This part adds support for sys_indirect on x86 and x86-64.
b/arch/x86/ia32/ia32entry.S | 1 +
b/arch/x86/kernel/syscall_table_32.S | 1 +
b/include/asm-x86/indirect.h | 5 +++++
b/include/asm-x86/indirect_32.h | 27 +++++++++++++++++++++++++++
b/include/asm-x86/indirect_64.h | 30 ++++++++++++++++++++++++++++++
b/include/asm-x86/unistd_32.h | 3 ++-
b/include/asm-x86/unistd_64.h | 2 ++
7 files changed, 68 insertions(+), 1 deletion(-)
--- a/arch/x86/ia32/ia32entry.S
+++ b/arch/x86/ia32/ia32entry.S
@@ -726,4 +726,5 @@ ia32_sys_call_table:
.quad compat_sys_timerfd
.quad sys_eventfd
.quad sys32_fallocate
+ .quad sys_indirect /* 325 */
ia32_syscall_end:
diff --git a/arch/x86/kernel/syscall_table_32.S b/arch/x86/kernel/syscall_table_32.S
index 8344c70..92095b2 100644
--- a/arch/x86/kernel/syscall_table_32.S
+++ b/arch/x86/kernel/syscall_table_32.S
@@ -324,3 +324,4 @@ ENTRY(sys_call_table)
.long sys_timerfd
.long sys_eventfd
.long sys_fallocate
+ .long sys_indirect /* 325 */
diff --git a/include/asm-x86/unistd_32.h b/include/asm-x86/unistd_32.h
index 9b15545..8ee0b20 100644
--- a/include/asm-x86/unistd_32.h
+++ b/include/asm-x86/unistd_32.h
@@ -330,10 +330,11 @@
#define __NR_timerfd 322
#define __NR_eventfd 323
#define __NR_fallocate 324
+#define __NR_indirect 325
#ifdef __KERNEL__
-#define NR_syscalls 325
+#define NR_syscalls 326
#define __ARCH_WANT_IPC_PARSE_VERSION
#define __ARCH_WANT_OLD_READDIR
diff --git a/include/asm-x86/unistd_64.h b/include/asm-x86/unistd_64.h
index 5ff4d3e..66eab33 100644
--- a/include/asm-x86/unistd_64.h
+++ b/include/asm-x86/unistd_64.h
@@ -635,6 +635,8 @@ __SYSCALL(__NR_timerfd, sys_timerfd)
__SYSCALL(__NR_eventfd, sys_eventfd)
#define __NR_fallocate 285
__SYSCALL(__NR_fallocate, sys_fallocate)
+#define __NR_indirect 286
+__SYSCALL(__NR_indirect, sys_indirect)
#ifndef __NO_STUBS
#define __ARCH_WANT_OLD_READDIR
diff --git a/include/linux/sched.h b/include/linux/sched.h
index ee800e7..e4e8a22 100644
--- /dev/null 2007-09-23 16:36:38.465394704 -0700
+++ b/include/asm-x86/indirect_32.h 2007-11-14 22:16:50.000000000 -0800
@@ -0,0 +1,27 @@
+#ifndef _ASM_X86_INDIRECT_32_H
+#define _ASM_X86_INDIRECT_32_H
+
+struct indirect_registers {
+ __u32 eax;
+ __u32 ebx;
+ __u32 ecx;
+ __u32 edx;
+ __u32 esi;
+ __u32 edi;
+ __u32 ebp;
+};
+
+#define CALL_INDIRECT(regs) \
+ ({ long res; \
+ if (unlikely((regs)->eax > NR_syscalls)) \
+ res = -ENOSYS; \
+ else if (unlikely((regs)->eax == __NR_indirect)) \
+ res = -EINVAL; \
+ else { \
+ extern long (*sys_call_table[]) (__u32, __u32, __u32, __u32, __u32, __u32); \
+ res = sys_call_table[(regs)->eax] ((regs)->ebx, (regs)->ecx, (regs)->edx, \
+ (regs)->esi, (regs)->edi, (regs)->ebp); \
+ } \
+ res; })
+
+#endif
--- /dev/null 2007-09-23 16:36:38.465394704 -0700
+++ b/include/asm-x86/indirect_64.h 2007-11-14 22:17:43.000000000 -0800
@@ -0,0 +1,30 @@
+#ifndef _ASM_X86_INDIRECT_64_H
+#define _ASM_X86_INDIRECT_64_H
+
+struct indirect_registers {
+ __u64 rax;
+ __u64 rdi;
+ __u64 rsi;
+ __u64 rdx;
+ __u64 r10;
+ __u64 r8;
+ __u64 r9;
+};
+
+
+#define CALL_INDIRECT(regs) \
+ CALL_INDIRECT_WMAX(regs, __NR_syscall_max)
+#define CALL_INDIRECT_WMAX(regs, max) \
+ ({ long res; \
+ if (unlikely((regs)->rax > max)) \
+ res = -ENOSYS; \
+ else if (unlikely((regs)->rax == __NR_indirect)) \
+ res = -EINVAL; \
+ else { \
+ extern long (*sys_call_table[]) (__u64, __u64, __u64, __u64, __u64, __u64); \
+ res = sys_call_table[(regs)->rax] ((regs)->rdi, (regs)->rsi, (regs)->rdx, \
+ (regs)->r10, (regs)->r8, (regs)->r9); \
+ } \
+ res; })
+
+#endif
--- /dev/null 2007-09-23 16:36:38.465394704 -0700
+++ b/include/asm-x86/indirect.h 2007-11-14 17:15:58.000000000 -0800
@@ -0,0 +1,5 @@
+#ifdef CONFIG_X86_32
+# include "indirect_32.h"
+#else
+# include "indirect_64.h"
+#endif
[ Ingo, Thomas - see the whole series on linux-kernel ]
On Thu, 15 Nov 2007, Ulrich Drepper wrote:
>
> This part adds support for sys_indirect on x86 and x86-64.
The thing is, not all system calls can do this.
Some system calls are magic, and don't just take the arguments in
registers: they also care about the actual stack pointer and the whole
pt_regs struct when returning to user mode.
So this does need more infrastructure: some way of marking which system
calls cannot be executed indirectly.
The magic system calls are things like:
- sys_iopl() - this one changes the eflags value restored on iret
- execve/clone/vfork() - need direct access to pt_regs
- vm86() - does magic with the stack, cares about pt_regs
- sigreturn - magic pt_regs accesses again
and there may be others I have forgotten about.
Calling these system calls from C code will just corrupt the kernel stack,
and is a big big no-no.
Linus