[PATCH 1/1] ptrace_vm: let us simplify the code for ptrace and add useful features for VM
From: Renzo Davoli
Date: Mon Jun 16 2008 - 04:21:24 EST
This patch adds the new PTRACE_VM_SKIPCALL and PTRACE_VM_SKIPEXIT
tags for ptrace's addr parameter.
In this way it is possible to (eventually) get rid of PTRACE_SYSEMU
PTRACE_SYSEMU_SINGLESTEP, while providing not only the same features
but a more general support for Virtual Machines.
renzo
Signed-off-by: Renzo Davoli <renzo@xxxxxxxxxxx>
---
diff -Naur linux-2.6.26-rc6/arch/powerpc/kernel/entry_32.S linux-2.6.26-rc6-ptrace_vm/arch/powerpc/kernel/entry_32.S
--- linux-2.6.26-rc6/arch/powerpc/kernel/entry_32.S 2008-06-13 13:33:26.000000000 +0200
+++ linux-2.6.26-rc6-ptrace_vm/arch/powerpc/kernel/entry_32.S 2008-06-14 16:42:36.000000000 +0200
@@ -299,6 +299,7 @@
stw r0,_TRAP(r1)
addi r3,r1,STACK_FRAME_OVERHEAD
bl do_syscall_trace_enter
+ mr r10,r3
lwz r0,GPR0(r1) /* Restore original registers */
lwz r3,GPR3(r1)
lwz r4,GPR4(r1)
@@ -307,6 +308,8 @@
lwz r7,GPR7(r1)
lwz r8,GPR8(r1)
REST_NVGPRS(r1)
+ cmpwi r10,0
+ bne- ret_from_syscall
b syscall_dotrace_cont
syscall_exit_work:
diff -Naur linux-2.6.26-rc6/arch/powerpc/kernel/entry_64.S linux-2.6.26-rc6-ptrace_vm/arch/powerpc/kernel/entry_64.S
--- linux-2.6.26-rc6/arch/powerpc/kernel/entry_64.S 2008-06-13 13:33:26.000000000 +0200
+++ linux-2.6.26-rc6-ptrace_vm/arch/powerpc/kernel/entry_64.S 2008-06-14 16:42:36.000000000 +0200
@@ -213,6 +213,7 @@
bl .save_nvgprs
addi r3,r1,STACK_FRAME_OVERHEAD
bl .do_syscall_trace_enter
+ mr r11,r3
ld r0,GPR0(r1) /* Restore original registers */
ld r3,GPR3(r1)
ld r4,GPR4(r1)
@@ -223,6 +224,8 @@
addi r9,r1,STACK_FRAME_OVERHEAD
clrrdi r10,r1,THREAD_SHIFT
ld r10,TI_FLAGS(r10)
+ cmpwi r11,0
+ bne- syscall_exit
b syscall_dotrace_cont
syscall_enosys:
diff -Naur linux-2.6.26-rc6/arch/powerpc/kernel/ptrace.c linux-2.6.26-rc6-ptrace_vm/arch/powerpc/kernel/ptrace.c
--- linux-2.6.26-rc6/arch/powerpc/kernel/ptrace.c 2008-04-17 04:49:44.000000000 +0200
+++ linux-2.6.26-rc6-ptrace_vm/arch/powerpc/kernel/ptrace.c 2008-06-14 16:42:36.000000000 +0200
@@ -867,7 +867,7 @@
}
}
-void do_syscall_trace_enter(struct pt_regs *regs)
+int do_syscall_trace_enter(struct pt_regs *regs)
{
secure_computing(regs->gpr[0]);
@@ -891,6 +891,7 @@
regs->gpr[5] & 0xffffffff,
regs->gpr[6] & 0xffffffff);
}
+ return (current->ptrace & PT_SYSCALL_SKIPCALL);
}
void do_syscall_trace_leave(struct pt_regs *regs)
@@ -900,7 +901,9 @@
regs->result);
if ((test_thread_flag(TIF_SYSCALL_TRACE)
- || test_thread_flag(TIF_SINGLESTEP))
- && (current->ptrace & PT_PTRACED))
+ || test_thread_flag(TIF_SINGLESTEP))
+ && (current->ptrace & PT_PTRACED) &&
+ ((current->ptrace & PT_SYSCALL_SKIPEXIT)==0))
+
do_syscall_trace();
}
diff -Naur linux-2.6.26-rc6/arch/ppc/kernel/entry.S linux-2.6.26-rc6-ptrace_vm/arch/ppc/kernel/entry.S
--- linux-2.6.26-rc6/arch/ppc/kernel/entry.S 2008-04-17 04:49:44.000000000 +0200
+++ linux-2.6.26-rc6-ptrace_vm/arch/ppc/kernel/entry.S 2008-06-14 16:42:36.000000000 +0200
@@ -292,6 +292,7 @@
stw r0,TRAP(r1)
addi r3,r1,STACK_FRAME_OVERHEAD
bl do_syscall_trace_enter
+ mr r10,r3
lwz r0,GPR0(r1) /* Restore original registers */
lwz r3,GPR3(r1)
lwz r4,GPR4(r1)
@@ -300,6 +301,8 @@
lwz r7,GPR7(r1)
lwz r8,GPR8(r1)
REST_NVGPRS(r1)
+ cmpwi r10,0
+ bne- ret_from_syscall
b syscall_dotrace_cont
syscall_exit_work:
diff -Naur linux-2.6.26-rc6/arch/um/include/kern_util.h linux-2.6.26-rc6-ptrace_vm/arch/um/include/kern_util.h
--- linux-2.6.26-rc6/arch/um/include/kern_util.h 2008-04-17 04:49:44.000000000 +0200
+++ linux-2.6.26-rc6-ptrace_vm/arch/um/include/kern_util.h 2008-06-14 16:42:36.000000000 +0200
@@ -57,7 +57,7 @@
extern unsigned long to_irq_stack(unsigned long *mask_out);
extern unsigned long from_irq_stack(int nested);
-extern void syscall_trace(struct uml_pt_regs *regs, int entryexit);
+extern int syscall_trace(struct uml_pt_regs *regs, int entryexit);
extern int singlestepping(void *t);
extern void segv_handler(int sig, struct uml_pt_regs *regs);
diff -Naur linux-2.6.26-rc6/arch/um/include/ptrace_user.h linux-2.6.26-rc6-ptrace_vm/arch/um/include/ptrace_user.h
--- linux-2.6.26-rc6/arch/um/include/ptrace_user.h 2008-04-17 04:49:44.000000000 +0200
+++ linux-2.6.26-rc6-ptrace_vm/arch/um/include/ptrace_user.h 2008-06-14 16:42:36.000000000 +0200
@@ -40,9 +40,20 @@
#define PTRACE_OLDSETOPTIONS PTRACE_SETOPTIONS
#endif
+/* these constant should eventually enter in sys/ptrace.h */
+#ifndef PTRACE_SYSCALL_SKIPCALL
+#define PTRACE_SYSCALL_SKIPCALL 0x6
+#endif
+#ifndef PTRACE_SYSCALL_SKIPEXIT
+#define PTRACE_SYSCALL_SKIPEXIT 0x2
+#endif
+
void set_using_sysemu(int value);
int get_using_sysemu(void);
extern int sysemu_supported;
+void set_using_sysptvm(int value);
+int get_using_sysptvm(void);
+extern int sysptvm_supported;
#define SELECT_PTRACE_OPERATION(sysemu_mode, singlestep_mode) \
(((int[3][3] ) { \
diff -Naur linux-2.6.26-rc6/arch/um/kernel/process.c linux-2.6.26-rc6-ptrace_vm/arch/um/kernel/process.c
--- linux-2.6.26-rc6/arch/um/kernel/process.c 2008-06-13 13:33:26.000000000 +0200
+++ linux-2.6.26-rc6-ptrace_vm/arch/um/kernel/process.c 2008-06-14 16:42:36.000000000 +0200
@@ -322,7 +322,9 @@
}
static atomic_t using_sysemu = ATOMIC_INIT(0);
+static atomic_t using_sysptvm = ATOMIC_INIT(0);
int sysemu_supported;
+int sysptvm_supported;
void set_using_sysemu(int value)
{
@@ -336,6 +338,16 @@
return atomic_read(&using_sysemu);
}
+void set_using_sysptvm(int value)
+{
+ atomic_set(&using_sysptvm, value);
+}
+
+int get_using_sysptvm(void)
+{
+ return atomic_read(&using_sysptvm);
+}
+
static int proc_read_sysemu(char *buf, char **start, off_t offset, int size,int *eof, void *data)
{
if (snprintf(buf, size, "%d\n", get_using_sysemu()) < size)
@@ -358,27 +370,62 @@
return count;
}
-int __init make_proc_sysemu(void)
+static int proc_read_sysptvm(char *buf, char **start, off_t offset, int size,int *eof, void *data)
{
- struct proc_dir_entry *ent;
- if (!sysemu_supported)
- return 0;
+ int sysptvm=(get_using_sysptvm() != 0);
+ if (snprintf(buf, size, "%d\n", sysptvm) < size)
+ /* No overflow */
+ *eof = 1;
- ent = create_proc_entry("sysemu", 0600, NULL);
+ return strlen(buf);
+}
- if (ent == NULL)
- {
- printk(KERN_WARNING "Failed to register /proc/sysemu\n");
- return 0;
- }
+static int proc_write_sysptvm(struct file *file,const char __user *buf, unsigned long count,void *data)
+{
+ char tmp[2];
+
+ if (copy_from_user(tmp, buf, 1))
+ return -EFAULT;
+
+ if (tmp[0] == '0')
+ set_using_sysptvm(0);
+ if (tmp[0] == '1')
+ set_using_sysemu(/* XXX */ 6);
+ /* We use the first char, but pretend to write everything */
+ return count;
+}
- ent->read_proc = proc_read_sysemu;
- ent->write_proc = proc_write_sysemu;
+int __init make_proc_sysemu_or_sysptvm(void)
+{
+ struct proc_dir_entry *ent;
+ if (sysptvm_supported) {
+ ent = create_proc_entry("sysptvm", 0600, NULL);
+
+ if (ent == NULL)
+ {
+ printk(KERN_WARNING "Failed to register /proc/sysptvm\n");
+ return 0;
+ }
+
+ ent->read_proc = proc_read_sysptvm;
+ ent->write_proc = proc_write_sysptvm;
+ } else if (sysemu_supported) {
+ ent = create_proc_entry("sysemu", 0600, NULL);
+
+ if (ent == NULL)
+ {
+ printk(KERN_WARNING "Failed to register /proc/sysemu\n");
+ return 0;
+ }
+
+ ent->read_proc = proc_read_sysemu;
+ ent->write_proc = proc_write_sysemu;
+ }
return 0;
}
-late_initcall(make_proc_sysemu);
+late_initcall(make_proc_sysemu_or_sysptvm);
int singlestepping(void * t)
{
diff -Naur linux-2.6.26-rc6/arch/um/kernel/ptrace.c linux-2.6.26-rc6-ptrace_vm/arch/um/kernel/ptrace.c
--- linux-2.6.26-rc6/arch/um/kernel/ptrace.c 2008-04-17 04:49:44.000000000 +0200
+++ linux-2.6.26-rc6-ptrace_vm/arch/um/kernel/ptrace.c 2008-06-14 16:42:36.000000000 +0200
@@ -76,6 +76,8 @@
if (request == PTRACE_SYSCALL)
set_tsk_thread_flag(child, TIF_SYSCALL_TRACE);
else clear_tsk_thread_flag(child, TIF_SYSCALL_TRACE);
+ child->ptrace &= ~PT_SYSCALL_MASK;
+ child->ptrace |= (addr & PTRACE_SYSCALL_MASK) << 28;
child->exit_code = data;
wake_up_process(child);
ret = 0;
@@ -102,7 +104,9 @@
ret = -EIO;
if (!valid_signal(data))
break;
+ child->ptrace &= ~PT_SYSCALL_MASK;
clear_tsk_thread_flag(child, TIF_SYSCALL_TRACE);
+ child->ptrace |= (addr & PTRACE_SYSCALL_MASK) << 28;
set_singlestepping(child, 1);
child->exit_code = data;
/* give it a chance to run. */
@@ -245,7 +249,7 @@
* XXX Check PT_DTRACE vs TIF_SINGLESTEP for singlestepping check and
* PT_PTRACED vs TIF_SYSCALL_TRACE for syscall tracing check
*/
-void syscall_trace(struct uml_pt_regs *regs, int entryexit)
+int syscall_trace(struct uml_pt_regs *regs, int entryexit)
{
int is_singlestep = (current->ptrace & PT_DTRACE) && entryexit;
int tracesysgood;
@@ -267,10 +271,13 @@
send_sigtrap(current, regs, 0);
if (!test_thread_flag(TIF_SYSCALL_TRACE))
- return;
+ return 0;
if (!(current->ptrace & PT_PTRACED))
- return;
+ return 0;
+
+ if (entryexit && (current->ptrace & PT_SYSCALL_SKIPEXIT))
+ return 0;
/*
* the 0x80 provides a way for the tracing parent to distinguish
@@ -291,4 +298,8 @@
send_sig(current->exit_code, current, 1);
current->exit_code = 0;
}
+ if (!entryexit && (current->ptrace & PT_SYSCALL_SKIPCALL))
+ return 1;
+ else
+ return 0;
}
diff -Naur linux-2.6.26-rc6/arch/um/kernel/skas/syscall.c linux-2.6.26-rc6-ptrace_vm/arch/um/kernel/skas/syscall.c
--- linux-2.6.26-rc6/arch/um/kernel/skas/syscall.c 2008-04-17 04:49:44.000000000 +0200
+++ linux-2.6.26-rc6-ptrace_vm/arch/um/kernel/skas/syscall.c 2008-06-14 16:42:36.000000000 +0200
@@ -17,8 +17,9 @@
struct pt_regs *regs = container_of(r, struct pt_regs, regs);
long result;
int syscall;
+ int skip_call;
- syscall_trace(r, 0);
+ skip_call=syscall_trace(r, 0);
/*
* This should go in the declaration of syscall, but when I do that,
@@ -29,12 +30,14 @@
* gcc version 4.0.1 20050727 (Red Hat 4.0.1-5)
* in case it's a compiler bug.
*/
- syscall = UPT_SYSCALL_NR(r);
- if ((syscall >= NR_syscalls) || (syscall < 0))
- result = -ENOSYS;
- else result = EXECUTE_SYSCALL(syscall, regs);
+ if (skip_call == 0) {
+ syscall = UPT_SYSCALL_NR(r);
+ if ((syscall >= NR_syscalls) || (syscall < 0))
+ result = -ENOSYS;
+ else result = EXECUTE_SYSCALL(syscall, regs);
- REGS_SET_SYSCALL_RETURN(r->gp, result);
+ REGS_SET_SYSCALL_RETURN(r->gp, result);
+ }
syscall_trace(r, 1);
}
diff -Naur linux-2.6.26-rc6/arch/um/os-Linux/skas/process.c linux-2.6.26-rc6-ptrace_vm/arch/um/os-Linux/skas/process.c
--- linux-2.6.26-rc6/arch/um/os-Linux/skas/process.c 2008-06-13 13:33:26.000000000 +0200
+++ linux-2.6.26-rc6-ptrace_vm/arch/um/os-Linux/skas/process.c 2008-06-14 16:42:36.000000000 +0200
@@ -157,7 +157,7 @@
* (in local_using_sysemu
*/
static void handle_trap(int pid, struct uml_pt_regs *regs,
- int local_using_sysemu)
+ int local_using_sysptvm_or_sysemu)
{
int err, status;
@@ -167,7 +167,7 @@
/* Mark this as a syscall */
UPT_SYSCALL_NR(regs) = PT_SYSCALL_NR(regs->gp);
- if (!local_using_sysemu)
+ if (!local_using_sysptvm_or_sysemu)
{
err = ptrace(PTRACE_POKEUSR, pid, PT_SYSCALL_NR_OFFSET,
__NR_getpid);
@@ -354,6 +354,7 @@
int err, status, op, pid = userspace_pid[0];
/* To prevent races if using_sysemu changes under us.*/
int local_using_sysemu;
+ int local_using_sysptvm;
if (getitimer(ITIMER_VIRTUAL, &timer))
printk(UM_KERN_ERR "Failed to get itimer, errno = %d\n", errno);
@@ -375,11 +376,12 @@
/* Now we set local_using_sysemu to be used for one loop */
local_using_sysemu = get_using_sysemu();
+ local_using_sysptvm = get_using_sysptvm();
op = SELECT_PTRACE_OPERATION(local_using_sysemu,
singlestepping(NULL));
- if (ptrace(op, pid, 0, 0)) {
+ if (ptrace(op, pid, local_using_sysptvm, 0)) {
printk(UM_KERN_ERR "userspace - ptrace continue "
"failed, op = %d, errno = %d\n", op, errno);
fatal_sigsegv();
diff -Naur linux-2.6.26-rc6/arch/um/os-Linux/start_up.c linux-2.6.26-rc6-ptrace_vm/arch/um/os-Linux/start_up.c
--- linux-2.6.26-rc6/arch/um/os-Linux/start_up.c 2008-06-13 13:33:26.000000000 +0200
+++ linux-2.6.26-rc6-ptrace_vm/arch/um/os-Linux/start_up.c 2008-06-14 16:42:36.000000000 +0200
@@ -196,6 +196,35 @@
" See http://perso.wanadoo.fr/laurent.vivier/UML/ for further \n"
" information.\n\n");
+/* Changed only during early boot */
+static int force_sysptvm_disabled = 0;
+
+static int __init nosysptvm_cmd_param(char *str, int* add)
+{
+ force_sysptvm_disabled = 1;
+ return 0;
+}
+
+__uml_setup("nosysptvm", nosysptvm_cmd_param,
+"nosysptvm\n"
+" Turns off syscall emulation tags for ptrace (ptrace_vm) on.\n"
+" Ptrace_vm is a feature introduced by Renzo Davoli. It changes\n"
+" behaviour of ptrace() and helps reducing host context switch rate.\n"
+"\n");
+
+static int use_sysemu = 0;
+
+static int __init usesysemu_cmd_param(char *str, int* add)
+{
+ use_sysemu = 1;
+ return 0;
+}
+
+__uml_setup("usesysemu", usesysemu_cmd_param,
+"usesysemu\n"
+" Use sysemu instead of sysptvm even when the kernel supports it.\n\n"
+);
+
static void __init check_sysemu(void)
{
unsigned long regs[MAX_REG_NR];
@@ -286,6 +315,100 @@
non_fatal("missing\n");
}
+/* test thread code. This thread is started only to test
+ * which features are provided by the linux kernel */
+static int sysptvm_child(void *arg)
+{
+ int *featurep=arg;
+ int p[2]={-1,-1};
+ pid_t pid=os_getpid();
+ if(ptrace(PTRACE_TRACEME, 0, 0, 0) < 0){
+ perror("ptrace test_ptracemulti");
+ kill(pid, SIGKILL);
+ }
+ kill(pid, SIGSTOP);
+ *featurep=0;
+ os_getpid();
+ /* if it reaches this point in 1 stop it means that
+ * PTRACE_SYSCALL_SKIPEXIT works */
+ *featurep=PTRACE_SYSCALL_SKIPEXIT;
+ pipe(p);
+ /* if after a PTRACE_SYSCALL_SKIPCALL p[0] is already <0
+ * pipe has been really skipped */
+ if (p[0] < 0)
+ *featurep=PTRACE_SYSCALL_SKIPCALL;
+ else { /* clean up everything */
+ close(p[0]);
+ close(p[1]);
+ }
+ return 0;
+}
+
+/* kernel feature test:
+ * it returns:
+ * -1 error
+ * 0 old PTRACE_SYSCALL (addr is ignored)
+ * PTRACE_SYSCALL_SKIPEXIT: just skip_exit is provided
+ * PTRACE_SYSCALL_SKIPCALL: the entire syntax is implemented
+ * by the running kernel */
+static int __init test_ptrace_sysptvm(void) {
+ int pid, status, rv, feature;
+ static char stack[1024];
+ feature=0;
+
+ if((pid = clone(sysptvm_child, &stack[1020], SIGCHLD | CLONE_VM, &feature)) < 0)
+ return 0;
+ if(waitpid(pid, &status, WUNTRACED) < 0){
+ kill(pid, SIGKILL);
+ return 0;
+ }
+ /* restart and wait for the next syscall (getpid)*/
+ rv=ptrace(PTRACE_SYSCALL, pid, 0, 0);
+ if(waitpid(pid, &status, WUNTRACED) < 0)
+ goto out;
+ /* try to skip the exit call */
+ rv=ptrace(PTRACE_SYSCALL, pid, PTRACE_SYSCALL_SKIPEXIT, 0);
+ if (rv < 0)
+ goto out;
+ /* wait for the next stop */
+ if(waitpid(pid, &status, WUNTRACED) < 0)
+ goto out;
+ /* if feature is already 0 it means that this is the exit call,
+ * and it has not been skipped, otherwise this is the
+ * entry call for the system call "time" */
+ if (feature<PTRACE_SYSCALL_SKIPEXIT)
+ goto out;
+ /* restart (time) and and try to skip the entire call */
+ rv=ptrace(PTRACE_SYSCALL, pid, PTRACE_SYSCALL_SKIPCALL, 0);
+ if(waitpid(pid, &status, WUNTRACED) < 0)
+ return 0;
+out:
+ ptrace(PTRACE_KILL,pid,0,0);
+ /* eliminate zombie */
+ if(waitpid(pid, &status, WUNTRACED) < 0)
+ return 0;
+ return feature;
+}
+
+static int __init check_sysptvm(void)
+{
+ int feature=test_ptrace_sysptvm();
+
+ non_fatal("Checking ptrace new tags for syscall emulation...");
+ if (feature==PTRACE_SYSCALL_SKIPCALL) {
+ sysptvm_supported=1;
+ non_fatal("OK");
+ if (!force_sysptvm_disabled)
+ set_using_sysptvm(PTRACE_SYSCALL_SKIPCALL);
+ else
+ non_fatal(" (disabled)");
+ non_fatal("\n");
+ return 1;
+ } else
+ non_fatal("unsupported\n");
+ return 0;
+}
+
static void __init check_ptrace(void)
{
int pid, syscall, n, status;
@@ -323,7 +446,8 @@
}
stop_ptraced_child(pid, 0, 1);
non_fatal("OK\n");
- check_sysemu();
+ if (use_sysemu || !check_sysptvm())
+ check_sysemu();
}
extern void check_tmpexec(void);
diff -Naur linux-2.6.26-rc6/arch/x86/ia32/ia32entry.S linux-2.6.26-rc6-ptrace_vm/arch/x86/ia32/ia32entry.S
--- linux-2.6.26-rc6/arch/x86/ia32/ia32entry.S 2008-06-13 13:33:26.000000000 +0200
+++ linux-2.6.26-rc6-ptrace_vm/arch/x86/ia32/ia32entry.S 2008-06-14 16:46:41.000000000 +0200
@@ -165,6 +165,8 @@
movq $-ENOSYS,RAX(%rsp)/* ptrace can change this for a bad syscall */
movq %rsp,%rdi /* &pt_regs -> arg1 */
call syscall_trace_enter
+ testq %rax,%rax
+ jnz ia32_skipcall
LOAD_ARGS32 ARGOFFSET /* reload args from stack in case ptrace changed it */
RESTORE_REST
xchgl %ebp,%r9d
@@ -266,6 +268,8 @@
movq $-ENOSYS,RAX(%rsp) /* ptrace can change this for a bad syscall */
movq %rsp,%rdi /* &pt_regs -> arg1 */
call syscall_trace_enter
+ testq %rax,%rax
+ jnz ia32_skipcall
LOAD_ARGS32 ARGOFFSET /* reload args from stack in case ptrace changed it */
RESTORE_REST
xchgl %ebp,%r9d
@@ -342,6 +346,8 @@
movq $-ENOSYS,RAX(%rsp) /* ptrace can change this for a bad syscall */
movq %rsp,%rdi /* &pt_regs -> arg1 */
call syscall_trace_enter
+ testq %rax,%rax
+ jnz ia32_skipcall
LOAD_ARGS32 ARGOFFSET /* reload args from stack in case ptrace changed it */
RESTORE_REST
jmp ia32_do_syscall
@@ -352,6 +358,11 @@
movq $-ENOSYS,RAX-ARGOFFSET(%rsp)
jmp int_ret_from_sys_call
+ia32_skipcall:
+ LOAD_ARGS32 ARGOFFSET /* reload args from stack in case ptrace changed it */
+ RESTORE_REST
+ jmp int_ret_from_sys_call
+
quiet_ni_syscall:
movq $-ENOSYS,%rax
ret
diff -Naur linux-2.6.26-rc6/arch/x86/kernel/entry_64.S linux-2.6.26-rc6-ptrace_vm/arch/x86/kernel/entry_64.S
--- linux-2.6.26-rc6/arch/x86/kernel/entry_64.S 2008-06-13 13:33:26.000000000 +0200
+++ linux-2.6.26-rc6-ptrace_vm/arch/x86/kernel/entry_64.S 2008-06-14 16:42:36.000000000 +0200
@@ -316,6 +316,11 @@
movq $-ENOSYS,RAX-ARGOFFSET(%rsp)
jmp ret_from_sys_call
+skipcall:
+ LOAD_ARGS ARGOFFSET /* reload args from stack in case ptrace changed it */
+ RESTORE_REST
+ jmp ret_from_sys_call
+
/* Do syscall tracing */
tracesys:
SAVE_REST
@@ -323,6 +328,8 @@
FIXUP_TOP_OF_STACK %rdi
movq %rsp,%rdi
call syscall_trace_enter
+ testq %rax,%rax
+ jnz skipcall
LOAD_ARGS ARGOFFSET /* reload args from stack in case ptrace changed it */
RESTORE_REST
cmpq $__NR_syscall_max,%rax
diff -Naur linux-2.6.26-rc6/arch/x86/kernel/ptrace.c linux-2.6.26-rc6-ptrace_vm/arch/x86/kernel/ptrace.c
--- linux-2.6.26-rc6/arch/x86/kernel/ptrace.c 2008-06-13 13:33:26.000000000 +0200
+++ linux-2.6.26-rc6-ptrace_vm/arch/x86/kernel/ptrace.c 2008-06-14 16:42:36.000000000 +0200
@@ -1415,6 +1415,9 @@
if (!(current->ptrace & PT_PTRACED))
goto out;
+ if (entryexit && (current->ptrace & PT_SYSCALL_SKIPEXIT))
+ return 0;
+
/* If a process stops on the 1st tracepoint with SYSCALL_TRACE
* and then is resumed with SYSEMU_SINGLESTEP, it will come in
* here. We have to check this and return */
@@ -1442,7 +1445,8 @@
send_sig(current->exit_code, current, 1);
current->exit_code = 0;
}
- ret = is_sysemu;
+ ret = (is_sysemu || (!entryexit && (current->ptrace & PT_SYSCALL_SKIPCALL)));
+
out:
if (unlikely(current->audit_context) && !entryexit)
audit_syscall_entry(AUDIT_ARCH_I386, regs->orig_ax,
@@ -1481,7 +1485,7 @@
}
}
-asmlinkage void syscall_trace_enter(struct pt_regs *regs)
+asmlinkage long syscall_trace_enter(struct pt_regs *regs)
{
/* do the secure computing check first */
secure_computing(regs->orig_ax);
@@ -1503,6 +1507,8 @@
regs->dx, regs->r10);
}
}
+
+ return (current->ptrace & PT_SYSCALL_SKIPCALL);
}
asmlinkage void syscall_trace_leave(struct pt_regs *regs)
@@ -1512,7 +1518,8 @@
if ((test_thread_flag(TIF_SYSCALL_TRACE)
|| test_thread_flag(TIF_SINGLESTEP))
- && (current->ptrace & PT_PTRACED))
+ && (current->ptrace & PT_PTRACED) &&
+ ((current->ptrace & PT_SYSCALL_SKIPEXIT)==0))
syscall_trace(regs);
}
diff -Naur linux-2.6.26-rc6/include/linux/ptrace.h linux-2.6.26-rc6-ptrace_vm/include/linux/ptrace.h
--- linux-2.6.26-rc6/include/linux/ptrace.h 2008-06-13 13:33:30.000000000 +0200
+++ linux-2.6.26-rc6-ptrace_vm/include/linux/ptrace.h 2008-06-14 16:42:36.000000000 +0200
@@ -46,6 +46,11 @@
#define PTRACE_EVENT_VFORK_DONE 5
#define PTRACE_EVENT_EXIT 6
+/* options for new PTRACE_SYSCALL syntax*/
+#define PTRACE_SYSCALL_SKIPEXIT 0x2
+#define PTRACE_SYSCALL_SKIPCALL 0x6
+#define PTRACE_SYSCALL_MASK 0x00000006
+
#include <asm/ptrace.h>
#ifdef __KERNEL__
@@ -68,6 +73,10 @@
#define PT_TRACE_VFORK_DONE 0x00000100
#define PT_TRACE_EXIT 0x00000200
+#define PT_SYSCALL_SKIPEXIT 0x60000000
+#define PT_SYSCALL_SKIPCALL 0x40000000
+#define PT_SYSCALL_MASK 0x60000000
+
#define PT_TRACE_MASK 0x000003f4
/* single stepping state bits (used on ARM and PA-RISC) */
diff -Naur linux-2.6.26-rc6/kernel/ptrace.c linux-2.6.26-rc6-ptrace_vm/kernel/ptrace.c
--- linux-2.6.26-rc6/kernel/ptrace.c 2008-06-13 13:33:31.000000000 +0200
+++ linux-2.6.26-rc6-ptrace_vm/kernel/ptrace.c 2008-06-14 16:42:36.000000000 +0200
@@ -375,7 +375,7 @@
#define is_sysemu_singlestep(request) 0
#endif
-static int ptrace_resume(struct task_struct *child, long request, long data)
+static int ptrace_resume(struct task_struct *child, long request, long addr, long data)
{
if (!valid_signal(data))
return -EIO;
@@ -404,6 +404,9 @@
else
user_disable_single_step(child);
+ child->ptrace &= ~PT_SYSCALL_MASK;
+ child->ptrace |= (addr & PTRACE_SYSCALL_MASK) << 28;
+
child->exit_code = data;
wake_up_process(child);
@@ -465,12 +468,12 @@
#endif
case PTRACE_SYSCALL:
case PTRACE_CONT:
- return ptrace_resume(child, request, data);
+ return ptrace_resume(child, request, addr, data);
case PTRACE_KILL:
if (child->exit_state) /* already dead */
return 0;
- return ptrace_resume(child, request, SIGKILL);
+ return ptrace_resume(child, request, addr, SIGKILL);
default:
break;
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/