aboutsummaryrefslogtreecommitdiffstats
path: root/arch/x86
diff options
context:
space:
mode:
authorLinus Torvalds <torvalds@linux-foundation.org>2011-11-02 12:45:39 -0400
committerLinus Torvalds <torvalds@linux-foundation.org>2011-11-02 12:45:39 -0400
commitde0a5345a55b8dd5a4695181275df0e691176830 (patch)
tree17530e824f7f46ce0b1757657179fb5957a6add5 /arch/x86
parent994c0e992522c123298b4a91b72f5e67ba2d1123 (diff)
parent8535639810e578960233ad39def3ac2157b0c3ec (diff)
Merge branch 'for-linus' of git://github.com/richardweinberger/linux
* 'for-linus' of git://github.com/richardweinberger/linux: (90 commits) um: fix ubd cow size um: Fix kmalloc argument order in um/vdso/vma.c um: switch to use of drivers/Kconfig UserModeLinux-HOWTO.txt: fix a typo UserModeLinux-HOWTO.txt: remove ^H characters um: we need sys/user.h only on i386 um: merge delay_{32,64}.c um: distribute exports to where exported stuff is defined um: kill system-um.h um: generic ftrace.h will do... um: segment.h is x86-only and needed only there um: asm/pda.h is not needed anymore um: hw_irq.h can go generic as well um: switch to generic-y um: clean Kconfig up a bit um: a couple of missing dependencies... um: kill useless argument of free_chan() and free_one_chan() um: unify ptrace_user.h um: unify KSTK_... um: fix gcov build breakage ...
Diffstat (limited to 'arch/x86')
-rw-r--r--arch/x86/Makefile.um61
-rw-r--r--arch/x86/um/Kconfig70
-rw-r--r--arch/x86/um/Makefile45
-rw-r--r--arch/x86/um/asm/apic.h4
-rw-r--r--arch/x86/um/asm/arch_hweight.h6
-rw-r--r--arch/x86/um/asm/archparam.h20
-rw-r--r--arch/x86/um/asm/checksum.h10
-rw-r--r--arch/x86/um/asm/checksum_32.h201
-rw-r--r--arch/x86/um/asm/checksum_64.h144
-rw-r--r--arch/x86/um/asm/desc.h16
-rw-r--r--arch/x86/um/asm/elf.h221
-rw-r--r--arch/x86/um/asm/irq_vectors.h10
-rw-r--r--arch/x86/um/asm/mm_context.h72
-rw-r--r--arch/x86/um/asm/module.h23
-rw-r--r--arch/x86/um/asm/processor.h22
-rw-r--r--arch/x86/um/asm/processor_32.h66
-rw-r--r--arch/x86/um/asm/processor_64.h45
-rw-r--r--arch/x86/um/asm/ptrace.h5
-rw-r--r--arch/x86/um/asm/ptrace_32.h51
-rw-r--r--arch/x86/um/asm/ptrace_64.h72
-rw-r--r--arch/x86/um/asm/required-features.h9
-rw-r--r--arch/x86/um/asm/segment.h10
-rw-r--r--arch/x86/um/asm/system.h135
-rw-r--r--arch/x86/um/asm/vm-flags.h25
-rw-r--r--arch/x86/um/bug.c21
-rw-r--r--arch/x86/um/bugs_32.c76
-rw-r--r--arch/x86/um/bugs_64.c15
-rw-r--r--arch/x86/um/checksum_32.S458
-rw-r--r--arch/x86/um/delay.c60
-rw-r--r--arch/x86/um/elfcore.c83
-rw-r--r--arch/x86/um/fault.c28
-rw-r--r--arch/x86/um/ksyms.c13
-rw-r--r--arch/x86/um/ldt.c502
-rw-r--r--arch/x86/um/mem_32.c62
-rw-r--r--arch/x86/um/mem_64.c26
-rw-r--r--arch/x86/um/os-Linux/Makefile13
-rw-r--r--arch/x86/um/os-Linux/mcontext.c31
-rw-r--r--arch/x86/um/os-Linux/prctl.c12
-rw-r--r--arch/x86/um/os-Linux/registers.c113
-rw-r--r--arch/x86/um/os-Linux/task_size.c150
-rw-r--r--arch/x86/um/os-Linux/tls.c67
-rw-r--r--arch/x86/um/ptrace_32.c273
-rw-r--r--arch/x86/um/ptrace_64.c271
-rw-r--r--arch/x86/um/ptrace_user.c21
-rw-r--r--arch/x86/um/setjmp_32.S58
-rw-r--r--arch/x86/um/setjmp_64.S54
-rw-r--r--arch/x86/um/shared/sysdep/archsetjmp.h5
-rw-r--r--arch/x86/um/shared/sysdep/archsetjmp_32.h22
-rw-r--r--arch/x86/um/shared/sysdep/archsetjmp_64.h24
-rw-r--r--arch/x86/um/shared/sysdep/faultinfo.h5
-rw-r--r--arch/x86/um/shared/sysdep/faultinfo_32.h35
-rw-r--r--arch/x86/um/shared/sysdep/faultinfo_64.h35
-rw-r--r--arch/x86/um/shared/sysdep/kernel-offsets.h21
-rw-r--r--arch/x86/um/shared/sysdep/mcontext.h31
-rw-r--r--arch/x86/um/shared/sysdep/ptrace.h5
-rw-r--r--arch/x86/um/shared/sysdep/ptrace_32.h114
-rw-r--r--arch/x86/um/shared/sysdep/ptrace_64.h157
-rw-r--r--arch/x86/um/shared/sysdep/ptrace_user.h27
-rw-r--r--arch/x86/um/shared/sysdep/skas_ptrace.h22
-rw-r--r--arch/x86/um/shared/sysdep/stub.h14
-rw-r--r--arch/x86/um/shared/sysdep/stub_32.h93
-rw-r--r--arch/x86/um/shared/sysdep/stub_64.h99
-rw-r--r--arch/x86/um/shared/sysdep/syscalls.h5
-rw-r--r--arch/x86/um/shared/sysdep/syscalls_32.h20
-rw-r--r--arch/x86/um/shared/sysdep/syscalls_64.h32
-rw-r--r--arch/x86/um/shared/sysdep/tls.h39
-rw-r--r--arch/x86/um/signal.c624
-rw-r--r--arch/x86/um/stub_32.S51
-rw-r--r--arch/x86/um/stub_64.S66
-rw-r--r--arch/x86/um/stub_segv.c19
-rw-r--r--arch/x86/um/sys_call_table_32.S26
-rw-r--r--arch/x86/um/sys_call_table_64.c64
-rw-r--r--arch/x86/um/syscalls_32.c66
-rw-r--r--arch/x86/um/syscalls_64.c102
-rw-r--r--arch/x86/um/sysrq_32.c101
-rw-r--r--arch/x86/um/sysrq_64.c41
-rw-r--r--arch/x86/um/tls_32.c396
-rw-r--r--arch/x86/um/tls_64.c17
-rw-r--r--arch/x86/um/user-offsets.c80
-rw-r--r--arch/x86/um/vdso/Makefile90
-rw-r--r--arch/x86/um/vdso/checkundef.sh10
-rw-r--r--arch/x86/um/vdso/um_vdso.c71
-rw-r--r--arch/x86/um/vdso/vdso-layout.lds.S64
-rw-r--r--arch/x86/um/vdso/vdso-note.S12
-rw-r--r--arch/x86/um/vdso/vdso.S10
-rw-r--r--arch/x86/um/vdso/vdso.lds.S32
-rw-r--r--arch/x86/um/vdso/vma.c74
87 files changed, 6571 insertions, 0 deletions
diff --git a/arch/x86/Makefile.um b/arch/x86/Makefile.um
new file mode 100644
index 000000000000..36ddec6a41c9
--- /dev/null
+++ b/arch/x86/Makefile.um
@@ -0,0 +1,61 @@
1core-y += arch/x86/crypto/
2
3ifeq ($(CONFIG_X86_32),y)
4START := 0x8048000
5
6LDFLAGS += -m elf_i386
7ELF_ARCH := i386
8ELF_FORMAT := elf32-i386
9CHECKFLAGS += -D__i386__
10
11ifeq ("$(origin SUBARCH)", "command line")
12ifneq ("$(shell uname -m | sed -e s/i.86/i386/)", "$(SUBARCH)")
13KBUILD_CFLAGS += $(call cc-option,-m32)
14KBUILD_AFLAGS += $(call cc-option,-m32)
15LINK-y += $(call cc-option,-m32)
16
17export LDFLAGS
18endif
19endif
20
21# First of all, tune CFLAGS for the specific CPU. This actually sets cflags-y.
22include $(srctree)/arch/x86/Makefile_32.cpu
23
24# prevent gcc from keeping the stack 16 byte aligned. Taken from i386.
25cflags-y += $(call cc-option,-mpreferred-stack-boundary=2)
26
27# Prevent sprintf in nfsd from being converted to strcpy and resulting in
28# an unresolved reference.
29cflags-y += -ffreestanding
30
31# Disable unit-at-a-time mode on pre-gcc-4.0 compilers, it makes gcc use
32# a lot more stack due to the lack of sharing of stacklots. Also, gcc
33# 4.3.0 needs -funit-at-a-time for extern inline functions.
34KBUILD_CFLAGS += $(shell if [ $(call cc-version) -lt 0400 ] ; then \
35 echo $(call cc-option,-fno-unit-at-a-time); \
36 else echo $(call cc-option,-funit-at-a-time); fi ;)
37
38KBUILD_CFLAGS += $(cflags-y)
39
40else
41
42START := 0x60000000
43
44KBUILD_CFLAGS += -fno-builtin -m64
45
46CHECKFLAGS += -m64 -D__x86_64__
47KBUILD_AFLAGS += -m64
48LDFLAGS += -m elf_x86_64
49KBUILD_CPPFLAGS += -m64
50
51ELF_ARCH := i386:x86-64
52ELF_FORMAT := elf64-x86-64
53
54# Not on all 64-bit distros /lib is a symlink to /lib64. PLD is an example.
55
56LINK-$(CONFIG_LD_SCRIPT_DYN) += -Wl,-rpath,/lib64
57LINK-y += -m64
58
59# Do unit-at-a-time unconditionally on x86_64, following the host
60KBUILD_CFLAGS += $(call cc-option,-funit-at-a-time)
61endif
diff --git a/arch/x86/um/Kconfig b/arch/x86/um/Kconfig
new file mode 100644
index 000000000000..21bebe63df66
--- /dev/null
+++ b/arch/x86/um/Kconfig
@@ -0,0 +1,70 @@
1mainmenu "User Mode Linux/$SUBARCH $KERNELVERSION Kernel Configuration"
2
3source "arch/um/Kconfig.common"
4
5menu "UML-specific options"
6
7menu "Host processor type and features"
8
9config CMPXCHG_LOCAL
10 bool
11 default n
12
13config CMPXCHG_DOUBLE
14 bool
15 default n
16
17source "arch/x86/Kconfig.cpu"
18
19endmenu
20
21config UML_X86
22 def_bool y
23 select GENERIC_FIND_FIRST_BIT
24
25config 64BIT
26 bool
27 default SUBARCH = "x86_64"
28
29config X86_32
30 def_bool !64BIT
31 select HAVE_AOUT
32
33config X86_64
34 def_bool 64BIT
35
36config RWSEM_XCHGADD_ALGORITHM
37 def_bool X86_XADD && 64BIT
38
39config RWSEM_GENERIC_SPINLOCK
40 def_bool !RWSEM_XCHGADD_ALGORITHM
41
42config 3_LEVEL_PGTABLES
43 bool "Three-level pagetables (EXPERIMENTAL)" if !64BIT
44 default 64BIT
45 depends on EXPERIMENTAL
46 help
47 Three-level pagetables will let UML have more than 4G of physical
48 memory. All the memory that can't be mapped directly will be treated
49 as high memory.
50
51 However, this it experimental on 32-bit architectures, so if unsure say
52 N (on x86-64 it's automatically enabled, instead, as it's safe there).
53
54config ARCH_HAS_SC_SIGNALS
55 def_bool !64BIT
56
57config ARCH_REUSE_HOST_VSYSCALL_AREA
58 def_bool !64BIT
59
60config SMP_BROKEN
61 def_bool 64BIT
62
63config GENERIC_HWEIGHT
64 def_bool y
65
66source "arch/um/Kconfig.um"
67
68endmenu
69
70source "arch/um/Kconfig.rest"
diff --git a/arch/x86/um/Makefile b/arch/x86/um/Makefile
new file mode 100644
index 000000000000..8fb58400e415
--- /dev/null
+++ b/arch/x86/um/Makefile
@@ -0,0 +1,45 @@
1#
2# Copyright (C) 2002 - 2007 Jeff Dike (jdike@{addtoit,linux.intel}.com)
3#
4
5ifeq ($(CONFIG_X86_32),y)
6 BITS := 32
7else
8 BITS := 64
9endif
10
11obj-y = bug.o bugs_$(BITS).o delay.o fault.o ksyms.o ldt.o \
12 ptrace_$(BITS).o ptrace_user.o setjmp_$(BITS).o signal.o \
13 stub_$(BITS).o stub_segv.o syscalls_$(BITS).o \
14 sys_call_table_$(BITS).o sysrq_$(BITS).o tls_$(BITS).o \
15 mem_$(BITS).o subarch.o os-$(OS)/
16
17ifeq ($(CONFIG_X86_32),y)
18
19obj-y += checksum_32.o
20obj-$(CONFIG_BINFMT_ELF) += elfcore.o
21
22subarch-y = ../lib/string_32.o ../lib/atomic64_32.o ../lib/atomic64_cx8_32.o
23subarch-$(CONFIG_RWSEM_XCHGADD_ALGORITHM) += ../lib/rwsem.o
24subarch-$(CONFIG_HIGHMEM) += ../mm/highmem_32.o
25
26else
27
28obj-y += vdso/
29
30subarch-y = ../lib/csum-partial_64.o ../lib/memcpy_64.o ../lib/thunk_64.o \
31 ../lib/rwsem.o
32
33endif
34
35subarch-$(CONFIG_MODULES) += ../kernel/module.o
36
37USER_OBJS := bugs_$(BITS).o ptrace_user.o fault.o
38
39extra-y += user-offsets.s
40$(obj)/user-offsets.s: c_flags = -Wp,-MD,$(depfile) $(USER_CFLAGS)
41
42UNPROFILE_OBJS := stub_segv.o
43CFLAGS_stub_segv.o := $(CFLAGS_NO_HARDENING)
44
45include arch/um/scripts/Makefile.rules
diff --git a/arch/x86/um/asm/apic.h b/arch/x86/um/asm/apic.h
new file mode 100644
index 000000000000..876dee84ab11
--- /dev/null
+++ b/arch/x86/um/asm/apic.h
@@ -0,0 +1,4 @@
1#ifndef __UM_APIC_H
2#define __UM_APIC_H
3
4#endif
diff --git a/arch/x86/um/asm/arch_hweight.h b/arch/x86/um/asm/arch_hweight.h
new file mode 100644
index 000000000000..c656cf443f4a
--- /dev/null
+++ b/arch/x86/um/asm/arch_hweight.h
@@ -0,0 +1,6 @@
1#ifndef _ASM_UM_HWEIGHT_H
2#define _ASM_UM_HWEIGHT_H
3
4#include <asm-generic/bitops/arch_hweight.h>
5
6#endif
diff --git a/arch/x86/um/asm/archparam.h b/arch/x86/um/asm/archparam.h
new file mode 100644
index 000000000000..c17cf68dda0f
--- /dev/null
+++ b/arch/x86/um/asm/archparam.h
@@ -0,0 +1,20 @@
1/*
2 * Copyright (C) 2000 - 2003 Jeff Dike (jdike@addtoit.com)
3 * Copyright 2003 PathScale, Inc.
4 * Licensed under the GPL
5 */
6
7#ifndef __UM_ARCHPARAM_H
8#define __UM_ARCHPARAM_H
9
10#ifdef CONFIG_X86_32
11
12#ifdef CONFIG_X86_PAE
13#define LAST_PKMAP 512
14#else
15#define LAST_PKMAP 1024
16#endif
17
18#endif
19
20#endif
diff --git a/arch/x86/um/asm/checksum.h b/arch/x86/um/asm/checksum.h
new file mode 100644
index 000000000000..b6efe2381b5d
--- /dev/null
+++ b/arch/x86/um/asm/checksum.h
@@ -0,0 +1,10 @@
1#ifndef __UM_CHECKSUM_H
2#define __UM_CHECKSUM_H
3
4#ifdef CONFIG_X86_32
5# include "checksum_32.h"
6#else
7# include "checksum_64.h"
8#endif
9
10#endif
diff --git a/arch/x86/um/asm/checksum_32.h b/arch/x86/um/asm/checksum_32.h
new file mode 100644
index 000000000000..caab74252e27
--- /dev/null
+++ b/arch/x86/um/asm/checksum_32.h
@@ -0,0 +1,201 @@
1/*
2 * Licensed under the GPL
3 */
4
5#ifndef __UM_SYSDEP_CHECKSUM_H
6#define __UM_SYSDEP_CHECKSUM_H
7
8#include "linux/in6.h"
9#include "linux/string.h"
10
11/*
12 * computes the checksum of a memory block at buff, length len,
13 * and adds in "sum" (32-bit)
14 *
15 * returns a 32-bit number suitable for feeding into itself
16 * or csum_tcpudp_magic
17 *
18 * this function must be called with even lengths, except
19 * for the last fragment, which may be odd
20 *
21 * it's best to have buff aligned on a 32-bit boundary
22 */
23__wsum csum_partial(const void *buff, int len, __wsum sum);
24
25/*
26 * Note: when you get a NULL pointer exception here this means someone
27 * passed in an incorrect kernel address to one of these functions.
28 *
29 * If you use these functions directly please don't forget the
30 * access_ok().
31 */
32
33static __inline__
34__wsum csum_partial_copy_nocheck(const void *src, void *dst,
35 int len, __wsum sum)
36{
37 memcpy(dst, src, len);
38 return csum_partial(dst, len, sum);
39}
40
41/*
42 * the same as csum_partial, but copies from src while it
43 * checksums, and handles user-space pointer exceptions correctly, when needed.
44 *
45 * here even more important to align src and dst on a 32-bit (or even
46 * better 64-bit) boundary
47 */
48
49static __inline__
50__wsum csum_partial_copy_from_user(const void __user *src, void *dst,
51 int len, __wsum sum, int *err_ptr)
52{
53 if (copy_from_user(dst, src, len)) {
54 *err_ptr = -EFAULT;
55 return (__force __wsum)-1;
56 }
57
58 return csum_partial(dst, len, sum);
59}
60
61/*
62 * This is a version of ip_compute_csum() optimized for IP headers,
63 * which always checksum on 4 octet boundaries.
64 *
65 * By Jorge Cwik <jorge@laser.satlink.net>, adapted for linux by
66 * Arnt Gulbrandsen.
67 */
68static inline __sum16 ip_fast_csum(const void *iph, unsigned int ihl)
69{
70 unsigned int sum;
71
72 __asm__ __volatile__(
73 "movl (%1), %0 ;\n"
74 "subl $4, %2 ;\n"
75 "jbe 2f ;\n"
76 "addl 4(%1), %0 ;\n"
77 "adcl 8(%1), %0 ;\n"
78 "adcl 12(%1), %0 ;\n"
79"1: adcl 16(%1), %0 ;\n"
80 "lea 4(%1), %1 ;\n"
81 "decl %2 ;\n"
82 "jne 1b ;\n"
83 "adcl $0, %0 ;\n"
84 "movl %0, %2 ;\n"
85 "shrl $16, %0 ;\n"
86 "addw %w2, %w0 ;\n"
87 "adcl $0, %0 ;\n"
88 "notl %0 ;\n"
89"2: ;\n"
90 /* Since the input registers which are loaded with iph and ipl
91 are modified, we must also specify them as outputs, or gcc
92 will assume they contain their original values. */
93 : "=r" (sum), "=r" (iph), "=r" (ihl)
94 : "1" (iph), "2" (ihl)
95 : "memory");
96 return (__force __sum16)sum;
97}
98
99/*
100 * Fold a partial checksum
101 */
102
103static inline __sum16 csum_fold(__wsum sum)
104{
105 __asm__(
106 "addl %1, %0 ;\n"
107 "adcl $0xffff, %0 ;\n"
108 : "=r" (sum)
109 : "r" ((__force u32)sum << 16),
110 "0" ((__force u32)sum & 0xffff0000)
111 );
112 return (__force __sum16)(~(__force u32)sum >> 16);
113}
114
115static inline __wsum csum_tcpudp_nofold(__be32 saddr, __be32 daddr,
116 unsigned short len,
117 unsigned short proto,
118 __wsum sum)
119{
120 __asm__(
121 "addl %1, %0 ;\n"
122 "adcl %2, %0 ;\n"
123 "adcl %3, %0 ;\n"
124 "adcl $0, %0 ;\n"
125 : "=r" (sum)
126 : "g" (daddr), "g"(saddr), "g"((len + proto) << 8), "0"(sum));
127 return sum;
128}
129
130/*
131 * computes the checksum of the TCP/UDP pseudo-header
132 * returns a 16-bit checksum, already complemented
133 */
134static inline __sum16 csum_tcpudp_magic(__be32 saddr, __be32 daddr,
135 unsigned short len,
136 unsigned short proto,
137 __wsum sum)
138{
139 return csum_fold(csum_tcpudp_nofold(saddr,daddr,len,proto,sum));
140}
141
142/*
143 * this routine is used for miscellaneous IP-like checksums, mainly
144 * in icmp.c
145 */
146
147static inline __sum16 ip_compute_csum(const void *buff, int len)
148{
149 return csum_fold (csum_partial(buff, len, 0));
150}
151
152#define _HAVE_ARCH_IPV6_CSUM
153static __inline__ __sum16 csum_ipv6_magic(const struct in6_addr *saddr,
154 const struct in6_addr *daddr,
155 __u32 len, unsigned short proto,
156 __wsum sum)
157{
158 __asm__(
159 "addl 0(%1), %0 ;\n"
160 "adcl 4(%1), %0 ;\n"
161 "adcl 8(%1), %0 ;\n"
162 "adcl 12(%1), %0 ;\n"
163 "adcl 0(%2), %0 ;\n"
164 "adcl 4(%2), %0 ;\n"
165 "adcl 8(%2), %0 ;\n"
166 "adcl 12(%2), %0 ;\n"
167 "adcl %3, %0 ;\n"
168 "adcl %4, %0 ;\n"
169 "adcl $0, %0 ;\n"
170 : "=&r" (sum)
171 : "r" (saddr), "r" (daddr),
172 "r"(htonl(len)), "r"(htonl(proto)), "0"(sum));
173
174 return csum_fold(sum);
175}
176
177/*
178 * Copy and checksum to user
179 */
180#define HAVE_CSUM_COPY_USER
181static __inline__ __wsum csum_and_copy_to_user(const void *src,
182 void __user *dst,
183 int len, __wsum sum, int *err_ptr)
184{
185 if (access_ok(VERIFY_WRITE, dst, len)) {
186 if (copy_to_user(dst, src, len)) {
187 *err_ptr = -EFAULT;
188 return (__force __wsum)-1;
189 }
190
191 return csum_partial(src, len, sum);
192 }
193
194 if (len)
195 *err_ptr = -EFAULT;
196
197 return (__force __wsum)-1; /* invalid checksum */
198}
199
200#endif
201
diff --git a/arch/x86/um/asm/checksum_64.h b/arch/x86/um/asm/checksum_64.h
new file mode 100644
index 000000000000..a5be9031ea85
--- /dev/null
+++ b/arch/x86/um/asm/checksum_64.h
@@ -0,0 +1,144 @@
1/*
2 * Licensed under the GPL
3 */
4
5#ifndef __UM_SYSDEP_CHECKSUM_H
6#define __UM_SYSDEP_CHECKSUM_H
7
8#include "linux/string.h"
9#include "linux/in6.h"
10#include "asm/uaccess.h"
11
12extern __wsum csum_partial(const void *buff, int len, __wsum sum);
13
14/*
15 * Note: when you get a NULL pointer exception here this means someone
16 * passed in an incorrect kernel address to one of these functions.
17 *
18 * If you use these functions directly please don't forget the
19 * access_ok().
20 */
21
22static __inline__
23__wsum csum_partial_copy_nocheck(const void *src, void *dst,
24 int len, __wsum sum)
25{
26 memcpy(dst, src, len);
27 return(csum_partial(dst, len, sum));
28}
29
30static __inline__
31__wsum csum_partial_copy_from_user(const void __user *src,
32 void *dst, int len, __wsum sum,
33 int *err_ptr)
34{
35 if (copy_from_user(dst, src, len)) {
36 *err_ptr = -EFAULT;
37 return (__force __wsum)-1;
38 }
39 return csum_partial(dst, len, sum);
40}
41
42/**
43 * csum_fold - Fold and invert a 32bit checksum.
44 * sum: 32bit unfolded sum
45 *
46 * Fold a 32bit running checksum to 16bit and invert it. This is usually
47 * the last step before putting a checksum into a packet.
48 * Make sure not to mix with 64bit checksums.
49 */
50static inline __sum16 csum_fold(__wsum sum)
51{
52 __asm__(
53 " addl %1,%0\n"
54 " adcl $0xffff,%0"
55 : "=r" (sum)
56 : "r" ((__force u32)sum << 16),
57 "0" ((__force u32)sum & 0xffff0000)
58 );
59 return (__force __sum16)(~(__force u32)sum >> 16);
60}
61
62/**
63 * csum_tcpup_nofold - Compute an IPv4 pseudo header checksum.
64 * @saddr: source address
65 * @daddr: destination address
66 * @len: length of packet
67 * @proto: ip protocol of packet
68 * @sum: initial sum to be added in (32bit unfolded)
69 *
70 * Returns the pseudo header checksum the input data. Result is
71 * 32bit unfolded.
72 */
73static inline __wsum
74csum_tcpudp_nofold(__be32 saddr, __be32 daddr, unsigned short len,
75 unsigned short proto, __wsum sum)
76{
77 asm(" addl %1, %0\n"
78 " adcl %2, %0\n"
79 " adcl %3, %0\n"
80 " adcl $0, %0\n"
81 : "=r" (sum)
82 : "g" (daddr), "g" (saddr), "g" ((len + proto) << 8), "0" (sum));
83 return sum;
84}
85
86/*
87 * computes the checksum of the TCP/UDP pseudo-header
88 * returns a 16-bit checksum, already complemented
89 */
90static inline __sum16 csum_tcpudp_magic(__be32 saddr, __be32 daddr,
91 unsigned short len,
92 unsigned short proto,
93 __wsum sum)
94{
95 return csum_fold(csum_tcpudp_nofold(saddr,daddr,len,proto,sum));
96}
97
98/**
99 * ip_fast_csum - Compute the IPv4 header checksum efficiently.
100 * iph: ipv4 header
101 * ihl: length of header / 4
102 */
103static inline __sum16 ip_fast_csum(const void *iph, unsigned int ihl)
104{
105 unsigned int sum;
106
107 asm( " movl (%1), %0\n"
108 " subl $4, %2\n"
109 " jbe 2f\n"
110 " addl 4(%1), %0\n"
111 " adcl 8(%1), %0\n"
112 " adcl 12(%1), %0\n"
113 "1: adcl 16(%1), %0\n"
114 " lea 4(%1), %1\n"
115 " decl %2\n"
116 " jne 1b\n"
117 " adcl $0, %0\n"
118 " movl %0, %2\n"
119 " shrl $16, %0\n"
120 " addw %w2, %w0\n"
121 " adcl $0, %0\n"
122 " notl %0\n"
123 "2:"
124 /* Since the input registers which are loaded with iph and ipl
125 are modified, we must also specify them as outputs, or gcc
126 will assume they contain their original values. */
127 : "=r" (sum), "=r" (iph), "=r" (ihl)
128 : "1" (iph), "2" (ihl)
129 : "memory");
130 return (__force __sum16)sum;
131}
132
133static inline unsigned add32_with_carry(unsigned a, unsigned b)
134{
135 asm("addl %2,%0\n\t"
136 "adcl $0,%0"
137 : "=r" (a)
138 : "0" (a), "r" (b));
139 return a;
140}
141
142extern __sum16 ip_compute_csum(const void *buff, int len);
143
144#endif
diff --git a/arch/x86/um/asm/desc.h b/arch/x86/um/asm/desc.h
new file mode 100644
index 000000000000..4ec34a51b62c
--- /dev/null
+++ b/arch/x86/um/asm/desc.h
@@ -0,0 +1,16 @@
1#ifndef __UM_DESC_H
2#define __UM_DESC_H
3
4/* Taken from asm-i386/desc.h, it's the only thing we need. The rest wouldn't
5 * compile, and has never been used. */
6#define LDT_empty(info) (\
7 (info)->base_addr == 0 && \
8 (info)->limit == 0 && \
9 (info)->contents == 0 && \
10 (info)->read_exec_only == 1 && \
11 (info)->seg_32bit == 0 && \
12 (info)->limit_in_pages == 0 && \
13 (info)->seg_not_present == 1 && \
14 (info)->useable == 0 )
15
16#endif
diff --git a/arch/x86/um/asm/elf.h b/arch/x86/um/asm/elf.h
new file mode 100644
index 000000000000..f3b0633b69a1
--- /dev/null
+++ b/arch/x86/um/asm/elf.h
@@ -0,0 +1,221 @@
1/*
2 * Copyright (C) 2000 - 2007 Jeff Dike (jdike@{addtoit,linux.intel}.com)
3 * Licensed under the GPL
4 */
5#ifndef __UM_ELF_X86_H
6#define __UM_ELF_X86_H
7
8#include <asm/user.h>
9#include "skas.h"
10
11#ifdef CONFIG_X86_32
12
13#define R_386_NONE 0
14#define R_386_32 1
15#define R_386_PC32 2
16#define R_386_GOT32 3
17#define R_386_PLT32 4
18#define R_386_COPY 5
19#define R_386_GLOB_DAT 6
20#define R_386_JMP_SLOT 7
21#define R_386_RELATIVE 8
22#define R_386_GOTOFF 9
23#define R_386_GOTPC 10
24#define R_386_NUM 11
25
26/*
27 * This is used to ensure we don't load something for the wrong architecture.
28 */
29#define elf_check_arch(x) \
30 (((x)->e_machine == EM_386) || ((x)->e_machine == EM_486))
31
32#define ELF_CLASS ELFCLASS32
33#define ELF_DATA ELFDATA2LSB
34#define ELF_ARCH EM_386
35
36#define ELF_PLAT_INIT(regs, load_addr) do { \
37 PT_REGS_EBX(regs) = 0; \
38 PT_REGS_ECX(regs) = 0; \
39 PT_REGS_EDX(regs) = 0; \
40 PT_REGS_ESI(regs) = 0; \
41 PT_REGS_EDI(regs) = 0; \
42 PT_REGS_EBP(regs) = 0; \
43 PT_REGS_EAX(regs) = 0; \
44} while (0)
45
46/* Shamelessly stolen from include/asm-i386/elf.h */
47
48#define ELF_CORE_COPY_REGS(pr_reg, regs) do { \
49 pr_reg[0] = PT_REGS_EBX(regs); \
50 pr_reg[1] = PT_REGS_ECX(regs); \
51 pr_reg[2] = PT_REGS_EDX(regs); \
52 pr_reg[3] = PT_REGS_ESI(regs); \
53 pr_reg[4] = PT_REGS_EDI(regs); \
54 pr_reg[5] = PT_REGS_EBP(regs); \
55 pr_reg[6] = PT_REGS_EAX(regs); \
56 pr_reg[7] = PT_REGS_DS(regs); \
57 pr_reg[8] = PT_REGS_ES(regs); \
58 /* fake once used fs and gs selectors? */ \
59 pr_reg[9] = PT_REGS_DS(regs); \
60 pr_reg[10] = PT_REGS_DS(regs); \
61 pr_reg[11] = PT_REGS_SYSCALL_NR(regs); \
62 pr_reg[12] = PT_REGS_IP(regs); \
63 pr_reg[13] = PT_REGS_CS(regs); \
64 pr_reg[14] = PT_REGS_EFLAGS(regs); \
65 pr_reg[15] = PT_REGS_SP(regs); \
66 pr_reg[16] = PT_REGS_SS(regs); \
67} while (0);
68
69extern char * elf_aux_platform;
70#define ELF_PLATFORM (elf_aux_platform)
71
72extern unsigned long vsyscall_ehdr;
73extern unsigned long vsyscall_end;
74extern unsigned long __kernel_vsyscall;
75
76/*
77 * This is the range that is readable by user mode, and things
78 * acting like user mode such as get_user_pages.
79 */
80#define FIXADDR_USER_START vsyscall_ehdr
81#define FIXADDR_USER_END vsyscall_end
82
83
84/*
85 * Architecture-neutral AT_ values in 0-17, leave some room
86 * for more of them, start the x86-specific ones at 32.
87 */
88#define AT_SYSINFO 32
89#define AT_SYSINFO_EHDR 33
90
91#define ARCH_DLINFO \
92do { \
93 if ( vsyscall_ehdr ) { \
94 NEW_AUX_ENT(AT_SYSINFO, __kernel_vsyscall); \
95 NEW_AUX_ENT(AT_SYSINFO_EHDR, vsyscall_ehdr); \
96 } \
97} while (0)
98
99#else
100
101/* x86-64 relocation types, taken from asm-x86_64/elf.h */
102#define R_X86_64_NONE 0 /* No reloc */
103#define R_X86_64_64 1 /* Direct 64 bit */
104#define R_X86_64_PC32 2 /* PC relative 32 bit signed */
105#define R_X86_64_GOT32 3 /* 32 bit GOT entry */
106#define R_X86_64_PLT32 4 /* 32 bit PLT address */
107#define R_X86_64_COPY 5 /* Copy symbol at runtime */
108#define R_X86_64_GLOB_DAT 6 /* Create GOT entry */
109#define R_X86_64_JUMP_SLOT 7 /* Create PLT entry */
110#define R_X86_64_RELATIVE 8 /* Adjust by program base */
111#define R_X86_64_GOTPCREL 9 /* 32 bit signed pc relative
112 offset to GOT */
113#define R_X86_64_32 10 /* Direct 32 bit zero extended */
114#define R_X86_64_32S 11 /* Direct 32 bit sign extended */
115#define R_X86_64_16 12 /* Direct 16 bit zero extended */
116#define R_X86_64_PC16 13 /* 16 bit sign extended pc relative */
117#define R_X86_64_8 14 /* Direct 8 bit sign extended */
118#define R_X86_64_PC8 15 /* 8 bit sign extended pc relative */
119
120#define R_X86_64_NUM 16
121
122/*
123 * This is used to ensure we don't load something for the wrong architecture.
124 */
125#define elf_check_arch(x) \
126 ((x)->e_machine == EM_X86_64)
127
128#define ELF_CLASS ELFCLASS64
129#define ELF_DATA ELFDATA2LSB
130#define ELF_ARCH EM_X86_64
131
132#define ELF_PLAT_INIT(regs, load_addr) do { \
133 PT_REGS_RBX(regs) = 0; \
134 PT_REGS_RCX(regs) = 0; \
135 PT_REGS_RDX(regs) = 0; \
136 PT_REGS_RSI(regs) = 0; \
137 PT_REGS_RDI(regs) = 0; \
138 PT_REGS_RBP(regs) = 0; \
139 PT_REGS_RAX(regs) = 0; \
140 PT_REGS_R8(regs) = 0; \
141 PT_REGS_R9(regs) = 0; \
142 PT_REGS_R10(regs) = 0; \
143 PT_REGS_R11(regs) = 0; \
144 PT_REGS_R12(regs) = 0; \
145 PT_REGS_R13(regs) = 0; \
146 PT_REGS_R14(regs) = 0; \
147 PT_REGS_R15(regs) = 0; \
148} while (0)
149
150#define ELF_CORE_COPY_REGS(pr_reg, _regs) \
151 (pr_reg)[0] = (_regs)->regs.gp[0]; \
152 (pr_reg)[1] = (_regs)->regs.gp[1]; \
153 (pr_reg)[2] = (_regs)->regs.gp[2]; \
154 (pr_reg)[3] = (_regs)->regs.gp[3]; \
155 (pr_reg)[4] = (_regs)->regs.gp[4]; \
156 (pr_reg)[5] = (_regs)->regs.gp[5]; \
157 (pr_reg)[6] = (_regs)->regs.gp[6]; \
158 (pr_reg)[7] = (_regs)->regs.gp[7]; \
159 (pr_reg)[8] = (_regs)->regs.gp[8]; \
160 (pr_reg)[9] = (_regs)->regs.gp[9]; \
161 (pr_reg)[10] = (_regs)->regs.gp[10]; \
162 (pr_reg)[11] = (_regs)->regs.gp[11]; \
163 (pr_reg)[12] = (_regs)->regs.gp[12]; \
164 (pr_reg)[13] = (_regs)->regs.gp[13]; \
165 (pr_reg)[14] = (_regs)->regs.gp[14]; \
166 (pr_reg)[15] = (_regs)->regs.gp[15]; \
167 (pr_reg)[16] = (_regs)->regs.gp[16]; \
168 (pr_reg)[17] = (_regs)->regs.gp[17]; \
169 (pr_reg)[18] = (_regs)->regs.gp[18]; \
170 (pr_reg)[19] = (_regs)->regs.gp[19]; \
171 (pr_reg)[20] = (_regs)->regs.gp[20]; \
172 (pr_reg)[21] = current->thread.arch.fs; \
173 (pr_reg)[22] = 0; \
174 (pr_reg)[23] = 0; \
175 (pr_reg)[24] = 0; \
176 (pr_reg)[25] = 0; \
177 (pr_reg)[26] = 0;
178
179#define ELF_PLATFORM "x86_64"
180
181/* No user-accessible fixmap addresses, i.e. vsyscall */
182#define FIXADDR_USER_START 0
183#define FIXADDR_USER_END 0
184
185#define ARCH_HAS_SETUP_ADDITIONAL_PAGES 1
186struct linux_binprm;
187extern int arch_setup_additional_pages(struct linux_binprm *bprm,
188 int uses_interp);
189
190extern unsigned long um_vdso_addr;
191#define AT_SYSINFO_EHDR 33
192#define ARCH_DLINFO NEW_AUX_ENT(AT_SYSINFO_EHDR, um_vdso_addr)
193
194#endif
195
196typedef unsigned long elf_greg_t;
197
198#define ELF_NGREG (sizeof (struct user_regs_struct) / sizeof(elf_greg_t))
199typedef elf_greg_t elf_gregset_t[ELF_NGREG];
200
201typedef struct user_i387_struct elf_fpregset_t;
202
203#define task_pt_regs(t) (&(t)->thread.regs)
204
205struct task_struct;
206
207extern int elf_core_copy_fpregs(struct task_struct *t, elf_fpregset_t *fpu);
208
209#define ELF_CORE_COPY_FPREGS(t, fpu) elf_core_copy_fpregs(t, fpu)
210
211#define ELF_EXEC_PAGESIZE 4096
212
213#define ELF_ET_DYN_BASE (2 * TASK_SIZE / 3)
214
215extern long elf_aux_hwcap;
216#define ELF_HWCAP (elf_aux_hwcap)
217
218#define SET_PERSONALITY(ex) do ; while(0)
219#define __HAVE_ARCH_GATE_AREA 1
220
221#endif
diff --git a/arch/x86/um/asm/irq_vectors.h b/arch/x86/um/asm/irq_vectors.h
new file mode 100644
index 000000000000..272a81e0ce14
--- /dev/null
+++ b/arch/x86/um/asm/irq_vectors.h
@@ -0,0 +1,10 @@
1/*
2 * Copyright (C) 2002 Jeff Dike (jdike@karaya.com)
3 * Licensed under the GPL
4 */
5
6#ifndef __UM_IRQ_VECTORS_H
7#define __UM_IRQ_VECTORS_H
8
9#endif
10
diff --git a/arch/x86/um/asm/mm_context.h b/arch/x86/um/asm/mm_context.h
new file mode 100644
index 000000000000..4a73d63e4760
--- /dev/null
+++ b/arch/x86/um/asm/mm_context.h
@@ -0,0 +1,72 @@
1/*
2 * Copyright (C) 2004 Fujitsu Siemens Computers GmbH
3 * Licensed under the GPL
4 *
5 * Author: Bodo Stroesser <bstroesser@fujitsu-siemens.com>
6 */
7
8#ifndef __ASM_LDT_H
9#define __ASM_LDT_H
10
11#include <linux/mutex.h>
12#include <asm/ldt.h>
13
14extern void ldt_host_info(void);
15
16#define LDT_PAGES_MAX \
17 ((LDT_ENTRIES * LDT_ENTRY_SIZE)/PAGE_SIZE)
18#define LDT_ENTRIES_PER_PAGE \
19 (PAGE_SIZE/LDT_ENTRY_SIZE)
20#define LDT_DIRECT_ENTRIES \
21 ((LDT_PAGES_MAX*sizeof(void *))/LDT_ENTRY_SIZE)
22
23struct ldt_entry {
24 __u32 a;
25 __u32 b;
26};
27
28typedef struct uml_ldt {
29 int entry_count;
30 struct mutex lock;
31 union {
32 struct ldt_entry * pages[LDT_PAGES_MAX];
33 struct ldt_entry entries[LDT_DIRECT_ENTRIES];
34 } u;
35} uml_ldt_t;
36
37#define LDT_entry_a(info) \
38 ((((info)->base_addr & 0x0000ffff) << 16) | ((info)->limit & 0x0ffff))
39
40#define LDT_entry_b(info) \
41 (((info)->base_addr & 0xff000000) | \
42 (((info)->base_addr & 0x00ff0000) >> 16) | \
43 ((info)->limit & 0xf0000) | \
44 (((info)->read_exec_only ^ 1) << 9) | \
45 ((info)->contents << 10) | \
46 (((info)->seg_not_present ^ 1) << 15) | \
47 ((info)->seg_32bit << 22) | \
48 ((info)->limit_in_pages << 23) | \
49 ((info)->useable << 20) | \
50 0x7000)
51
52#define _LDT_empty(info) (\
53 (info)->base_addr == 0 && \
54 (info)->limit == 0 && \
55 (info)->contents == 0 && \
56 (info)->read_exec_only == 1 && \
57 (info)->seg_32bit == 0 && \
58 (info)->limit_in_pages == 0 && \
59 (info)->seg_not_present == 1 && \
60 (info)->useable == 0 )
61
62#ifdef CONFIG_X86_64
63#define LDT_empty(info) (_LDT_empty(info) && ((info)->lm == 0))
64#else
65#define LDT_empty(info) (_LDT_empty(info))
66#endif
67
68struct uml_arch_mm_context {
69 uml_ldt_t ldt;
70};
71
72#endif
diff --git a/arch/x86/um/asm/module.h b/arch/x86/um/asm/module.h
new file mode 100644
index 000000000000..61af80e932eb
--- /dev/null
+++ b/arch/x86/um/asm/module.h
@@ -0,0 +1,23 @@
1#ifndef __UM_MODULE_H
2#define __UM_MODULE_H
3
4/* UML is simple */
5struct mod_arch_specific
6{
7};
8
9#ifdef CONFIG_X86_32
10
11#define Elf_Shdr Elf32_Shdr
12#define Elf_Sym Elf32_Sym
13#define Elf_Ehdr Elf32_Ehdr
14
15#else
16
17#define Elf_Shdr Elf64_Shdr
18#define Elf_Sym Elf64_Sym
19#define Elf_Ehdr Elf64_Ehdr
20
21#endif
22
23#endif
diff --git a/arch/x86/um/asm/processor.h b/arch/x86/um/asm/processor.h
new file mode 100644
index 000000000000..118c143a9cb4
--- /dev/null
+++ b/arch/x86/um/asm/processor.h
@@ -0,0 +1,22 @@
1#ifndef __UM_PROCESSOR_H
2#define __UM_PROCESSOR_H
3
4/* include faultinfo structure */
5#include <sysdep/faultinfo.h>
6
7#ifdef CONFIG_X86_32
8# include "processor_32.h"
9#else
10# include "processor_64.h"
11#endif
12
13#define KSTK_EIP(tsk) KSTK_REG(tsk, HOST_IP)
14#define KSTK_ESP(tsk) KSTK_REG(tsk, HOST_IP)
15#define KSTK_EBP(tsk) KSTK_REG(tsk, HOST_BP)
16
17#define ARCH_IS_STACKGROW(address) \
18 (address + 65536 + 32 * sizeof(unsigned long) >= UPT_SP(&current->thread.regs.regs))
19
20#include <asm/processor-generic.h>
21
22#endif
diff --git a/arch/x86/um/asm/processor_32.h b/arch/x86/um/asm/processor_32.h
new file mode 100644
index 000000000000..018f732704dd
--- /dev/null
+++ b/arch/x86/um/asm/processor_32.h
@@ -0,0 +1,66 @@
1/*
2 * Copyright (C) 2002 Jeff Dike (jdike@karaya.com)
3 * Licensed under the GPL
4 */
5
6#ifndef __UM_PROCESSOR_I386_H
7#define __UM_PROCESSOR_I386_H
8
9#include <linux/string.h>
10#include <asm/segment.h>
11#include <asm/ldt.h>
12
13extern int host_has_cmov;
14
15struct uml_tls_struct {
16 struct user_desc tls;
17 unsigned flushed:1;
18 unsigned present:1;
19};
20
21struct arch_thread {
22 struct uml_tls_struct tls_array[GDT_ENTRY_TLS_ENTRIES];
23 unsigned long debugregs[8];
24 int debugregs_seq;
25 struct faultinfo faultinfo;
26};
27
28#define INIT_ARCH_THREAD { \
29 .tls_array = { [ 0 ... GDT_ENTRY_TLS_ENTRIES - 1 ] = \
30 { .present = 0, .flushed = 0 } }, \
31 .debugregs = { [ 0 ... 7 ] = 0 }, \
32 .debugregs_seq = 0, \
33 .faultinfo = { 0, 0, 0 } \
34}
35
36static inline void arch_flush_thread(struct arch_thread *thread)
37{
38 /* Clear any TLS still hanging */
39 memset(&thread->tls_array, 0, sizeof(thread->tls_array));
40}
41
42static inline void arch_copy_thread(struct arch_thread *from,
43 struct arch_thread *to)
44{
45 memcpy(&to->tls_array, &from->tls_array, sizeof(from->tls_array));
46}
47
48#include <asm/user.h>
49
50/* REP NOP (PAUSE) is a good thing to insert into busy-wait loops. */
51static inline void rep_nop(void)
52{
53 __asm__ __volatile__("rep;nop": : :"memory");
54}
55
56#define cpu_relax() rep_nop()
57
58/*
59 * Default implementation of macro that returns current
60 * instruction pointer ("program counter"). Stolen
61 * from asm-i386/processor.h
62 */
63#define current_text_addr() \
64 ({ void *pc; __asm__("movl $1f,%0\n1:":"=g" (pc)); pc; })
65
66#endif
diff --git a/arch/x86/um/asm/processor_64.h b/arch/x86/um/asm/processor_64.h
new file mode 100644
index 000000000000..61de92d916c3
--- /dev/null
+++ b/arch/x86/um/asm/processor_64.h
@@ -0,0 +1,45 @@
1/*
2 * Copyright 2003 PathScale, Inc.
3 *
4 * Licensed under the GPL
5 */
6
7#ifndef __UM_PROCESSOR_X86_64_H
8#define __UM_PROCESSOR_X86_64_H
9
10struct arch_thread {
11 unsigned long debugregs[8];
12 int debugregs_seq;
13 unsigned long fs;
14 struct faultinfo faultinfo;
15};
16
17/* REP NOP (PAUSE) is a good thing to insert into busy-wait loops. */
18static inline void rep_nop(void)
19{
20 __asm__ __volatile__("rep;nop": : :"memory");
21}
22
23#define cpu_relax() rep_nop()
24
25#define INIT_ARCH_THREAD { .debugregs = { [ 0 ... 7 ] = 0 }, \
26 .debugregs_seq = 0, \
27 .fs = 0, \
28 .faultinfo = { 0, 0, 0 } }
29
30static inline void arch_flush_thread(struct arch_thread *thread)
31{
32}
33
34static inline void arch_copy_thread(struct arch_thread *from,
35 struct arch_thread *to)
36{
37 to->fs = from->fs;
38}
39
40#include <asm/user.h>
41
42#define current_text_addr() \
43 ({ void *pc; __asm__("movq $1f,%0\n1:":"=g" (pc)); pc; })
44
45#endif
diff --git a/arch/x86/um/asm/ptrace.h b/arch/x86/um/asm/ptrace.h
new file mode 100644
index 000000000000..c8aca8c501b0
--- /dev/null
+++ b/arch/x86/um/asm/ptrace.h
@@ -0,0 +1,5 @@
1#ifdef CONFIG_X86_32
2# include "ptrace_32.h"
3#else
4# include "ptrace_64.h"
5#endif
diff --git a/arch/x86/um/asm/ptrace_32.h b/arch/x86/um/asm/ptrace_32.h
new file mode 100644
index 000000000000..5d2a59112537
--- /dev/null
+++ b/arch/x86/um/asm/ptrace_32.h
@@ -0,0 +1,51 @@
1/*
2 * Copyright (C) 2000 - 2007 Jeff Dike (jdike@{addtoit,linux.intel}.com)
3 * Licensed under the GPL
4 */
5
6#ifndef __UM_PTRACE_I386_H
7#define __UM_PTRACE_I386_H
8
9#define HOST_AUDIT_ARCH AUDIT_ARCH_I386
10
11#include "linux/compiler.h"
12#include "asm/ptrace-generic.h"
13
14#define PT_REGS_EAX(r) UPT_EAX(&(r)->regs)
15#define PT_REGS_EBX(r) UPT_EBX(&(r)->regs)
16#define PT_REGS_ECX(r) UPT_ECX(&(r)->regs)
17#define PT_REGS_EDX(r) UPT_EDX(&(r)->regs)
18#define PT_REGS_ESI(r) UPT_ESI(&(r)->regs)
19#define PT_REGS_EDI(r) UPT_EDI(&(r)->regs)
20#define PT_REGS_EBP(r) UPT_EBP(&(r)->regs)
21
22#define PT_REGS_CS(r) UPT_CS(&(r)->regs)
23#define PT_REGS_SS(r) UPT_SS(&(r)->regs)
24#define PT_REGS_DS(r) UPT_DS(&(r)->regs)
25#define PT_REGS_ES(r) UPT_ES(&(r)->regs)
26#define PT_REGS_FS(r) UPT_FS(&(r)->regs)
27#define PT_REGS_GS(r) UPT_GS(&(r)->regs)
28
29#define PT_REGS_EFLAGS(r) UPT_EFLAGS(&(r)->regs)
30
31#define PT_REGS_ORIG_SYSCALL(r) PT_REGS_EAX(r)
32#define PT_REGS_SYSCALL_RET(r) PT_REGS_EAX(r)
33#define PT_FIX_EXEC_STACK(sp) do ; while(0)
34
35#define profile_pc(regs) PT_REGS_IP(regs)
36
37#define user_mode(r) UPT_IS_USER(&(r)->regs)
38
39/*
40 * Forward declaration to avoid including sysdep/tls.h, which causes a
41 * circular include, and compilation failures.
42 */
43struct user_desc;
44
45extern int ptrace_get_thread_area(struct task_struct *child, int idx,
46 struct user_desc __user *user_desc);
47
48extern int ptrace_set_thread_area(struct task_struct *child, int idx,
49 struct user_desc __user *user_desc);
50
51#endif
diff --git a/arch/x86/um/asm/ptrace_64.h b/arch/x86/um/asm/ptrace_64.h
new file mode 100644
index 000000000000..706a0d80545c
--- /dev/null
+++ b/arch/x86/um/asm/ptrace_64.h
@@ -0,0 +1,72 @@
1/*
2 * Copyright 2003 PathScale, Inc.
3 *
4 * Licensed under the GPL
5 */
6
7#ifndef __UM_PTRACE_X86_64_H
8#define __UM_PTRACE_X86_64_H
9
10#include "linux/compiler.h"
11#include "asm/errno.h"
12
13#define __FRAME_OFFSETS /* Needed to get the R* macros */
14#include "asm/ptrace-generic.h"
15
16#define HOST_AUDIT_ARCH AUDIT_ARCH_X86_64
17
18#define PT_REGS_RBX(r) UPT_RBX(&(r)->regs)
19#define PT_REGS_RCX(r) UPT_RCX(&(r)->regs)
20#define PT_REGS_RDX(r) UPT_RDX(&(r)->regs)
21#define PT_REGS_RSI(r) UPT_RSI(&(r)->regs)
22#define PT_REGS_RDI(r) UPT_RDI(&(r)->regs)
23#define PT_REGS_RBP(r) UPT_RBP(&(r)->regs)
24#define PT_REGS_RAX(r) UPT_RAX(&(r)->regs)
25#define PT_REGS_R8(r) UPT_R8(&(r)->regs)
26#define PT_REGS_R9(r) UPT_R9(&(r)->regs)
27#define PT_REGS_R10(r) UPT_R10(&(r)->regs)
28#define PT_REGS_R11(r) UPT_R11(&(r)->regs)
29#define PT_REGS_R12(r) UPT_R12(&(r)->regs)
30#define PT_REGS_R13(r) UPT_R13(&(r)->regs)
31#define PT_REGS_R14(r) UPT_R14(&(r)->regs)
32#define PT_REGS_R15(r) UPT_R15(&(r)->regs)
33
34#define PT_REGS_FS(r) UPT_FS(&(r)->regs)
35#define PT_REGS_GS(r) UPT_GS(&(r)->regs)
36#define PT_REGS_DS(r) UPT_DS(&(r)->regs)
37#define PT_REGS_ES(r) UPT_ES(&(r)->regs)
38#define PT_REGS_SS(r) UPT_SS(&(r)->regs)
39#define PT_REGS_CS(r) UPT_CS(&(r)->regs)
40
41#define PT_REGS_ORIG_RAX(r) UPT_ORIG_RAX(&(r)->regs)
42#define PT_REGS_RIP(r) UPT_IP(&(r)->regs)
43#define PT_REGS_SP(r) UPT_SP(&(r)->regs)
44
45#define PT_REGS_EFLAGS(r) UPT_EFLAGS(&(r)->regs)
46
47/* XXX */
48#define user_mode(r) UPT_IS_USER(&(r)->regs)
49#define PT_REGS_ORIG_SYSCALL(r) PT_REGS_RAX(r)
50#define PT_REGS_SYSCALL_RET(r) PT_REGS_RAX(r)
51
52#define PT_FIX_EXEC_STACK(sp) do ; while(0)
53
54#define profile_pc(regs) PT_REGS_IP(regs)
55
56struct user_desc;
57
58static inline int ptrace_get_thread_area(struct task_struct *child, int idx,
59 struct user_desc __user *user_desc)
60{
61 return -ENOSYS;
62}
63
64static inline int ptrace_set_thread_area(struct task_struct *child, int idx,
65 struct user_desc __user *user_desc)
66{
67 return -ENOSYS;
68}
69
70extern long arch_prctl(struct task_struct *task, int code,
71 unsigned long __user *addr);
72#endif
diff --git a/arch/x86/um/asm/required-features.h b/arch/x86/um/asm/required-features.h
new file mode 100644
index 000000000000..dfb967b2d2f3
--- /dev/null
+++ b/arch/x86/um/asm/required-features.h
@@ -0,0 +1,9 @@
1#ifndef __UM_REQUIRED_FEATURES_H
2#define __UM_REQUIRED_FEATURES_H
3
4/*
5 * Nothing to see, just need something for the i386 and x86_64 asm
6 * headers to include.
7 */
8
9#endif
diff --git a/arch/x86/um/asm/segment.h b/arch/x86/um/asm/segment.h
new file mode 100644
index 000000000000..45183fcd10b6
--- /dev/null
+++ b/arch/x86/um/asm/segment.h
@@ -0,0 +1,10 @@
1#ifndef __UM_SEGMENT_H
2#define __UM_SEGMENT_H
3
4extern int host_gdt_entry_tls_min;
5
6#define GDT_ENTRY_TLS_ENTRIES 3
7#define GDT_ENTRY_TLS_MIN host_gdt_entry_tls_min
8#define GDT_ENTRY_TLS_MAX (GDT_ENTRY_TLS_MIN + GDT_ENTRY_TLS_ENTRIES - 1)
9
10#endif
diff --git a/arch/x86/um/asm/system.h b/arch/x86/um/asm/system.h
new file mode 100644
index 000000000000..a459fd9b7598
--- /dev/null
+++ b/arch/x86/um/asm/system.h
@@ -0,0 +1,135 @@
1#ifndef _ASM_X86_SYSTEM_H_
2#define _ASM_X86_SYSTEM_H_
3
4#include <asm/asm.h>
5#include <asm/segment.h>
6#include <asm/cpufeature.h>
7#include <asm/cmpxchg.h>
8#include <asm/nops.h>
9
10#include <linux/kernel.h>
11#include <linux/irqflags.h>
12
13/* entries in ARCH_DLINFO: */
14#ifdef CONFIG_IA32_EMULATION
15# define AT_VECTOR_SIZE_ARCH 2
16#else
17# define AT_VECTOR_SIZE_ARCH 1
18#endif
19
20extern unsigned long arch_align_stack(unsigned long sp);
21
22void default_idle(void);
23
24/*
25 * Force strict CPU ordering.
26 * And yes, this is required on UP too when we're talking
27 * to devices.
28 */
29#ifdef CONFIG_X86_32
30/*
31 * Some non-Intel clones support out of order store. wmb() ceases to be a
32 * nop for these.
33 */
34#define mb() alternative("lock; addl $0,0(%%esp)", "mfence", X86_FEATURE_XMM2)
35#define rmb() alternative("lock; addl $0,0(%%esp)", "lfence", X86_FEATURE_XMM2)
36#define wmb() alternative("lock; addl $0,0(%%esp)", "sfence", X86_FEATURE_XMM)
37#else
38#define mb() asm volatile("mfence":::"memory")
39#define rmb() asm volatile("lfence":::"memory")
40#define wmb() asm volatile("sfence" ::: "memory")
41#endif
42
43/**
44 * read_barrier_depends - Flush all pending reads that subsequents reads
45 * depend on.
46 *
47 * No data-dependent reads from memory-like regions are ever reordered
48 * over this barrier. All reads preceding this primitive are guaranteed
49 * to access memory (but not necessarily other CPUs' caches) before any
50 * reads following this primitive that depend on the data return by
51 * any of the preceding reads. This primitive is much lighter weight than
52 * rmb() on most CPUs, and is never heavier weight than is
53 * rmb().
54 *
55 * These ordering constraints are respected by both the local CPU
56 * and the compiler.
57 *
58 * Ordering is not guaranteed by anything other than these primitives,
59 * not even by data dependencies. See the documentation for
60 * memory_barrier() for examples and URLs to more information.
61 *
62 * For example, the following code would force ordering (the initial
63 * value of "a" is zero, "b" is one, and "p" is "&a"):
64 *
65 * <programlisting>
66 * CPU 0 CPU 1
67 *
68 * b = 2;
69 * memory_barrier();
70 * p = &b; q = p;
71 * read_barrier_depends();
72 * d = *q;
73 * </programlisting>
74 *
75 * because the read of "*q" depends on the read of "p" and these
76 * two reads are separated by a read_barrier_depends(). However,
77 * the following code, with the same initial values for "a" and "b":
78 *
79 * <programlisting>
80 * CPU 0 CPU 1
81 *
82 * a = 2;
83 * memory_barrier();
84 * b = 3; y = b;
85 * read_barrier_depends();
86 * x = a;
87 * </programlisting>
88 *
89 * does not enforce ordering, since there is no data dependency between
90 * the read of "a" and the read of "b". Therefore, on some CPUs, such
91 * as Alpha, "y" could be set to 3 and "x" to 0. Use rmb()
92 * in cases like this where there are no data dependencies.
93 **/
94
95#define read_barrier_depends() do { } while (0)
96
97#ifdef CONFIG_SMP
98#define smp_mb() mb()
99#ifdef CONFIG_X86_PPRO_FENCE
100# define smp_rmb() rmb()
101#else
102# define smp_rmb() barrier()
103#endif
104#ifdef CONFIG_X86_OOSTORE
105# define smp_wmb() wmb()
106#else
107# define smp_wmb() barrier()
108#endif
109#define smp_read_barrier_depends() read_barrier_depends()
110#define set_mb(var, value) do { (void)xchg(&var, value); } while (0)
111#else
112#define smp_mb() barrier()
113#define smp_rmb() barrier()
114#define smp_wmb() barrier()
115#define smp_read_barrier_depends() do { } while (0)
116#define set_mb(var, value) do { var = value; barrier(); } while (0)
117#endif
118
119/*
120 * Stop RDTSC speculation. This is needed when you need to use RDTSC
121 * (or get_cycles or vread that possibly accesses the TSC) in a defined
122 * code region.
123 *
124 * (Could use an alternative three way for this if there was one.)
125 */
126static inline void rdtsc_barrier(void)
127{
128 alternative(ASM_NOP3, "mfence", X86_FEATURE_MFENCE_RDTSC);
129 alternative(ASM_NOP3, "lfence", X86_FEATURE_LFENCE_RDTSC);
130}
131
132extern void *_switch_to(void *prev, void *next, void *last);
133#define switch_to(prev, next, last) prev = _switch_to(prev, next, last)
134
135#endif
diff --git a/arch/x86/um/asm/vm-flags.h b/arch/x86/um/asm/vm-flags.h
new file mode 100644
index 000000000000..7c297e9e2413
--- /dev/null
+++ b/arch/x86/um/asm/vm-flags.h
@@ -0,0 +1,25 @@
1/*
2 * Copyright (C) 2004 Jeff Dike (jdike@addtoit.com)
3 * Copyright 2003 PathScale, Inc.
4 * Licensed under the GPL
5 */
6
7#ifndef __VM_FLAGS_X86_H
8#define __VM_FLAGS_X86_H
9
10#ifdef CONFIG_X86_32
11
12#define VM_DATA_DEFAULT_FLAGS \
13 (VM_READ | VM_WRITE | \
14 ((current->personality & READ_IMPLIES_EXEC) ? VM_EXEC : 0 ) | \
15 VM_MAYREAD | VM_MAYWRITE | VM_MAYEXEC)
16
17#else
18
19#define VM_DATA_DEFAULT_FLAGS (VM_READ | VM_WRITE | VM_EXEC | \
20 VM_MAYREAD | VM_MAYWRITE | VM_MAYEXEC)
21#define VM_STACK_DEFAULT_FLAGS (VM_GROWSDOWN | VM_READ | VM_WRITE | \
22 VM_EXEC | VM_MAYREAD | VM_MAYWRITE | VM_MAYEXEC)
23
24#endif
25#endif
diff --git a/arch/x86/um/bug.c b/arch/x86/um/bug.c
new file mode 100644
index 000000000000..e8034e363d83
--- /dev/null
+++ b/arch/x86/um/bug.c
@@ -0,0 +1,21 @@
1/*
2 * Copyright (C) 2006 Jeff Dike (jdike@addtoit.com)
3 * Licensed under the GPL V2
4 */
5
6#include <linux/uaccess.h>
7
8/*
9 * Mostly copied from i386/x86_86 - eliminated the eip < PAGE_OFFSET because
10 * that's not relevant in skas mode.
11 */
12
13int is_valid_bugaddr(unsigned long eip)
14{
15 unsigned short ud2;
16
17 if (probe_kernel_address((unsigned short __user *)eip, ud2))
18 return 0;
19
20 return ud2 == 0x0b0f;
21}
diff --git a/arch/x86/um/bugs_32.c b/arch/x86/um/bugs_32.c
new file mode 100644
index 000000000000..a1fba5fb9dbe
--- /dev/null
+++ b/arch/x86/um/bugs_32.c
@@ -0,0 +1,76 @@
1/*
2 * Copyright (C) 2002 - 2007 Jeff Dike (jdike@{addtoit,linux.intel}.com)
3 * Licensed under the GPL
4 */
5
6#include <signal.h>
7#include "kern_util.h"
8#include "longjmp.h"
9#include "sysdep/ptrace.h"
10#include <generated/asm-offsets.h>
11
12/* Set during early boot */
13static int host_has_cmov = 1;
14static jmp_buf cmov_test_return;
15
16#define TASK_PID(task) *((int *) &(((char *) (task))[HOST_TASK_PID]))
17
18static void cmov_sigill_test_handler(int sig)
19{
20 host_has_cmov = 0;
21 longjmp(cmov_test_return, 1);
22}
23
24void arch_check_bugs(void)
25{
26 struct sigaction old, new;
27
28 printk(UM_KERN_INFO "Checking for host processor cmov support...");
29 new.sa_handler = cmov_sigill_test_handler;
30
31 /* Make sure that SIGILL is enabled after the handler longjmps back */
32 new.sa_flags = SA_NODEFER;
33 sigemptyset(&new.sa_mask);
34 sigaction(SIGILL, &new, &old);
35
36 if (setjmp(cmov_test_return) == 0) {
37 unsigned long foo = 0;
38 __asm__ __volatile__("cmovz %0, %1" : "=r" (foo) : "0" (foo));
39 printk(UM_KERN_CONT "Yes\n");
40 } else
41 printk(UM_KERN_CONT "No\n");
42
43 sigaction(SIGILL, &old, &new);
44}
45
46void arch_examine_signal(int sig, struct uml_pt_regs *regs)
47{
48 unsigned char tmp[2];
49
50 /*
51 * This is testing for a cmov (0x0f 0x4x) instruction causing a
52 * SIGILL in init.
53 */
54 if ((sig != SIGILL) || (TASK_PID(get_current()) != 1))
55 return;
56
57 if (copy_from_user_proc(tmp, (void *) UPT_IP(regs), 2)) {
58 printk(UM_KERN_ERR "SIGILL in init, could not read "
59 "instructions!\n");
60 return;
61 }
62
63 if ((tmp[0] != 0x0f) || ((tmp[1] & 0xf0) != 0x40))
64 return;
65
66 if (host_has_cmov == 0)
67 printk(UM_KERN_ERR "SIGILL caused by cmov, which this "
68 "processor doesn't implement. Boot a filesystem "
69 "compiled for older processors");
70 else if (host_has_cmov == 1)
71 printk(UM_KERN_ERR "SIGILL caused by cmov, which this "
72 "processor claims to implement");
73 else
74 printk(UM_KERN_ERR "Bad value for host_has_cmov (%d)",
75 host_has_cmov);
76}
diff --git a/arch/x86/um/bugs_64.c b/arch/x86/um/bugs_64.c
new file mode 100644
index 000000000000..44e02ba2a265
--- /dev/null
+++ b/arch/x86/um/bugs_64.c
@@ -0,0 +1,15 @@
1/*
2 * Copyright 2003 PathScale, Inc.
3 *
4 * Licensed under the GPL
5 */
6
7#include "sysdep/ptrace.h"
8
9void arch_check_bugs(void)
10{
11}
12
13void arch_examine_signal(int sig, struct uml_pt_regs *regs)
14{
15}
diff --git a/arch/x86/um/checksum_32.S b/arch/x86/um/checksum_32.S
new file mode 100644
index 000000000000..f058d2f82e18
--- /dev/null
+++ b/arch/x86/um/checksum_32.S
@@ -0,0 +1,458 @@
1/*
2 * INET An implementation of the TCP/IP protocol suite for the LINUX
3 * operating system. INET is implemented using the BSD Socket
4 * interface as the means of communication with the user level.
5 *
6 * IP/TCP/UDP checksumming routines
7 *
8 * Authors: Jorge Cwik, <jorge@laser.satlink.net>
9 * Arnt Gulbrandsen, <agulbra@nvg.unit.no>
10 * Tom May, <ftom@netcom.com>
11 * Pentium Pro/II routines:
12 * Alexander Kjeldaas <astor@guardian.no>
13 * Finn Arne Gangstad <finnag@guardian.no>
14 * Lots of code moved from tcp.c and ip.c; see those files
15 * for more names.
16 *
17 * Changes: Ingo Molnar, converted csum_partial_copy() to 2.1 exception
18 * handling.
19 * Andi Kleen, add zeroing on error
20 * converted to pure assembler
21 *
22 * This program is free software; you can redistribute it and/or
23 * modify it under the terms of the GNU General Public License
24 * as published by the Free Software Foundation; either version
25 * 2 of the License, or (at your option) any later version.
26 */
27
28#include <asm/errno.h>
29
30/*
31 * computes a partial checksum, e.g. for TCP/UDP fragments
32 */
33
34/*
35unsigned int csum_partial(const unsigned char * buff, int len, unsigned int sum)
36 */
37
38.text
39.align 4
40.globl csum_partial
41
42#ifndef CONFIG_X86_USE_PPRO_CHECKSUM
43
44 /*
45 * Experiments with Ethernet and SLIP connections show that buff
46 * is aligned on either a 2-byte or 4-byte boundary. We get at
47 * least a twofold speedup on 486 and Pentium if it is 4-byte aligned.
48 * Fortunately, it is easy to convert 2-byte alignment to 4-byte
49 * alignment for the unrolled loop.
50 */
51csum_partial:
52 pushl %esi
53 pushl %ebx
54 movl 20(%esp),%eax # Function arg: unsigned int sum
55 movl 16(%esp),%ecx # Function arg: int len
56 movl 12(%esp),%esi # Function arg: unsigned char *buff
57 testl $2, %esi # Check alignment.
58 jz 2f # Jump if alignment is ok.
59 subl $2, %ecx # Alignment uses up two bytes.
60 jae 1f # Jump if we had at least two bytes.
61 addl $2, %ecx # ecx was < 2. Deal with it.
62 jmp 4f
631: movw (%esi), %bx
64 addl $2, %esi
65 addw %bx, %ax
66 adcl $0, %eax
672:
68 movl %ecx, %edx
69 shrl $5, %ecx
70 jz 2f
71 testl %esi, %esi
721: movl (%esi), %ebx
73 adcl %ebx, %eax
74 movl 4(%esi), %ebx
75 adcl %ebx, %eax
76 movl 8(%esi), %ebx
77 adcl %ebx, %eax
78 movl 12(%esi), %ebx
79 adcl %ebx, %eax
80 movl 16(%esi), %ebx
81 adcl %ebx, %eax
82 movl 20(%esi), %ebx
83 adcl %ebx, %eax
84 movl 24(%esi), %ebx
85 adcl %ebx, %eax
86 movl 28(%esi), %ebx
87 adcl %ebx, %eax
88 lea 32(%esi), %esi
89 dec %ecx
90 jne 1b
91 adcl $0, %eax
922: movl %edx, %ecx
93 andl $0x1c, %edx
94 je 4f
95 shrl $2, %edx # This clears CF
963: adcl (%esi), %eax
97 lea 4(%esi), %esi
98 dec %edx
99 jne 3b
100 adcl $0, %eax
1014: andl $3, %ecx
102 jz 7f
103 cmpl $2, %ecx
104 jb 5f
105 movw (%esi),%cx
106 leal 2(%esi),%esi
107 je 6f
108 shll $16,%ecx
1095: movb (%esi),%cl
1106: addl %ecx,%eax
111 adcl $0, %eax
1127:
113 popl %ebx
114 popl %esi
115 ret
116
117#else
118
119/* Version for PentiumII/PPro */
120
121csum_partial:
122 pushl %esi
123 pushl %ebx
124 movl 20(%esp),%eax # Function arg: unsigned int sum
125 movl 16(%esp),%ecx # Function arg: int len
126 movl 12(%esp),%esi # Function arg: const unsigned char *buf
127
128 testl $2, %esi
129 jnz 30f
13010:
131 movl %ecx, %edx
132 movl %ecx, %ebx
133 andl $0x7c, %ebx
134 shrl $7, %ecx
135 addl %ebx,%esi
136 shrl $2, %ebx
137 negl %ebx
138 lea 45f(%ebx,%ebx,2), %ebx
139 testl %esi, %esi
140 jmp *%ebx
141
142 # Handle 2-byte-aligned regions
14320: addw (%esi), %ax
144 lea 2(%esi), %esi
145 adcl $0, %eax
146 jmp 10b
147
14830: subl $2, %ecx
149 ja 20b
150 je 32f
151 movzbl (%esi),%ebx # csumming 1 byte, 2-aligned
152 addl %ebx, %eax
153 adcl $0, %eax
154 jmp 80f
15532:
156 addw (%esi), %ax # csumming 2 bytes, 2-aligned
157 adcl $0, %eax
158 jmp 80f
159
16040:
161 addl -128(%esi), %eax
162 adcl -124(%esi), %eax
163 adcl -120(%esi), %eax
164 adcl -116(%esi), %eax
165 adcl -112(%esi), %eax
166 adcl -108(%esi), %eax
167 adcl -104(%esi), %eax
168 adcl -100(%esi), %eax
169 adcl -96(%esi), %eax
170 adcl -92(%esi), %eax
171 adcl -88(%esi), %eax
172 adcl -84(%esi), %eax
173 adcl -80(%esi), %eax
174 adcl -76(%esi), %eax
175 adcl -72(%esi), %eax
176 adcl -68(%esi), %eax
177 adcl -64(%esi), %eax
178 adcl -60(%esi), %eax
179 adcl -56(%esi), %eax
180 adcl -52(%esi), %eax
181 adcl -48(%esi), %eax
182 adcl -44(%esi), %eax
183 adcl -40(%esi), %eax
184 adcl -36(%esi), %eax
185 adcl -32(%esi), %eax
186 adcl -28(%esi), %eax
187 adcl -24(%esi), %eax
188 adcl -20(%esi), %eax
189 adcl -16(%esi), %eax
190 adcl -12(%esi), %eax
191 adcl -8(%esi), %eax
192 adcl -4(%esi), %eax
19345:
194 lea 128(%esi), %esi
195 adcl $0, %eax
196 dec %ecx
197 jge 40b
198 movl %edx, %ecx
19950: andl $3, %ecx
200 jz 80f
201
202 # Handle the last 1-3 bytes without jumping
203 notl %ecx # 1->2, 2->1, 3->0, higher bits are masked
204 movl $0xffffff,%ebx # by the shll and shrl instructions
205 shll $3,%ecx
206 shrl %cl,%ebx
207 andl -128(%esi),%ebx # esi is 4-aligned so should be ok
208 addl %ebx,%eax
209 adcl $0,%eax
21080:
211 popl %ebx
212 popl %esi
213 ret
214
215#endif
216
217/*
218unsigned int csum_partial_copy_generic (const char *src, char *dst,
219 int len, int sum, int *src_err_ptr, int *dst_err_ptr)
220 */
221
222/*
223 * Copy from ds while checksumming, otherwise like csum_partial
224 *
225 * The macros SRC and DST specify the type of access for the instruction.
226 * thus we can call a custom exception handler for all access types.
227 *
228 * FIXME: could someone double-check whether I haven't mixed up some SRC and
229 * DST definitions? It's damn hard to trigger all cases. I hope I got
230 * them all but there's no guarantee.
231 */
232
233#define SRC(y...) \
234 9999: y; \
235 .section __ex_table, "a"; \
236 .long 9999b, 6001f ; \
237 .previous
238
239#define DST(y...) \
240 9999: y; \
241 .section __ex_table, "a"; \
242 .long 9999b, 6002f ; \
243 .previous
244
245.align 4
246
247#ifndef CONFIG_X86_USE_PPRO_CHECKSUM
248
249#define ARGBASE 16
250#define FP 12
251
252csum_partial_copy_generic_i386:
253 subl $4,%esp
254 pushl %edi
255 pushl %esi
256 pushl %ebx
257 movl ARGBASE+16(%esp),%eax # sum
258 movl ARGBASE+12(%esp),%ecx # len
259 movl ARGBASE+4(%esp),%esi # src
260 movl ARGBASE+8(%esp),%edi # dst
261
262 testl $2, %edi # Check alignment.
263 jz 2f # Jump if alignment is ok.
264 subl $2, %ecx # Alignment uses up two bytes.
265 jae 1f # Jump if we had at least two bytes.
266 addl $2, %ecx # ecx was < 2. Deal with it.
267 jmp 4f
268SRC(1: movw (%esi), %bx )
269 addl $2, %esi
270DST( movw %bx, (%edi) )
271 addl $2, %edi
272 addw %bx, %ax
273 adcl $0, %eax
2742:
275 movl %ecx, FP(%esp)
276 shrl $5, %ecx
277 jz 2f
278 testl %esi, %esi
279SRC(1: movl (%esi), %ebx )
280SRC( movl 4(%esi), %edx )
281 adcl %ebx, %eax
282DST( movl %ebx, (%edi) )
283 adcl %edx, %eax
284DST( movl %edx, 4(%edi) )
285
286SRC( movl 8(%esi), %ebx )
287SRC( movl 12(%esi), %edx )
288 adcl %ebx, %eax
289DST( movl %ebx, 8(%edi) )
290 adcl %edx, %eax
291DST( movl %edx, 12(%edi) )
292
293SRC( movl 16(%esi), %ebx )
294SRC( movl 20(%esi), %edx )
295 adcl %ebx, %eax
296DST( movl %ebx, 16(%edi) )
297 adcl %edx, %eax
298DST( movl %edx, 20(%edi) )
299
300SRC( movl 24(%esi), %ebx )
301SRC( movl 28(%esi), %edx )
302 adcl %ebx, %eax
303DST( movl %ebx, 24(%edi) )
304 adcl %edx, %eax
305DST( movl %edx, 28(%edi) )
306
307 lea 32(%esi), %esi
308 lea 32(%edi), %edi
309 dec %ecx
310 jne 1b
311 adcl $0, %eax
3122: movl FP(%esp), %edx
313 movl %edx, %ecx
314 andl $0x1c, %edx
315 je 4f
316 shrl $2, %edx # This clears CF
317SRC(3: movl (%esi), %ebx )
318 adcl %ebx, %eax
319DST( movl %ebx, (%edi) )
320 lea 4(%esi), %esi
321 lea 4(%edi), %edi
322 dec %edx
323 jne 3b
324 adcl $0, %eax
3254: andl $3, %ecx
326 jz 7f
327 cmpl $2, %ecx
328 jb 5f
329SRC( movw (%esi), %cx )
330 leal 2(%esi), %esi
331DST( movw %cx, (%edi) )
332 leal 2(%edi), %edi
333 je 6f
334 shll $16,%ecx
335SRC(5: movb (%esi), %cl )
336DST( movb %cl, (%edi) )
3376: addl %ecx, %eax
338 adcl $0, %eax
3397:
3405000:
341
342# Exception handler:
343.section .fixup, "ax"
344
3456001:
346 movl ARGBASE+20(%esp), %ebx # src_err_ptr
347 movl $-EFAULT, (%ebx)
348
349 # zero the complete destination - computing the rest
350 # is too much work
351 movl ARGBASE+8(%esp), %edi # dst
352 movl ARGBASE+12(%esp), %ecx # len
353 xorl %eax,%eax
354 rep ; stosb
355
356 jmp 5000b
357
3586002:
359 movl ARGBASE+24(%esp), %ebx # dst_err_ptr
360 movl $-EFAULT,(%ebx)
361 jmp 5000b
362
363.previous
364
365 popl %ebx
366 popl %esi
367 popl %edi
368 popl %ecx # equivalent to addl $4,%esp
369 ret
370
371#else
372
373/* Version for PentiumII/PPro */
374
375#define ROUND1(x) \
376 SRC(movl x(%esi), %ebx ) ; \
377 addl %ebx, %eax ; \
378 DST(movl %ebx, x(%edi) ) ;
379
380#define ROUND(x) \
381 SRC(movl x(%esi), %ebx ) ; \
382 adcl %ebx, %eax ; \
383 DST(movl %ebx, x(%edi) ) ;
384
385#define ARGBASE 12
386
387csum_partial_copy_generic_i386:
388 pushl %ebx
389 pushl %edi
390 pushl %esi
391 movl ARGBASE+4(%esp),%esi #src
392 movl ARGBASE+8(%esp),%edi #dst
393 movl ARGBASE+12(%esp),%ecx #len
394 movl ARGBASE+16(%esp),%eax #sum
395# movl %ecx, %edx
396 movl %ecx, %ebx
397 movl %esi, %edx
398 shrl $6, %ecx
399 andl $0x3c, %ebx
400 negl %ebx
401 subl %ebx, %esi
402 subl %ebx, %edi
403 lea -1(%esi),%edx
404 andl $-32,%edx
405 lea 3f(%ebx,%ebx), %ebx
406 testl %esi, %esi
407 jmp *%ebx
4081: addl $64,%esi
409 addl $64,%edi
410 SRC(movb -32(%edx),%bl) ; SRC(movb (%edx),%bl)
411 ROUND1(-64) ROUND(-60) ROUND(-56) ROUND(-52)
412 ROUND (-48) ROUND(-44) ROUND(-40) ROUND(-36)
413 ROUND (-32) ROUND(-28) ROUND(-24) ROUND(-20)
414 ROUND (-16) ROUND(-12) ROUND(-8) ROUND(-4)
4153: adcl $0,%eax
416 addl $64, %edx
417 dec %ecx
418 jge 1b
4194: movl ARGBASE+12(%esp),%edx #len
420 andl $3, %edx
421 jz 7f
422 cmpl $2, %edx
423 jb 5f
424SRC( movw (%esi), %dx )
425 leal 2(%esi), %esi
426DST( movw %dx, (%edi) )
427 leal 2(%edi), %edi
428 je 6f
429 shll $16,%edx
4305:
431SRC( movb (%esi), %dl )
432DST( movb %dl, (%edi) )
4336: addl %edx, %eax
434 adcl $0, %eax
4357:
436.section .fixup, "ax"
4376001: movl ARGBASE+20(%esp), %ebx # src_err_ptr
438 movl $-EFAULT, (%ebx)
439 # zero the complete destination (computing the rest is too much work)
440 movl ARGBASE+8(%esp),%edi # dst
441 movl ARGBASE+12(%esp),%ecx # len
442 xorl %eax,%eax
443 rep; stosb
444 jmp 7b
4456002: movl ARGBASE+24(%esp), %ebx # dst_err_ptr
446 movl $-EFAULT, (%ebx)
447 jmp 7b
448.previous
449
450 popl %esi
451 popl %edi
452 popl %ebx
453 ret
454
455#undef ROUND
456#undef ROUND1
457
458#endif
diff --git a/arch/x86/um/delay.c b/arch/x86/um/delay.c
new file mode 100644
index 000000000000..f3fe1a688f7e
--- /dev/null
+++ b/arch/x86/um/delay.c
@@ -0,0 +1,60 @@
1/*
2 * Copyright (C) 2011 Richard Weinberger <richrd@nod.at>
3 * Mostly copied from arch/x86/lib/delay.c
4 *
5 * This program is free software; you can redistribute it and/or modify
6 * it under the terms of the GNU General Public License version 2 as
7 * published by the Free Software Foundation.
8 */
9
10#include <linux/module.h>
11#include <linux/kernel.h>
12#include <linux/delay.h>
13#include <asm/param.h>
14
15void __delay(unsigned long loops)
16{
17 asm volatile(
18 "test %0,%0\n"
19 "jz 3f\n"
20 "jmp 1f\n"
21
22 ".align 16\n"
23 "1: jmp 2f\n"
24
25 ".align 16\n"
26 "2: dec %0\n"
27 " jnz 2b\n"
28 "3: dec %0\n"
29
30 : /* we don't need output */
31 : "a" (loops)
32 );
33}
34EXPORT_SYMBOL(__delay);
35
36inline void __const_udelay(unsigned long xloops)
37{
38 int d0;
39
40 xloops *= 4;
41 asm("mull %%edx"
42 : "=d" (xloops), "=&a" (d0)
43 : "1" (xloops), "0"
44 (loops_per_jiffy * (HZ/4)));
45
46 __delay(++xloops);
47}
48EXPORT_SYMBOL(__const_udelay);
49
50void __udelay(unsigned long usecs)
51{
52 __const_udelay(usecs * 0x000010c7); /* 2**32 / 1000000 (rounded up) */
53}
54EXPORT_SYMBOL(__udelay);
55
56void __ndelay(unsigned long nsecs)
57{
58 __const_udelay(nsecs * 0x00005); /* 2**32 / 1000000000 (rounded up) */
59}
60EXPORT_SYMBOL(__ndelay);
diff --git a/arch/x86/um/elfcore.c b/arch/x86/um/elfcore.c
new file mode 100644
index 000000000000..6bb49b687c97
--- /dev/null
+++ b/arch/x86/um/elfcore.c
@@ -0,0 +1,83 @@
1#include <linux/elf.h>
2#include <linux/coredump.h>
3#include <linux/fs.h>
4#include <linux/mm.h>
5
6#include <asm/elf.h>
7
8
9Elf32_Half elf_core_extra_phdrs(void)
10{
11 return vsyscall_ehdr ? (((struct elfhdr *)vsyscall_ehdr)->e_phnum) : 0;
12}
13
14int elf_core_write_extra_phdrs(struct file *file, loff_t offset, size_t *size,
15 unsigned long limit)
16{
17 if ( vsyscall_ehdr ) {
18 const struct elfhdr *const ehdrp =
19 (struct elfhdr *) vsyscall_ehdr;
20 const struct elf_phdr *const phdrp =
21 (const struct elf_phdr *) (vsyscall_ehdr + ehdrp->e_phoff);
22 int i;
23 Elf32_Off ofs = 0;
24
25 for (i = 0; i < ehdrp->e_phnum; ++i) {
26 struct elf_phdr phdr = phdrp[i];
27
28 if (phdr.p_type == PT_LOAD) {
29 ofs = phdr.p_offset = offset;
30 offset += phdr.p_filesz;
31 } else {
32 phdr.p_offset += ofs;
33 }
34 phdr.p_paddr = 0; /* match other core phdrs */
35 *size += sizeof(phdr);
36 if (*size > limit
37 || !dump_write(file, &phdr, sizeof(phdr)))
38 return 0;
39 }
40 }
41 return 1;
42}
43
44int elf_core_write_extra_data(struct file *file, size_t *size,
45 unsigned long limit)
46{
47 if ( vsyscall_ehdr ) {
48 const struct elfhdr *const ehdrp =
49 (struct elfhdr *) vsyscall_ehdr;
50 const struct elf_phdr *const phdrp =
51 (const struct elf_phdr *) (vsyscall_ehdr + ehdrp->e_phoff);
52 int i;
53
54 for (i = 0; i < ehdrp->e_phnum; ++i) {
55 if (phdrp[i].p_type == PT_LOAD) {
56 void *addr = (void *) phdrp[i].p_vaddr;
57 size_t filesz = phdrp[i].p_filesz;
58
59 *size += filesz;
60 if (*size > limit
61 || !dump_write(file, addr, filesz))
62 return 0;
63 }
64 }
65 }
66 return 1;
67}
68
69size_t elf_core_extra_data_size(void)
70{
71 if ( vsyscall_ehdr ) {
72 const struct elfhdr *const ehdrp =
73 (struct elfhdr *)vsyscall_ehdr;
74 const struct elf_phdr *const phdrp =
75 (const struct elf_phdr *) (vsyscall_ehdr + ehdrp->e_phoff);
76 int i;
77
78 for (i = 0; i < ehdrp->e_phnum; ++i)
79 if (phdrp[i].p_type == PT_LOAD)
80 return (size_t) phdrp[i].p_filesz;
81 }
82 return 0;
83}
diff --git a/arch/x86/um/fault.c b/arch/x86/um/fault.c
new file mode 100644
index 000000000000..d670f68532f4
--- /dev/null
+++ b/arch/x86/um/fault.c
@@ -0,0 +1,28 @@
1/*
2 * Copyright (C) 2002 - 2007 Jeff Dike (jdike@{addtoit,linux.intel}.com)
3 * Licensed under the GPL
4 */
5
6#include "sysdep/ptrace.h"
7
8/* These two are from asm-um/uaccess.h and linux/module.h, check them. */
9struct exception_table_entry
10{
11 unsigned long insn;
12 unsigned long fixup;
13};
14
15const struct exception_table_entry *search_exception_tables(unsigned long add);
16
17/* Compare this to arch/i386/mm/extable.c:fixup_exception() */
18int arch_fixup(unsigned long address, struct uml_pt_regs *regs)
19{
20 const struct exception_table_entry *fixup;
21
22 fixup = search_exception_tables(address);
23 if (fixup != 0) {
24 UPT_IP(regs) = fixup->fixup;
25 return 1;
26 }
27 return 0;
28}
diff --git a/arch/x86/um/ksyms.c b/arch/x86/um/ksyms.c
new file mode 100644
index 000000000000..2e8f43ec6214
--- /dev/null
+++ b/arch/x86/um/ksyms.c
@@ -0,0 +1,13 @@
1#include <linux/module.h>
2#include <asm/string.h>
3#include <asm/checksum.h>
4
5#ifndef CONFIG_X86_32
6/*XXX: we need them because they would be exported by x86_64 */
7#if (__GNUC__ == 4 && __GNUC_MINOR__ >= 3) || __GNUC__ > 4
8EXPORT_SYMBOL(memcpy);
9#else
10EXPORT_SYMBOL(__memcpy);
11#endif
12#endif
13EXPORT_SYMBOL(csum_partial);
diff --git a/arch/x86/um/ldt.c b/arch/x86/um/ldt.c
new file mode 100644
index 000000000000..26b0e39d2ce9
--- /dev/null
+++ b/arch/x86/um/ldt.c
@@ -0,0 +1,502 @@
1/*
2 * Copyright (C) 2001 - 2007 Jeff Dike (jdike@{addtoit,linux.intel}.com)
3 * Licensed under the GPL
4 */
5
6#include <linux/mm.h>
7#include <linux/sched.h>
8#include <linux/slab.h>
9#include <asm/unistd.h>
10#include "os.h"
11#include "proc_mm.h"
12#include "skas.h"
13#include "skas_ptrace.h"
14#include "sysdep/tls.h"
15
16extern int modify_ldt(int func, void *ptr, unsigned long bytecount);
17
18static long write_ldt_entry(struct mm_id *mm_idp, int func,
19 struct user_desc *desc, void **addr, int done)
20{
21 long res;
22
23 if (proc_mm) {
24 /*
25 * This is a special handling for the case, that the mm to
26 * modify isn't current->active_mm.
27 * If this is called directly by modify_ldt,
28 * (current->active_mm->context.skas.u == mm_idp)
29 * will be true. So no call to __switch_mm(mm_idp) is done.
30 * If this is called in case of init_new_ldt or PTRACE_LDT,
31 * mm_idp won't belong to current->active_mm, but child->mm.
32 * So we need to switch child's mm into our userspace, then
33 * later switch back.
34 *
35 * Note: I'm unsure: should interrupts be disabled here?
36 */
37 if (!current->active_mm || current->active_mm == &init_mm ||
38 mm_idp != &current->active_mm->context.id)
39 __switch_mm(mm_idp);
40 }
41
42 if (ptrace_ldt) {
43 struct ptrace_ldt ldt_op = (struct ptrace_ldt) {
44 .func = func,
45 .ptr = desc,
46 .bytecount = sizeof(*desc)};
47 u32 cpu;
48 int pid;
49
50 if (!proc_mm)
51 pid = mm_idp->u.pid;
52 else {
53 cpu = get_cpu();
54 pid = userspace_pid[cpu];
55 }
56
57 res = os_ptrace_ldt(pid, 0, (unsigned long) &ldt_op);
58
59 if (proc_mm)
60 put_cpu();
61 }
62 else {
63 void *stub_addr;
64 res = syscall_stub_data(mm_idp, (unsigned long *)desc,
65 (sizeof(*desc) + sizeof(long) - 1) &
66 ~(sizeof(long) - 1),
67 addr, &stub_addr);
68 if (!res) {
69 unsigned long args[] = { func,
70 (unsigned long)stub_addr,
71 sizeof(*desc),
72 0, 0, 0 };
73 res = run_syscall_stub(mm_idp, __NR_modify_ldt, args,
74 0, addr, done);
75 }
76 }
77
78 if (proc_mm) {
79 /*
80 * This is the second part of special handling, that makes
81 * PTRACE_LDT possible to implement.
82 */
83 if (current->active_mm && current->active_mm != &init_mm &&
84 mm_idp != &current->active_mm->context.id)
85 __switch_mm(&current->active_mm->context.id);
86 }
87
88 return res;
89}
90
91static long read_ldt_from_host(void __user * ptr, unsigned long bytecount)
92{
93 int res, n;
94 struct ptrace_ldt ptrace_ldt = (struct ptrace_ldt) {
95 .func = 0,
96 .bytecount = bytecount,
97 .ptr = kmalloc(bytecount, GFP_KERNEL)};
98 u32 cpu;
99
100 if (ptrace_ldt.ptr == NULL)
101 return -ENOMEM;
102
103 /*
104 * This is called from sys_modify_ldt only, so userspace_pid gives
105 * us the right number
106 */
107
108 cpu = get_cpu();
109 res = os_ptrace_ldt(userspace_pid[cpu], 0, (unsigned long) &ptrace_ldt);
110 put_cpu();
111 if (res < 0)
112 goto out;
113
114 n = copy_to_user(ptr, ptrace_ldt.ptr, res);
115 if (n != 0)
116 res = -EFAULT;
117
118 out:
119 kfree(ptrace_ldt.ptr);
120
121 return res;
122}
123
124/*
125 * In skas mode, we hold our own ldt data in UML.
126 * Thus, the code implementing sys_modify_ldt_skas
127 * is very similar to (and mostly stolen from) sys_modify_ldt
128 * for arch/i386/kernel/ldt.c
129 * The routines copied and modified in part are:
130 * - read_ldt
131 * - read_default_ldt
132 * - write_ldt
133 * - sys_modify_ldt_skas
134 */
135
136static int read_ldt(void __user * ptr, unsigned long bytecount)
137{
138 int i, err = 0;
139 unsigned long size;
140 uml_ldt_t *ldt = &current->mm->context.arch.ldt;
141
142 if (!ldt->entry_count)
143 goto out;
144 if (bytecount > LDT_ENTRY_SIZE*LDT_ENTRIES)
145 bytecount = LDT_ENTRY_SIZE*LDT_ENTRIES;
146 err = bytecount;
147
148 if (ptrace_ldt)
149 return read_ldt_from_host(ptr, bytecount);
150
151 mutex_lock(&ldt->lock);
152 if (ldt->entry_count <= LDT_DIRECT_ENTRIES) {
153 size = LDT_ENTRY_SIZE*LDT_DIRECT_ENTRIES;
154 if (size > bytecount)
155 size = bytecount;
156 if (copy_to_user(ptr, ldt->u.entries, size))
157 err = -EFAULT;
158 bytecount -= size;
159 ptr += size;
160 }
161 else {
162 for (i=0; i<ldt->entry_count/LDT_ENTRIES_PER_PAGE && bytecount;
163 i++) {
164 size = PAGE_SIZE;
165 if (size > bytecount)
166 size = bytecount;
167 if (copy_to_user(ptr, ldt->u.pages[i], size)) {
168 err = -EFAULT;
169 break;
170 }
171 bytecount -= size;
172 ptr += size;
173 }
174 }
175 mutex_unlock(&ldt->lock);
176
177 if (bytecount == 0 || err == -EFAULT)
178 goto out;
179
180 if (clear_user(ptr, bytecount))
181 err = -EFAULT;
182
183out:
184 return err;
185}
186
187static int read_default_ldt(void __user * ptr, unsigned long bytecount)
188{
189 int err;
190
191 if (bytecount > 5*LDT_ENTRY_SIZE)
192 bytecount = 5*LDT_ENTRY_SIZE;
193
194 err = bytecount;
195 /*
196 * UML doesn't support lcall7 and lcall27.
197 * So, we don't really have a default ldt, but emulate
198 * an empty ldt of common host default ldt size.
199 */
200 if (clear_user(ptr, bytecount))
201 err = -EFAULT;
202
203 return err;
204}
205
206static int write_ldt(void __user * ptr, unsigned long bytecount, int func)
207{
208 uml_ldt_t *ldt = &current->mm->context.arch.ldt;
209 struct mm_id * mm_idp = &current->mm->context.id;
210 int i, err;
211 struct user_desc ldt_info;
212 struct ldt_entry entry0, *ldt_p;
213 void *addr = NULL;
214
215 err = -EINVAL;
216 if (bytecount != sizeof(ldt_info))
217 goto out;
218 err = -EFAULT;
219 if (copy_from_user(&ldt_info, ptr, sizeof(ldt_info)))
220 goto out;
221
222 err = -EINVAL;
223 if (ldt_info.entry_number >= LDT_ENTRIES)
224 goto out;
225 if (ldt_info.contents == 3) {
226 if (func == 1)
227 goto out;
228 if (ldt_info.seg_not_present == 0)
229 goto out;
230 }
231
232 if (!ptrace_ldt)
233 mutex_lock(&ldt->lock);
234
235 err = write_ldt_entry(mm_idp, func, &ldt_info, &addr, 1);
236 if (err)
237 goto out_unlock;
238 else if (ptrace_ldt) {
239 /* With PTRACE_LDT available, this is used as a flag only */
240 ldt->entry_count = 1;
241 goto out;
242 }
243
244 if (ldt_info.entry_number >= ldt->entry_count &&
245 ldt_info.entry_number >= LDT_DIRECT_ENTRIES) {
246 for (i=ldt->entry_count/LDT_ENTRIES_PER_PAGE;
247 i*LDT_ENTRIES_PER_PAGE <= ldt_info.entry_number;
248 i++) {
249 if (i == 0)
250 memcpy(&entry0, ldt->u.entries,
251 sizeof(entry0));
252 ldt->u.pages[i] = (struct ldt_entry *)
253 __get_free_page(GFP_KERNEL|__GFP_ZERO);
254 if (!ldt->u.pages[i]) {
255 err = -ENOMEM;
256 /* Undo the change in host */
257 memset(&ldt_info, 0, sizeof(ldt_info));
258 write_ldt_entry(mm_idp, 1, &ldt_info, &addr, 1);
259 goto out_unlock;
260 }
261 if (i == 0) {
262 memcpy(ldt->u.pages[0], &entry0,
263 sizeof(entry0));
264 memcpy(ldt->u.pages[0]+1, ldt->u.entries+1,
265 sizeof(entry0)*(LDT_DIRECT_ENTRIES-1));
266 }
267 ldt->entry_count = (i + 1) * LDT_ENTRIES_PER_PAGE;
268 }
269 }
270 if (ldt->entry_count <= ldt_info.entry_number)
271 ldt->entry_count = ldt_info.entry_number + 1;
272
273 if (ldt->entry_count <= LDT_DIRECT_ENTRIES)
274 ldt_p = ldt->u.entries + ldt_info.entry_number;
275 else
276 ldt_p = ldt->u.pages[ldt_info.entry_number/LDT_ENTRIES_PER_PAGE] +
277 ldt_info.entry_number%LDT_ENTRIES_PER_PAGE;
278
279 if (ldt_info.base_addr == 0 && ldt_info.limit == 0 &&
280 (func == 1 || LDT_empty(&ldt_info))) {
281 ldt_p->a = 0;
282 ldt_p->b = 0;
283 }
284 else{
285 if (func == 1)
286 ldt_info.useable = 0;
287 ldt_p->a = LDT_entry_a(&ldt_info);
288 ldt_p->b = LDT_entry_b(&ldt_info);
289 }
290 err = 0;
291
292out_unlock:
293 mutex_unlock(&ldt->lock);
294out:
295 return err;
296}
297
298static long do_modify_ldt_skas(int func, void __user *ptr,
299 unsigned long bytecount)
300{
301 int ret = -ENOSYS;
302
303 switch (func) {
304 case 0:
305 ret = read_ldt(ptr, bytecount);
306 break;
307 case 1:
308 case 0x11:
309 ret = write_ldt(ptr, bytecount, func);
310 break;
311 case 2:
312 ret = read_default_ldt(ptr, bytecount);
313 break;
314 }
315 return ret;
316}
317
318static DEFINE_SPINLOCK(host_ldt_lock);
319static short dummy_list[9] = {0, -1};
320static short * host_ldt_entries = NULL;
321
322static void ldt_get_host_info(void)
323{
324 long ret;
325 struct ldt_entry * ldt;
326 short *tmp;
327 int i, size, k, order;
328
329 spin_lock(&host_ldt_lock);
330
331 if (host_ldt_entries != NULL) {
332 spin_unlock(&host_ldt_lock);
333 return;
334 }
335 host_ldt_entries = dummy_list+1;
336
337 spin_unlock(&host_ldt_lock);
338
339 for (i = LDT_PAGES_MAX-1, order=0; i; i>>=1, order++)
340 ;
341
342 ldt = (struct ldt_entry *)
343 __get_free_pages(GFP_KERNEL|__GFP_ZERO, order);
344 if (ldt == NULL) {
345 printk(KERN_ERR "ldt_get_host_info: couldn't allocate buffer "
346 "for host ldt\n");
347 return;
348 }
349
350 ret = modify_ldt(0, ldt, (1<<order)*PAGE_SIZE);
351 if (ret < 0) {
352 printk(KERN_ERR "ldt_get_host_info: couldn't read host ldt\n");
353 goto out_free;
354 }
355 if (ret == 0) {
356 /* default_ldt is active, simply write an empty entry 0 */
357 host_ldt_entries = dummy_list;
358 goto out_free;
359 }
360
361 for (i=0, size=0; i<ret/LDT_ENTRY_SIZE; i++) {
362 if (ldt[i].a != 0 || ldt[i].b != 0)
363 size++;
364 }
365
366 if (size < ARRAY_SIZE(dummy_list))
367 host_ldt_entries = dummy_list;
368 else {
369 size = (size + 1) * sizeof(dummy_list[0]);
370 tmp = kmalloc(size, GFP_KERNEL);
371 if (tmp == NULL) {
372 printk(KERN_ERR "ldt_get_host_info: couldn't allocate "
373 "host ldt list\n");
374 goto out_free;
375 }
376 host_ldt_entries = tmp;
377 }
378
379 for (i=0, k=0; i<ret/LDT_ENTRY_SIZE; i++) {
380 if (ldt[i].a != 0 || ldt[i].b != 0)
381 host_ldt_entries[k++] = i;
382 }
383 host_ldt_entries[k] = -1;
384
385out_free:
386 free_pages((unsigned long)ldt, order);
387}
388
389long init_new_ldt(struct mm_context *new_mm, struct mm_context *from_mm)
390{
391 struct user_desc desc;
392 short * num_p;
393 int i;
394 long page, err=0;
395 void *addr = NULL;
396 struct proc_mm_op copy;
397
398
399 if (!ptrace_ldt)
400 mutex_init(&new_mm->arch.ldt.lock);
401
402 if (!from_mm) {
403 memset(&desc, 0, sizeof(desc));
404 /*
405 * We have to initialize a clean ldt.
406 */
407 if (proc_mm) {
408 /*
409 * If the new mm was created using proc_mm, host's
410 * default-ldt currently is assigned, which normally
411 * contains the call-gates for lcall7 and lcall27.
412 * To remove these gates, we simply write an empty
413 * entry as number 0 to the host.
414 */
415 err = write_ldt_entry(&new_mm->id, 1, &desc, &addr, 1);
416 }
417 else{
418 /*
419 * Now we try to retrieve info about the ldt, we
420 * inherited from the host. All ldt-entries found
421 * will be reset in the following loop
422 */
423 ldt_get_host_info();
424 for (num_p=host_ldt_entries; *num_p != -1; num_p++) {
425 desc.entry_number = *num_p;
426 err = write_ldt_entry(&new_mm->id, 1, &desc,
427 &addr, *(num_p + 1) == -1);
428 if (err)
429 break;
430 }
431 }
432 new_mm->arch.ldt.entry_count = 0;
433
434 goto out;
435 }
436
437 if (proc_mm) {
438 /*
439 * We have a valid from_mm, so we now have to copy the LDT of
440 * from_mm to new_mm, because using proc_mm an new mm with
441 * an empty/default LDT was created in new_mm()
442 */
443 copy = ((struct proc_mm_op) { .op = MM_COPY_SEGMENTS,
444 .u =
445 { .copy_segments =
446 from_mm->id.u.mm_fd } } );
447 i = os_write_file(new_mm->id.u.mm_fd, &copy, sizeof(copy));
448 if (i != sizeof(copy))
449 printk(KERN_ERR "new_mm : /proc/mm copy_segments "
450 "failed, err = %d\n", -i);
451 }
452
453 if (!ptrace_ldt) {
454 /*
455 * Our local LDT is used to supply the data for
456 * modify_ldt(READLDT), if PTRACE_LDT isn't available,
457 * i.e., we have to use the stub for modify_ldt, which
458 * can't handle the big read buffer of up to 64kB.
459 */
460 mutex_lock(&from_mm->arch.ldt.lock);
461 if (from_mm->arch.ldt.entry_count <= LDT_DIRECT_ENTRIES)
462 memcpy(new_mm->arch.ldt.u.entries, from_mm->arch.ldt.u.entries,
463 sizeof(new_mm->arch.ldt.u.entries));
464 else {
465 i = from_mm->arch.ldt.entry_count / LDT_ENTRIES_PER_PAGE;
466 while (i-->0) {
467 page = __get_free_page(GFP_KERNEL|__GFP_ZERO);
468 if (!page) {
469 err = -ENOMEM;
470 break;
471 }
472 new_mm->arch.ldt.u.pages[i] =
473 (struct ldt_entry *) page;
474 memcpy(new_mm->arch.ldt.u.pages[i],
475 from_mm->arch.ldt.u.pages[i], PAGE_SIZE);
476 }
477 }
478 new_mm->arch.ldt.entry_count = from_mm->arch.ldt.entry_count;
479 mutex_unlock(&from_mm->arch.ldt.lock);
480 }
481
482 out:
483 return err;
484}
485
486
487void free_ldt(struct mm_context *mm)
488{
489 int i;
490
491 if (!ptrace_ldt && mm->arch.ldt.entry_count > LDT_DIRECT_ENTRIES) {
492 i = mm->arch.ldt.entry_count / LDT_ENTRIES_PER_PAGE;
493 while (i-- > 0)
494 free_page((long) mm->arch.ldt.u.pages[i]);
495 }
496 mm->arch.ldt.entry_count = 0;
497}
498
499int sys_modify_ldt(int func, void __user *ptr, unsigned long bytecount)
500{
501 return do_modify_ldt_skas(func, ptr, bytecount);
502}
diff --git a/arch/x86/um/mem_32.c b/arch/x86/um/mem_32.c
new file mode 100644
index 000000000000..639900a6fde9
--- /dev/null
+++ b/arch/x86/um/mem_32.c
@@ -0,0 +1,62 @@
1/*
2 * Copyright (C) 2011 Richard Weinberger <richrd@nod.at>
3 *
4 * This program is free software; you can redistribute it and/or modify
5 * it under the terms of the GNU General Public License version 2 as
6 * published by the Free Software Foundation.
7 */
8
9#include <linux/mm.h>
10#include <asm/page.h>
11#include <asm/mman.h>
12
13static struct vm_area_struct gate_vma;
14
15static int __init gate_vma_init(void)
16{
17 if (!FIXADDR_USER_START)
18 return 0;
19
20 gate_vma.vm_mm = NULL;
21 gate_vma.vm_start = FIXADDR_USER_START;
22 gate_vma.vm_end = FIXADDR_USER_END;
23 gate_vma.vm_flags = VM_READ | VM_MAYREAD | VM_EXEC | VM_MAYEXEC;
24 gate_vma.vm_page_prot = __P101;
25
26 /*
27 * Make sure the vDSO gets into every core dump.
28 * Dumping its contents makes post-mortem fully interpretable later
29 * without matching up the same kernel and hardware config to see
30 * what PC values meant.
31 */
32 gate_vma.vm_flags |= VM_ALWAYSDUMP;
33
34 return 0;
35}
36__initcall(gate_vma_init);
37
38struct vm_area_struct *get_gate_vma(struct mm_struct *mm)
39{
40 return FIXADDR_USER_START ? &gate_vma : NULL;
41}
42
43int in_gate_area_no_mm(unsigned long addr)
44{
45 if (!FIXADDR_USER_START)
46 return 0;
47
48 if ((addr >= FIXADDR_USER_START) && (addr < FIXADDR_USER_END))
49 return 1;
50
51 return 0;
52}
53
54int in_gate_area(struct mm_struct *mm, unsigned long addr)
55{
56 struct vm_area_struct *vma = get_gate_vma(mm);
57
58 if (!vma)
59 return 0;
60
61 return (addr >= vma->vm_start) && (addr < vma->vm_end);
62}
diff --git a/arch/x86/um/mem_64.c b/arch/x86/um/mem_64.c
new file mode 100644
index 000000000000..546518727a73
--- /dev/null
+++ b/arch/x86/um/mem_64.c
@@ -0,0 +1,26 @@
1#include "linux/mm.h"
2#include "asm/page.h"
3#include "asm/mman.h"
4
5const char *arch_vma_name(struct vm_area_struct *vma)
6{
7 if (vma->vm_mm && vma->vm_start == um_vdso_addr)
8 return "[vdso]";
9
10 return NULL;
11}
12
13struct vm_area_struct *get_gate_vma(struct mm_struct *mm)
14{
15 return NULL;
16}
17
18int in_gate_area(struct mm_struct *mm, unsigned long addr)
19{
20 return 0;
21}
22
23int in_gate_area_no_mm(unsigned long addr)
24{
25 return 0;
26}
diff --git a/arch/x86/um/os-Linux/Makefile b/arch/x86/um/os-Linux/Makefile
new file mode 100644
index 000000000000..253bfb8cb702
--- /dev/null
+++ b/arch/x86/um/os-Linux/Makefile
@@ -0,0 +1,13 @@
1#
2# Copyright (C) 2000 - 2007 Jeff Dike (jdike@{addtoit,linux.intel}.com)
3# Licensed under the GPL
4#
5
6obj-y = registers.o task_size.o mcontext.o
7
8obj-$(CONFIG_X86_32) += tls.o
9obj-$(CONFIG_64BIT) += prctl.o
10
11USER_OBJS := $(obj-y)
12
13include arch/um/scripts/Makefile.rules
diff --git a/arch/x86/um/os-Linux/mcontext.c b/arch/x86/um/os-Linux/mcontext.c
new file mode 100644
index 000000000000..1d33d72c6284
--- /dev/null
+++ b/arch/x86/um/os-Linux/mcontext.c
@@ -0,0 +1,31 @@
1#include <sys/ucontext.h>
2#define __FRAME_OFFSETS
3#include <asm/ptrace.h>
4#include <sysdep/ptrace.h>
5
6void get_regs_from_mc(struct uml_pt_regs *regs, mcontext_t *mc)
7{
8#ifdef __i386__
9#define COPY2(X,Y) regs->gp[X] = mc->gregs[REG_##Y]
10#define COPY(X) regs->gp[X] = mc->gregs[REG_##X]
11#define COPY_SEG(X) regs->gp[X] = mc->gregs[REG_##X] & 0xffff;
12#define COPY_SEG_CPL3(X) regs->gp[X] = (mc->gregs[REG_##X] & 0xffff) | 3;
13 COPY_SEG(GS); COPY_SEG(FS); COPY_SEG(ES); COPY_SEG(DS);
14 COPY(EDI); COPY(ESI); COPY(EBP);
15 COPY2(UESP, ESP); /* sic */
16 COPY(EBX); COPY(EDX); COPY(ECX); COPY(EAX);
17 COPY(EIP); COPY_SEG_CPL3(CS); COPY(EFL); COPY_SEG_CPL3(SS);
18#else
19#define COPY2(X,Y) regs->gp[X/sizeof(unsigned long)] = mc->gregs[REG_##Y]
20#define COPY(X) regs->gp[X/sizeof(unsigned long)] = mc->gregs[REG_##X]
21 COPY(R8); COPY(R9); COPY(R10); COPY(R11);
22 COPY(R12); COPY(R13); COPY(R14); COPY(R15);
23 COPY(RDI); COPY(RSI); COPY(RBP); COPY(RBX);
24 COPY(RDX); COPY(RAX); COPY(RCX); COPY(RSP);
25 COPY(RIP);
26 COPY2(EFLAGS, EFL);
27 COPY2(CS, CSGSFS);
28 regs->gp[CS / sizeof(unsigned long)] &= 0xffff;
29 regs->gp[CS / sizeof(unsigned long)] |= 3;
30#endif
31}
diff --git a/arch/x86/um/os-Linux/prctl.c b/arch/x86/um/os-Linux/prctl.c
new file mode 100644
index 000000000000..9d34eddb517f
--- /dev/null
+++ b/arch/x86/um/os-Linux/prctl.c
@@ -0,0 +1,12 @@
1/*
2 * Copyright (C) 2007 Jeff Dike (jdike@{addtoit.com,linux.intel.com})
3 * Licensed under the GPL
4 */
5
6#include <sys/ptrace.h>
7#include <linux/ptrace.h>
8
9int os_arch_prctl(int pid, int code, unsigned long *addr)
10{
11 return ptrace(PTRACE_ARCH_PRCTL, pid, (unsigned long) addr, code);
12}
diff --git a/arch/x86/um/os-Linux/registers.c b/arch/x86/um/os-Linux/registers.c
new file mode 100644
index 000000000000..0cdbb86b012b
--- /dev/null
+++ b/arch/x86/um/os-Linux/registers.c
@@ -0,0 +1,113 @@
1/*
2 * Copyright (C) 2004 PathScale, Inc
3 * Copyright (C) 2004 - 2007 Jeff Dike (jdike@{addtoit,linux.intel}.com)
4 * Licensed under the GPL
5 */
6
7#include <errno.h>
8#include <sys/ptrace.h>
9#ifdef __i386__
10#include <sys/user.h>
11#endif
12#include "longjmp.h"
13#include "sysdep/ptrace_user.h"
14
15int save_fp_registers(int pid, unsigned long *fp_regs)
16{
17 if (ptrace(PTRACE_GETFPREGS, pid, 0, fp_regs) < 0)
18 return -errno;
19 return 0;
20}
21
22int restore_fp_registers(int pid, unsigned long *fp_regs)
23{
24 if (ptrace(PTRACE_SETFPREGS, pid, 0, fp_regs) < 0)
25 return -errno;
26 return 0;
27}
28
29#ifdef __i386__
30int have_fpx_regs = 1;
31int save_fpx_registers(int pid, unsigned long *fp_regs)
32{
33 if (ptrace(PTRACE_GETFPXREGS, pid, 0, fp_regs) < 0)
34 return -errno;
35 return 0;
36}
37
38int restore_fpx_registers(int pid, unsigned long *fp_regs)
39{
40 if (ptrace(PTRACE_SETFPXREGS, pid, 0, fp_regs) < 0)
41 return -errno;
42 return 0;
43}
44
45int get_fp_registers(int pid, unsigned long *regs)
46{
47 if (have_fpx_regs)
48 return save_fpx_registers(pid, regs);
49 else
50 return save_fp_registers(pid, regs);
51}
52
53int put_fp_registers(int pid, unsigned long *regs)
54{
55 if (have_fpx_regs)
56 return restore_fpx_registers(pid, regs);
57 else
58 return restore_fp_registers(pid, regs);
59}
60
61void arch_init_registers(int pid)
62{
63 struct user_fpxregs_struct fpx_regs;
64 int err;
65
66 err = ptrace(PTRACE_GETFPXREGS, pid, 0, &fpx_regs);
67 if (!err)
68 return;
69
70 if (errno != EIO)
71 panic("check_ptrace : PTRACE_GETFPXREGS failed, errno = %d",
72 errno);
73
74 have_fpx_regs = 0;
75}
76#else
77
78int get_fp_registers(int pid, unsigned long *regs)
79{
80 return save_fp_registers(pid, regs);
81}
82
83int put_fp_registers(int pid, unsigned long *regs)
84{
85 return restore_fp_registers(pid, regs);
86}
87
88#endif
89
90unsigned long get_thread_reg(int reg, jmp_buf *buf)
91{
92 switch (reg) {
93#ifdef __i386__
94 case HOST_IP:
95 return buf[0]->__eip;
96 case HOST_SP:
97 return buf[0]->__esp;
98 case HOST_BP:
99 return buf[0]->__ebp;
100#else
101 case HOST_IP:
102 return buf[0]->__rip;
103 case HOST_SP:
104 return buf[0]->__rsp;
105 case HOST_BP:
106 return buf[0]->__rbp;
107#endif
108 default:
109 printk(UM_KERN_ERR "get_thread_regs - unknown register %d\n",
110 reg);
111 return 0;
112 }
113}
diff --git a/arch/x86/um/os-Linux/task_size.c b/arch/x86/um/os-Linux/task_size.c
new file mode 100644
index 000000000000..efb16c5c9bcf
--- /dev/null
+++ b/arch/x86/um/os-Linux/task_size.c
@@ -0,0 +1,150 @@
1#include <stdio.h>
2#include <stdlib.h>
3#include <signal.h>
4#include <sys/mman.h>
5#include "longjmp.h"
6
7#ifdef __i386__
8
9static jmp_buf buf;
10
11static void segfault(int sig)
12{
13 longjmp(buf, 1);
14}
15
16static int page_ok(unsigned long page)
17{
18 unsigned long *address = (unsigned long *) (page << UM_KERN_PAGE_SHIFT);
19 unsigned long n = ~0UL;
20 void *mapped = NULL;
21 int ok = 0;
22
23 /*
24 * First see if the page is readable. If it is, it may still
25 * be a VDSO, so we go on to see if it's writable. If not
26 * then try mapping memory there. If that fails, then we're
27 * still in the kernel area. As a sanity check, we'll fail if
28 * the mmap succeeds, but gives us an address different from
29 * what we wanted.
30 */
31 if (setjmp(buf) == 0)
32 n = *address;
33 else {
34 mapped = mmap(address, UM_KERN_PAGE_SIZE,
35 PROT_READ | PROT_WRITE,
36 MAP_FIXED | MAP_PRIVATE | MAP_ANONYMOUS, -1, 0);
37 if (mapped == MAP_FAILED)
38 return 0;
39 if (mapped != address)
40 goto out;
41 }
42
43 /*
44 * Now, is it writeable? If so, then we're in user address
45 * space. If not, then try mprotecting it and try the write
46 * again.
47 */
48 if (setjmp(buf) == 0) {
49 *address = n;
50 ok = 1;
51 goto out;
52 } else if (mprotect(address, UM_KERN_PAGE_SIZE,
53 PROT_READ | PROT_WRITE) != 0)
54 goto out;
55
56 if (setjmp(buf) == 0) {
57 *address = n;
58 ok = 1;
59 }
60
61 out:
62 if (mapped != NULL)
63 munmap(mapped, UM_KERN_PAGE_SIZE);
64 return ok;
65}
66
67unsigned long os_get_top_address(void)
68{
69 struct sigaction sa, old;
70 unsigned long bottom = 0;
71 /*
72 * A 32-bit UML on a 64-bit host gets confused about the VDSO at
73 * 0xffffe000. It is mapped, is readable, can be reprotected writeable
74 * and written. However, exec discovers later that it can't be
75 * unmapped. So, just set the highest address to be checked to just
76 * below it. This might waste some address space on 4G/4G 32-bit
77 * hosts, but shouldn't hurt otherwise.
78 */
79 unsigned long top = 0xffffd000 >> UM_KERN_PAGE_SHIFT;
80 unsigned long test, original;
81
82 printf("Locating the bottom of the address space ... ");
83 fflush(stdout);
84
85 /*
86 * We're going to be longjmping out of the signal handler, so
87 * SA_DEFER needs to be set.
88 */
89 sa.sa_handler = segfault;
90 sigemptyset(&sa.sa_mask);
91 sa.sa_flags = SA_NODEFER;
92 if (sigaction(SIGSEGV, &sa, &old)) {
93 perror("os_get_top_address");
94 exit(1);
95 }
96
97 /* Manually scan the address space, bottom-up, until we find
98 * the first valid page (or run out of them).
99 */
100 for (bottom = 0; bottom < top; bottom++) {
101 if (page_ok(bottom))
102 break;
103 }
104
105 /* If we've got this far, we ran out of pages. */
106 if (bottom == top) {
107 fprintf(stderr, "Unable to determine bottom of address "
108 "space.\n");
109 exit(1);
110 }
111
112 printf("0x%x\n", bottom << UM_KERN_PAGE_SHIFT);
113 printf("Locating the top of the address space ... ");
114 fflush(stdout);
115
116 original = bottom;
117
118 /* This could happen with a 4G/4G split */
119 if (page_ok(top))
120 goto out;
121
122 do {
123 test = bottom + (top - bottom) / 2;
124 if (page_ok(test))
125 bottom = test;
126 else
127 top = test;
128 } while (top - bottom > 1);
129
130out:
131 /* Restore the old SIGSEGV handling */
132 if (sigaction(SIGSEGV, &old, NULL)) {
133 perror("os_get_top_address");
134 exit(1);
135 }
136 top <<= UM_KERN_PAGE_SHIFT;
137 printf("0x%x\n", top);
138
139 return top;
140}
141
142#else
143
144unsigned long os_get_top_address(void)
145{
146 /* The old value of CONFIG_TOP_ADDR */
147 return 0x7fc0000000;
148}
149
150#endif
diff --git a/arch/x86/um/os-Linux/tls.c b/arch/x86/um/os-Linux/tls.c
new file mode 100644
index 000000000000..82276b6071af
--- /dev/null
+++ b/arch/x86/um/os-Linux/tls.c
@@ -0,0 +1,67 @@
1#include <errno.h>
2#include <linux/unistd.h>
3
4#include <sys/ptrace.h>
5#include <sys/syscall.h>
6#include <unistd.h>
7
8#include "sysdep/tls.h"
9
10#ifndef PTRACE_GET_THREAD_AREA
11#define PTRACE_GET_THREAD_AREA 25
12#endif
13
14#ifndef PTRACE_SET_THREAD_AREA
15#define PTRACE_SET_THREAD_AREA 26
16#endif
17
18/* Checks whether host supports TLS, and sets *tls_min according to the value
19 * valid on the host.
20 * i386 host have it == 6; x86_64 host have it == 12, for i386 emulation. */
21void check_host_supports_tls(int *supports_tls, int *tls_min)
22{
23 /* Values for x86 and x86_64.*/
24 int val[] = {GDT_ENTRY_TLS_MIN_I386, GDT_ENTRY_TLS_MIN_X86_64};
25 int i;
26
27 for (i = 0; i < ARRAY_SIZE(val); i++) {
28 user_desc_t info;
29 info.entry_number = val[i];
30
31 if (syscall(__NR_get_thread_area, &info) == 0) {
32 *tls_min = val[i];
33 *supports_tls = 1;
34 return;
35 } else {
36 if (errno == EINVAL)
37 continue;
38 else if (errno == ENOSYS)
39 *supports_tls = 0;
40 return;
41 }
42 }
43
44 *supports_tls = 0;
45}
46
47int os_set_thread_area(user_desc_t *info, int pid)
48{
49 int ret;
50
51 ret = ptrace(PTRACE_SET_THREAD_AREA, pid, info->entry_number,
52 (unsigned long) info);
53 if (ret < 0)
54 ret = -errno;
55 return ret;
56}
57
58int os_get_thread_area(user_desc_t *info, int pid)
59{
60 int ret;
61
62 ret = ptrace(PTRACE_GET_THREAD_AREA, pid, info->entry_number,
63 (unsigned long) info);
64 if (ret < 0)
65 ret = -errno;
66 return ret;
67}
diff --git a/arch/x86/um/ptrace_32.c b/arch/x86/um/ptrace_32.c
new file mode 100644
index 000000000000..3b949daa095c
--- /dev/null
+++ b/arch/x86/um/ptrace_32.c
@@ -0,0 +1,273 @@
1/*
2 * Copyright (C) 2000 - 2007 Jeff Dike (jdike@{addtoit,linux.intel}.com)
3 * Licensed under the GPL
4 */
5
6#include "linux/mm.h"
7#include "linux/sched.h"
8#include "asm/uaccess.h"
9#include "skas.h"
10
11extern int arch_switch_tls(struct task_struct *to);
12
13void arch_switch_to(struct task_struct *to)
14{
15 int err = arch_switch_tls(to);
16 if (!err)
17 return;
18
19 if (err != -EINVAL)
20 printk(KERN_WARNING "arch_switch_tls failed, errno %d, "
21 "not EINVAL\n", -err);
22 else
23 printk(KERN_WARNING "arch_switch_tls failed, errno = EINVAL\n");
24}
25
26int is_syscall(unsigned long addr)
27{
28 unsigned short instr;
29 int n;
30
31 n = copy_from_user(&instr, (void __user *) addr, sizeof(instr));
32 if (n) {
33 /* access_process_vm() grants access to vsyscall and stub,
34 * while copy_from_user doesn't. Maybe access_process_vm is
35 * slow, but that doesn't matter, since it will be called only
36 * in case of singlestepping, if copy_from_user failed.
37 */
38 n = access_process_vm(current, addr, &instr, sizeof(instr), 0);
39 if (n != sizeof(instr)) {
40 printk(KERN_ERR "is_syscall : failed to read "
41 "instruction from 0x%lx\n", addr);
42 return 1;
43 }
44 }
45 /* int 0x80 or sysenter */
46 return (instr == 0x80cd) || (instr == 0x340f);
47}
48
49/* determines which flags the user has access to. */
50/* 1 = access 0 = no access */
51#define FLAG_MASK 0x00044dd5
52
53static const int reg_offsets[] = {
54 [EBX] = HOST_BX,
55 [ECX] = HOST_CX,
56 [EDX] = HOST_DX,
57 [ESI] = HOST_SI,
58 [EDI] = HOST_DI,
59 [EBP] = HOST_BP,
60 [EAX] = HOST_AX,
61 [DS] = HOST_DS,
62 [ES] = HOST_ES,
63 [FS] = HOST_FS,
64 [GS] = HOST_GS,
65 [EIP] = HOST_IP,
66 [CS] = HOST_CS,
67 [EFL] = HOST_EFLAGS,
68 [UESP] = HOST_SP,
69 [SS] = HOST_SS,
70};
71
72int putreg(struct task_struct *child, int regno, unsigned long value)
73{
74 regno >>= 2;
75 switch (regno) {
76 case EBX:
77 case ECX:
78 case EDX:
79 case ESI:
80 case EDI:
81 case EBP:
82 case EAX:
83 case EIP:
84 case UESP:
85 break;
86 case FS:
87 if (value && (value & 3) != 3)
88 return -EIO;
89 break;
90 case GS:
91 if (value && (value & 3) != 3)
92 return -EIO;
93 break;
94 case DS:
95 case ES:
96 if (value && (value & 3) != 3)
97 return -EIO;
98 value &= 0xffff;
99 break;
100 case SS:
101 case CS:
102 if ((value & 3) != 3)
103 return -EIO;
104 value &= 0xffff;
105 break;
106 case EFL:
107 value &= FLAG_MASK;
108 child->thread.regs.regs.gp[HOST_EFLAGS] |= value;
109 return 0;
110 case ORIG_EAX:
111 child->thread.regs.regs.syscall = value;
112 return 0;
113 default :
114 panic("Bad register in putreg() : %d\n", regno);
115 }
116 child->thread.regs.regs.gp[reg_offsets[regno]] = value;
117 return 0;
118}
119
120int poke_user(struct task_struct *child, long addr, long data)
121{
122 if ((addr & 3) || addr < 0)
123 return -EIO;
124
125 if (addr < MAX_REG_OFFSET)
126 return putreg(child, addr, data);
127 else if ((addr >= offsetof(struct user, u_debugreg[0])) &&
128 (addr <= offsetof(struct user, u_debugreg[7]))) {
129 addr -= offsetof(struct user, u_debugreg[0]);
130 addr = addr >> 2;
131 if ((addr == 4) || (addr == 5))
132 return -EIO;
133 child->thread.arch.debugregs[addr] = data;
134 return 0;
135 }
136 return -EIO;
137}
138
139unsigned long getreg(struct task_struct *child, int regno)
140{
141 unsigned long mask = ~0UL;
142
143 regno >>= 2;
144 switch (regno) {
145 case ORIG_EAX:
146 return child->thread.regs.regs.syscall;
147 case FS:
148 case GS:
149 case DS:
150 case ES:
151 case SS:
152 case CS:
153 mask = 0xffff;
154 break;
155 case EIP:
156 case UESP:
157 case EAX:
158 case EBX:
159 case ECX:
160 case EDX:
161 case ESI:
162 case EDI:
163 case EBP:
164 case EFL:
165 break;
166 default:
167 panic("Bad register in getreg() : %d\n", regno);
168 }
169 return mask & child->thread.regs.regs.gp[reg_offsets[regno]];
170}
171
172/* read the word at location addr in the USER area. */
173int peek_user(struct task_struct *child, long addr, long data)
174{
175 unsigned long tmp;
176
177 if ((addr & 3) || addr < 0)
178 return -EIO;
179
180 tmp = 0; /* Default return condition */
181 if (addr < MAX_REG_OFFSET) {
182 tmp = getreg(child, addr);
183 }
184 else if ((addr >= offsetof(struct user, u_debugreg[0])) &&
185 (addr <= offsetof(struct user, u_debugreg[7]))) {
186 addr -= offsetof(struct user, u_debugreg[0]);
187 addr = addr >> 2;
188 tmp = child->thread.arch.debugregs[addr];
189 }
190 return put_user(tmp, (unsigned long __user *) data);
191}
192
193static int get_fpregs(struct user_i387_struct __user *buf, struct task_struct *child)
194{
195 int err, n, cpu = ((struct thread_info *) child->stack)->cpu;
196 struct user_i387_struct fpregs;
197
198 err = save_fp_registers(userspace_pid[cpu], (unsigned long *) &fpregs);
199 if (err)
200 return err;
201
202 n = copy_to_user(buf, &fpregs, sizeof(fpregs));
203 if(n > 0)
204 return -EFAULT;
205
206 return n;
207}
208
209static int set_fpregs(struct user_i387_struct __user *buf, struct task_struct *child)
210{
211 int n, cpu = ((struct thread_info *) child->stack)->cpu;
212 struct user_i387_struct fpregs;
213
214 n = copy_from_user(&fpregs, buf, sizeof(fpregs));
215 if (n > 0)
216 return -EFAULT;
217
218 return restore_fp_registers(userspace_pid[cpu],
219 (unsigned long *) &fpregs);
220}
221
222static int get_fpxregs(struct user_fxsr_struct __user *buf, struct task_struct *child)
223{
224 int err, n, cpu = ((struct thread_info *) child->stack)->cpu;
225 struct user_fxsr_struct fpregs;
226
227 err = save_fpx_registers(userspace_pid[cpu], (unsigned long *) &fpregs);
228 if (err)
229 return err;
230
231 n = copy_to_user(buf, &fpregs, sizeof(fpregs));
232 if(n > 0)
233 return -EFAULT;
234
235 return n;
236}
237
238static int set_fpxregs(struct user_fxsr_struct __user *buf, struct task_struct *child)
239{
240 int n, cpu = ((struct thread_info *) child->stack)->cpu;
241 struct user_fxsr_struct fpregs;
242
243 n = copy_from_user(&fpregs, buf, sizeof(fpregs));
244 if (n > 0)
245 return -EFAULT;
246
247 return restore_fpx_registers(userspace_pid[cpu],
248 (unsigned long *) &fpregs);
249}
250
251long subarch_ptrace(struct task_struct *child, long request,
252 unsigned long addr, unsigned long data)
253{
254 int ret = -EIO;
255 void __user *datap = (void __user *) data;
256 switch (request) {
257 case PTRACE_GETFPREGS: /* Get the child FPU state. */
258 ret = get_fpregs(datap, child);
259 break;
260 case PTRACE_SETFPREGS: /* Set the child FPU state. */
261 ret = set_fpregs(datap, child);
262 break;
263 case PTRACE_GETFPXREGS: /* Get the child FPU state. */
264 ret = get_fpxregs(datap, child);
265 break;
266 case PTRACE_SETFPXREGS: /* Set the child FPU state. */
267 ret = set_fpxregs(datap, child);
268 break;
269 default:
270 ret = -EIO;
271 }
272 return ret;
273}
diff --git a/arch/x86/um/ptrace_64.c b/arch/x86/um/ptrace_64.c
new file mode 100644
index 000000000000..3b52bf0b418a
--- /dev/null
+++ b/arch/x86/um/ptrace_64.c
@@ -0,0 +1,271 @@
1/*
2 * Copyright 2003 PathScale, Inc.
3 * Copyright (C) 2003 - 2007 Jeff Dike (jdike@{addtoit,linux.intel}.com)
4 *
5 * Licensed under the GPL
6 */
7
8#include <linux/mm.h>
9#include <linux/sched.h>
10#include <linux/errno.h>
11#define __FRAME_OFFSETS
12#include <asm/ptrace.h>
13#include <asm/uaccess.h>
14
15/*
16 * determines which flags the user has access to.
17 * 1 = access 0 = no access
18 */
19#define FLAG_MASK 0x44dd5UL
20
21static const int reg_offsets[] =
22{
23 [R8 >> 3] = HOST_R8,
24 [R9 >> 3] = HOST_R9,
25 [R10 >> 3] = HOST_R10,
26 [R11 >> 3] = HOST_R11,
27 [R12 >> 3] = HOST_R12,
28 [R13 >> 3] = HOST_R13,
29 [R14 >> 3] = HOST_R14,
30 [R15 >> 3] = HOST_R15,
31 [RIP >> 3] = HOST_IP,
32 [RSP >> 3] = HOST_SP,
33 [RAX >> 3] = HOST_AX,
34 [RBX >> 3] = HOST_BX,
35 [RCX >> 3] = HOST_CX,
36 [RDX >> 3] = HOST_DX,
37 [RSI >> 3] = HOST_SI,
38 [RDI >> 3] = HOST_DI,
39 [RBP >> 3] = HOST_BP,
40 [CS >> 3] = HOST_CS,
41 [SS >> 3] = HOST_SS,
42 [FS_BASE >> 3] = HOST_FS_BASE,
43 [GS_BASE >> 3] = HOST_GS_BASE,
44 [DS >> 3] = HOST_DS,
45 [ES >> 3] = HOST_ES,
46 [FS >> 3] = HOST_FS,
47 [GS >> 3] = HOST_GS,
48 [EFLAGS >> 3] = HOST_EFLAGS,
49 [ORIG_RAX >> 3] = HOST_ORIG_AX,
50};
51
52int putreg(struct task_struct *child, int regno, unsigned long value)
53{
54#ifdef TIF_IA32
55 /*
56 * Some code in the 64bit emulation may not be 64bit clean.
57 * Don't take any chances.
58 */
59 if (test_tsk_thread_flag(child, TIF_IA32))
60 value &= 0xffffffff;
61#endif
62 switch (regno) {
63 case R8:
64 case R9:
65 case R10:
66 case R11:
67 case R12:
68 case R13:
69 case R14:
70 case R15:
71 case RIP:
72 case RSP:
73 case RAX:
74 case RBX:
75 case RCX:
76 case RDX:
77 case RSI:
78 case RDI:
79 case RBP:
80 case ORIG_RAX:
81 break;
82
83 case FS:
84 case GS:
85 case DS:
86 case ES:
87 case SS:
88 case CS:
89 if (value && (value & 3) != 3)
90 return -EIO;
91 value &= 0xffff;
92 break;
93
94 case FS_BASE:
95 case GS_BASE:
96 if (!((value >> 48) == 0 || (value >> 48) == 0xffff))
97 return -EIO;
98 break;
99
100 case EFLAGS:
101 value &= FLAG_MASK;
102 child->thread.regs.regs.gp[HOST_EFLAGS] |= value;
103 return 0;
104
105 default:
106 panic("Bad register in putreg(): %d\n", regno);
107 }
108
109 child->thread.regs.regs.gp[reg_offsets[regno >> 3]] = value;
110 return 0;
111}
112
113int poke_user(struct task_struct *child, long addr, long data)
114{
115 if ((addr & 3) || addr < 0)
116 return -EIO;
117
118 if (addr < MAX_REG_OFFSET)
119 return putreg(child, addr, data);
120 else if ((addr >= offsetof(struct user, u_debugreg[0])) &&
121 (addr <= offsetof(struct user, u_debugreg[7]))) {
122 addr -= offsetof(struct user, u_debugreg[0]);
123 addr = addr >> 2;
124 if ((addr == 4) || (addr == 5))
125 return -EIO;
126 child->thread.arch.debugregs[addr] = data;
127 return 0;
128 }
129 return -EIO;
130}
131
132unsigned long getreg(struct task_struct *child, int regno)
133{
134 unsigned long mask = ~0UL;
135#ifdef TIF_IA32
136 if (test_tsk_thread_flag(child, TIF_IA32))
137 mask = 0xffffffff;
138#endif
139 switch (regno) {
140 case R8:
141 case R9:
142 case R10:
143 case R11:
144 case R12:
145 case R13:
146 case R14:
147 case R15:
148 case RIP:
149 case RSP:
150 case RAX:
151 case RBX:
152 case RCX:
153 case RDX:
154 case RSI:
155 case RDI:
156 case RBP:
157 case ORIG_RAX:
158 case EFLAGS:
159 case FS_BASE:
160 case GS_BASE:
161 break;
162 case FS:
163 case GS:
164 case DS:
165 case ES:
166 case SS:
167 case CS:
168 mask = 0xffff;
169 break;
170 default:
171 panic("Bad register in getreg: %d\n", regno);
172 }
173 return mask & child->thread.regs.regs.gp[reg_offsets[regno >> 3]];
174}
175
176int peek_user(struct task_struct *child, long addr, long data)
177{
178 /* read the word at location addr in the USER area. */
179 unsigned long tmp;
180
181 if ((addr & 3) || addr < 0)
182 return -EIO;
183
184 tmp = 0; /* Default return condition */
185 if (addr < MAX_REG_OFFSET)
186 tmp = getreg(child, addr);
187 else if ((addr >= offsetof(struct user, u_debugreg[0])) &&
188 (addr <= offsetof(struct user, u_debugreg[7]))) {
189 addr -= offsetof(struct user, u_debugreg[0]);
190 addr = addr >> 2;
191 tmp = child->thread.arch.debugregs[addr];
192 }
193 return put_user(tmp, (unsigned long *) data);
194}
195
196/* XXX Mostly copied from sys-i386 */
197int is_syscall(unsigned long addr)
198{
199 unsigned short instr;
200 int n;
201
202 n = copy_from_user(&instr, (void __user *) addr, sizeof(instr));
203 if (n) {
204 /*
205 * access_process_vm() grants access to vsyscall and stub,
206 * while copy_from_user doesn't. Maybe access_process_vm is
207 * slow, but that doesn't matter, since it will be called only
208 * in case of singlestepping, if copy_from_user failed.
209 */
210 n = access_process_vm(current, addr, &instr, sizeof(instr), 0);
211 if (n != sizeof(instr)) {
212 printk("is_syscall : failed to read instruction from "
213 "0x%lx\n", addr);
214 return 1;
215 }
216 }
217 /* sysenter */
218 return instr == 0x050f;
219}
220
221static int get_fpregs(struct user_i387_struct __user *buf, struct task_struct *child)
222{
223 int err, n, cpu = ((struct thread_info *) child->stack)->cpu;
224 long fpregs[HOST_FP_SIZE];
225
226 BUG_ON(sizeof(*buf) != sizeof(fpregs));
227 err = save_fp_registers(userspace_pid[cpu], fpregs);
228 if (err)
229 return err;
230
231 n = copy_to_user(buf, fpregs, sizeof(fpregs));
232 if (n > 0)
233 return -EFAULT;
234
235 return n;
236}
237
238static int set_fpregs(struct user_i387_struct __user *buf, struct task_struct *child)
239{
240 int n, cpu = ((struct thread_info *) child->stack)->cpu;
241 long fpregs[HOST_FP_SIZE];
242
243 BUG_ON(sizeof(*buf) != sizeof(fpregs));
244 n = copy_from_user(fpregs, buf, sizeof(fpregs));
245 if (n > 0)
246 return -EFAULT;
247
248 return restore_fp_registers(userspace_pid[cpu], fpregs);
249}
250
251long subarch_ptrace(struct task_struct *child, long request,
252 unsigned long addr, unsigned long data)
253{
254 int ret = -EIO;
255 void __user *datap = (void __user *) data;
256
257 switch (request) {
258 case PTRACE_GETFPREGS: /* Get the child FPU state. */
259 ret = get_fpregs(datap, child);
260 break;
261 case PTRACE_SETFPREGS: /* Set the child FPU state. */
262 ret = set_fpregs(datap, child);
263 break;
264 case PTRACE_ARCH_PRCTL:
265 /* XXX Calls ptrace on the host - needs some SMP thinking */
266 ret = arch_prctl(child, data, (void __user *) addr);
267 break;
268 }
269
270 return ret;
271}
diff --git a/arch/x86/um/ptrace_user.c b/arch/x86/um/ptrace_user.c
new file mode 100644
index 000000000000..3960ca1dd35a
--- /dev/null
+++ b/arch/x86/um/ptrace_user.c
@@ -0,0 +1,21 @@
1/*
2 * Copyright (C) 2002 - 2007 Jeff Dike (jdike@{addtoit,linux.intel}.com)
3 * Licensed under the GPL
4 */
5
6#include <errno.h>
7#include "ptrace_user.h"
8
9int ptrace_getregs(long pid, unsigned long *regs_out)
10{
11 if (ptrace(PTRACE_GETREGS, pid, 0, regs_out) < 0)
12 return -errno;
13 return 0;
14}
15
16int ptrace_setregs(long pid, unsigned long *regs)
17{
18 if (ptrace(PTRACE_SETREGS, pid, 0, regs) < 0)
19 return -errno;
20 return 0;
21}
diff --git a/arch/x86/um/setjmp_32.S b/arch/x86/um/setjmp_32.S
new file mode 100644
index 000000000000..b766792c9933
--- /dev/null
+++ b/arch/x86/um/setjmp_32.S
@@ -0,0 +1,58 @@
1#
2# arch/i386/setjmp.S
3#
4# setjmp/longjmp for the i386 architecture
5#
6
7#
8# The jmp_buf is assumed to contain the following, in order:
9# %ebx
10# %esp
11# %ebp
12# %esi
13# %edi
14# <return address>
15#
16
17 .text
18 .align 4
19 .globl setjmp
20 .type setjmp, @function
21setjmp:
22#ifdef _REGPARM
23 movl %eax,%edx
24#else
25 movl 4(%esp),%edx
26#endif
27 popl %ecx # Return address, and adjust the stack
28 xorl %eax,%eax # Return value
29 movl %ebx,(%edx)
30 movl %esp,4(%edx) # Post-return %esp!
31 pushl %ecx # Make the call/return stack happy
32 movl %ebp,8(%edx)
33 movl %esi,12(%edx)
34 movl %edi,16(%edx)
35 movl %ecx,20(%edx) # Return address
36 ret
37
38 .size setjmp,.-setjmp
39
40 .text
41 .align 4
42 .globl longjmp
43 .type longjmp, @function
44longjmp:
45#ifdef _REGPARM
46 xchgl %eax,%edx
47#else
48 movl 4(%esp),%edx # jmp_ptr address
49 movl 8(%esp),%eax # Return value
50#endif
51 movl (%edx),%ebx
52 movl 4(%edx),%esp
53 movl 8(%edx),%ebp
54 movl 12(%edx),%esi
55 movl 16(%edx),%edi
56 jmp *20(%edx)
57
58 .size longjmp,.-longjmp
diff --git a/arch/x86/um/setjmp_64.S b/arch/x86/um/setjmp_64.S
new file mode 100644
index 000000000000..45f547b4043e
--- /dev/null
+++ b/arch/x86/um/setjmp_64.S
@@ -0,0 +1,54 @@
1#
2# arch/x86_64/setjmp.S
3#
4# setjmp/longjmp for the x86-64 architecture
5#
6
7#
8# The jmp_buf is assumed to contain the following, in order:
9# %rbx
10# %rsp (post-return)
11# %rbp
12# %r12
13# %r13
14# %r14
15# %r15
16# <return address>
17#
18
19 .text
20 .align 4
21 .globl setjmp
22 .type setjmp, @function
23setjmp:
24 pop %rsi # Return address, and adjust the stack
25 xorl %eax,%eax # Return value
26 movq %rbx,(%rdi)
27 movq %rsp,8(%rdi) # Post-return %rsp!
28 push %rsi # Make the call/return stack happy
29 movq %rbp,16(%rdi)
30 movq %r12,24(%rdi)
31 movq %r13,32(%rdi)
32 movq %r14,40(%rdi)
33 movq %r15,48(%rdi)
34 movq %rsi,56(%rdi) # Return address
35 ret
36
37 .size setjmp,.-setjmp
38
39 .text
40 .align 4
41 .globl longjmp
42 .type longjmp, @function
43longjmp:
44 movl %esi,%eax # Return value (int)
45 movq (%rdi),%rbx
46 movq 8(%rdi),%rsp
47 movq 16(%rdi),%rbp
48 movq 24(%rdi),%r12
49 movq 32(%rdi),%r13
50 movq 40(%rdi),%r14
51 movq 48(%rdi),%r15
52 jmp *56(%rdi)
53
54 .size longjmp,.-longjmp
diff --git a/arch/x86/um/shared/sysdep/archsetjmp.h b/arch/x86/um/shared/sysdep/archsetjmp.h
new file mode 100644
index 000000000000..ff7766d28226
--- /dev/null
+++ b/arch/x86/um/shared/sysdep/archsetjmp.h
@@ -0,0 +1,5 @@
1#ifdef __i386__
2#include "archsetjmp_32.h"
3#else
4#include "archsetjmp_64.h"
5#endif
diff --git a/arch/x86/um/shared/sysdep/archsetjmp_32.h b/arch/x86/um/shared/sysdep/archsetjmp_32.h
new file mode 100644
index 000000000000..0f312085ce1d
--- /dev/null
+++ b/arch/x86/um/shared/sysdep/archsetjmp_32.h
@@ -0,0 +1,22 @@
1/*
2 * arch/um/include/sysdep-i386/archsetjmp.h
3 */
4
5#ifndef _KLIBC_ARCHSETJMP_H
6#define _KLIBC_ARCHSETJMP_H
7
8struct __jmp_buf {
9 unsigned int __ebx;
10 unsigned int __esp;
11 unsigned int __ebp;
12 unsigned int __esi;
13 unsigned int __edi;
14 unsigned int __eip;
15};
16
17typedef struct __jmp_buf jmp_buf[1];
18
19#define JB_IP __eip
20#define JB_SP __esp
21
22#endif /* _SETJMP_H */
diff --git a/arch/x86/um/shared/sysdep/archsetjmp_64.h b/arch/x86/um/shared/sysdep/archsetjmp_64.h
new file mode 100644
index 000000000000..2af8f12ca161
--- /dev/null
+++ b/arch/x86/um/shared/sysdep/archsetjmp_64.h
@@ -0,0 +1,24 @@
1/*
2 * arch/um/include/sysdep-x86_64/archsetjmp.h
3 */
4
5#ifndef _KLIBC_ARCHSETJMP_H
6#define _KLIBC_ARCHSETJMP_H
7
8struct __jmp_buf {
9 unsigned long __rbx;
10 unsigned long __rsp;
11 unsigned long __rbp;
12 unsigned long __r12;
13 unsigned long __r13;
14 unsigned long __r14;
15 unsigned long __r15;
16 unsigned long __rip;
17};
18
19typedef struct __jmp_buf jmp_buf[1];
20
21#define JB_IP __rip
22#define JB_SP __rsp
23
24#endif /* _SETJMP_H */
diff --git a/arch/x86/um/shared/sysdep/faultinfo.h b/arch/x86/um/shared/sysdep/faultinfo.h
new file mode 100644
index 000000000000..862ecb1c7781
--- /dev/null
+++ b/arch/x86/um/shared/sysdep/faultinfo.h
@@ -0,0 +1,5 @@
1#ifdef __i386__
2#include "faultinfo_32.h"
3#else
4#include "faultinfo_64.h"
5#endif
diff --git a/arch/x86/um/shared/sysdep/faultinfo_32.h b/arch/x86/um/shared/sysdep/faultinfo_32.h
new file mode 100644
index 000000000000..a26086b8a800
--- /dev/null
+++ b/arch/x86/um/shared/sysdep/faultinfo_32.h
@@ -0,0 +1,35 @@
1/*
2 * Copyright (C) 2004 Fujitsu Siemens Computers GmbH
3 * Author: Bodo Stroesser <bstroesser@fujitsu-siemens.com>
4 * Licensed under the GPL
5 */
6
7#ifndef __FAULTINFO_I386_H
8#define __FAULTINFO_I386_H
9
10/* this structure contains the full arch-specific faultinfo
11 * from the traps.
12 * On i386, ptrace_faultinfo unfortunately doesn't provide
13 * all the info, since trap_no is missing.
14 * All common elements are defined at the same position in
15 * both structures, thus making it easy to copy the
16 * contents without knowledge about the structure elements.
17 */
18struct faultinfo {
19 int error_code; /* in ptrace_faultinfo misleadingly called is_write */
20 unsigned long cr2; /* in ptrace_faultinfo called addr */
21 int trap_no; /* missing in ptrace_faultinfo */
22};
23
24#define FAULT_WRITE(fi) ((fi).error_code & 2)
25#define FAULT_ADDRESS(fi) ((fi).cr2)
26
27/* This is Page Fault */
28#define SEGV_IS_FIXABLE(fi) ((fi)->trap_no == 14)
29
30/* SKAS3 has no trap_no on i386, but get_skas_faultinfo() sets it to 0. */
31#define SEGV_MAYBE_FIXABLE(fi) ((fi)->trap_no == 0 && ptrace_faultinfo)
32
33#define PTRACE_FULL_FAULTINFO 0
34
35#endif
diff --git a/arch/x86/um/shared/sysdep/faultinfo_64.h b/arch/x86/um/shared/sysdep/faultinfo_64.h
new file mode 100644
index 000000000000..f811cbe15d62
--- /dev/null
+++ b/arch/x86/um/shared/sysdep/faultinfo_64.h
@@ -0,0 +1,35 @@
1/*
2 * Copyright (C) 2004 Fujitsu Siemens Computers GmbH
3 * Author: Bodo Stroesser <bstroesser@fujitsu-siemens.com>
4 * Licensed under the GPL
5 */
6
7#ifndef __FAULTINFO_X86_64_H
8#define __FAULTINFO_X86_64_H
9
10/* this structure contains the full arch-specific faultinfo
11 * from the traps.
12 * On i386, ptrace_faultinfo unfortunately doesn't provide
13 * all the info, since trap_no is missing.
14 * All common elements are defined at the same position in
15 * both structures, thus making it easy to copy the
16 * contents without knowledge about the structure elements.
17 */
18struct faultinfo {
19 int error_code; /* in ptrace_faultinfo misleadingly called is_write */
20 unsigned long cr2; /* in ptrace_faultinfo called addr */
21 int trap_no; /* missing in ptrace_faultinfo */
22};
23
24#define FAULT_WRITE(fi) ((fi).error_code & 2)
25#define FAULT_ADDRESS(fi) ((fi).cr2)
26
27/* This is Page Fault */
28#define SEGV_IS_FIXABLE(fi) ((fi)->trap_no == 14)
29
30/* No broken SKAS API, which doesn't pass trap_no, here. */
31#define SEGV_MAYBE_FIXABLE(fi) 0
32
33#define PTRACE_FULL_FAULTINFO 1
34
35#endif
diff --git a/arch/x86/um/shared/sysdep/kernel-offsets.h b/arch/x86/um/shared/sysdep/kernel-offsets.h
new file mode 100644
index 000000000000..5868526b5eef
--- /dev/null
+++ b/arch/x86/um/shared/sysdep/kernel-offsets.h
@@ -0,0 +1,21 @@
1#include <linux/stddef.h>
2#include <linux/sched.h>
3#include <linux/elf.h>
4#include <linux/crypto.h>
5#include <asm/mman.h>
6
7#define DEFINE(sym, val) \
8 asm volatile("\n->" #sym " %0 " #val : : "i" (val))
9
10#define STR(x) #x
11#define DEFINE_STR(sym, val) asm volatile("\n->" #sym " " STR(val) " " #val: : )
12
13#define BLANK() asm volatile("\n->" : : )
14
15#define OFFSET(sym, str, mem) \
16 DEFINE(sym, offsetof(struct str, mem));
17
18void foo(void)
19{
20#include <common-offsets.h>
21}
diff --git a/arch/x86/um/shared/sysdep/mcontext.h b/arch/x86/um/shared/sysdep/mcontext.h
new file mode 100644
index 000000000000..b724c54da316
--- /dev/null
+++ b/arch/x86/um/shared/sysdep/mcontext.h
@@ -0,0 +1,31 @@
1/*
2 * Copyright (C) 2000 - 2007 Jeff Dike (jdike@{addtoit,linux.intel}.com)
3 * Licensed under the GPL
4 */
5
6#ifndef __SYS_SIGCONTEXT_X86_H
7#define __SYS_SIGCONTEXT_X86_H
8
9extern void get_regs_from_mc(struct uml_pt_regs *, mcontext_t *);
10
11#ifdef __i386__
12
13#define GET_FAULTINFO_FROM_MC(fi, mc) \
14 { \
15 (fi).cr2 = (mc)->cr2; \
16 (fi).error_code = (mc)->gregs[REG_ERR]; \
17 (fi).trap_no = (mc)->gregs[REG_TRAPNO]; \
18 }
19
20#else
21
22#define GET_FAULTINFO_FROM_MC(fi, mc) \
23 { \
24 (fi).cr2 = (mc)->gregs[REG_CR2]; \
25 (fi).error_code = (mc)->gregs[REG_ERR]; \
26 (fi).trap_no = (mc)->gregs[REG_TRAPNO]; \
27 }
28
29#endif
30
31#endif
diff --git a/arch/x86/um/shared/sysdep/ptrace.h b/arch/x86/um/shared/sysdep/ptrace.h
new file mode 100644
index 000000000000..711b1621747f
--- /dev/null
+++ b/arch/x86/um/shared/sysdep/ptrace.h
@@ -0,0 +1,5 @@
1#ifdef __i386__
2#include "ptrace_32.h"
3#else
4#include "ptrace_64.h"
5#endif
diff --git a/arch/x86/um/shared/sysdep/ptrace_32.h b/arch/x86/um/shared/sysdep/ptrace_32.h
new file mode 100644
index 000000000000..befd1df32ed0
--- /dev/null
+++ b/arch/x86/um/shared/sysdep/ptrace_32.h
@@ -0,0 +1,114 @@
1/*
2 * Copyright (C) 2000 - 2007 Jeff Dike (jdike@{addtoit,linux.intel}.com)
3 * Licensed under the GPL
4 */
5
6#ifndef __SYSDEP_I386_PTRACE_H
7#define __SYSDEP_I386_PTRACE_H
8
9#include <generated/user_constants.h>
10#include "sysdep/faultinfo.h"
11
12#define MAX_REG_NR (UM_FRAME_SIZE / sizeof(unsigned long))
13#define MAX_REG_OFFSET (UM_FRAME_SIZE)
14
15static inline void update_debugregs(int seq) {}
16
17/* syscall emulation path in ptrace */
18
19#ifndef PTRACE_SYSEMU
20#define PTRACE_SYSEMU 31
21#endif
22
23void set_using_sysemu(int value);
24int get_using_sysemu(void);
25extern int sysemu_supported;
26
27#define REGS_IP(r) ((r)[HOST_IP])
28#define REGS_SP(r) ((r)[HOST_SP])
29#define REGS_EFLAGS(r) ((r)[HOST_EFLAGS])
30#define REGS_EAX(r) ((r)[HOST_AX])
31#define REGS_EBX(r) ((r)[HOST_BX])
32#define REGS_ECX(r) ((r)[HOST_CX])
33#define REGS_EDX(r) ((r)[HOST_DX])
34#define REGS_ESI(r) ((r)[HOST_SI])
35#define REGS_EDI(r) ((r)[HOST_DI])
36#define REGS_EBP(r) ((r)[HOST_BP])
37#define REGS_CS(r) ((r)[HOST_CS])
38#define REGS_SS(r) ((r)[HOST_SS])
39#define REGS_DS(r) ((r)[HOST_DS])
40#define REGS_ES(r) ((r)[HOST_ES])
41#define REGS_FS(r) ((r)[HOST_FS])
42#define REGS_GS(r) ((r)[HOST_GS])
43
44#define REGS_SET_SYSCALL_RETURN(r, res) REGS_EAX(r) = (res)
45
46#define IP_RESTART_SYSCALL(ip) ((ip) -= 2)
47#define REGS_RESTART_SYSCALL(r) IP_RESTART_SYSCALL(REGS_IP(r))
48
49#ifndef PTRACE_SYSEMU_SINGLESTEP
50#define PTRACE_SYSEMU_SINGLESTEP 32
51#endif
52
53struct uml_pt_regs {
54 unsigned long gp[MAX_REG_NR];
55 unsigned long fp[HOST_FPX_SIZE];
56 struct faultinfo faultinfo;
57 long syscall;
58 int is_user;
59};
60
61#define EMPTY_UML_PT_REGS { }
62
63#define UPT_IP(r) REGS_IP((r)->gp)
64#define UPT_SP(r) REGS_SP((r)->gp)
65#define UPT_EFLAGS(r) REGS_EFLAGS((r)->gp)
66#define UPT_EAX(r) REGS_EAX((r)->gp)
67#define UPT_EBX(r) REGS_EBX((r)->gp)
68#define UPT_ECX(r) REGS_ECX((r)->gp)
69#define UPT_EDX(r) REGS_EDX((r)->gp)
70#define UPT_ESI(r) REGS_ESI((r)->gp)
71#define UPT_EDI(r) REGS_EDI((r)->gp)
72#define UPT_EBP(r) REGS_EBP((r)->gp)
73#define UPT_ORIG_EAX(r) ((r)->syscall)
74#define UPT_CS(r) REGS_CS((r)->gp)
75#define UPT_SS(r) REGS_SS((r)->gp)
76#define UPT_DS(r) REGS_DS((r)->gp)
77#define UPT_ES(r) REGS_ES((r)->gp)
78#define UPT_FS(r) REGS_FS((r)->gp)
79#define UPT_GS(r) REGS_GS((r)->gp)
80
81#define UPT_SYSCALL_ARG1(r) UPT_EBX(r)
82#define UPT_SYSCALL_ARG2(r) UPT_ECX(r)
83#define UPT_SYSCALL_ARG3(r) UPT_EDX(r)
84#define UPT_SYSCALL_ARG4(r) UPT_ESI(r)
85#define UPT_SYSCALL_ARG5(r) UPT_EDI(r)
86#define UPT_SYSCALL_ARG6(r) UPT_EBP(r)
87
88extern int user_context(unsigned long sp);
89
90#define UPT_IS_USER(r) ((r)->is_user)
91
92struct syscall_args {
93 unsigned long args[6];
94};
95
96#define SYSCALL_ARGS(r) ((struct syscall_args) \
97 { .args = { UPT_SYSCALL_ARG1(r), \
98 UPT_SYSCALL_ARG2(r), \
99 UPT_SYSCALL_ARG3(r), \
100 UPT_SYSCALL_ARG4(r), \
101 UPT_SYSCALL_ARG5(r), \
102 UPT_SYSCALL_ARG6(r) } } )
103
104#define UPT_RESTART_SYSCALL(r) REGS_RESTART_SYSCALL((r)->gp)
105
106#define UPT_ORIG_SYSCALL(r) UPT_EAX(r)
107#define UPT_SYSCALL_NR(r) UPT_ORIG_EAX(r)
108#define UPT_SYSCALL_RET(r) UPT_EAX(r)
109
110#define UPT_FAULTINFO(r) (&(r)->faultinfo)
111
112extern void arch_init_registers(int pid);
113
114#endif
diff --git a/arch/x86/um/shared/sysdep/ptrace_64.h b/arch/x86/um/shared/sysdep/ptrace_64.h
new file mode 100644
index 000000000000..031edc53ac57
--- /dev/null
+++ b/arch/x86/um/shared/sysdep/ptrace_64.h
@@ -0,0 +1,157 @@
1/*
2 * Copyright 2003 PathScale, Inc.
3 * Copyright (C) 2003 - 2007 Jeff Dike (jdike@{addtoit,linux.intel}.com)
4 *
5 * Licensed under the GPL
6 */
7
8#ifndef __SYSDEP_X86_64_PTRACE_H
9#define __SYSDEP_X86_64_PTRACE_H
10
11#include <generated/user_constants.h>
12#include "sysdep/faultinfo.h"
13
14#define MAX_REG_OFFSET (UM_FRAME_SIZE)
15#define MAX_REG_NR ((MAX_REG_OFFSET) / sizeof(unsigned long))
16
17#define REGS_IP(r) ((r)[HOST_IP])
18#define REGS_SP(r) ((r)[HOST_SP])
19
20#define REGS_RBX(r) ((r)[HOST_BX])
21#define REGS_RCX(r) ((r)[HOST_CX])
22#define REGS_RDX(r) ((r)[HOST_DX])
23#define REGS_RSI(r) ((r)[HOST_SI])
24#define REGS_RDI(r) ((r)[HOST_DI])
25#define REGS_RBP(r) ((r)[HOST_BP])
26#define REGS_RAX(r) ((r)[HOST_AX])
27#define REGS_R8(r) ((r)[HOST_R8])
28#define REGS_R9(r) ((r)[HOST_R9])
29#define REGS_R10(r) ((r)[HOST_R10])
30#define REGS_R11(r) ((r)[HOST_R11])
31#define REGS_R12(r) ((r)[HOST_R12])
32#define REGS_R13(r) ((r)[HOST_R13])
33#define REGS_R14(r) ((r)[HOST_R14])
34#define REGS_R15(r) ((r)[HOST_R15])
35#define REGS_CS(r) ((r)[HOST_CS])
36#define REGS_EFLAGS(r) ((r)[HOST_EFLAGS])
37#define REGS_SS(r) ((r)[HOST_SS])
38
39#define HOST_FS_BASE 21
40#define HOST_GS_BASE 22
41#define HOST_DS 23
42#define HOST_ES 24
43#define HOST_FS 25
44#define HOST_GS 26
45
46/* Also defined in asm/ptrace-x86_64.h, but not in libc headers. So, these
47 * are already defined for kernel code, but not for userspace code.
48 */
49#ifndef FS_BASE
50/* These aren't defined in ptrace.h, but exist in struct user_regs_struct,
51 * which is what x86_64 ptrace actually uses.
52 */
53#define FS_BASE (HOST_FS_BASE * sizeof(long))
54#define GS_BASE (HOST_GS_BASE * sizeof(long))
55#define DS (HOST_DS * sizeof(long))
56#define ES (HOST_ES * sizeof(long))
57#define FS (HOST_FS * sizeof(long))
58#define GS (HOST_GS * sizeof(long))
59#endif
60
61#define REGS_FS_BASE(r) ((r)[HOST_FS_BASE])
62#define REGS_GS_BASE(r) ((r)[HOST_GS_BASE])
63#define REGS_DS(r) ((r)[HOST_DS])
64#define REGS_ES(r) ((r)[HOST_ES])
65#define REGS_FS(r) ((r)[HOST_FS])
66#define REGS_GS(r) ((r)[HOST_GS])
67
68#define REGS_ORIG_RAX(r) ((r)[HOST_ORIG_AX])
69
70#define REGS_SET_SYSCALL_RETURN(r, res) REGS_RAX(r) = (res)
71
72#define IP_RESTART_SYSCALL(ip) ((ip) -= 2)
73#define REGS_RESTART_SYSCALL(r) IP_RESTART_SYSCALL(REGS_IP(r))
74
75#define REGS_FAULT_ADDR(r) ((r)->fault_addr)
76
77#define REGS_FAULT_WRITE(r) FAULT_WRITE((r)->fault_type)
78
79#define REGS_TRAP(r) ((r)->trap_type)
80
81#define REGS_ERR(r) ((r)->fault_type)
82
83struct uml_pt_regs {
84 unsigned long gp[MAX_REG_NR];
85 unsigned long fp[HOST_FP_SIZE];
86 struct faultinfo faultinfo;
87 long syscall;
88 int is_user;
89};
90
91#define EMPTY_UML_PT_REGS { }
92
93#define UPT_RBX(r) REGS_RBX((r)->gp)
94#define UPT_RCX(r) REGS_RCX((r)->gp)
95#define UPT_RDX(r) REGS_RDX((r)->gp)
96#define UPT_RSI(r) REGS_RSI((r)->gp)
97#define UPT_RDI(r) REGS_RDI((r)->gp)
98#define UPT_RBP(r) REGS_RBP((r)->gp)
99#define UPT_RAX(r) REGS_RAX((r)->gp)
100#define UPT_R8(r) REGS_R8((r)->gp)
101#define UPT_R9(r) REGS_R9((r)->gp)
102#define UPT_R10(r) REGS_R10((r)->gp)
103#define UPT_R11(r) REGS_R11((r)->gp)
104#define UPT_R12(r) REGS_R12((r)->gp)
105#define UPT_R13(r) REGS_R13((r)->gp)
106#define UPT_R14(r) REGS_R14((r)->gp)
107#define UPT_R15(r) REGS_R15((r)->gp)
108#define UPT_CS(r) REGS_CS((r)->gp)
109#define UPT_FS_BASE(r) REGS_FS_BASE((r)->gp)
110#define UPT_FS(r) REGS_FS((r)->gp)
111#define UPT_GS_BASE(r) REGS_GS_BASE((r)->gp)
112#define UPT_GS(r) REGS_GS((r)->gp)
113#define UPT_DS(r) REGS_DS((r)->gp)
114#define UPT_ES(r) REGS_ES((r)->gp)
115#define UPT_CS(r) REGS_CS((r)->gp)
116#define UPT_SS(r) REGS_SS((r)->gp)
117#define UPT_ORIG_RAX(r) REGS_ORIG_RAX((r)->gp)
118
119#define UPT_IP(r) REGS_IP((r)->gp)
120#define UPT_SP(r) REGS_SP((r)->gp)
121
122#define UPT_EFLAGS(r) REGS_EFLAGS((r)->gp)
123#define UPT_SYSCALL_NR(r) ((r)->syscall)
124#define UPT_SYSCALL_RET(r) UPT_RAX(r)
125
126extern int user_context(unsigned long sp);
127
128#define UPT_IS_USER(r) ((r)->is_user)
129
130#define UPT_SYSCALL_ARG1(r) UPT_RDI(r)
131#define UPT_SYSCALL_ARG2(r) UPT_RSI(r)
132#define UPT_SYSCALL_ARG3(r) UPT_RDX(r)
133#define UPT_SYSCALL_ARG4(r) UPT_R10(r)
134#define UPT_SYSCALL_ARG5(r) UPT_R8(r)
135#define UPT_SYSCALL_ARG6(r) UPT_R9(r)
136
137struct syscall_args {
138 unsigned long args[6];
139};
140
141#define SYSCALL_ARGS(r) ((struct syscall_args) \
142 { .args = { UPT_SYSCALL_ARG1(r), \
143 UPT_SYSCALL_ARG2(r), \
144 UPT_SYSCALL_ARG3(r), \
145 UPT_SYSCALL_ARG4(r), \
146 UPT_SYSCALL_ARG5(r), \
147 UPT_SYSCALL_ARG6(r) } } )
148
149#define UPT_RESTART_SYSCALL(r) REGS_RESTART_SYSCALL((r)->gp)
150
151#define UPT_FAULTINFO(r) (&(r)->faultinfo)
152
153static inline void arch_init_registers(int pid)
154{
155}
156
157#endif
diff --git a/arch/x86/um/shared/sysdep/ptrace_user.h b/arch/x86/um/shared/sysdep/ptrace_user.h
new file mode 100644
index 000000000000..16cd6b5e71f7
--- /dev/null
+++ b/arch/x86/um/shared/sysdep/ptrace_user.h
@@ -0,0 +1,27 @@
1#include <generated/user_constants.h>
2
3#define PT_OFFSET(r) ((r) * sizeof(long))
4
5#define PT_SYSCALL_NR(regs) ((regs)[HOST_ORIG_AX])
6#define PT_SYSCALL_NR_OFFSET PT_OFFSET(HOST_ORIG_AX)
7
8#define PT_SYSCALL_RET_OFFSET PT_OFFSET(HOST_AX)
9
10#define REGS_IP_INDEX HOST_IP
11#define REGS_SP_INDEX HOST_SP
12
13#ifdef __i386__
14#define FP_SIZE ((HOST_FPX_SIZE > HOST_FP_SIZE) ? HOST_FPX_SIZE : HOST_FP_SIZE)
15#else
16#define FP_SIZE HOST_FP_SIZE
17
18/*
19 * x86_64 FC3 doesn't define this in /usr/include/linux/ptrace.h even though
20 * it's defined in the kernel's include/linux/ptrace.h. Additionally, use the
21 * 2.4 name and value for 2.4 host compatibility.
22 */
23#ifndef PTRACE_OLDSETOPTIONS
24#define PTRACE_OLDSETOPTIONS 21
25#endif
26
27#endif
diff --git a/arch/x86/um/shared/sysdep/skas_ptrace.h b/arch/x86/um/shared/sysdep/skas_ptrace.h
new file mode 100644
index 000000000000..453febe98993
--- /dev/null
+++ b/arch/x86/um/shared/sysdep/skas_ptrace.h
@@ -0,0 +1,22 @@
1/*
2 * Copyright (C) 2000, 2001, 2002 Jeff Dike (jdike@karaya.com)
3 * Licensed under the GPL
4 */
5
6#ifndef __SYSDEP_X86_SKAS_PTRACE_H
7#define __SYSDEP_X86_SKAS_PTRACE_H
8
9struct ptrace_faultinfo {
10 int is_write;
11 unsigned long addr;
12};
13
14struct ptrace_ldt {
15 int func;
16 void *ptr;
17 unsigned long bytecount;
18};
19
20#define PTRACE_LDT 54
21
22#endif
diff --git a/arch/x86/um/shared/sysdep/stub.h b/arch/x86/um/shared/sysdep/stub.h
new file mode 100644
index 000000000000..bd161e300102
--- /dev/null
+++ b/arch/x86/um/shared/sysdep/stub.h
@@ -0,0 +1,14 @@
1#include <asm/unistd.h>
2#include <sys/mman.h>
3#include <signal.h>
4#include "as-layout.h"
5#include "stub-data.h"
6
7#ifdef __i386__
8#include "stub_32.h"
9#else
10#include "stub_64.h"
11#endif
12
13extern void stub_segv_handler(int, siginfo_t *, void *);
14extern void stub_clone_handler(void);
diff --git a/arch/x86/um/shared/sysdep/stub_32.h b/arch/x86/um/shared/sysdep/stub_32.h
new file mode 100644
index 000000000000..51fd256c75f0
--- /dev/null
+++ b/arch/x86/um/shared/sysdep/stub_32.h
@@ -0,0 +1,93 @@
1/*
2 * Copyright (C) 2004 Jeff Dike (jdike@addtoit.com)
3 * Licensed under the GPL
4 */
5
6#ifndef __SYSDEP_STUB_H
7#define __SYSDEP_STUB_H
8
9#include <asm/ptrace.h>
10
11#define STUB_SYSCALL_RET EAX
12#define STUB_MMAP_NR __NR_mmap2
13#define MMAP_OFFSET(o) ((o) >> UM_KERN_PAGE_SHIFT)
14
15static inline long stub_syscall0(long syscall)
16{
17 long ret;
18
19 __asm__ volatile ("int $0x80" : "=a" (ret) : "0" (syscall));
20
21 return ret;
22}
23
24static inline long stub_syscall1(long syscall, long arg1)
25{
26 long ret;
27
28 __asm__ volatile ("int $0x80" : "=a" (ret) : "0" (syscall), "b" (arg1));
29
30 return ret;
31}
32
33static inline long stub_syscall2(long syscall, long arg1, long arg2)
34{
35 long ret;
36
37 __asm__ volatile ("int $0x80" : "=a" (ret) : "0" (syscall), "b" (arg1),
38 "c" (arg2));
39
40 return ret;
41}
42
43static inline long stub_syscall3(long syscall, long arg1, long arg2, long arg3)
44{
45 long ret;
46
47 __asm__ volatile ("int $0x80" : "=a" (ret) : "0" (syscall), "b" (arg1),
48 "c" (arg2), "d" (arg3));
49
50 return ret;
51}
52
53static inline long stub_syscall4(long syscall, long arg1, long arg2, long arg3,
54 long arg4)
55{
56 long ret;
57
58 __asm__ volatile ("int $0x80" : "=a" (ret) : "0" (syscall), "b" (arg1),
59 "c" (arg2), "d" (arg3), "S" (arg4));
60
61 return ret;
62}
63
64static inline long stub_syscall5(long syscall, long arg1, long arg2, long arg3,
65 long arg4, long arg5)
66{
67 long ret;
68
69 __asm__ volatile ("int $0x80" : "=a" (ret) : "0" (syscall), "b" (arg1),
70 "c" (arg2), "d" (arg3), "S" (arg4), "D" (arg5));
71
72 return ret;
73}
74
75static inline void trap_myself(void)
76{
77 __asm("int3");
78}
79
80static inline void remap_stack(int fd, unsigned long offset)
81{
82 __asm__ volatile ("movl %%eax,%%ebp ; movl %0,%%eax ; int $0x80 ;"
83 "movl %7, %%ebx ; movl %%eax, (%%ebx)"
84 : : "g" (STUB_MMAP_NR), "b" (STUB_DATA),
85 "c" (UM_KERN_PAGE_SIZE),
86 "d" (PROT_READ | PROT_WRITE),
87 "S" (MAP_FIXED | MAP_SHARED), "D" (fd),
88 "a" (offset),
89 "i" (&((struct stub_data *) STUB_DATA)->err)
90 : "memory");
91}
92
93#endif
diff --git a/arch/x86/um/shared/sysdep/stub_64.h b/arch/x86/um/shared/sysdep/stub_64.h
new file mode 100644
index 000000000000..994df93c5ed3
--- /dev/null
+++ b/arch/x86/um/shared/sysdep/stub_64.h
@@ -0,0 +1,99 @@
1/*
2 * Copyright (C) 2004 Jeff Dike (jdike@addtoit.com)
3 * Licensed under the GPL
4 */
5
6#ifndef __SYSDEP_STUB_H
7#define __SYSDEP_STUB_H
8
9#include <sysdep/ptrace_user.h>
10
11#define STUB_SYSCALL_RET PT_INDEX(RAX)
12#define STUB_MMAP_NR __NR_mmap
13#define MMAP_OFFSET(o) (o)
14
15#define __syscall_clobber "r11","rcx","memory"
16#define __syscall "syscall"
17
18static inline long stub_syscall0(long syscall)
19{
20 long ret;
21
22 __asm__ volatile (__syscall
23 : "=a" (ret)
24 : "0" (syscall) : __syscall_clobber );
25
26 return ret;
27}
28
29static inline long stub_syscall2(long syscall, long arg1, long arg2)
30{
31 long ret;
32
33 __asm__ volatile (__syscall
34 : "=a" (ret)
35 : "0" (syscall), "D" (arg1), "S" (arg2) : __syscall_clobber );
36
37 return ret;
38}
39
40static inline long stub_syscall3(long syscall, long arg1, long arg2, long arg3)
41{
42 long ret;
43
44 __asm__ volatile (__syscall
45 : "=a" (ret)
46 : "0" (syscall), "D" (arg1), "S" (arg2), "d" (arg3)
47 : __syscall_clobber );
48
49 return ret;
50}
51
52static inline long stub_syscall4(long syscall, long arg1, long arg2, long arg3,
53 long arg4)
54{
55 long ret;
56
57 __asm__ volatile ("movq %5,%%r10 ; " __syscall
58 : "=a" (ret)
59 : "0" (syscall), "D" (arg1), "S" (arg2), "d" (arg3),
60 "g" (arg4)
61 : __syscall_clobber, "r10" );
62
63 return ret;
64}
65
66static inline long stub_syscall5(long syscall, long arg1, long arg2, long arg3,
67 long arg4, long arg5)
68{
69 long ret;
70
71 __asm__ volatile ("movq %5,%%r10 ; movq %6,%%r8 ; " __syscall
72 : "=a" (ret)
73 : "0" (syscall), "D" (arg1), "S" (arg2), "d" (arg3),
74 "g" (arg4), "g" (arg5)
75 : __syscall_clobber, "r10", "r8" );
76
77 return ret;
78}
79
80static inline void trap_myself(void)
81{
82 __asm("int3");
83}
84
85static inline void remap_stack(long fd, unsigned long offset)
86{
87 __asm__ volatile ("movq %4,%%r10 ; movq %5,%%r8 ; "
88 "movq %6, %%r9; " __syscall "; movq %7, %%rbx ; "
89 "movq %%rax, (%%rbx)":
90 : "a" (STUB_MMAP_NR), "D" (STUB_DATA),
91 "S" (UM_KERN_PAGE_SIZE),
92 "d" (PROT_READ | PROT_WRITE),
93 "g" (MAP_FIXED | MAP_SHARED), "g" (fd),
94 "g" (offset),
95 "i" (&((struct stub_data *) STUB_DATA)->err)
96 : __syscall_clobber, "r10", "r8", "r9" );
97}
98
99#endif
diff --git a/arch/x86/um/shared/sysdep/syscalls.h b/arch/x86/um/shared/sysdep/syscalls.h
new file mode 100644
index 000000000000..bd9a89b67e41
--- /dev/null
+++ b/arch/x86/um/shared/sysdep/syscalls.h
@@ -0,0 +1,5 @@
1#ifdef __i386__
2#include "syscalls_32.h"
3#else
4#include "syscalls_64.h"
5#endif
diff --git a/arch/x86/um/shared/sysdep/syscalls_32.h b/arch/x86/um/shared/sysdep/syscalls_32.h
new file mode 100644
index 000000000000..05cb796aecb5
--- /dev/null
+++ b/arch/x86/um/shared/sysdep/syscalls_32.h
@@ -0,0 +1,20 @@
1/*
2 * Copyright (C) 2000 - 2008 Jeff Dike (jdike@{addtoit,linux.intel}.com)
3 * Licensed under the GPL
4 */
5
6#include "asm/unistd.h"
7#include "sysdep/ptrace.h"
8
9typedef long syscall_handler_t(struct pt_regs);
10
11/* Not declared on x86, incompatible declarations on x86_64, so these have
12 * to go here rather than in sys_call_table.c
13 */
14extern syscall_handler_t sys_rt_sigaction;
15
16extern syscall_handler_t *sys_call_table[];
17
18#define EXECUTE_SYSCALL(syscall, regs) \
19 ((long (*)(struct syscall_args)) \
20 (*sys_call_table[syscall]))(SYSCALL_ARGS(&regs->regs))
diff --git a/arch/x86/um/shared/sysdep/syscalls_64.h b/arch/x86/um/shared/sysdep/syscalls_64.h
new file mode 100644
index 000000000000..8a7d5e1da98e
--- /dev/null
+++ b/arch/x86/um/shared/sysdep/syscalls_64.h
@@ -0,0 +1,32 @@
1/*
2 * Copyright 2003 PathScale, Inc.
3 *
4 * Licensed under the GPL
5 */
6
7#ifndef __SYSDEP_X86_64_SYSCALLS_H__
8#define __SYSDEP_X86_64_SYSCALLS_H__
9
10#include <linux/msg.h>
11#include <linux/shm.h>
12
13typedef long syscall_handler_t(void);
14
15extern syscall_handler_t *sys_call_table[];
16
17#define EXECUTE_SYSCALL(syscall, regs) \
18 (((long (*)(long, long, long, long, long, long)) \
19 (*sys_call_table[syscall]))(UPT_SYSCALL_ARG1(&regs->regs), \
20 UPT_SYSCALL_ARG2(&regs->regs), \
21 UPT_SYSCALL_ARG3(&regs->regs), \
22 UPT_SYSCALL_ARG4(&regs->regs), \
23 UPT_SYSCALL_ARG5(&regs->regs), \
24 UPT_SYSCALL_ARG6(&regs->regs)))
25
26extern long old_mmap(unsigned long addr, unsigned long len,
27 unsigned long prot, unsigned long flags,
28 unsigned long fd, unsigned long pgoff);
29extern syscall_handler_t sys_modify_ldt;
30extern syscall_handler_t sys_arch_prctl;
31
32#endif
diff --git a/arch/x86/um/shared/sysdep/tls.h b/arch/x86/um/shared/sysdep/tls.h
new file mode 100644
index 000000000000..27cce00c6b30
--- /dev/null
+++ b/arch/x86/um/shared/sysdep/tls.h
@@ -0,0 +1,39 @@
1#ifndef _SYSDEP_TLS_H
2#define _SYSDEP_TLS_H
3
4# ifndef __KERNEL__
5
6/* Change name to avoid conflicts with the original one from <asm/ldt.h>, which
7 * may be named user_desc (but in 2.4 and in header matching its API was named
8 * modify_ldt_ldt_s). */
9
10typedef struct um_dup_user_desc {
11 unsigned int entry_number;
12 unsigned int base_addr;
13 unsigned int limit;
14 unsigned int seg_32bit:1;
15 unsigned int contents:2;
16 unsigned int read_exec_only:1;
17 unsigned int limit_in_pages:1;
18 unsigned int seg_not_present:1;
19 unsigned int useable:1;
20#ifdef __x86_64__
21 unsigned int lm:1;
22#endif
23} user_desc_t;
24
25# else /* __KERNEL__ */
26
27typedef struct user_desc user_desc_t;
28
29# endif /* __KERNEL__ */
30
31extern int os_set_thread_area(user_desc_t *info, int pid);
32extern int os_get_thread_area(user_desc_t *info, int pid);
33
34#ifdef __i386__
35#define GDT_ENTRY_TLS_MIN_I386 6
36#define GDT_ENTRY_TLS_MIN_X86_64 12
37#endif
38
39#endif /* _SYSDEP_TLS_H */
diff --git a/arch/x86/um/signal.c b/arch/x86/um/signal.c
new file mode 100644
index 000000000000..4883b9546016
--- /dev/null
+++ b/arch/x86/um/signal.c
@@ -0,0 +1,624 @@
1/*
2 * Copyright (C) 2003 PathScale, Inc.
3 * Copyright (C) 2003 - 2007 Jeff Dike (jdike@{addtoit,linux.intel}.com)
4 * Licensed under the GPL
5 */
6
7
8#include <linux/personality.h>
9#include <linux/ptrace.h>
10#include <linux/kernel.h>
11#include <asm/unistd.h>
12#include <asm/uaccess.h>
13#include <asm/ucontext.h>
14#include "frame_kern.h"
15#include "skas.h"
16
17#ifdef CONFIG_X86_32
18
19/*
20 * FPU tag word conversions.
21 */
22
23static inline unsigned short twd_i387_to_fxsr(unsigned short twd)
24{
25 unsigned int tmp; /* to avoid 16 bit prefixes in the code */
26
27 /* Transform each pair of bits into 01 (valid) or 00 (empty) */
28 tmp = ~twd;
29 tmp = (tmp | (tmp>>1)) & 0x5555; /* 0V0V0V0V0V0V0V0V */
30 /* and move the valid bits to the lower byte. */
31 tmp = (tmp | (tmp >> 1)) & 0x3333; /* 00VV00VV00VV00VV */
32 tmp = (tmp | (tmp >> 2)) & 0x0f0f; /* 0000VVVV0000VVVV */
33 tmp = (tmp | (tmp >> 4)) & 0x00ff; /* 00000000VVVVVVVV */
34 return tmp;
35}
36
37static inline unsigned long twd_fxsr_to_i387(struct user_fxsr_struct *fxsave)
38{
39 struct _fpxreg *st = NULL;
40 unsigned long twd = (unsigned long) fxsave->twd;
41 unsigned long tag;
42 unsigned long ret = 0xffff0000;
43 int i;
44
45#define FPREG_ADDR(f, n) ((char *)&(f)->st_space + (n) * 16)
46
47 for (i = 0; i < 8; i++) {
48 if (twd & 0x1) {
49 st = (struct _fpxreg *) FPREG_ADDR(fxsave, i);
50
51 switch (st->exponent & 0x7fff) {
52 case 0x7fff:
53 tag = 2; /* Special */
54 break;
55 case 0x0000:
56 if ( !st->significand[0] &&
57 !st->significand[1] &&
58 !st->significand[2] &&
59 !st->significand[3] ) {
60 tag = 1; /* Zero */
61 } else {
62 tag = 2; /* Special */
63 }
64 break;
65 default:
66 if (st->significand[3] & 0x8000) {
67 tag = 0; /* Valid */
68 } else {
69 tag = 2; /* Special */
70 }
71 break;
72 }
73 } else {
74 tag = 3; /* Empty */
75 }
76 ret |= (tag << (2 * i));
77 twd = twd >> 1;
78 }
79 return ret;
80}
81
82static int convert_fxsr_to_user(struct _fpstate __user *buf,
83 struct user_fxsr_struct *fxsave)
84{
85 unsigned long env[7];
86 struct _fpreg __user *to;
87 struct _fpxreg *from;
88 int i;
89
90 env[0] = (unsigned long)fxsave->cwd | 0xffff0000ul;
91 env[1] = (unsigned long)fxsave->swd | 0xffff0000ul;
92 env[2] = twd_fxsr_to_i387(fxsave);
93 env[3] = fxsave->fip;
94 env[4] = fxsave->fcs | ((unsigned long)fxsave->fop << 16);
95 env[5] = fxsave->foo;
96 env[6] = fxsave->fos;
97
98 if (__copy_to_user(buf, env, 7 * sizeof(unsigned long)))
99 return 1;
100
101 to = &buf->_st[0];
102 from = (struct _fpxreg *) &fxsave->st_space[0];
103 for (i = 0; i < 8; i++, to++, from++) {
104 unsigned long __user *t = (unsigned long __user *)to;
105 unsigned long *f = (unsigned long *)from;
106
107 if (__put_user(*f, t) ||
108 __put_user(*(f + 1), t + 1) ||
109 __put_user(from->exponent, &to->exponent))
110 return 1;
111 }
112 return 0;
113}
114
115static int convert_fxsr_from_user(struct user_fxsr_struct *fxsave,
116 struct _fpstate __user *buf)
117{
118 unsigned long env[7];
119 struct _fpxreg *to;
120 struct _fpreg __user *from;
121 int i;
122
123 if (copy_from_user( env, buf, 7 * sizeof(long)))
124 return 1;
125
126 fxsave->cwd = (unsigned short)(env[0] & 0xffff);
127 fxsave->swd = (unsigned short)(env[1] & 0xffff);
128 fxsave->twd = twd_i387_to_fxsr((unsigned short)(env[2] & 0xffff));
129 fxsave->fip = env[3];
130 fxsave->fop = (unsigned short)((env[4] & 0xffff0000ul) >> 16);
131 fxsave->fcs = (env[4] & 0xffff);
132 fxsave->foo = env[5];
133 fxsave->fos = env[6];
134
135 to = (struct _fpxreg *) &fxsave->st_space[0];
136 from = &buf->_st[0];
137 for (i = 0; i < 8; i++, to++, from++) {
138 unsigned long *t = (unsigned long *)to;
139 unsigned long __user *f = (unsigned long __user *)from;
140
141 if (__get_user(*t, f) ||
142 __get_user(*(t + 1), f + 1) ||
143 __get_user(to->exponent, &from->exponent))
144 return 1;
145 }
146 return 0;
147}
148
149extern int have_fpx_regs;
150
151#endif
152
153static int copy_sc_from_user(struct pt_regs *regs,
154 struct sigcontext __user *from)
155{
156 struct sigcontext sc;
157 int err, pid;
158
159 err = copy_from_user(&sc, from, sizeof(sc));
160 if (err)
161 return err;
162
163#define GETREG(regno, regname) regs->regs.gp[HOST_##regno] = sc.regname
164
165#ifdef CONFIG_X86_32
166 GETREG(GS, gs);
167 GETREG(FS, fs);
168 GETREG(ES, es);
169 GETREG(DS, ds);
170#endif
171 GETREG(DI, di);
172 GETREG(SI, si);
173 GETREG(BP, bp);
174 GETREG(SP, sp);
175 GETREG(BX, bx);
176 GETREG(DX, dx);
177 GETREG(CX, cx);
178 GETREG(AX, ax);
179 GETREG(IP, ip);
180
181#ifdef CONFIG_X86_64
182 GETREG(R8, r8);
183 GETREG(R9, r9);
184 GETREG(R10, r10);
185 GETREG(R11, r11);
186 GETREG(R12, r12);
187 GETREG(R13, r13);
188 GETREG(R14, r14);
189 GETREG(R15, r15);
190#endif
191
192 GETREG(CS, cs);
193 GETREG(EFLAGS, flags);
194#ifdef CONFIG_X86_32
195 GETREG(SS, ss);
196#endif
197
198#undef GETREG
199
200 pid = userspace_pid[current_thread_info()->cpu];
201#ifdef CONFIG_X86_32
202 if (have_fpx_regs) {
203 struct user_fxsr_struct fpx;
204
205 err = copy_from_user(&fpx,
206 &((struct _fpstate __user *)sc.fpstate)->_fxsr_env[0],
207 sizeof(struct user_fxsr_struct));
208 if (err)
209 return 1;
210
211 err = convert_fxsr_from_user(&fpx, sc.fpstate);
212 if (err)
213 return 1;
214
215 err = restore_fpx_registers(pid, (unsigned long *) &fpx);
216 if (err < 0) {
217 printk(KERN_ERR "copy_sc_from_user - "
218 "restore_fpx_registers failed, errno = %d\n",
219 -err);
220 return 1;
221 }
222 } else
223#endif
224 {
225 struct user_i387_struct fp;
226
227 err = copy_from_user(&fp, sc.fpstate,
228 sizeof(struct user_i387_struct));
229 if (err)
230 return 1;
231
232 err = restore_fp_registers(pid, (unsigned long *) &fp);
233 if (err < 0) {
234 printk(KERN_ERR "copy_sc_from_user - "
235 "restore_fp_registers failed, errno = %d\n",
236 -err);
237 return 1;
238 }
239 }
240 return 0;
241}
242
243static int copy_sc_to_user(struct sigcontext __user *to,
244 struct _fpstate __user *to_fp, struct pt_regs *regs,
245 unsigned long mask)
246{
247 struct sigcontext sc;
248 struct faultinfo * fi = &current->thread.arch.faultinfo;
249 int err, pid;
250 memset(&sc, 0, sizeof(struct sigcontext));
251
252#define PUTREG(regno, regname) sc.regname = regs->regs.gp[HOST_##regno]
253
254#ifdef CONFIG_X86_32
255 PUTREG(GS, gs);
256 PUTREG(FS, fs);
257 PUTREG(ES, es);
258 PUTREG(DS, ds);
259#endif
260 PUTREG(DI, di);
261 PUTREG(SI, si);
262 PUTREG(BP, bp);
263 PUTREG(SP, sp);
264 PUTREG(BX, bx);
265 PUTREG(DX, dx);
266 PUTREG(CX, cx);
267 PUTREG(AX, ax);
268#ifdef CONFIG_X86_64
269 PUTREG(R8, r8);
270 PUTREG(R9, r9);
271 PUTREG(R10, r10);
272 PUTREG(R11, r11);
273 PUTREG(R12, r12);
274 PUTREG(R13, r13);
275 PUTREG(R14, r14);
276 PUTREG(R15, r15);
277#endif
278
279 sc.cr2 = fi->cr2;
280 sc.err = fi->error_code;
281 sc.trapno = fi->trap_no;
282 PUTREG(IP, ip);
283 PUTREG(CS, cs);
284 PUTREG(EFLAGS, flags);
285#ifdef CONFIG_X86_32
286 PUTREG(SP, sp_at_signal);
287 PUTREG(SS, ss);
288#endif
289#undef PUTREG
290 sc.oldmask = mask;
291 sc.fpstate = to_fp;
292
293 err = copy_to_user(to, &sc, sizeof(struct sigcontext));
294 if (err)
295 return 1;
296
297 pid = userspace_pid[current_thread_info()->cpu];
298
299#ifdef CONFIG_X86_32
300 if (have_fpx_regs) {
301 struct user_fxsr_struct fpx;
302
303 err = save_fpx_registers(pid, (unsigned long *) &fpx);
304 if (err < 0){
305 printk(KERN_ERR "copy_sc_to_user - save_fpx_registers "
306 "failed, errno = %d\n", err);
307 return 1;
308 }
309
310 err = convert_fxsr_to_user(to_fp, &fpx);
311 if (err)
312 return 1;
313
314 err |= __put_user(fpx.swd, &to_fp->status);
315 err |= __put_user(X86_FXSR_MAGIC, &to_fp->magic);
316 if (err)
317 return 1;
318
319 if (copy_to_user(&to_fp->_fxsr_env[0], &fpx,
320 sizeof(struct user_fxsr_struct)))
321 return 1;
322 } else
323#endif
324 {
325 struct user_i387_struct fp;
326
327 err = save_fp_registers(pid, (unsigned long *) &fp);
328 if (copy_to_user(to_fp, &fp, sizeof(struct user_i387_struct)))
329 return 1;
330 }
331
332 return 0;
333}
334
335#ifdef CONFIG_X86_32
336static int copy_ucontext_to_user(struct ucontext __user *uc,
337 struct _fpstate __user *fp, sigset_t *set,
338 unsigned long sp)
339{
340 int err = 0;
341
342 err |= put_user(current->sas_ss_sp, &uc->uc_stack.ss_sp);
343 err |= put_user(sas_ss_flags(sp), &uc->uc_stack.ss_flags);
344 err |= put_user(current->sas_ss_size, &uc->uc_stack.ss_size);
345 err |= copy_sc_to_user(&uc->uc_mcontext, fp, &current->thread.regs, 0);
346 err |= copy_to_user(&uc->uc_sigmask, set, sizeof(*set));
347 return err;
348}
349
350struct sigframe
351{
352 char __user *pretcode;
353 int sig;
354 struct sigcontext sc;
355 struct _fpstate fpstate;
356 unsigned long extramask[_NSIG_WORDS-1];
357 char retcode[8];
358};
359
360struct rt_sigframe
361{
362 char __user *pretcode;
363 int sig;
364 struct siginfo __user *pinfo;
365 void __user *puc;
366 struct siginfo info;
367 struct ucontext uc;
368 struct _fpstate fpstate;
369 char retcode[8];
370};
371
372int setup_signal_stack_sc(unsigned long stack_top, int sig,
373 struct k_sigaction *ka, struct pt_regs *regs,
374 sigset_t *mask)
375{
376 struct sigframe __user *frame;
377 void __user *restorer;
378 int err = 0;
379
380 /* This is the same calculation as i386 - ((sp + 4) & 15) == 0 */
381 stack_top = ((stack_top + 4) & -16UL) - 4;
382 frame = (struct sigframe __user *) stack_top - 1;
383 if (!access_ok(VERIFY_WRITE, frame, sizeof(*frame)))
384 return 1;
385
386 restorer = frame->retcode;
387 if (ka->sa.sa_flags & SA_RESTORER)
388 restorer = ka->sa.sa_restorer;
389
390 err |= __put_user(restorer, &frame->pretcode);
391 err |= __put_user(sig, &frame->sig);
392 err |= copy_sc_to_user(&frame->sc, &frame->fpstate, regs, mask->sig[0]);
393 if (_NSIG_WORDS > 1)
394 err |= __copy_to_user(&frame->extramask, &mask->sig[1],
395 sizeof(frame->extramask));
396
397 /*
398 * This is popl %eax ; movl $,%eax ; int $0x80
399 *
400 * WE DO NOT USE IT ANY MORE! It's only left here for historical
401 * reasons and because gdb uses it as a signature to notice
402 * signal handler stack frames.
403 */
404 err |= __put_user(0xb858, (short __user *)(frame->retcode+0));
405 err |= __put_user(__NR_sigreturn, (int __user *)(frame->retcode+2));
406 err |= __put_user(0x80cd, (short __user *)(frame->retcode+6));
407
408 if (err)
409 return err;
410
411 PT_REGS_SP(regs) = (unsigned long) frame;
412 PT_REGS_IP(regs) = (unsigned long) ka->sa.sa_handler;
413 PT_REGS_EAX(regs) = (unsigned long) sig;
414 PT_REGS_EDX(regs) = (unsigned long) 0;
415 PT_REGS_ECX(regs) = (unsigned long) 0;
416
417 if ((current->ptrace & PT_DTRACE) && (current->ptrace & PT_PTRACED))
418 ptrace_notify(SIGTRAP);
419 return 0;
420}
421
422int setup_signal_stack_si(unsigned long stack_top, int sig,
423 struct k_sigaction *ka, struct pt_regs *regs,
424 siginfo_t *info, sigset_t *mask)
425{
426 struct rt_sigframe __user *frame;
427 void __user *restorer;
428 int err = 0;
429
430 stack_top &= -8UL;
431 frame = (struct rt_sigframe __user *) stack_top - 1;
432 if (!access_ok(VERIFY_WRITE, frame, sizeof(*frame)))
433 return 1;
434
435 restorer = frame->retcode;
436 if (ka->sa.sa_flags & SA_RESTORER)
437 restorer = ka->sa.sa_restorer;
438
439 err |= __put_user(restorer, &frame->pretcode);
440 err |= __put_user(sig, &frame->sig);
441 err |= __put_user(&frame->info, &frame->pinfo);
442 err |= __put_user(&frame->uc, &frame->puc);
443 err |= copy_siginfo_to_user(&frame->info, info);
444 err |= copy_ucontext_to_user(&frame->uc, &frame->fpstate, mask,
445 PT_REGS_SP(regs));
446
447 /*
448 * This is movl $,%eax ; int $0x80
449 *
450 * WE DO NOT USE IT ANY MORE! It's only left here for historical
451 * reasons and because gdb uses it as a signature to notice
452 * signal handler stack frames.
453 */
454 err |= __put_user(0xb8, (char __user *)(frame->retcode+0));
455 err |= __put_user(__NR_rt_sigreturn, (int __user *)(frame->retcode+1));
456 err |= __put_user(0x80cd, (short __user *)(frame->retcode+5));
457
458 if (err)
459 return err;
460
461 PT_REGS_SP(regs) = (unsigned long) frame;
462 PT_REGS_IP(regs) = (unsigned long) ka->sa.sa_handler;
463 PT_REGS_EAX(regs) = (unsigned long) sig;
464 PT_REGS_EDX(regs) = (unsigned long) &frame->info;
465 PT_REGS_ECX(regs) = (unsigned long) &frame->uc;
466
467 if ((current->ptrace & PT_DTRACE) && (current->ptrace & PT_PTRACED))
468 ptrace_notify(SIGTRAP);
469 return 0;
470}
471
472long sys_sigreturn(struct pt_regs *regs)
473{
474 unsigned long sp = PT_REGS_SP(&current->thread.regs);
475 struct sigframe __user *frame = (struct sigframe __user *)(sp - 8);
476 sigset_t set;
477 struct sigcontext __user *sc = &frame->sc;
478 unsigned long __user *oldmask = &sc->oldmask;
479 unsigned long __user *extramask = frame->extramask;
480 int sig_size = (_NSIG_WORDS - 1) * sizeof(unsigned long);
481
482 if (copy_from_user(&set.sig[0], oldmask, sizeof(set.sig[0])) ||
483 copy_from_user(&set.sig[1], extramask, sig_size))
484 goto segfault;
485
486 sigdelsetmask(&set, ~_BLOCKABLE);
487 set_current_blocked(&set);
488
489 if (copy_sc_from_user(&current->thread.regs, sc))
490 goto segfault;
491
492 /* Avoid ERESTART handling */
493 PT_REGS_SYSCALL_NR(&current->thread.regs) = -1;
494 return PT_REGS_SYSCALL_RET(&current->thread.regs);
495
496 segfault:
497 force_sig(SIGSEGV, current);
498 return 0;
499}
500
501#else
502
503struct rt_sigframe
504{
505 char __user *pretcode;
506 struct ucontext uc;
507 struct siginfo info;
508 struct _fpstate fpstate;
509};
510
511int setup_signal_stack_si(unsigned long stack_top, int sig,
512 struct k_sigaction *ka, struct pt_regs * regs,
513 siginfo_t *info, sigset_t *set)
514{
515 struct rt_sigframe __user *frame;
516 int err = 0;
517 struct task_struct *me = current;
518
519 frame = (struct rt_sigframe __user *)
520 round_down(stack_top - sizeof(struct rt_sigframe), 16);
521 /* Subtract 128 for a red zone and 8 for proper alignment */
522 frame = (struct rt_sigframe __user *) ((unsigned long) frame - 128 - 8);
523
524 if (!access_ok(VERIFY_WRITE, frame, sizeof(*frame)))
525 goto out;
526
527 if (ka->sa.sa_flags & SA_SIGINFO) {
528 err |= copy_siginfo_to_user(&frame->info, info);
529 if (err)
530 goto out;
531 }
532
533 /* Create the ucontext. */
534 err |= __put_user(0, &frame->uc.uc_flags);
535 err |= __put_user(0, &frame->uc.uc_link);
536 err |= __put_user(me->sas_ss_sp, &frame->uc.uc_stack.ss_sp);
537 err |= __put_user(sas_ss_flags(PT_REGS_SP(regs)),
538 &frame->uc.uc_stack.ss_flags);
539 err |= __put_user(me->sas_ss_size, &frame->uc.uc_stack.ss_size);
540 err |= copy_sc_to_user(&frame->uc.uc_mcontext, &frame->fpstate, regs,
541 set->sig[0]);
542 err |= __put_user(&frame->fpstate, &frame->uc.uc_mcontext.fpstate);
543 if (sizeof(*set) == 16) {
544 __put_user(set->sig[0], &frame->uc.uc_sigmask.sig[0]);
545 __put_user(set->sig[1], &frame->uc.uc_sigmask.sig[1]);
546 }
547 else
548 err |= __copy_to_user(&frame->uc.uc_sigmask, set,
549 sizeof(*set));
550
551 /*
552 * Set up to return from userspace. If provided, use a stub
553 * already in userspace.
554 */
555 /* x86-64 should always use SA_RESTORER. */
556 if (ka->sa.sa_flags & SA_RESTORER)
557 err |= __put_user(ka->sa.sa_restorer, &frame->pretcode);
558 else
559 /* could use a vstub here */
560 return err;
561
562 if (err)
563 return err;
564
565 /* Set up registers for signal handler */
566 {
567 struct exec_domain *ed = current_thread_info()->exec_domain;
568 if (unlikely(ed && ed->signal_invmap && sig < 32))
569 sig = ed->signal_invmap[sig];
570 }
571
572 PT_REGS_SP(regs) = (unsigned long) frame;
573 PT_REGS_RDI(regs) = sig;
574 /* In case the signal handler was declared without prototypes */
575 PT_REGS_RAX(regs) = 0;
576
577 /*
578 * This also works for non SA_SIGINFO handlers because they expect the
579 * next argument after the signal number on the stack.
580 */
581 PT_REGS_RSI(regs) = (unsigned long) &frame->info;
582 PT_REGS_RDX(regs) = (unsigned long) &frame->uc;
583 PT_REGS_RIP(regs) = (unsigned long) ka->sa.sa_handler;
584 out:
585 return err;
586}
587#endif
588
589long sys_rt_sigreturn(struct pt_regs *regs)
590{
591 unsigned long sp = PT_REGS_SP(&current->thread.regs);
592 struct rt_sigframe __user *frame =
593 (struct rt_sigframe __user *)(sp - sizeof(long));
594 struct ucontext __user *uc = &frame->uc;
595 sigset_t set;
596
597 if (copy_from_user(&set, &uc->uc_sigmask, sizeof(set)))
598 goto segfault;
599
600 sigdelsetmask(&set, ~_BLOCKABLE);
601 set_current_blocked(&set);
602
603 if (copy_sc_from_user(&current->thread.regs, &uc->uc_mcontext))
604 goto segfault;
605
606 /* Avoid ERESTART handling */
607 PT_REGS_SYSCALL_NR(&current->thread.regs) = -1;
608 return PT_REGS_SYSCALL_RET(&current->thread.regs);
609
610 segfault:
611 force_sig(SIGSEGV, current);
612 return 0;
613}
614
615#ifdef CONFIG_X86_32
616long ptregs_sigreturn(void)
617{
618 return sys_sigreturn(NULL);
619}
620long ptregs_rt_sigreturn(void)
621{
622 return sys_rt_sigreturn(NULL);
623}
624#endif
diff --git a/arch/x86/um/stub_32.S b/arch/x86/um/stub_32.S
new file mode 100644
index 000000000000..54a36ec20cb7
--- /dev/null
+++ b/arch/x86/um/stub_32.S
@@ -0,0 +1,51 @@
1#include "as-layout.h"
2
3 .globl syscall_stub
4.section .__syscall_stub, "ax"
5
6 .globl batch_syscall_stub
7batch_syscall_stub:
8 /* load pointer to first operation */
9 mov $(STUB_DATA+8), %esp
10
11again:
12 /* load length of additional data */
13 mov 0x0(%esp), %eax
14
15 /* if(length == 0) : end of list */
16 /* write possible 0 to header */
17 mov %eax, STUB_DATA+4
18 cmpl $0, %eax
19 jz done
20
21 /* save current pointer */
22 mov %esp, STUB_DATA+4
23
24 /* skip additional data */
25 add %eax, %esp
26
27 /* load syscall-# */
28 pop %eax
29
30 /* load syscall params */
31 pop %ebx
32 pop %ecx
33 pop %edx
34 pop %esi
35 pop %edi
36 pop %ebp
37
38 /* execute syscall */
39 int $0x80
40
41 /* check return value */
42 pop %ebx
43 cmp %ebx, %eax
44 je again
45
46done:
47 /* save return value */
48 mov %eax, STUB_DATA
49
50 /* stop */
51 int3
diff --git a/arch/x86/um/stub_64.S b/arch/x86/um/stub_64.S
new file mode 100644
index 000000000000..20e4a96a6dcb
--- /dev/null
+++ b/arch/x86/um/stub_64.S
@@ -0,0 +1,66 @@
1#include "as-layout.h"
2
3 .globl syscall_stub
4.section .__syscall_stub, "ax"
5syscall_stub:
6 syscall
7 /* We don't have 64-bit constants, so this constructs the address
8 * we need.
9 */
10 movq $(STUB_DATA >> 32), %rbx
11 salq $32, %rbx
12 movq $(STUB_DATA & 0xffffffff), %rcx
13 or %rcx, %rbx
14 movq %rax, (%rbx)
15 int3
16
17 .globl batch_syscall_stub
18batch_syscall_stub:
19 mov $(STUB_DATA >> 32), %rbx
20 sal $32, %rbx
21 mov $(STUB_DATA & 0xffffffff), %rax
22 or %rax, %rbx
23 /* load pointer to first operation */
24 mov %rbx, %rsp
25 add $0x10, %rsp
26again:
27 /* load length of additional data */
28 mov 0x0(%rsp), %rax
29
30 /* if(length == 0) : end of list */
31 /* write possible 0 to header */
32 mov %rax, 8(%rbx)
33 cmp $0, %rax
34 jz done
35
36 /* save current pointer */
37 mov %rsp, 8(%rbx)
38
39 /* skip additional data */
40 add %rax, %rsp
41
42 /* load syscall-# */
43 pop %rax
44
45 /* load syscall params */
46 pop %rdi
47 pop %rsi
48 pop %rdx
49 pop %r10
50 pop %r8
51 pop %r9
52
53 /* execute syscall */
54 syscall
55
56 /* check return value */
57 pop %rcx
58 cmp %rcx, %rax
59 je again
60
61done:
62 /* save return value */
63 mov %rax, (%rbx)
64
65 /* stop */
66 int3
diff --git a/arch/x86/um/stub_segv.c b/arch/x86/um/stub_segv.c
new file mode 100644
index 000000000000..b7450bd22e7d
--- /dev/null
+++ b/arch/x86/um/stub_segv.c
@@ -0,0 +1,19 @@
1/*
2 * Copyright (C) 2004 - 2007 Jeff Dike (jdike@{addtoit,linux.intel}.com)
3 * Licensed under the GPL
4 */
5
6#include "sysdep/stub.h"
7#include "sysdep/faultinfo.h"
8#include "sysdep/mcontext.h"
9
10void __attribute__ ((__section__ (".__syscall_stub")))
11stub_segv_handler(int sig, siginfo_t *info, void *p)
12{
13 struct ucontext *uc = p;
14
15 GET_FAULTINFO_FROM_MC(*((struct faultinfo *) STUB_DATA),
16 &uc->uc_mcontext);
17 trap_myself();
18}
19
diff --git a/arch/x86/um/sys_call_table_32.S b/arch/x86/um/sys_call_table_32.S
new file mode 100644
index 000000000000..a7ca80d2dceb
--- /dev/null
+++ b/arch/x86/um/sys_call_table_32.S
@@ -0,0 +1,26 @@
1#include <linux/linkage.h>
2/* Steal i386 syscall table for our purposes, but with some slight changes.*/
3
4#define sys_iopl sys_ni_syscall
5#define sys_ioperm sys_ni_syscall
6
7#define sys_vm86old sys_ni_syscall
8#define sys_vm86 sys_ni_syscall
9
10#define old_mmap sys_old_mmap
11
12#define ptregs_fork sys_fork
13#define ptregs_execve sys_execve
14#define ptregs_iopl sys_iopl
15#define ptregs_vm86old sys_vm86old
16#define ptregs_clone sys_clone
17#define ptregs_vm86 sys_vm86
18#define ptregs_sigaltstack sys_sigaltstack
19#define ptregs_vfork sys_vfork
20
21.section .rodata,"a"
22
23#include "../kernel/syscall_table_32.S"
24
25ENTRY(syscall_table_size)
26.long .-sys_call_table
diff --git a/arch/x86/um/sys_call_table_64.c b/arch/x86/um/sys_call_table_64.c
new file mode 100644
index 000000000000..99522f78b162
--- /dev/null
+++ b/arch/x86/um/sys_call_table_64.c
@@ -0,0 +1,64 @@
1/*
2 * System call table for UML/x86-64, copied from arch/x86_64/kernel/syscall.c
3 * with some changes for UML.
4 */
5
6#include <linux/linkage.h>
7#include <linux/sys.h>
8#include <linux/cache.h>
9
10#define __NO_STUBS
11
12/*
13 * Below you can see, in terms of #define's, the differences between the x86-64
14 * and the UML syscall table.
15 */
16
17/* Not going to be implemented by UML, since we have no hardware. */
18#define stub_iopl sys_ni_syscall
19#define sys_ioperm sys_ni_syscall
20
21/*
22 * The UML TLS problem. Note that x86_64 does not implement this, so the below
23 * is needed only for the ia32 compatibility.
24 */
25
26/* On UML we call it this way ("old" means it's not mmap2) */
27#define sys_mmap old_mmap
28
29#define stub_clone sys_clone
30#define stub_fork sys_fork
31#define stub_vfork sys_vfork
32#define stub_execve sys_execve
33#define stub_rt_sigsuspend sys_rt_sigsuspend
34#define stub_sigaltstack sys_sigaltstack
35#define stub_rt_sigreturn sys_rt_sigreturn
36
37#define __SYSCALL(nr, sym) extern asmlinkage void sym(void) ;
38#undef _ASM_X86_UNISTD_64_H
39#include "../../x86/include/asm/unistd_64.h"
40
41#undef __SYSCALL
42#define __SYSCALL(nr, sym) [ nr ] = sym,
43#undef _ASM_X86_UNISTD_64_H
44
45typedef void (*sys_call_ptr_t)(void);
46
47extern void sys_ni_syscall(void);
48
49/*
50 * We used to have a trick here which made sure that holes in the
51 * x86_64 table were filled in with sys_ni_syscall, but a comment in
52 * unistd_64.h says that holes aren't allowed, so the trick was
53 * removed.
54 * The trick looked like this
55 * [0 ... UM_NR_syscall_max] = &sys_ni_syscall
56 * before including unistd_64.h - the later initializations overwrote
57 * the sys_ni_syscall filler.
58 */
59
60sys_call_ptr_t sys_call_table[] __cacheline_aligned = {
61#include <asm/unistd_64.h>
62};
63
64int syscall_table_size = sizeof(sys_call_table);
diff --git a/arch/x86/um/syscalls_32.c b/arch/x86/um/syscalls_32.c
new file mode 100644
index 000000000000..70ca357393b8
--- /dev/null
+++ b/arch/x86/um/syscalls_32.c
@@ -0,0 +1,66 @@
1/*
2 * Copyright (C) 2000 - 2003 Jeff Dike (jdike@addtoit.com)
3 * Licensed under the GPL
4 */
5
6#include "linux/sched.h"
7#include "linux/shm.h"
8#include "linux/ipc.h"
9#include "linux/syscalls.h"
10#include "asm/mman.h"
11#include "asm/uaccess.h"
12#include "asm/unistd.h"
13
14/*
15 * The prototype on i386 is:
16 *
17 * int clone(int flags, void * child_stack, int * parent_tidptr, struct user_desc * newtls, int * child_tidptr)
18 *
19 * and the "newtls" arg. on i386 is read by copy_thread directly from the
20 * register saved on the stack.
21 */
22long sys_clone(unsigned long clone_flags, unsigned long newsp,
23 int __user *parent_tid, void *newtls, int __user *child_tid)
24{
25 long ret;
26
27 if (!newsp)
28 newsp = UPT_SP(&current->thread.regs.regs);
29
30 current->thread.forking = 1;
31 ret = do_fork(clone_flags, newsp, &current->thread.regs, 0, parent_tid,
32 child_tid);
33 current->thread.forking = 0;
34 return ret;
35}
36
37long sys_sigaction(int sig, const struct old_sigaction __user *act,
38 struct old_sigaction __user *oact)
39{
40 struct k_sigaction new_ka, old_ka;
41 int ret;
42
43 if (act) {
44 old_sigset_t mask;
45 if (!access_ok(VERIFY_READ, act, sizeof(*act)) ||
46 __get_user(new_ka.sa.sa_handler, &act->sa_handler) ||
47 __get_user(new_ka.sa.sa_restorer, &act->sa_restorer))
48 return -EFAULT;
49 __get_user(new_ka.sa.sa_flags, &act->sa_flags);
50 __get_user(mask, &act->sa_mask);
51 siginitset(&new_ka.sa.sa_mask, mask);
52 }
53
54 ret = do_sigaction(sig, act ? &new_ka : NULL, oact ? &old_ka : NULL);
55
56 if (!ret && oact) {
57 if (!access_ok(VERIFY_WRITE, oact, sizeof(*oact)) ||
58 __put_user(old_ka.sa.sa_handler, &oact->sa_handler) ||
59 __put_user(old_ka.sa.sa_restorer, &oact->sa_restorer))
60 return -EFAULT;
61 __put_user(old_ka.sa.sa_flags, &oact->sa_flags);
62 __put_user(old_ka.sa.sa_mask.sig[0], &oact->sa_mask);
63 }
64
65 return ret;
66}
diff --git a/arch/x86/um/syscalls_64.c b/arch/x86/um/syscalls_64.c
new file mode 100644
index 000000000000..f3d82bb6e15a
--- /dev/null
+++ b/arch/x86/um/syscalls_64.c
@@ -0,0 +1,102 @@
1/*
2 * Copyright (C) 2003 - 2007 Jeff Dike (jdike@{addtoit,linux.intel}.com)
3 * Copyright 2003 PathScale, Inc.
4 *
5 * Licensed under the GPL
6 */
7
8#include "linux/linkage.h"
9#include "linux/personality.h"
10#include "linux/utsname.h"
11#include "asm/prctl.h" /* XXX This should get the constants from libc */
12#include "asm/uaccess.h"
13#include "os.h"
14
15long arch_prctl(struct task_struct *task, int code, unsigned long __user *addr)
16{
17 unsigned long *ptr = addr, tmp;
18 long ret;
19 int pid = task->mm->context.id.u.pid;
20
21 /*
22 * With ARCH_SET_FS (and ARCH_SET_GS is treated similarly to
23 * be safe), we need to call arch_prctl on the host because
24 * setting %fs may result in something else happening (like a
25 * GDT or thread.fs being set instead). So, we let the host
26 * fiddle the registers and thread struct and restore the
27 * registers afterwards.
28 *
29 * So, the saved registers are stored to the process (this
30 * needed because a stub may have been the last thing to run),
31 * arch_prctl is run on the host, then the registers are read
32 * back.
33 */
34 switch (code) {
35 case ARCH_SET_FS:
36 case ARCH_SET_GS:
37 ret = restore_registers(pid, &current->thread.regs.regs);
38 if (ret)
39 return ret;
40 break;
41 case ARCH_GET_FS:
42 case ARCH_GET_GS:
43 /*
44 * With these two, we read to a local pointer and
45 * put_user it to the userspace pointer that we were
46 * given. If addr isn't valid (because it hasn't been
47 * faulted in or is just bogus), we want put_user to
48 * fault it in (or return -EFAULT) instead of having
49 * the host return -EFAULT.
50 */
51 ptr = &tmp;
52 }
53
54 ret = os_arch_prctl(pid, code, ptr);
55 if (ret)
56 return ret;
57
58 switch (code) {
59 case ARCH_SET_FS:
60 current->thread.arch.fs = (unsigned long) ptr;
61 ret = save_registers(pid, &current->thread.regs.regs);
62 break;
63 case ARCH_SET_GS:
64 ret = save_registers(pid, &current->thread.regs.regs);
65 break;
66 case ARCH_GET_FS:
67 ret = put_user(tmp, addr);
68 break;
69 case ARCH_GET_GS:
70 ret = put_user(tmp, addr);
71 break;
72 }
73
74 return ret;
75}
76
77long sys_arch_prctl(int code, unsigned long addr)
78{
79 return arch_prctl(current, code, (unsigned long __user *) addr);
80}
81
82long sys_clone(unsigned long clone_flags, unsigned long newsp,
83 void __user *parent_tid, void __user *child_tid)
84{
85 long ret;
86
87 if (!newsp)
88 newsp = UPT_SP(&current->thread.regs.regs);
89 current->thread.forking = 1;
90 ret = do_fork(clone_flags, newsp, &current->thread.regs, 0, parent_tid,
91 child_tid);
92 current->thread.forking = 0;
93 return ret;
94}
95
96void arch_switch_to(struct task_struct *to)
97{
98 if ((to->thread.arch.fs == 0) || (to->mm == NULL))
99 return;
100
101 arch_prctl(to, ARCH_SET_FS, (void __user *) to->thread.arch.fs);
102}
diff --git a/arch/x86/um/sysrq_32.c b/arch/x86/um/sysrq_32.c
new file mode 100644
index 000000000000..171b3e9dc867
--- /dev/null
+++ b/arch/x86/um/sysrq_32.c
@@ -0,0 +1,101 @@
1/*
2 * Copyright (C) 2001 - 2003 Jeff Dike (jdike@addtoit.com)
3 * Licensed under the GPL
4 */
5
6#include "linux/kernel.h"
7#include "linux/smp.h"
8#include "linux/sched.h"
9#include "linux/kallsyms.h"
10#include "asm/ptrace.h"
11#include "sysrq.h"
12
13/* This is declared by <linux/sched.h> */
14void show_regs(struct pt_regs *regs)
15{
16 printk("\n");
17 printk("EIP: %04lx:[<%08lx>] CPU: %d %s",
18 0xffff & PT_REGS_CS(regs), PT_REGS_IP(regs),
19 smp_processor_id(), print_tainted());
20 if (PT_REGS_CS(regs) & 3)
21 printk(" ESP: %04lx:%08lx", 0xffff & PT_REGS_SS(regs),
22 PT_REGS_SP(regs));
23 printk(" EFLAGS: %08lx\n %s\n", PT_REGS_EFLAGS(regs),
24 print_tainted());
25 printk("EAX: %08lx EBX: %08lx ECX: %08lx EDX: %08lx\n",
26 PT_REGS_EAX(regs), PT_REGS_EBX(regs),
27 PT_REGS_ECX(regs),
28 PT_REGS_EDX(regs));
29 printk("ESI: %08lx EDI: %08lx EBP: %08lx",
30 PT_REGS_ESI(regs), PT_REGS_EDI(regs),
31 PT_REGS_EBP(regs));
32 printk(" DS: %04lx ES: %04lx\n",
33 0xffff & PT_REGS_DS(regs),
34 0xffff & PT_REGS_ES(regs));
35
36 show_trace(NULL, (unsigned long *) &regs);
37}
38
39/* Copied from i386. */
40static inline int valid_stack_ptr(struct thread_info *tinfo, void *p)
41{
42 return p > (void *)tinfo &&
43 p < (void *)tinfo + THREAD_SIZE - 3;
44}
45
46/* Adapted from i386 (we also print the address we read from). */
47static inline unsigned long print_context_stack(struct thread_info *tinfo,
48 unsigned long *stack, unsigned long ebp)
49{
50 unsigned long addr;
51
52#ifdef CONFIG_FRAME_POINTER
53 while (valid_stack_ptr(tinfo, (void *)ebp)) {
54 addr = *(unsigned long *)(ebp + 4);
55 printk("%08lx: [<%08lx>]", ebp + 4, addr);
56 print_symbol(" %s", addr);
57 printk("\n");
58 ebp = *(unsigned long *)ebp;
59 }
60#else
61 while (valid_stack_ptr(tinfo, stack)) {
62 addr = *stack;
63 if (__kernel_text_address(addr)) {
64 printk("%08lx: [<%08lx>]", (unsigned long) stack, addr);
65 print_symbol(" %s", addr);
66 printk("\n");
67 }
68 stack++;
69 }
70#endif
71 return ebp;
72}
73
74void show_trace(struct task_struct* task, unsigned long * stack)
75{
76 unsigned long ebp;
77 struct thread_info *context;
78
79 /* Turn this into BUG_ON if possible. */
80 if (!stack) {
81 stack = (unsigned long*) &stack;
82 printk("show_trace: got NULL stack, implicit assumption task == current");
83 WARN_ON(1);
84 }
85
86 if (!task)
87 task = current;
88
89 if (task != current) {
90 ebp = (unsigned long) KSTK_EBP(task);
91 } else {
92 asm ("movl %%ebp, %0" : "=r" (ebp) : );
93 }
94
95 context = (struct thread_info *)
96 ((unsigned long)stack & (~(THREAD_SIZE - 1)));
97 print_context_stack(context, stack, ebp);
98
99 printk("\n");
100}
101
diff --git a/arch/x86/um/sysrq_64.c b/arch/x86/um/sysrq_64.c
new file mode 100644
index 000000000000..e8913436d7dc
--- /dev/null
+++ b/arch/x86/um/sysrq_64.c
@@ -0,0 +1,41 @@
1/*
2 * Copyright 2003 PathScale, Inc.
3 *
4 * Licensed under the GPL
5 */
6
7#include <linux/kernel.h>
8#include <linux/module.h>
9#include <linux/sched.h>
10#include <linux/utsname.h>
11#include <asm/current.h>
12#include <asm/ptrace.h>
13#include "sysrq.h"
14
15void __show_regs(struct pt_regs *regs)
16{
17 printk("\n");
18 print_modules();
19 printk(KERN_INFO "Pid: %d, comm: %.20s %s %s\n", task_pid_nr(current),
20 current->comm, print_tainted(), init_utsname()->release);
21 printk(KERN_INFO "RIP: %04lx:[<%016lx>]\n", PT_REGS_CS(regs) & 0xffff,
22 PT_REGS_RIP(regs));
23 printk(KERN_INFO "RSP: %016lx EFLAGS: %08lx\n", PT_REGS_SP(regs),
24 PT_REGS_EFLAGS(regs));
25 printk(KERN_INFO "RAX: %016lx RBX: %016lx RCX: %016lx\n",
26 PT_REGS_RAX(regs), PT_REGS_RBX(regs), PT_REGS_RCX(regs));
27 printk(KERN_INFO "RDX: %016lx RSI: %016lx RDI: %016lx\n",
28 PT_REGS_RDX(regs), PT_REGS_RSI(regs), PT_REGS_RDI(regs));
29 printk(KERN_INFO "RBP: %016lx R08: %016lx R09: %016lx\n",
30 PT_REGS_RBP(regs), PT_REGS_R8(regs), PT_REGS_R9(regs));
31 printk(KERN_INFO "R10: %016lx R11: %016lx R12: %016lx\n",
32 PT_REGS_R10(regs), PT_REGS_R11(regs), PT_REGS_R12(regs));
33 printk(KERN_INFO "R13: %016lx R14: %016lx R15: %016lx\n",
34 PT_REGS_R13(regs), PT_REGS_R14(regs), PT_REGS_R15(regs));
35}
36
37void show_regs(struct pt_regs *regs)
38{
39 __show_regs(regs);
40 show_trace(current, (unsigned long *) &regs);
41}
diff --git a/arch/x86/um/tls_32.c b/arch/x86/um/tls_32.c
new file mode 100644
index 000000000000..c6c7131e563b
--- /dev/null
+++ b/arch/x86/um/tls_32.c
@@ -0,0 +1,396 @@
1/*
2 * Copyright (C) 2005 Paolo 'Blaisorblade' Giarrusso <blaisorblade@yahoo.it>
3 * Licensed under the GPL
4 */
5
6#include "linux/percpu.h"
7#include "linux/sched.h"
8#include "asm/uaccess.h"
9#include "os.h"
10#include "skas.h"
11#include "sysdep/tls.h"
12
13/*
14 * If needed we can detect when it's uninitialized.
15 *
16 * These are initialized in an initcall and unchanged thereafter.
17 */
18static int host_supports_tls = -1;
19int host_gdt_entry_tls_min;
20
21int do_set_thread_area(struct user_desc *info)
22{
23 int ret;
24 u32 cpu;
25
26 cpu = get_cpu();
27 ret = os_set_thread_area(info, userspace_pid[cpu]);
28 put_cpu();
29
30 if (ret)
31 printk(KERN_ERR "PTRACE_SET_THREAD_AREA failed, err = %d, "
32 "index = %d\n", ret, info->entry_number);
33
34 return ret;
35}
36
37int do_get_thread_area(struct user_desc *info)
38{
39 int ret;
40 u32 cpu;
41
42 cpu = get_cpu();
43 ret = os_get_thread_area(info, userspace_pid[cpu]);
44 put_cpu();
45
46 if (ret)
47 printk(KERN_ERR "PTRACE_GET_THREAD_AREA failed, err = %d, "
48 "index = %d\n", ret, info->entry_number);
49
50 return ret;
51}
52
53/*
54 * sys_get_thread_area: get a yet unused TLS descriptor index.
55 * XXX: Consider leaving one free slot for glibc usage at first place. This must
56 * be done here (and by changing GDT_ENTRY_TLS_* macros) and nowhere else.
57 *
58 * Also, this must be tested when compiling in SKAS mode with dynamic linking
59 * and running against NPTL.
60 */
61static int get_free_idx(struct task_struct* task)
62{
63 struct thread_struct *t = &task->thread;
64 int idx;
65
66 if (!t->arch.tls_array)
67 return GDT_ENTRY_TLS_MIN;
68
69 for (idx = 0; idx < GDT_ENTRY_TLS_ENTRIES; idx++)
70 if (!t->arch.tls_array[idx].present)
71 return idx + GDT_ENTRY_TLS_MIN;
72 return -ESRCH;
73}
74
75static inline void clear_user_desc(struct user_desc* info)
76{
77 /* Postcondition: LDT_empty(info) returns true. */
78 memset(info, 0, sizeof(*info));
79
80 /*
81 * Check the LDT_empty or the i386 sys_get_thread_area code - we obtain
82 * indeed an empty user_desc.
83 */
84 info->read_exec_only = 1;
85 info->seg_not_present = 1;
86}
87
88#define O_FORCE 1
89
90static int load_TLS(int flags, struct task_struct *to)
91{
92 int ret = 0;
93 int idx;
94
95 for (idx = GDT_ENTRY_TLS_MIN; idx < GDT_ENTRY_TLS_MAX; idx++) {
96 struct uml_tls_struct* curr =
97 &to->thread.arch.tls_array[idx - GDT_ENTRY_TLS_MIN];
98
99 /*
100 * Actually, now if it wasn't flushed it gets cleared and
101 * flushed to the host, which will clear it.
102 */
103 if (!curr->present) {
104 if (!curr->flushed) {
105 clear_user_desc(&curr->tls);
106 curr->tls.entry_number = idx;
107 } else {
108 WARN_ON(!LDT_empty(&curr->tls));
109 continue;
110 }
111 }
112
113 if (!(flags & O_FORCE) && curr->flushed)
114 continue;
115
116 ret = do_set_thread_area(&curr->tls);
117 if (ret)
118 goto out;
119
120 curr->flushed = 1;
121 }
122out:
123 return ret;
124}
125
126/*
127 * Verify if we need to do a flush for the new process, i.e. if there are any
128 * present desc's, only if they haven't been flushed.
129 */
130static inline int needs_TLS_update(struct task_struct *task)
131{
132 int i;
133 int ret = 0;
134
135 for (i = GDT_ENTRY_TLS_MIN; i < GDT_ENTRY_TLS_MAX; i++) {
136 struct uml_tls_struct* curr =
137 &task->thread.arch.tls_array[i - GDT_ENTRY_TLS_MIN];
138
139 /*
140 * Can't test curr->present, we may need to clear a descriptor
141 * which had a value.
142 */
143 if (curr->flushed)
144 continue;
145 ret = 1;
146 break;
147 }
148 return ret;
149}
150
151/*
152 * On a newly forked process, the TLS descriptors haven't yet been flushed. So
153 * we mark them as such and the first switch_to will do the job.
154 */
155void clear_flushed_tls(struct task_struct *task)
156{
157 int i;
158
159 for (i = GDT_ENTRY_TLS_MIN; i < GDT_ENTRY_TLS_MAX; i++) {
160 struct uml_tls_struct* curr =
161 &task->thread.arch.tls_array[i - GDT_ENTRY_TLS_MIN];
162
163 /*
164 * Still correct to do this, if it wasn't present on the host it
165 * will remain as flushed as it was.
166 */
167 if (!curr->present)
168 continue;
169
170 curr->flushed = 0;
171 }
172}
173
174/*
175 * In SKAS0 mode, currently, multiple guest threads sharing the same ->mm have a
176 * common host process. So this is needed in SKAS0 too.
177 *
178 * However, if each thread had a different host process (and this was discussed
179 * for SMP support) this won't be needed.
180 *
181 * And this will not need be used when (and if) we'll add support to the host
182 * SKAS patch.
183 */
184
185int arch_switch_tls(struct task_struct *to)
186{
187 if (!host_supports_tls)
188 return 0;
189
190 /*
191 * We have no need whatsoever to switch TLS for kernel threads; beyond
192 * that, that would also result in us calling os_set_thread_area with
193 * userspace_pid[cpu] == 0, which gives an error.
194 */
195 if (likely(to->mm))
196 return load_TLS(O_FORCE, to);
197
198 return 0;
199}
200
201static int set_tls_entry(struct task_struct* task, struct user_desc *info,
202 int idx, int flushed)
203{
204 struct thread_struct *t = &task->thread;
205
206 if (idx < GDT_ENTRY_TLS_MIN || idx > GDT_ENTRY_TLS_MAX)
207 return -EINVAL;
208
209 t->arch.tls_array[idx - GDT_ENTRY_TLS_MIN].tls = *info;
210 t->arch.tls_array[idx - GDT_ENTRY_TLS_MIN].present = 1;
211 t->arch.tls_array[idx - GDT_ENTRY_TLS_MIN].flushed = flushed;
212
213 return 0;
214}
215
216int arch_copy_tls(struct task_struct *new)
217{
218 struct user_desc info;
219 int idx, ret = -EFAULT;
220
221 if (copy_from_user(&info,
222 (void __user *) UPT_ESI(&new->thread.regs.regs),
223 sizeof(info)))
224 goto out;
225
226 ret = -EINVAL;
227 if (LDT_empty(&info))
228 goto out;
229
230 idx = info.entry_number;
231
232 ret = set_tls_entry(new, &info, idx, 0);
233out:
234 return ret;
235}
236
237/* XXX: use do_get_thread_area to read the host value? I'm not at all sure! */
238static int get_tls_entry(struct task_struct *task, struct user_desc *info,
239 int idx)
240{
241 struct thread_struct *t = &task->thread;
242
243 if (!t->arch.tls_array)
244 goto clear;
245
246 if (idx < GDT_ENTRY_TLS_MIN || idx > GDT_ENTRY_TLS_MAX)
247 return -EINVAL;
248
249 if (!t->arch.tls_array[idx - GDT_ENTRY_TLS_MIN].present)
250 goto clear;
251
252 *info = t->arch.tls_array[idx - GDT_ENTRY_TLS_MIN].tls;
253
254out:
255 /*
256 * Temporary debugging check, to make sure that things have been
257 * flushed. This could be triggered if load_TLS() failed.
258 */
259 if (unlikely(task == current &&
260 !t->arch.tls_array[idx - GDT_ENTRY_TLS_MIN].flushed)) {
261 printk(KERN_ERR "get_tls_entry: task with pid %d got here "
262 "without flushed TLS.", current->pid);
263 }
264
265 return 0;
266clear:
267 /*
268 * When the TLS entry has not been set, the values read to user in the
269 * tls_array are 0 (because it's cleared at boot, see
270 * arch/i386/kernel/head.S:cpu_gdt_table). Emulate that.
271 */
272 clear_user_desc(info);
273 info->entry_number = idx;
274 goto out;
275}
276
277int sys_set_thread_area(struct user_desc __user *user_desc)
278{
279 struct user_desc info;
280 int idx, ret;
281
282 if (!host_supports_tls)
283 return -ENOSYS;
284
285 if (copy_from_user(&info, user_desc, sizeof(info)))
286 return -EFAULT;
287
288 idx = info.entry_number;
289
290 if (idx == -1) {
291 idx = get_free_idx(current);
292 if (idx < 0)
293 return idx;
294 info.entry_number = idx;
295 /* Tell the user which slot we chose for him.*/
296 if (put_user(idx, &user_desc->entry_number))
297 return -EFAULT;
298 }
299
300 ret = do_set_thread_area(&info);
301 if (ret)
302 return ret;
303 return set_tls_entry(current, &info, idx, 1);
304}
305
306/*
307 * Perform set_thread_area on behalf of the traced child.
308 * Note: error handling is not done on the deferred load, and this differ from
309 * i386. However the only possible error are caused by bugs.
310 */
311int ptrace_set_thread_area(struct task_struct *child, int idx,
312 struct user_desc __user *user_desc)
313{
314 struct user_desc info;
315
316 if (!host_supports_tls)
317 return -EIO;
318
319 if (copy_from_user(&info, user_desc, sizeof(info)))
320 return -EFAULT;
321
322 return set_tls_entry(child, &info, idx, 0);
323}
324
325int sys_get_thread_area(struct user_desc __user *user_desc)
326{
327 struct user_desc info;
328 int idx, ret;
329
330 if (!host_supports_tls)
331 return -ENOSYS;
332
333 if (get_user(idx, &user_desc->entry_number))
334 return -EFAULT;
335
336 ret = get_tls_entry(current, &info, idx);
337 if (ret < 0)
338 goto out;
339
340 if (copy_to_user(user_desc, &info, sizeof(info)))
341 ret = -EFAULT;
342
343out:
344 return ret;
345}
346
347/*
348 * Perform get_thread_area on behalf of the traced child.
349 */
350int ptrace_get_thread_area(struct task_struct *child, int idx,
351 struct user_desc __user *user_desc)
352{
353 struct user_desc info;
354 int ret;
355
356 if (!host_supports_tls)
357 return -EIO;
358
359 ret = get_tls_entry(child, &info, idx);
360 if (ret < 0)
361 goto out;
362
363 if (copy_to_user(user_desc, &info, sizeof(info)))
364 ret = -EFAULT;
365out:
366 return ret;
367}
368
369/*
370 * This code is really i386-only, but it detects and logs x86_64 GDT indexes
371 * if a 32-bit UML is running on a 64-bit host.
372 */
373static int __init __setup_host_supports_tls(void)
374{
375 check_host_supports_tls(&host_supports_tls, &host_gdt_entry_tls_min);
376 if (host_supports_tls) {
377 printk(KERN_INFO "Host TLS support detected\n");
378 printk(KERN_INFO "Detected host type: ");
379 switch (host_gdt_entry_tls_min) {
380 case GDT_ENTRY_TLS_MIN_I386:
381 printk(KERN_CONT "i386");
382 break;
383 case GDT_ENTRY_TLS_MIN_X86_64:
384 printk(KERN_CONT "x86_64");
385 break;
386 }
387 printk(KERN_CONT " (GDT indexes %d to %d)\n",
388 host_gdt_entry_tls_min,
389 host_gdt_entry_tls_min + GDT_ENTRY_TLS_ENTRIES);
390 } else
391 printk(KERN_ERR " Host TLS support NOT detected! "
392 "TLS support inside UML will not work\n");
393 return 0;
394}
395
396__initcall(__setup_host_supports_tls);
diff --git a/arch/x86/um/tls_64.c b/arch/x86/um/tls_64.c
new file mode 100644
index 000000000000..f7ba46200ecd
--- /dev/null
+++ b/arch/x86/um/tls_64.c
@@ -0,0 +1,17 @@
1#include "linux/sched.h"
2
3void clear_flushed_tls(struct task_struct *task)
4{
5}
6
7int arch_copy_tls(struct task_struct *t)
8{
9 /*
10 * If CLONE_SETTLS is set, we need to save the thread id
11 * (which is argument 5, child_tid, of clone) so it can be set
12 * during context switches.
13 */
14 t->thread.arch.fs = t->thread.regs.regs.gp[R8 / sizeof(long)];
15
16 return 0;
17}
diff --git a/arch/x86/um/user-offsets.c b/arch/x86/um/user-offsets.c
new file mode 100644
index 000000000000..ca49be8ddd0c
--- /dev/null
+++ b/arch/x86/um/user-offsets.c
@@ -0,0 +1,80 @@
1#include <stdio.h>
2#include <stddef.h>
3#include <signal.h>
4#include <sys/poll.h>
5#include <sys/mman.h>
6#include <sys/user.h>
7#define __FRAME_OFFSETS
8#include <asm/ptrace.h>
9#include <asm/types.h>
10
11#define DEFINE(sym, val) \
12 asm volatile("\n->" #sym " %0 " #val : : "i" (val))
13
14#define DEFINE_LONGS(sym, val) \
15 asm volatile("\n->" #sym " %0 " #val : : "i" (val/sizeof(unsigned long)))
16
17void foo(void)
18{
19#ifdef __i386__
20 DEFINE_LONGS(HOST_FP_SIZE, sizeof(struct user_fpregs_struct));
21 DEFINE_LONGS(HOST_FPX_SIZE, sizeof(struct user_fpxregs_struct));
22
23 DEFINE(HOST_IP, EIP);
24 DEFINE(HOST_SP, UESP);
25 DEFINE(HOST_EFLAGS, EFL);
26 DEFINE(HOST_AX, EAX);
27 DEFINE(HOST_BX, EBX);
28 DEFINE(HOST_CX, ECX);
29 DEFINE(HOST_DX, EDX);
30 DEFINE(HOST_SI, ESI);
31 DEFINE(HOST_DI, EDI);
32 DEFINE(HOST_BP, EBP);
33 DEFINE(HOST_CS, CS);
34 DEFINE(HOST_SS, SS);
35 DEFINE(HOST_DS, DS);
36 DEFINE(HOST_FS, FS);
37 DEFINE(HOST_ES, ES);
38 DEFINE(HOST_GS, GS);
39 DEFINE(HOST_ORIG_AX, ORIG_EAX);
40#else
41 DEFINE(HOST_FP_SIZE, sizeof(struct _fpstate) / sizeof(unsigned long));
42 DEFINE_LONGS(HOST_BX, RBX);
43 DEFINE_LONGS(HOST_CX, RCX);
44 DEFINE_LONGS(HOST_DI, RDI);
45 DEFINE_LONGS(HOST_SI, RSI);
46 DEFINE_LONGS(HOST_DX, RDX);
47 DEFINE_LONGS(HOST_BP, RBP);
48 DEFINE_LONGS(HOST_AX, RAX);
49 DEFINE_LONGS(HOST_R8, R8);
50 DEFINE_LONGS(HOST_R9, R9);
51 DEFINE_LONGS(HOST_R10, R10);
52 DEFINE_LONGS(HOST_R11, R11);
53 DEFINE_LONGS(HOST_R12, R12);
54 DEFINE_LONGS(HOST_R13, R13);
55 DEFINE_LONGS(HOST_R14, R14);
56 DEFINE_LONGS(HOST_R15, R15);
57 DEFINE_LONGS(HOST_ORIG_AX, ORIG_RAX);
58 DEFINE_LONGS(HOST_CS, CS);
59 DEFINE_LONGS(HOST_SS, SS);
60 DEFINE_LONGS(HOST_EFLAGS, EFLAGS);
61#if 0
62 DEFINE_LONGS(HOST_FS, FS);
63 DEFINE_LONGS(HOST_GS, GS);
64 DEFINE_LONGS(HOST_DS, DS);
65 DEFINE_LONGS(HOST_ES, ES);
66#endif
67
68 DEFINE_LONGS(HOST_IP, RIP);
69 DEFINE_LONGS(HOST_SP, RSP);
70#endif
71
72 DEFINE(UM_FRAME_SIZE, sizeof(struct user_regs_struct));
73 DEFINE(UM_POLLIN, POLLIN);
74 DEFINE(UM_POLLPRI, POLLPRI);
75 DEFINE(UM_POLLOUT, POLLOUT);
76
77 DEFINE(UM_PROT_READ, PROT_READ);
78 DEFINE(UM_PROT_WRITE, PROT_WRITE);
79 DEFINE(UM_PROT_EXEC, PROT_EXEC);
80}
diff --git a/arch/x86/um/vdso/Makefile b/arch/x86/um/vdso/Makefile
new file mode 100644
index 000000000000..6c803ca49b5d
--- /dev/null
+++ b/arch/x86/um/vdso/Makefile
@@ -0,0 +1,90 @@
1#
2# Building vDSO images for x86.
3#
4
5VDSO64-y := y
6
7vdso-install-$(VDSO64-y) += vdso.so
8
9
10# files to link into the vdso
11vobjs-y := vdso-note.o um_vdso.o
12
13# files to link into kernel
14obj-$(VDSO64-y) += vdso.o vma.o
15
16vobjs := $(foreach F,$(vobjs-y),$(obj)/$F)
17
18$(obj)/vdso.o: $(obj)/vdso.so
19
20targets += vdso.so vdso.so.dbg vdso.lds $(vobjs-y)
21
22export CPPFLAGS_vdso.lds += -P -C
23
24VDSO_LDFLAGS_vdso.lds = -m64 -Wl,-soname=linux-vdso.so.1 \
25 -Wl,-z,max-page-size=4096 -Wl,-z,common-page-size=4096
26
27$(obj)/vdso.o: $(src)/vdso.S $(obj)/vdso.so
28
29$(obj)/vdso.so.dbg: $(src)/vdso.lds $(vobjs) FORCE
30 $(call if_changed,vdso)
31
32$(obj)/%.so: OBJCOPYFLAGS := -S
33$(obj)/%.so: $(obj)/%.so.dbg FORCE
34 $(call if_changed,objcopy)
35
36#
37# Don't omit frame pointers for ease of userspace debugging, but do
38# optimize sibling calls.
39#
40CFL := $(PROFILING) -mcmodel=small -fPIC -O2 -fasynchronous-unwind-tables -m64 \
41 $(filter -g%,$(KBUILD_CFLAGS)) $(call cc-option, -fno-stack-protector) \
42 -fno-omit-frame-pointer -foptimize-sibling-calls
43
44$(vobjs): KBUILD_CFLAGS += $(CFL)
45
46#
47# vDSO code runs in userspace and -pg doesn't help with profiling anyway.
48#
49CFLAGS_REMOVE_vdso-note.o = -pg -fprofile-arcs -ftest-coverage
50CFLAGS_REMOVE_um_vdso.o = -pg -fprofile-arcs -ftest-coverage
51
52targets += vdso-syms.lds
53obj-$(VDSO64-y) += vdso-syms.lds
54
55#
56# Match symbols in the DSO that look like VDSO*; produce a file of constants.
57#
58sed-vdsosym := -e 's/^00*/0/' \
59 -e 's/^\([0-9a-fA-F]*\) . \(VDSO[a-zA-Z0-9_]*\)$$/\2 = 0x\1;/p'
60quiet_cmd_vdsosym = VDSOSYM $@
61define cmd_vdsosym
62 $(NM) $< | LC_ALL=C sed -n $(sed-vdsosym) | LC_ALL=C sort > $@
63endef
64
65$(obj)/%-syms.lds: $(obj)/%.so.dbg FORCE
66 $(call if_changed,vdsosym)
67
68#
69# The DSO images are built using a special linker script.
70#
71quiet_cmd_vdso = VDSO $@
72 cmd_vdso = $(CC) -nostdlib -o $@ \
73 $(VDSO_LDFLAGS) $(VDSO_LDFLAGS_$(filter %.lds,$(^F))) \
74 -Wl,-T,$(filter %.lds,$^) $(filter %.o,$^) && \
75 sh $(srctree)/$(src)/checkundef.sh '$(NM)' '$@'
76
77VDSO_LDFLAGS = -fPIC -shared $(call cc-ldoption, -Wl$(comma)--hash-style=sysv)
78GCOV_PROFILE := n
79
80#
81# Install the unstripped copy of vdso*.so listed in $(vdso-install-y).
82#
83quiet_cmd_vdso_install = INSTALL $@
84 cmd_vdso_install = cp $(obj)/$@.dbg $(MODLIB)/vdso/$@
85$(vdso-install-y): %.so: $(obj)/%.so.dbg FORCE
86 @mkdir -p $(MODLIB)/vdso
87 $(call cmd,vdso_install)
88
89PHONY += vdso_install $(vdso-install-y)
90vdso_install: $(vdso-install-y)
diff --git a/arch/x86/um/vdso/checkundef.sh b/arch/x86/um/vdso/checkundef.sh
new file mode 100644
index 000000000000..7ee90a9b549d
--- /dev/null
+++ b/arch/x86/um/vdso/checkundef.sh
@@ -0,0 +1,10 @@
1#!/bin/sh
2nm="$1"
3file="$2"
4$nm "$file" | grep '^ *U' > /dev/null 2>&1
5if [ $? -eq 1 ]; then
6 exit 0
7else
8 echo "$file: undefined symbols found" >&2
9 exit 1
10fi
diff --git a/arch/x86/um/vdso/um_vdso.c b/arch/x86/um/vdso/um_vdso.c
new file mode 100644
index 000000000000..7c441b59d375
--- /dev/null
+++ b/arch/x86/um/vdso/um_vdso.c
@@ -0,0 +1,71 @@
1/*
2 * Copyright (C) 2011 Richard Weinberger <richrd@nod.at>
3 *
4 * This program is free software; you can redistribute it and/or modify
5 * it under the terms of the GNU General Public License version 2 as
6 * published by the Free Software Foundation.
7 *
8 * This vDSO turns all calls into a syscall so that UML can trap them.
9 */
10
11
12/* Disable profiling for userspace code */
13#define DISABLE_BRANCH_PROFILING
14
15#include <linux/time.h>
16#include <linux/getcpu.h>
17#include <asm/unistd.h>
18
19int __vdso_clock_gettime(clockid_t clock, struct timespec *ts)
20{
21 long ret;
22
23 asm("syscall" : "=a" (ret) :
24 "0" (__NR_clock_gettime), "D" (clock), "S" (ts) : "memory");
25
26 return ret;
27}
28int clock_gettime(clockid_t, struct timespec *)
29 __attribute__((weak, alias("__vdso_clock_gettime")));
30
31int __vdso_gettimeofday(struct timeval *tv, struct timezone *tz)
32{
33 long ret;
34
35 asm("syscall" : "=a" (ret) :
36 "0" (__NR_gettimeofday), "D" (tv), "S" (tz) : "memory");
37
38 return ret;
39}
40int gettimeofday(struct timeval *, struct timezone *)
41 __attribute__((weak, alias("__vdso_gettimeofday")));
42
43time_t __vdso_time(time_t *t)
44{
45 long secs;
46
47 asm volatile("syscall"
48 : "=a" (secs)
49 : "0" (__NR_time), "D" (t) : "cc", "r11", "cx", "memory");
50
51 return secs;
52}
53int time(time_t *t) __attribute__((weak, alias("__vdso_time")));
54
55long
56__vdso_getcpu(unsigned *cpu, unsigned *node, struct getcpu_cache *unused)
57{
58 /*
59 * UML does not support SMP, we can cheat here. :)
60 */
61
62 if (cpu)
63 *cpu = 0;
64 if (node)
65 *node = 0;
66
67 return 0;
68}
69
70long getcpu(unsigned *cpu, unsigned *node, struct getcpu_cache *tcache)
71 __attribute__((weak, alias("__vdso_getcpu")));
diff --git a/arch/x86/um/vdso/vdso-layout.lds.S b/arch/x86/um/vdso/vdso-layout.lds.S
new file mode 100644
index 000000000000..634a2cf62046
--- /dev/null
+++ b/arch/x86/um/vdso/vdso-layout.lds.S
@@ -0,0 +1,64 @@
1/*
2 * Linker script for vDSO. This is an ELF shared object prelinked to
3 * its virtual address, and with only one read-only segment.
4 * This script controls its layout.
5 */
6
7SECTIONS
8{
9 . = VDSO_PRELINK + SIZEOF_HEADERS;
10
11 .hash : { *(.hash) } :text
12 .gnu.hash : { *(.gnu.hash) }
13 .dynsym : { *(.dynsym) }
14 .dynstr : { *(.dynstr) }
15 .gnu.version : { *(.gnu.version) }
16 .gnu.version_d : { *(.gnu.version_d) }
17 .gnu.version_r : { *(.gnu.version_r) }
18
19 .note : { *(.note.*) } :text :note
20
21 .eh_frame_hdr : { *(.eh_frame_hdr) } :text :eh_frame_hdr
22 .eh_frame : { KEEP (*(.eh_frame)) } :text
23
24 .dynamic : { *(.dynamic) } :text :dynamic
25
26 .rodata : { *(.rodata*) } :text
27 .data : {
28 *(.data*)
29 *(.sdata*)
30 *(.got.plt) *(.got)
31 *(.gnu.linkonce.d.*)
32 *(.bss*)
33 *(.dynbss*)
34 *(.gnu.linkonce.b.*)
35 }
36
37 .altinstructions : { *(.altinstructions) }
38 .altinstr_replacement : { *(.altinstr_replacement) }
39
40 /*
41 * Align the actual code well away from the non-instruction data.
42 * This is the best thing for the I-cache.
43 */
44 . = ALIGN(0x100);
45
46 .text : { *(.text*) } :text =0x90909090
47}
48
49/*
50 * Very old versions of ld do not recognize this name token; use the constant.
51 */
52#define PT_GNU_EH_FRAME 0x6474e550
53
54/*
55 * We must supply the ELF program headers explicitly to get just one
56 * PT_LOAD segment, and set the flags explicitly to make segments read-only.
57 */
58PHDRS
59{
60 text PT_LOAD FLAGS(5) FILEHDR PHDRS; /* PF_R|PF_X */
61 dynamic PT_DYNAMIC FLAGS(4); /* PF_R */
62 note PT_NOTE FLAGS(4); /* PF_R */
63 eh_frame_hdr PT_GNU_EH_FRAME;
64}
diff --git a/arch/x86/um/vdso/vdso-note.S b/arch/x86/um/vdso/vdso-note.S
new file mode 100644
index 000000000000..79a071e4357e
--- /dev/null
+++ b/arch/x86/um/vdso/vdso-note.S
@@ -0,0 +1,12 @@
1/*
2 * This supplies .note.* sections to go into the PT_NOTE inside the vDSO text.
3 * Here we can supply some information useful to userland.
4 */
5
6#include <linux/uts.h>
7#include <linux/version.h>
8#include <linux/elfnote.h>
9
10ELFNOTE_START(Linux, 0, "a")
11 .long LINUX_VERSION_CODE
12ELFNOTE_END
diff --git a/arch/x86/um/vdso/vdso.S b/arch/x86/um/vdso/vdso.S
new file mode 100644
index 000000000000..1cb468adacbb
--- /dev/null
+++ b/arch/x86/um/vdso/vdso.S
@@ -0,0 +1,10 @@
1#include <linux/init.h>
2
3__INITDATA
4
5 .globl vdso_start, vdso_end
6vdso_start:
7 .incbin "arch/x86/um/vdso/vdso.so"
8vdso_end:
9
10__FINIT
diff --git a/arch/x86/um/vdso/vdso.lds.S b/arch/x86/um/vdso/vdso.lds.S
new file mode 100644
index 000000000000..b96b2677cad8
--- /dev/null
+++ b/arch/x86/um/vdso/vdso.lds.S
@@ -0,0 +1,32 @@
1/*
2 * Linker script for 64-bit vDSO.
3 * We #include the file to define the layout details.
4 * Here we only choose the prelinked virtual address.
5 *
6 * This file defines the version script giving the user-exported symbols in
7 * the DSO. We can define local symbols here called VDSO* to make their
8 * values visible using the asm-x86/vdso.h macros from the kernel proper.
9 */
10
11#define VDSO_PRELINK 0xffffffffff700000
12#include "vdso-layout.lds.S"
13
14/*
15 * This controls what userland symbols we export from the vDSO.
16 */
17VERSION {
18 LINUX_2.6 {
19 global:
20 clock_gettime;
21 __vdso_clock_gettime;
22 gettimeofday;
23 __vdso_gettimeofday;
24 getcpu;
25 __vdso_getcpu;
26 time;
27 __vdso_time;
28 local: *;
29 };
30}
31
32VDSO64_PRELINK = VDSO_PRELINK;
diff --git a/arch/x86/um/vdso/vma.c b/arch/x86/um/vdso/vma.c
new file mode 100644
index 000000000000..91f4ec9a0a56
--- /dev/null
+++ b/arch/x86/um/vdso/vma.c
@@ -0,0 +1,74 @@
1/*
2 * Copyright (C) 2011 Richard Weinberger <richrd@nod.at>
3 *
4 * This program is free software; you can redistribute it and/or modify
5 * it under the terms of the GNU General Public License version 2 as
6 * published by the Free Software Foundation.
7 */
8
9#include <linux/slab.h>
10#include <linux/sched.h>
11#include <linux/mm.h>
12#include <asm/page.h>
13#include <linux/init.h>
14
15unsigned int __read_mostly vdso_enabled = 1;
16unsigned long um_vdso_addr;
17
18extern unsigned long task_size;
19extern char vdso_start[], vdso_end[];
20
21static struct page **vdsop;
22
23static int __init init_vdso(void)
24{
25 struct page *um_vdso;
26
27 BUG_ON(vdso_end - vdso_start > PAGE_SIZE);
28
29 um_vdso_addr = task_size - PAGE_SIZE;
30
31 vdsop = kmalloc(sizeof(struct page *), GFP_KERNEL);
32 if (!vdsop)
33 goto oom;
34
35 um_vdso = alloc_page(GFP_KERNEL);
36 if (!um_vdso) {
37 kfree(vdsop);
38
39 goto oom;
40 }
41
42 copy_page(page_address(um_vdso), vdso_start);
43 *vdsop = um_vdso;
44
45 return 0;
46
47oom:
48 printk(KERN_ERR "Cannot allocate vdso\n");
49 vdso_enabled = 0;
50
51 return -ENOMEM;
52}
53subsys_initcall(init_vdso);
54
55int arch_setup_additional_pages(struct linux_binprm *bprm, int uses_interp)
56{
57 int err;
58 struct mm_struct *mm = current->mm;
59
60 if (!vdso_enabled)
61 return 0;
62
63 down_write(&mm->mmap_sem);
64
65 err = install_special_mapping(mm, um_vdso_addr, PAGE_SIZE,
66 VM_READ|VM_EXEC|
67 VM_MAYREAD|VM_MAYWRITE|VM_MAYEXEC|
68 VM_ALWAYSDUMP,
69 vdsop);
70
71 up_write(&mm->mmap_sem);
72
73 return err;
74}