
+		return 1;
+	}
+	return 0;
+}
+
+int testu_qnan(long flags)
+{
+	feclearexcept(FE_DIVBYZERO|FE_INEXACT|FE_INVALID|FE_OVERFLOW|FE_UNDERFLOW);
+
+	asm ("\n"
+	"	push	%0""\n"
+	"	popf""\n"
+	"	flds	qnan""\n"
+	"	fld1""\n"
+	"	fnclex""\n"		// fld of a qnan raised FE_INVALID, clear it
+	"	fucomi	%%st(1), %%st" "\n"
+	"	ffree	%%st(0)" "\n"
+	"	ffree	%%st(1)" "\n"
+	"	pushf""\n"
+	"	pop	res_fcomi_nan_1""\n"
+	:
+	: "r" (flags)
+	);
+	if ((res_fcomi_nan_1 & ARITH) != (ZF|CF|PF)) {
+		printf("[BAD]\tfcomi_qnan_1 with flags:%lx\n", flags);
+		return 1;
+	}
+	if (fetestexcept(FE_INVALID) != 0) {
+		printf("[BAD]\tFE_INVALID is set in %s\n", __func__);
+		return 1;
+	}
+	return 0;
+}
+
+int testu_snan(long flags)
+{
+	feclearexcept(FE_DIVBYZERO|FE_INEXACT|FE_INVALID|FE_OVERFLOW|FE_UNDERFLOW);
+
+	asm ("\n"
+	"	push	%0""\n"
+	"	popf""\n"
+//	"	flds	snan""\n"	// WRONG, this will convert 32-bit fp snan to a *qnan* in 80-bit fp register!
+//	"	fstpt	snan1""\n"	// if uncommented, it prints "snan1:7fff c111 1100 0000 0000" - c111, not 8111!
+//	"	fnclex""\n"		// flds of a snan raised FE_INVALID, clear it
+	"	fldt	snan80""\n"	// fldt never raise FE_INVALID
+	"	fld1""\n"
+	"	fucomi	%%st(1), %%st" "\n"
+	"	ffree	%%st(0)" "\n"
+	"	ffree	%%st(1)" "\n"
+	"	pushf""\n"
+	"	pop	res_fcomi_nan_1""\n"
+	:
+	: "r" (flags)
+	);
+	if ((res_fcomi_nan_1 & ARITH) != (ZF|CF|PF)) {
+		printf("[BAD]\tfcomi_qnan_1 with flags:%lx\n", flags);
+		return 1;
+	}
+//	printf("snan:%x snan1:%04x %04x %04x %04x %04x\n", snan, snan1[4], snan1[3], snan1[2], snan1[1], snan1[0]);
+	if (fetestexcept(FE_INVALID) != FE_INVALID) {
+		printf("[BAD]\tFE_INVALID is not set in %s\n", __func__);
+		return 1;
+	}
+	return 0;
+}
+
+int testp(long flags)
+{
+	feclearexcept(FE_DIVBYZERO|FE_INEXACT|FE_INVALID|FE_OVERFLOW|FE_UNDERFLOW);
+
+	asm ("\n"
+
+	"	push	%0""\n"
+	"	popf""\n"
+	"	fld1""\n"
+	"	fldpi""\n"
+	"	fcomip	%%st(1), %%st" "\n"
+	"	ffree	%%st(0)" "\n"
+	"	pushf""\n"
+	"	pop	res_fcomi_1_pi""\n"
+
+	"	push	%0""\n"
+	"	popf""\n"
+	"	fldpi""\n"
+	"	fld1""\n"
+	"	fcomip	%%st(1), %%st" "\n"
+	"	ffree	%%st(0)" "\n"
+	"	pushf""\n"
+	"	pop	res_fcomi_pi_1""\n"
+
+	"	push	%0""\n"
+	"	popf""\n"
+	"	fld1""\n"
+	"	fld1""\n"
+	"	fcomip	%%st(1), %%st" "\n"
+	"	ffree	%%st(0)" "\n"
+	"	pushf""\n"
+	"	pop	res_fcomi_1_1""\n"
+	:
+	: "r" (flags)
+	);
+	if ((res_fcomi_1_pi & ARITH) != (0)) {
+		printf("[BAD]\tfcomi_1_pi with flags:%lx\n", flags);
+		return 1;
+	}
+	if ((res_fcomi_pi_1 & ARITH) != (CF)) {
+		printf("[BAD]\tfcomi_pi_1 with flags:%lx->%lx\n", flags, res_fcomi_pi_1 & ARITH);
+		return 1;
+	}
+	if ((res_fcomi_1_1 & ARITH) != (ZF)) {
+		printf("[BAD]\tfcomi_1_1 with flags:%lx\n", flags);
+		return 1;
+	}
+	if (fetestexcept(FE_INVALID) != 0) {
+		printf("[BAD]\tFE_INVALID is set in %s\n", __func__);
+		return 1;
+	}
+	return 0;
+}
+
+int testp_qnan(long flags)
+{
+	feclearexcept(FE_DIVBYZERO|FE_INEXACT|FE_INVALID|FE_OVERFLOW|FE_UNDERFLOW);
+
+	asm ("\n"
+	"	push	%0""\n"
+	"	popf""\n"
+	"	flds	qnan""\n"
+	"	fld1""\n"
+	"	fnclex""\n"		// fld of a qnan raised FE_INVALID, clear it
+	"	fcomip	%%st(1), %%st" "\n"
+	"	ffree	%%st(0)" "\n"
+	"	pushf""\n"
+	"	pop	res_fcomi_nan_1""\n"
+	:
+	: "r" (flags)
+	);
+	if ((res_fcomi_nan_1 & ARITH) != (ZF|CF|PF)) {
+		printf("[BAD]\tfcomi_qnan_1 with flags:%lx\n", flags);
+		return 1;
+	}
+	if (fetestexcept(FE_INVALID) != FE_INVALID) {
+		printf("[BAD]\tFE_INVALID is not set in %s\n", __func__);
+		return 1;
+	}
+	return 0;
+}
+
+int testup_qnan(long flags)
+{
+	feclearexcept(FE_DIVBYZERO|FE_INEXACT|FE_INVALID|FE_OVERFLOW|FE_UNDERFLOW);
+
+	asm ("\n"
+	"	push	%0""\n"
+	"	popf""\n"
+	"	flds	qnan""\n"
+	"	fld1""\n"
+	"	fnclex""\n"		// fld of a qnan raised FE_INVALID, clear it
+	"	fucomip	%%st(1), %%st" "\n"
+	"	ffree	%%st(0)" "\n"
+	"	pushf""\n"
+	"	pop	res_fcomi_nan_1""\n"
+	:
+	: "r" (flags)
+	);
+	if ((res_fcomi_nan_1 & ARITH) != (ZF|CF|PF)) {
+		printf("[BAD]\tfcomi_qnan_1 with flags:%lx\n", flags);
+		return 1;
+	}
+	if (fetestexcept(FE_INVALID) != 0) {
+		printf("[BAD]\tFE_INVALID is set in %s\n", __func__);
+		return 1;
+	}
+	return 0;
+}
+
+void sighandler(int sig)
+{
+	printf("[FAIL]\tGot signal %d, exiting\n", sig);
+	exit(1);
+}
+
+int main(int argc, char **argv, char **envp)
+{
+	int err = 0;
+
+	/* SIGILL triggers on 32-bit kernels w/o fcomi emulation
+	 * when run with "no387 nofxsr". Other signals are caught
+	 * just in case.
+	 */
+	signal(SIGILL, sighandler);
+	signal(SIGFPE, sighandler);
+	signal(SIGSEGV, sighandler);
+
+	printf("[RUN]\tTesting f[u]comi[p] instructions\n");
+	err |= test(0);
+	err |= test_qnan(0);
+	err |= testu_qnan(0);
+	err |= testu_snan(0);
+	err |= test(CF|ZF|PF);
+	err |= test_qnan(CF|ZF|PF);
+	err |= testu_qnan(CF|ZF|PF);
+	err |= testu_snan(CF|ZF|PF);
+	err |= testp(0);
+	err |= testp_qnan(0);
+	err |= testup_qnan(0);
+	err |= testp(CF|ZF|PF);
+	err |= testp_qnan(CF|ZF|PF);
+	err |= testup_qnan(CF|ZF|PF);
+	if (!err)
+		printf("[OK]\tf[u]comi[p]\n");
+	else
+		printf("[FAIL]\tf[u]comi[p] errors: %d\n", err);
+
+	return err;
+}
diff --git a/x86/test_FISTTP.c b/x86/test_FISTTP.c
new file mode 100644
index 0000000..b8e61a0
--- /dev/null
+++ b/x86/test_FISTTP.c
@@ -0,0 +1,137 @@
+#undef _GNU_SOURCE
+#define _GNU_SOURCE 1
+#undef __USE_GNU
+#define __USE_GNU 1
+#include <unistd.h>
+#include <stdlib.h>
+#include <string.h>
+#include <stdio.h>
+#include <signal.h>
+#include <sys/types.h>
+#include <sys/select.h>
+#include <sys/time.h>
+#include <sys/wait.h>
+#include <fenv.h>
+
+unsigned long long res64 = -1;
+unsigned int res32 = -1;
+unsigned short res16 = -1;
+
+int test(void)
+{
+	int ex;
+
+	feclearexcept(FE_DIVBYZERO|FE_INEXACT|FE_INVALID|FE_OVERFLOW|FE_UNDERFLOW);
+	asm volatile ("\n"
+	"	fld1""\n"
+	"	fisttp	res16""\n"
+	"	fld1""\n"
+	"	fisttpl	res32""\n"
+	"	fld1""\n"
+	"	fisttpll res64""\n"
+	: : : "memory"
+	);
+	if (res16 != 1 || res32 != 1 || res64 != 1) {
+		printf("[BAD]\tfisttp 1\n");
+		return 1;
+	}
+	ex = fetestexcept(FE_DIVBYZERO|FE_INEXACT|FE_INVALID|FE_OVERFLOW|FE_UNDERFLOW);
+	if (ex != 0) {
+		printf("[BAD]\tfisttp 1: wrong exception state\n");
+		return 1;
+	}
+
+	feclearexcept(FE_DIVBYZERO|FE_INEXACT|FE_INVALID|FE_OVERFLOW|FE_UNDERFLOW);
+	asm volatile ("\n"
+	"	fldpi""\n"
+	"	fisttp	res16""\n"
+	"	fldpi""\n"
+	"	fisttpl	res32""\n"
+	"	fldpi""\n"
+	"	fisttpll res64""\n"
+	: : : "memory"
+	);
+	if (res16 != 3 || res32 != 3 || res64 != 3) {
+		printf("[BAD]\tfisttp pi\n");
+		return 1;
+	}
+	ex = fetestexcept(FE_DIVBYZERO|FE_INEXACT|FE_INVALID|FE_OVERFLOW|FE_UNDERFLOW);
+	if (ex != FE_INEXACT) {
+		printf("[BAD]\tfisttp pi: wrong exception state\n");
+		return 1;
+	}
+
+	feclearexcept(FE_DIVBYZERO|FE_INEXACT|FE_INVALID|FE_OVERFLOW|FE_UNDERFLOW);
+	asm volatile ("\n"
+	"	fldpi""\n"
+	"	fchs""\n"
+	"	fisttp	res16""\n"
+	"	fldpi""\n"
+	"	fchs""\n"
+	"	fisttpl	res32""\n"
+	"	fldpi""\n"
+	"	fchs""\n"
+	"	fisttpll res64""\n"
+	: : : "memory"
+	);
+	if (res16 != 0xfffd || res32 != 0xfffffffd || res64 != 0xfffffffffffffffdULL) {
+		printf("[BAD]\tfisttp -pi\n");
+		return 1;
+	}
+	ex = fetestexcept(FE_DIVBYZERO|FE_INEXACT|FE_INVALID|FE_OVERFLOW|FE_UNDERFLOW);
+	if (ex != FE_INEXACT) {
+		printf("[BAD]\tfisttp -pi: wrong exception state\n");
+		return 1;
+	}
+
+	feclearexcept(FE_DIVBYZERO|FE_INEXACT|FE_INVALID|FE_OVERFLOW|FE_UNDERFLOW);
+	asm volatile ("\n"
+	"	fldln2""\n"
+	"	fisttp	res16""\n"
+	"	fldln2""\n"
+	"	fisttpl	res32""\n"
+	"	fldln2""\n"
+	"	fisttpll res64""\n"
+	: : : "memory"
+	);
+	/* Test truncation to zero (round-to-nearest would give 1 here) */
+	if (res16 != 0 || res32 != 0 || res64 != 0) {
+		printf("[BAD]\tfisttp ln2\n");
+		return 1;
+	}
+	ex = fetestexcept(FE_DIVBYZERO|FE_INEXACT|FE_INVALID|FE_OVERFLOW|FE_UNDERFLOW);
+	if (ex != FE_INEXACT) {
+		printf("[BAD]\tfisttp ln2: wrong exception state\n");
+		return 1;
+	}
+
+	return 0;
+}
+
+void sighandler(int sig)
+{
+	printf("[FAIL]\tGot signal %d, exiting\n", sig);
+	exit(1);
+}
+
+int main(int argc, char **argv, char **envp)
+{
+	int err = 0;
+
+	/* SIGILL triggers on 32-bit kernels w/o fisttp emulation
+	 * when run with "no387 nofxsr". Other signals are caught
+	 * just in case.
+	 */
+	signal(SIGILL, sighandler);
+	signal(SIGFPE, sighandler);
+	signal(SIGSEGV, sighandler);
+
+	printf("[RUN]\tTesting fisttp instructions\n");
+	err |= test();
+	if (!err)
+		printf("[OK]\tfisttp\n");
+	else
+		printf("[FAIL]\tfisttp errors: %d\n", err);
+
+	return err;
+}
diff --git a/x86/test_syscall_vdso.c b/x86/test_syscall_vdso.c
new file mode 100644
index 0000000..4037035
--- /dev/null
+++ b/x86/test_syscall_vdso.c
@@ -0,0 +1,401 @@
+/*
+ * 32-bit syscall ABI conformance test.
+ *
+ * Copyright (c) 2015 Denys Vlasenko
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms and conditions of the GNU General Public License,
+ * version 2, as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope it will be useful, but
+ * WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ */
+/*
+ * Can be built statically:
+ * gcc -Os -Wall -static -m32 test_syscall_vdso.c thunks_32.S
+ */
+#undef _GNU_SOURCE
+#define _GNU_SOURCE 1
+#undef __USE_GNU
+#define __USE_GNU 1
+#include <unistd.h>
+#include <stdlib.h>
+#include <string.h>
+#include <stdio.h>
+#include <signal.h>
+#include <sys/types.h>
+#include <sys/select.h>
+#include <sys/time.h>
+#include <elf.h>
+#include <sys/ptrace.h>
+#include <sys/wait.h>
+
+#if !defined(__i386__)
+int main(int argc, char **argv, char **envp)
+{
+	printf("[SKIP]\tNot a 32-bit x86 userspace\n");
+	return 0;
+}
+#else
+
+long syscall_addr;
+long get_syscall(char **envp)
+{
+	Elf32_auxv_t *auxv;
+	while (*envp++ != NULL)
+		continue;
+	for (auxv = (void *)envp; auxv->a_type != AT_NULL; auxv++)
+		if (auxv->a_type == AT_SYSINFO)
+			return auxv->a_un.a_val;
+	printf("[WARN]\tAT_SYSINFO not supplied\n");
+	return 0;
+}
+
+asm (
+	"	.pushsection .text\n"
+	"	.global	int80\n"
+	"int80:\n"
+	"	int	$0x80\n"
+	"	ret\n"
+	"	.popsection\n"
+);
+extern char int80;
+
+struct regs64 {
+	uint64_t rax, rbx, rcx, rdx;
+	uint64_t rsi, rdi, rbp, rsp;
+	uint64_t r8,  r9,  r10, r11;
+	uint64_t r12, r13, r14, r15;
+};
+struct regs64 regs64;
+int kernel_is_64bit;
+
+asm (
+	"	.pushsection .text\n"
+	"	.code64\n"
+	"get_regs64:\n"
+	"	push	%rax\n"
+	"	mov	$regs64, %eax\n"
+	"	pop	0*8(%rax)\n"
+	"	movq	%rbx, 1*8(%rax)\n"
+	"	movq	%rcx, 2*8(%rax)\n"
+	"	movq	%rdx, 3*8(%rax)\n"
+	"	movq	%rsi, 4*8(%rax)\n"
+	"	movq	%rdi, 5*8(%rax)\n"
+	"	movq	%rbp, 6*8(%rax)\n"
+	"	movq	%rsp, 7*8(%rax)\n"
+	"	movq	%r8,  8*8(%rax)\n"
+	"	movq	%r9,  9*8(%rax)\n"
+	"	movq	%r10, 10*8(%rax)\n"
+	"	movq	%r11, 11*8(%rax)\n"
+	"	movq	%r12, 12*8(%rax)\n"
+	"	movq	%r13, 13*8(%rax)\n"
+	"	movq	%r14, 14*8(%rax)\n"
+	"	movq	%r15, 15*8(%rax)\n"
+	"	ret\n"
+	"poison_regs64:\n"
+	"	movq	$0x7f7f7f7f, %r8\n"
+	"	shl	$32, %r8\n"
+	"	orq	$0x7f7f7f7f, %r8\n"
+	"	movq	%r8, %r9\n"
+	"	movq	%r8, %r10\n"
+	"	movq	%r8, %r11\n"
+	"	movq	%r8, %r12\n"
+	"	movq	%r8, %r13\n"
+	"	movq	%r8, %r14\n"
+	"	movq	%r8, %r15\n"
+	"	ret\n"
+	"	.code32\n"
+	"	.popsection\n"
+);
+extern void get_regs64(void);
+extern void poison_regs64(void);
+extern unsigned long call64_from_32(void (*function)(void));
+void print_regs64(void)
+{
+	if (!kernel_is_64bit)
+		return;
+	printf("ax:%016llx bx:%016llx cx:%016llx dx:%016llx\n", regs64.rax,  regs64.rbx,  regs64.rcx,  regs64.rdx);
+	printf("si:%016llx di:%016llx bp:%016llx sp:%016llx\n", regs64.rsi,  regs64.rdi,  regs64.rbp,  regs64.rsp);
+	printf(" 8:%016llx  9:%016llx 10:%016llx 11:%016llx\n", regs64.r8 ,  regs64.r9 ,  regs64.r10,  regs64.r11);
+	printf("12:%016llx 13:%016llx 14:%016llx 15:%016llx\n", regs64.r12,  regs64.r13,  regs64.r14,  regs64.r15);
+}
+
+int check_regs64(void)
+{
+	int err = 0;
+	int num = 8;
+	uint64_t *r64 = &regs64.r8;
+
+	if (!kernel_is_64bit)
+		return 0;
+
+	do {
+		if (*r64 == 0x7f7f7f7f7f7f7f7fULL)
+			continue; /* register did not change */
+		if (syscall_addr != (long)&int80) {
+			/*
+			 * Non-INT80 syscall entrypoints are allowed to clobber R8+ regs:
+			 * either clear them to 0, or for R11, load EFLAGS.
+			 */
+			if (*r64 == 0)
+				continue;
+			if (num == 11) {
+				printf("[NOTE]\tR11 has changed:%016llx - assuming clobbered by SYSRET insn\n", *r64);
+				continue;
+			}
+		} else {
+			/* INT80 syscall entrypoint can be used by
+			 * 64-bit programs too, unlike SYSCALL/SYSENTER.
+			 * Therefore it must preserve R12+
+			 * (they are callee-saved registers in 64-bit C ABI).
+			 *
+			 * This was probably historically not intended,
+			 * but R8..11 are clobbered (cleared to 0).
+			 * IOW: they are the only registers which aren't
+			 * preserved across INT80 syscall.
+			 */
+			if (*r64 == 0 && num <= 11)
+				continue;
+		}
+		printf("[FAIL]\tR%d has changed:%016llx\n", num, *r64);
+		err++;
+	} while (r64++, ++num < 16);
+
+	if (!err)
+		printf("[OK]\tR8..R15 did not leak kernel data\n");
+	return err;
+}
+
+int nfds;
+fd_set rfds;
+fd_set wfds;
+fd_set efds;
+struct timespec timeout;
+sigset_t sigmask;
+struct {
+	sigset_t *sp;
+	int sz;
+} sigmask_desc;
+
+void prep_args()
+{
+	nfds = 42;
+	FD_ZERO(&rfds);
+	FD_ZERO(&wfds);
+	FD_ZERO(&efds);
+	FD_SET(0, &rfds);
+	FD_SET(1, &wfds);
+	FD_SET(2, &efds);
+	timeout.tv_sec = 0;
+	timeout.tv_nsec = 123;
+	sigemptyset(&sigmask);
+	sigaddset(&sigmask, SIGINT);
+	sigaddset(&sigmask, SIGUSR2);
+	sigaddset(&sigmask, SIGRTMAX);
+	sigmask_desc.sp = &sigmask;
+	sigmask_desc.sz = 8; /* bytes */
+}
+
+static void print_flags(const char *name, unsigned long r)
+{
+	static const char *bitarray[] = {
+	"\n" ,"c\n" ,/* Carry Flag */
+	"0 " ,"1 "  ,/* Bit 1 - always on */
+	""   ,"p "  ,/* Parity Flag */
+	"0 " ,"3? " ,
+	""   ,"a "  ,/* Auxiliary carry Flag */
+	"0 " ,"5? " ,
+	""   ,"z "  ,/* Zero Flag */
+	""   ,"s "  ,/* Sign Flag */
+	""   ,"t "  ,/* Trap Flag */
+	""   ,"i "  ,/* Interrupt Flag */
+	""   ,"d "  ,/* Direction Flag */
+	""   ,"o "  ,/* Overflow Flag */
+	"0 " ,"1 "  ,/* I/O Privilege Level (2 bits) */
+	"0"  ,"1"   ,/* I/O Privilege Level (2 bits) */
+	""   ,"n "  ,/* Nested Task */
+	"0 " ,"15? ",
+	""   ,"r "  ,/* Resume Flag */
+	""   ,"v "  ,/* Virtual Mode */
+	""   ,"ac " ,/* Alignment Check/Access Control */
+	""   ,"vif ",/* Virtual Interrupt Flag */
+	""   ,"vip ",/* Virtual Interrupt Pending */
+	""   ,"id " ,/* CPUID detection */
+	NULL
+	};
+	const char **bitstr;
+	int bit;
+
+	printf("%s=%016lx ", name, r);
+	bitstr = bitarray + 42;
+	bit = 21;
+	if ((r >> 22) != 0)
+		printf("(extra bits are set) ");
+	do {
+		if (bitstr[(r >> bit) & 1][0])
+			fputs(bitstr[(r >> bit) & 1], stdout);
+		bitstr -= 2;
+		bit--;
+	} while (bit >= 0);
+}
+
+int run_syscall(void)
+{
+	long flags, bad_arg;
+
+	prep_args();
+
+	if (kernel_is_64bit)
+		call64_from_32(poison_regs64);
+	/*print_regs64();*/
+
+	asm("\n"
+	/* Try 6-arg syscall: pselect. It should return quickly */
+	"	push	%%ebp\n"
+	"	mov	$308, %%eax\n"     /* PSELECT */
+	"	mov	nfds, %%ebx\n"     /* ebx  arg1 */
+	"	mov	$rfds, %%ecx\n"    /* ecx  arg2 */
+	"	mov	$wfds, %%edx\n"    /* edx  arg3 */
+	"	mov	$efds, %%esi\n"    /* esi  arg4 */
+	"	mov	$timeout, %%edi\n" /* edi  arg5 */
+	"	mov	$sigmask_desc, %%ebp\n" /* %ebp arg6 */
+	"	push	$0x200ed7\n"      /* set almost all flags */
+	"	popf\n"		/* except TF, IOPL, NT, RF, VM, AC, VIF, VIP */
+	"	call	*syscall_addr\n"
+	/* Check that registers are not clobbered */
+	"	pushf\n"
+	"	pop	%%eax\n"
+	"	cld\n"
+	"	cmp	nfds, %%ebx\n"     /* ebx  arg1 */
+	"	mov	$1, %%ebx\n"
+	"	jne	1f\n"
+	"	cmp	$rfds, %%ecx\n"    /* ecx  arg2 */
+	"	mov	$2, %%ebx\n"
+	"	jne	1f\n"
+	"	cmp	$wfds, %%edx\n"    /* edx  arg3 */
+	"	mov	$3, %%ebx\n"
+	"	jne	1f\n"
+	"	cmp	$efds, %%esi\n"    /* esi  arg4 */
+	"	mov	$4, %%ebx\n"
+	"	jne	1f\n"
+	"	cmp	$timeout, %%edi\n" /* edi  arg5 */
+	"	mov	$5, %%ebx\n"
+	"	jne	1f\n"
+	"	cmpl	$sigmask_desc, %%ebp\n" /* %ebp arg6 */
+	"	mov	$6, %%ebx\n"
+	"	jne	1f\n"
+	"	mov	$0, %%ebx\n"
+	"1:\n"
+	"	pop	%%ebp\n"
+	: "=a" (flags), "=b" (bad_arg)
+	:
+	: "cx", "dx", "si", "di"
+	);
+
+	if (kernel_is_64bit) {
+		memset(&regs64, 0x77, sizeof(regs64));
+		call64_from_32(get_regs64);
+		/*print_regs64();*/
+	}
+
+	/*
+	 * On paravirt kernels, flags are not preserved across syscalls.
+	 * Thus, we do not consider it a bug if some are changed.
+	 * We just show ones which do.
+	 */
+	if ((0x200ed7 ^ flags) != 0) {
+		print_flags("[WARN]\tFlags before", 0x200ed7);
+		print_flags("[WARN]\tFlags  after", flags);
+		print_flags("[WARN]\tFlags change", (0x200ed7 ^ flags));
+	}
+
+	if (bad_arg) {
+		printf("[FAIL]\targ#%ld clobbered\n", bad_arg);
+		return 1;
+	}
+	printf("[OK]\tArguments are preserved across syscall\n");
+
+	return check_regs64();
+}
+
+int run_syscall_twice()
+{
+	int exitcode = 0;
+	long sv;
+
+	if (syscall_addr) {
+		printf("[RUN]\tExecuting 6-argument 32-bit syscall via VDSO\n");
+		exitcode = run_syscall();
+	}
+	sv = syscall_addr;
+	syscall_addr = (long)&int80;
+	printf("[RUN]\tExecuting 6-argument 32-bit syscall via INT 80\n");
+	exitcode += run_syscall();
+	syscall_addr = sv;
+	return exitcode;
+}
+
+void ptrace_me()
+{
+	pid_t pid;
+
+	fflush(NULL);
+	pid = fork();
+	if (pid < 0)
+		exit(1);
+	if (pid == 0) {
+		/* child */
+		if (ptrace(PTRACE_TRACEME, 0L, 0L, 0L) != 0)
+			exit(0);
+		raise(SIGSTOP);
+		return;
+	}
+	/* parent */
+	printf("[RUN]\tRunning tests under ptrace\n");
+	while (1) {
+		int status;
+		pid = waitpid(-1, &status, __WALL);
+		if (WIFEXITED(status))
+			exit(WEXITSTATUS(status));
+		if (WIFSIGNALED(status))
+			exit(WTERMSIG(status));
+		if (pid <= 0 || !WIFSTOPPED(status)) /* paranoia */
+			exit(255);
+		/*
+		 * Note: we do not inject sig = WSTOPSIG(status).
+		 * We probably should, but careful: do not inject SIGTRAP
+		 * generated by syscall entry/exit stops.
+		 * That kills the child.
+		 */
+		ptrace(PTRACE_SYSCALL, pid, 0L, 0L /*sig*/);
+	}
+}
+
+int main(int argc, char **argv, char **envp)
+{
+	int exitcode = 0;
+	int cs;
+
+	asm("\n"
+	"	movl	%%cs, %%eax\n"
+	: "=a" (cs)
+	);
+	kernel_is_64bit = (cs == 0x23);
+	if (!kernel_is_64bit)
+		printf("[NOTE]\tNot a 64-bit kernel, won't test R8..R15 leaks\n");
+
+	/* This only works for non-static builds:
+	 * syscall_addr = dlsym(dlopen("linux-gate.so.1", RTLD_NOW), "__kernel_vsyscall");
+	 */
+	syscall_addr = get_syscall(envp);
+
+	exitcode += run_syscall_twice();
+	ptrace_me();
+	exitcode += run_syscall_twice();
+
+	return exitcode;
+}
+#endif
diff --git a/x86/thunks.S b/x86/thunks.S
new file mode 100644
index 0000000..ce8a995
--- /dev/null
+++ b/x86/thunks.S
@@ -0,0 +1,67 @@
+/*
+ * thunks.S - assembly helpers for mixed-bitness code
+ * Copyright (c) 2015 Andrew Lutomirski
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms and conditions of the GNU General Public License,
+ * version 2, as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope it will be useful, but
+ * WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * These are little helpers that make it easier to switch bitness on
+ * the fly.
+ */
+
+	.text
+
+	.global call32_from_64
+	.type call32_from_64, @function
+call32_from_64:
+	// rdi: stack to use
+	// esi: function to call
+
+	// Save registers
+	pushq %rbx
+	pushq %rbp
+	pushq %r12
+	pushq %r13
+	pushq %r14
+	pushq %r15
+	pushfq
+
+	// Switch stacks
+	mov %rsp,(%rdi)
+	mov %rdi,%rsp
+
+	// Switch to compatibility mode
+	pushq $0x23  /* USER32_CS */
+	pushq $1f
+	lretq
+
+1:
+	.code32
+	// Call the function
+	call *%esi
+	// Switch back to long mode
+	jmp $0x33,$1f
+	.code64
+
+1:
+	// Restore the stack
+	mov (%rsp),%rsp
+
+	// Restore registers
+	popfq
+	popq %r15
+	popq %r14
+	popq %r13
+	popq %r12
+	popq %rbp
+	popq %rbx
+
+	ret
+
+.size call32_from_64, .-call32_from_64
diff --git a/x86/thunks_32.S b/x86/thunks_32.S
new file mode 100644
index 0000000..29b644b
--- /dev/null
+++ b/x86/thunks_32.S
@@ -0,0 +1,55 @@
+/*
+ * thunks_32.S - assembly helpers for mixed-bitness code
+ * Copyright (c) 2015 Denys Vlasenko
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms and conditions of the GNU General Public License,
+ * version 2, as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope it will be useful, but
+ * WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * These are little helpers that make it easier to switch bitness on
+ * the fly.
+ */
+
+	.text
+	.code32
+
+	.global call64_from_32
+	.type call32_from_64, @function
+
+	// 4(%esp): function to call
+call64_from_32:
+	// Fetch function address
+	mov	4(%esp), %eax
+
+	// Save registers which are callee-clobbered by 64-bit ABI
+	push	%ecx
+	push	%edx
+	push	%esi
+	push	%edi
+
+	// Switch to long mode
+	jmp	$0x33,$1f
+1:	.code64
+
+	// Call the function
+	call	*%rax
+
+	// Switch to compatibility mode
+	push	$0x23  /* USER32_CS */
+	.code32; push $1f; .code64 /* hack: can't have X86_64_32S relocation in 32-bit ELF */
+	lretq
+1:	.code32
+
+	pop	%edi
+	pop	%esi
+	pop	%edx
+	pop	%ecx
+
+	ret
+
+.size call64_from_32, .-call64_from_32
diff --git a/x86/trivial_32bit_program.c b/x86/trivial_32bit_program.c
new file mode 100644
index 0000000..fabdf0f
--- /dev/null
+++ b/x86/trivial_32bit_program.c
@@ -0,0 +1,18 @@
+/*
+ * Trivial program to check that we have a valid 32-bit build environment.
+ * Copyright (c) 2015 Andy Lutomirski
+ * GPL v2
+ */
+
+#ifndef __i386__
+# error wrong architecture
+#endif
+
+#include <stdio.h>
+
+int main()
+{
+	printf("\n");
+
+	return 0;
+}
diff --git a/x86/trivial_64bit_program.c b/x86/trivial_64bit_program.c
new file mode 100644
index 0000000..05c6a41
--- /dev/null
+++ b/x86/trivial_64bit_program.c
@@ -0,0 +1,18 @@
+/*
+ * Trivial program to check that we have a valid 64-bit build environment.
+ * Copyright (c) 2015 Andy Lutomirski
+ * GPL v2
+ */
+
+#ifndef __x86_64__
+# error wrong architecture
+#endif
+
+#include <stdio.h>
+
+int main()
+{
+	printf("\n");
+
+	return 0;
+}
diff --git a/x86/unwind_vdso.c b/x86/unwind_vdso.c
new file mode 100644
index 0000000..00a26a8
--- /dev/null
+++ b/x86/unwind_vdso.c
@@ -0,0 +1,211 @@
+/*
+ * unwind_vdso.c - tests unwind info for AT_SYSINFO in the vDSO
+ * Copyright (c) 2014-2015 Andrew Lutomirski
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms and conditions of the GNU General Public License,
+ * version 2, as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope it will be useful, but
+ * WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * This tests __kernel_vsyscall's unwind info.
+ */
+
+#define _GNU_SOURCE
+
+#include <features.h>
+#include <stdio.h>
+
+#if defined(__GLIBC__) && __GLIBC__ == 2 && __GLIBC_MINOR__ < 16
+
+int main()
+{
+	/* We need getauxval(). */
+	printf("[SKIP]\tGLIBC before 2.16 cannot compile this test\n");
+	return 0;
+}
+
+#else
+
+#include <sys/time.h>
+#include <stdlib.h>
+#include <syscall.h>
+#include <unistd.h>
+#include <string.h>
+#include <inttypes.h>
+#include <sys/mman.h>
+#include <signal.h>
+#include <sys/ucontext.h>
+#include <err.h>
+#include <stddef.h>
+#include <stdbool.h>
+#include <sys/ptrace.h>
+#include <sys/user.h>
+#include <sys/ucontext.h>
+#include <link.h>
+#include <sys/auxv.h>
+#include <dlfcn.h>
+#include <unwind.h>
+
+static void sethandler(int sig, void (*handler)(int, siginfo_t *, void *),
+		       int flags)
+{
+	struct sigaction sa;
+	memset(&sa, 0, sizeof(sa));
+	sa.sa_sigaction = handler;
+	sa.sa_flags = SA_SIGINFO | flags;
+	sigemptyset(&sa.sa_mask);
+	if (sigaction(sig, &sa, 0))
+		err(1, "sigaction");
+}
+
+#ifdef __x86_64__
+# define WIDTH "q"
+#else
+# define WIDTH "l"
+#endif
+
+static unsigned long get_eflags(void)
+{
+	unsigned long eflags;
+	asm volatile ("pushf" WIDTH "\n\tpop" WIDTH " %0" : "=rm" (eflags));
+	return eflags;
+}
+
+static void set_eflags(unsigned long eflags)
+{
+	asm volatile ("push" WIDTH " %0\n\tpopf" WIDTH
+		      : : "rm" (eflags) : "flags");
+}
+
+#define X86_EFLAGS_TF (1UL << 8)
+
+static volatile sig_atomic_t nerrs;
+static unsigned long sysinfo;
+static bool got_sysinfo = false;
+static unsigned long return_address;
+
+struct unwind_state {
+	unsigned long ip;	/* trap source */
+	int depth;		/* -1 until we hit the trap source */
+};
+
+_Unwind_Reason_Code trace_fn(struct _Unwind_Context * ctx, void *opaque)
+{
+	struct unwind_state *state = opaque;
+	unsigned long ip = _Unwind_GetIP(ctx);
+
+	if (state->depth == -1) {
+		if (ip == state->ip)
+			state->depth = 0;
+		else
+			return _URC_NO_REASON;	/* Not there yet */
+	}
+	printf("\t  0x%lx\n", ip);
+
+	if (ip == return_address) {
+		/* Here we are. */
+		unsigned long eax = _Unwind_GetGR(ctx, 0);
+		unsigned long ecx = _Unwind_GetGR(ctx, 1);
+		unsigned long edx = _Unwind_GetGR(ctx, 2);
+		unsigned long ebx = _Unwind_GetGR(ctx, 3);
+		unsigned long ebp = _Unwind_GetGR(ctx, 5);
+		unsigned long esi = _Unwind_GetGR(ctx, 6);
+		unsigned long edi = _Unwind_GetGR(ctx, 7);
+		bool ok = (eax == SYS_getpid || eax == getpid()) &&
+			ebx == 1 && ecx == 2 && edx == 3 &&
+			esi == 4 && edi == 5 && ebp == 6;
+
+		if (!ok)
+			nerrs++;
+		printf("[%s]\t  NR = %ld, args = %ld, %ld, %ld, %ld, %ld, %ld\n",
+		       (ok ? "OK" : "FAIL"),
+		       eax, ebx, ecx, edx, esi, edi, ebp);
+
+		return _URC_NORMAL_STOP;
+	} else {
+		state->depth++;
+		return _URC_NO_REASON;
+	}
+}
+
+static void sigtrap(int sig, siginfo_t *info, void *ctx_void)
+{
+	ucontext_t *ctx = (ucontext_t *)ctx_void;
+	struct unwind_state state;
+	unsigned long ip = ctx->uc_mcontext.gregs[REG_EIP];
+
+	if (!got_sysinfo && ip == sysinfo) {
+		got_sysinfo = true;
+
+		/* Find the return address. */
+		return_address = *(unsigned long *)(unsigned long)ctx->uc_mcontext.gregs[REG_ESP];
+
+		printf("\tIn vsyscall at 0x%lx, returning to 0x%lx\n",
+		       ip, return_address);
+	}
+
+	if (!got_sysinfo)
+		return;		/* Not there yet */
+
+	if (ip == return_address) {
+		ctx->uc_mcontext.gregs[REG_EFL] &= ~X86_EFLAGS_TF;
+		printf("\tVsyscall is done\n");
+		return;
+	}
+
+	printf("\tSIGTRAP at 0x%lx\n", ip);
+
+	state.ip = ip;
+	state.depth = -1;
+	_Unwind_Backtrace(trace_fn, &state);
+}
+
+int main()
+{
+	sysinfo = getauxval(AT_SYSINFO);
+	printf("\tAT_SYSINFO is 0x%lx\n", sysinfo);
+
+	Dl_info info;
+	if (!dladdr((void *)sysinfo, &info)) {
+		printf("[WARN]\tdladdr failed on AT_SYSINFO\n");
+	} else {
+		printf("[OK]\tAT_SYSINFO maps to %s, loaded at 0x%p\n",
+		       info.dli_fname, info.dli_fbase);
+	}
+
+	sethandler(SIGTRAP, sigtrap, 0);
+
+	syscall(SYS_getpid);  /* Force symbol binding without TF set. */
+	printf("[RUN]\tSet TF and check a fast syscall\n");
+	set_eflags(get_eflags() | X86_EFLAGS_TF);
+	syscall(SYS_getpid, 1, 2, 3, 4, 5, 6);
+	if (!got_sysinfo) {
+		set_eflags(get_eflags() & ~X86_EFLAGS_TF);
+
+		/*
+		 * The most likely cause of this is that you're on Debian or
+		 * a Debian-based distro, you're missing libc6-i686, and you're
+		 * affected by libc/19006 (https://sourceware.org/PR19006).
+		 */
+		printf("[WARN]\tsyscall(2) didn't enter AT_SYSINFO\n");
+	}
+
+	if (get_eflags() & X86_EFLAGS_TF) {
+		printf("[FAIL]\tTF is still set\n");
+		nerrs++;
+	}
+
+	if (nerrs) {
+		printf("[FAIL]\tThere were errors\n");
+		return 1;
+	} else {
+		printf("[OK]\tAll is well\n");
+		return 0;
+	}
+}
+
+#endif	/* New enough libc */