* [uml-devel] [RFC PATCH 10/10] SKAS4 - Guest get_mm and switch_mm support
@ 2008-01-14 21:36 Jeff Dike
0 siblings, 0 replies; 3+ messages in thread
From: Jeff Dike @ 2008-01-14 21:36 UTC (permalink / raw)
To: UML-user, uml-devel
diff --git a/arch/um/include/as-layout.h b/arch/um/include/as-layout.h
index a5cdf95..90ee798 100644
--- a/arch/um/include/as-layout.h
+++ b/arch/um/include/as-layout.h
@@ -17,6 +17,7 @@
#define ASM_STUB_CODE (UML_CONFIG_TOP_ADDR - 2 * UM_KERN_PAGE_SIZE)
#define ASM_STUB_DATA (UML_CONFIG_TOP_ADDR - UM_KERN_PAGE_SIZE)
#define ASM_STUB_START ASM_STUB_CODE
+#define ASM_STUB_END UML_CONFIG_TOP_ADDR
/*
* This file is included by the assembly stubs, which just want the
@@ -27,6 +28,7 @@
#define STUB_CODE ((unsigned long) ASM_STUB_CODE)
#define STUB_DATA ((unsigned long) ASM_STUB_DATA)
#define STUB_START ((unsigned long) ASM_STUB_START)
+#define STUB_END ((unsigned long) ASM_STUB_END)
#include "sysdep/ptrace.h"
diff --git a/arch/um/include/skas/mm_id.h b/arch/um/include/skas/mm_id.h
index 48dd098..a2e7643 100644
--- a/arch/um/include/skas/mm_id.h
+++ b/arch/um/include/skas/mm_id.h
@@ -7,7 +7,7 @@
#define __MM_ID_H
struct mm_id {
- union {
+ struct {
int mm_fd;
int pid;
} u;
diff --git a/arch/um/include/skas/skas.h b/arch/um/include/skas/skas.h
index 061a362..bd3af6c 100644
--- a/arch/um/include/skas/skas.h
+++ b/arch/um/include/skas/skas.h
@@ -6,14 +6,41 @@
#ifndef __SKAS_H
#define __SKAS_H
-#include "sysdep/ptrace.h"
+#ifndef __KERNEL__
+#include <unistd.h>
+#include <sys/syscall.h>
+#endif
+#include "uml-config.h"
-extern int have_siginfo_segv;
+#ifdef UML_CONFIG_X86_32
+#define __NR_new_mm 325
+#define __NR_switch_mm 326
+#else
+#define __NR_new_mm 286
+#define __NR_switch_mm 287
+#endif
+
+#define MM_COPY 0
+
+#define MM_ALL_REGS 0
+#define MM_SP_IP 1
+#define MM_SAME 2
+
+#define PTRACE_SWITCH_MM 33
+
+#ifndef __ASSEMBLY__
+
+#include "sysdep/ptrace.h"
extern int userspace_pid[];
extern int proc_mm, ptrace_faultinfo, ptrace_ldt;
extern int skas_needs_stub;
+extern int have_switch_mm;
+extern int have_ptrace_switch_mm;
+extern int have_siginfo_segv;
+extern int self_mm_fd;
+
extern int user_thread(unsigned long stack, int flags);
extern void new_thread_handler(void);
extern void handle_syscall(struct uml_pt_regs *regs);
@@ -22,4 +49,19 @@ extern void get_skas_faultinfo(int pid, struct faultinfo * fi);
extern long execute_syscall_skas(void *r);
extern unsigned long current_stub_stack(void);
+#ifndef __KERNEL__
+static inline long new_mm(int flags)
+{
+ return syscall(__NR_new_mm, MM_COPY, 0, 0, 0, 0, 0, 0);
+}
+
+static inline long switch_mm(int mm_fd, unsigned long flags,
+ unsigned long *new_regs, unsigned long *save_regs)
+{
+ return syscall(__NR_switch_mm, mm_fd, flags, new_regs, save_regs, 0, 0);
+}
+#endif
+
+#endif
+
#endif
diff --git a/arch/um/kernel/reboot.c b/arch/um/kernel/reboot.c
index 04cebcf..0a5468e 100644
--- a/arch/um/kernel/reboot.c
+++ b/arch/um/kernel/reboot.c
@@ -11,7 +11,7 @@ void (*pm_power_off)(void);
static void kill_off_processes(void)
{
- if(proc_mm)
+ if(proc_mm || have_switch_mm)
/*
* FIXME: need to loop over userspace_pids
*/
diff --git a/arch/um/kernel/skas/mmu.c b/arch/um/kernel/skas/mmu.c
index 3426711..3155263 100644
--- a/arch/um/kernel/skas/mmu.c
+++ b/arch/um/kernel/skas/mmu.c
@@ -65,6 +65,9 @@ static int init_stub_pte(struct mm_struct *mm, unsigned long proc,
return -ENOMEM;
}
+extern int copy_context_skas4(struct mm_id *id);
+extern int get_new_mm(void);
+
int init_new_context(struct task_struct *task, struct mm_struct *mm)
{
struct mm_context *from_mm = NULL;
@@ -109,6 +112,20 @@ int init_new_context(struct task_struct *task, struct mm_struct *mm)
}
to_mm->id.u.mm_fd = ret;
}
+ else if (have_switch_mm) {
+ to_mm->id.u.mm_fd = get_new_mm();
+ if(to_mm->id.u.mm_fd < 0) {
+ ret = to_mm->id.u.mm_fd;
+ goto out_free;
+ }
+
+ ret = copy_context_skas4(&to_mm->id);
+ if (ret < 0) {
+ os_close_file(to_mm->id.u.mm_fd);
+ to_mm->id.u.mm_fd = -1;
+ goto out_free;
+ }
+ }
else {
if (from_mm)
to_mm->id.u.pid = copy_context_skas0(stack,
@@ -136,11 +153,15 @@ void destroy_context(struct mm_struct *mm)
{
struct mm_context *mmu = &mm->context;
- if (proc_mm)
+ if (proc_mm || have_switch_mm)
os_close_file(mmu->id.u.mm_fd);
- else
+ else {
os_kill_ptraced_process(mmu->id.u.pid, 1);
+ if (have_switch_mm)
+ os_close_file(mmu->id.u.mm_fd);
+ }
+
if (!proc_mm || !ptrace_faultinfo) {
free_page(mmu->id.stack);
pte_lock_deinit(virt_to_page(mmu->last_page_table));
diff --git a/arch/um/kernel/skas/process.c b/arch/um/kernel/skas/process.c
index 2c6de0a..e5e8613 100644
--- a/arch/um/kernel/skas/process.c
+++ b/arch/um/kernel/skas/process.c
@@ -49,7 +49,7 @@ int __init start_uml(void)
{
stack_protections((unsigned long) &cpu0_irqstack);
set_sigstack(cpu0_irqstack, THREAD_SIZE);
- if (proc_mm)
+ if (proc_mm || have_switch_mm)
userspace_pid[0] = start_userspace(0);
init_new_thread_signals();
diff --git a/arch/um/kernel/um_arch.c b/arch/um/kernel/um_arch.c
index f1c7139..d92108b 100644
--- a/arch/um/kernel/um_arch.c
+++ b/arch/um/kernel/um_arch.c
@@ -271,7 +271,9 @@ int __init linux_main(int argc, char **argv)
can_do_skas();
- if (proc_mm && ptrace_faultinfo)
+ if (have_switch_mm)
+ mode = "SKAS4";
+ else if (proc_mm && ptrace_faultinfo)
mode = "SKAS3";
else
mode = "SKAS0";
diff --git a/arch/um/os-Linux/skas/mem.c b/arch/um/os-Linux/skas/mem.c
index efda5e1..cc86c0b 100644
--- a/arch/um/os-Linux/skas/mem.c
+++ b/arch/um/os-Linux/skas/mem.c
@@ -6,6 +6,7 @@
#include <stddef.h>
#include <unistd.h>
#include <errno.h>
+#include <signal.h>
#include <string.h>
#include <sys/mman.h>
#include "init.h"
@@ -22,7 +23,7 @@
#include "sysdep/stub.h"
#include "uml-config.h"
-extern unsigned long batch_syscall_stub, __syscall_stub_start;
+extern unsigned long batch_syscall_stub, switch_mm_stub, __syscall_stub_start;
extern void wait_stub_done(int pid);
@@ -43,7 +44,10 @@ static int __init init_syscall_regs(void)
unsigned long *stub_entry;
get_safe_registers(syscall_regs);
- stub_entry = &batch_syscall_stub;
+ if (have_switch_mm)
+ stub_entry = &switch_mm_stub;
+ else
+ stub_entry = &batch_syscall_stub;
syscall_regs[REGS_IP_INDEX] = STUB_CODE +
((unsigned long) stub_entry -
@@ -126,6 +130,32 @@ static long do_syscall_stub(struct mm_id *mm_idp, void **addr)
return ret;
}
+static long do_syscall_stub_skas4(struct mm_id *mm_idp, void **addr)
+{
+ long ret;
+ unsigned long *return_regs;
+ int err;
+ sigset_t sigs, old;
+
+ return_regs = (unsigned long *) (mm_idp->stack + UM_KERN_PAGE_SIZE) -
+ MAX_REG_NR;
+ *(return_regs - 1) = self_mm_fd;
+
+ return_regs = (unsigned long *) (STUB_DATA + UM_KERN_PAGE_SIZE) -
+ MAX_REG_NR;
+
+ sigfillset(&sigs);
+ sigprocmask(SIG_SETMASK, &sigs, &old);
+ err = switch_mm(mm_idp->u.mm_fd, MM_SP_IP, syscall_regs, return_regs);
+ sigprocmask(SIG_SETMASK, &old, NULL);
+
+ ret = syscall_stub_done(mm_idp->stack);
+
+ *addr = check_init_stack(mm_idp, NULL);
+
+ return ret;
+}
+
static int flush_syscalls(struct mm_id *mm_idp, void **addr, int extra)
{
unsigned long *stack = check_init_stack(mm_idp, *addr);
@@ -134,10 +164,16 @@ static int flush_syscalls(struct mm_id *mm_idp, void **addr, int extra)
current = ((unsigned long) stack) & ~UM_KERN_PAGE_MASK;
end = UM_KERN_PAGE_SIZE;
+ if(have_switch_mm)
+ end -= (MAX_REG_NR + 1) * sizeof(long);
+
if (current + (10 + extra) * sizeof(long) < end)
return 0;
- return do_syscall_stub(mm_idp, addr);
+ if (have_switch_mm)
+ return do_syscall_stub_skas4(mm_idp, addr);
+ else
+ return do_syscall_stub(mm_idp, addr);
}
long run_syscall_stub(struct mm_id *mm_idp, int syscall,
@@ -166,8 +202,12 @@ long run_syscall_stub(struct mm_id *mm_idp, int syscall,
*stack++ = expected;
*stack = 0;
- if (done)
- return do_syscall_stub(mm_idp, addr);
+ if (done) {
+ if (have_switch_mm)
+ return do_syscall_stub_skas4(mm_idp, addr);
+ else
+ return do_syscall_stub(mm_idp, addr);
+ }
*addr = stack;
diff --git a/arch/um/os-Linux/skas/process.c b/arch/um/os-Linux/skas/process.c
index 522d0f1..9c2c086 100644
--- a/arch/um/os-Linux/skas/process.c
+++ b/arch/um/os-Linux/skas/process.c
@@ -227,7 +227,7 @@ static int userspace_tramp(void *stack)
}
}
}
- if (!ptrace_faultinfo && (stack != NULL)) {
+ if (!ptrace_faultinfo) {
struct sigaction sa;
unsigned long v = STUB_CODE +
@@ -271,7 +271,7 @@ int start_userspace(unsigned long stub_stack)
sp = (unsigned long) stack + UM_KERN_PAGE_SIZE - sizeof(void *);
flags = CLONE_FILES;
- if (proc_mm)
+ if (proc_mm || have_switch_mm)
flags |= CLONE_VM;
else
flags |= SIGCHLD;
@@ -384,8 +384,14 @@ void userspace(struct uml_pt_regs *regs)
printk(UM_KERN_ERR "userspace - child stopped "
"with signal %d\n", sig);
}
- pid = userspace_pid[0];
+
+ /*
+ * userspace_pid can change in in_interrupt since
+ * PTRACE_SWITCH_MM can cause a process to change
+ * address spaces
+ */
interrupt_end();
+ pid = userspace_pid[0];
/* Avoid -ERESTARTSYS handling in host */
if (PT_SYSCALL_NR_OFFSET != PT_SYSCALL_RET_OFFSET)
@@ -473,6 +479,69 @@ int copy_context_skas0(unsigned long new_stack, int pid)
return pid;
}
+extern unsigned long switch_mm_stub;
+extern long task_size;
+
+static void unmap_new_as(void)
+{
+ void (*p)(void);
+ void *addr;
+ unsigned long stack = (unsigned long) &stack & ~(UM_KERN_PAGE_SIZE - 1);
+ unsigned long long data_offset, code_offset;
+ int fd = phys_mapping(to_phys((void *) stack), &data_offset);
+
+ addr = mmap((void *) STUB_DATA, UM_KERN_PAGE_SIZE,
+ PROT_READ | PROT_WRITE, MAP_SHARED | MAP_FIXED, fd,
+ data_offset);
+ if (addr == MAP_FAILED)
+ panic("Failed to remap stack");
+
+ fd = phys_mapping(to_phys(&__syscall_stub_start), &code_offset);
+ addr = mmap((void *) STUB_CODE, UM_KERN_PAGE_SIZE,
+ PROT_READ | PROT_WRITE, MAP_SHARED | MAP_FIXED, fd,
+ code_offset);
+ if (addr == MAP_FAILED)
+ panic("Failed to remap code");
+
+ p = (void (*)(void)) (STUB_CODE +
+ ((unsigned long) &switch_mm_stub -
+ (unsigned long) &__syscall_stub_start));
+ (*p)();
+}
+
+int copy_context_skas4(struct mm_id *id)
+{
+ void *data = NULL;
+ unsigned long *return_regs, *fd_ptr, regs[MAX_REG_NR];
+ sigset_t sigs, old;
+ int err;
+
+ err = unmap(id, 0, STUB_START, 0, &data);
+ if (err)
+ return err;
+
+ err = unmap(id, STUB_END, task_size - STUB_END, 0, &data);
+ if (err)
+ return err;
+
+ return_regs = (unsigned long *) (id->stack + UM_KERN_PAGE_SIZE -
+ MAX_REG_NR * sizeof(long));
+ fd_ptr = return_regs - 1;
+ *fd_ptr = self_mm_fd;
+
+ regs[REGS_IP_INDEX] = (unsigned long) unmap_new_as;
+ regs[REGS_SP_INDEX] = id->stack + UM_KERN_PAGE_SIZE / 2;
+
+ sigfillset(&sigs);
+ sigprocmask(SIG_SETMASK, &sigs, &old);
+
+ err = switch_mm(id->u.mm_fd, MM_SP_IP, regs, return_regs);
+
+ sigprocmask(SIG_SETMASK, &old, NULL);
+
+ return err;
+}
+
/*
* This is used only, if stub pages are needed, while proc_mm is
* available. Opening /proc/mm creates a new mm_context, which lacks
@@ -630,8 +699,15 @@ void __switch_mm(struct mm_id *mm_idp)
err = ptrace(OLD_PTRACE_SWITCH_MM, userspace_pid[0], 0,
mm_idp->u.mm_fd);
if (err)
- panic("__switch_mm - PTRACE_SWITCH_MM failed, "
+ panic("__switch_mm - OLD_PTRACE_SWITCH_MM failed, "
"errno = %d\n", errno);
}
+ else if (have_ptrace_switch_mm) {
+ err = ptrace(PTRACE_SWITCH_MM, userspace_pid[0], 0,
+ mm_idp->u.mm_fd);
+ if (err)
+ panic("__switch_mm - PTRACE_SWITCH_MM "
+ "failed, errno = %d\n", errno);
+ }
else userspace_pid[0] = mm_idp->u.pid;
}
diff --git a/arch/um/os-Linux/start_up.c b/arch/um/os-Linux/start_up.c
index 0e0f738..a0d45e7 100644
--- a/arch/um/os-Linux/start_up.c
+++ b/arch/um/os-Linux/start_up.c
@@ -161,6 +161,9 @@ static int disable_switch_mm;
int have_siginfo_segv;
static int disable_siginfo_segv;
+int have_ptrace_switch_mm;
+static int disable_ptrace_switch_mm;
+
int skas_needs_stub;
static int __init skas0_cmd_param(char *str, int* add)
@@ -168,8 +171,10 @@ static int __init skas0_cmd_param(char *str, int* add)
disable_ptrace_faultinfo = 1;
disable_ptrace_ldt = 1;
disable_proc_mm = 1;
+
disable_switch_mm = 1;
disable_siginfo_segv = 1;
+ disable_ptrace_switch_mm = 1;
return 0;
}
@@ -483,6 +488,18 @@ static inline void check_skas3_proc_mm(void)
}
}
+static void can_do_skas3(void)
+{
+ non_fatal("Checking for the skas3 patch in the host:\n");
+
+ check_skas3_proc_mm();
+ check_skas3_ptrace_faultinfo();
+ check_skas3_ptrace_ldt();
+
+ if (!proc_mm || !ptrace_faultinfo || !ptrace_ldt)
+ skas_needs_stub = 1;
+}
+
static void *fault_address;
static int check_fault_info(struct faultinfo *fi)
@@ -614,17 +631,207 @@ static int check_siginfo(void)
return ok;
}
-void can_do_skas(void)
+static char *mm_stack;
+static unsigned long return_regs[MAX_REG_NR];
+int self_mm_fd;
+
+static int switch_mm_works;
+
+static void after_switch(void)
{
- non_fatal("Checking for the skas3 patch in the host:\n");
+ /*
+ * If we are really in a new address space, setting this to
+ * zero won't affect the value of 1 already set in the old
+ * address space.
+ */
+ switch_mm_works = 0;
- check_skas3_proc_mm();
- check_skas3_ptrace_faultinfo();
- check_skas3_ptrace_ldt();
- check_siginfo();
+ switch_mm(self_mm_fd, MM_ALL_REGS, return_regs, NULL);
+}
- if (!proc_mm || !ptrace_faultinfo || !ptrace_ldt)
+static int check_switch_mm(void)
+{
+ unsigned long regs[MAX_REG_NR];
+ int err, there = -1;
+
+ non_fatal("\t/proc/self/mm ... ");
+ self_mm_fd = open("/proc/self/mm", O_RDONLY);
+ if (self_mm_fd < 0)
+ goto bad;
+ non_fatal("OK\n");
+
+ mm_stack = mmap(NULL, UM_KERN_PAGE_SIZE, PROT_READ | PROT_WRITE,
+ MAP_PRIVATE | MAP_ANONYMOUS, -1, 0);
+ if(mm_stack == MAP_FAILED)
+ goto bad;
+
+ non_fatal("\tnew_mm ... ");
+ there = new_mm(MM_COPY);
+ if(there < 0)
+ goto bad;
+ non_fatal("OK\n");
+
+ regs[REGS_IP_INDEX] = (unsigned long) after_switch;
+ regs[REGS_SP_INDEX] = ((unsigned long) &mm_stack[UM_KERN_PAGE_SIZE]) -
+ sizeof(void *);
+
+ non_fatal("\tswitching over ... ");
+ err = switch_mm(there, MM_SP_IP, regs, return_regs);
+ if (err < 0)
+ goto bad;
+ non_fatal("switched back ... ");
+ switch_mm_works = 1;
+ if(!switch_mm_works)
+ goto bad;
+ else
+ non_fatal("OK\n");
+
+ munmap(mm_stack, UM_KERN_PAGE_SIZE);
+ close(there);
+
+ if (disable_switch_mm)
+ non_fatal("switch_mm support disabled on command line\n");
+ else
+ have_switch_mm = 1;
+
+ return 1;
+ bad:
+ if (there > 0)
+ close(there);
+ munmap(mm_stack, UM_KERN_PAGE_SIZE);
+ non_fatal("Failed - \n");
+ perror("");
+ return 0;
+}
+
+static int ptrace_switch_mm_works;
+
+static int after_ptrace_switch(void)
+{
+ ptrace_switch_mm_works = 1;
+ exit(0);
+}
+
+static int check_ptrace_switch_mm(void)
+{
+ void *stack;
+ unsigned long regs[MAX_REG_NR];
+ int pid, here, err, status;
+
+ non_fatal("\tPTRACE_SWITCH_MM ... ");
+ pid = fork();
+ if(pid == 0){
+ ptrace(PTRACE_TRACEME, 0, 0, 0);
+ kill(getpid(), SIGSTOP);
+
+ exit(0);
+ }
+ else if(pid < 0)
+ goto bad;
+
+ stack = mmap(NULL, UM_KERN_PAGE_SIZE, PROT_READ | PROT_WRITE,
+ MAP_SHARED | MAP_ANONYMOUS, -1, 0);
+ if(stack == MAP_FAILED)
+ goto bad;
+
+ here = open("/proc/self/mm", O_RDONLY);
+ if(here < 0)
+ goto bad_unmap;
+
+ err = waitpid(pid, &status, WUNTRACED);
+ if (err < 0)
+ goto bad_close;
+ else if (err != pid) {
+ non_fatal("waitpid returned %d, expected %d\n", err, pid);
+ goto bad_close;
+ }
+ else if (!WIFSTOPPED(status) || (WSTOPSIG(status) != SIGSTOP)) {
+ non_fatal("waitpid returned status 0x%d\n", status);
+ goto bad_close;
+ }
+
+ err = ptrace(PTRACE_GETREGS, pid, 0, regs);
+ if (err < 0)
+ goto bad_close;
+
+ regs[REGS_IP_INDEX] = (unsigned long) after_ptrace_switch;
+ regs[REGS_SP_INDEX] = (unsigned long) stack + UM_KERN_PAGE_SIZE -
+ sizeof(void *);
+
+ if (ptrace(PTRACE_SETREGS, pid, 0, regs) < 0)
+ goto bad_close;
+
+ if (ptrace(PTRACE_SWITCH_MM, pid, NULL, here) < 0)
+ goto bad_close;
+
+ if (ptrace(PTRACE_CONT, pid, NULL, 0) < 0)
+ goto bad_close;
+
+ err = waitpid(pid, &status, WUNTRACED);
+ if (err < 0)
+ goto bad_close;
+ else if(err != pid) {
+ non_fatal("waitpid returned %d, expected %d\n", err, pid);
+ goto bad_close;
+ }
+ else if (!WIFEXITED(status) || (WEXITSTATUS(status) != 0)) {
+ non_fatal("waitpid returned status 0x%d\n", status);
+ goto bad_close;
+ }
+
+ if (!ptrace_switch_mm_works)
+ goto bad_close;
+ else non_fatal("OK\n");
+
+ if (disable_ptrace_switch_mm)
+ non_fatal("PTRACE_SWITCH_MM support disabled on command "
+ "line\n");
+ else
+ have_ptrace_switch_mm = 1;
+
+ close(here);
+ munmap(stack, UM_KERN_PAGE_SIZE);
+
+ return 1;
+
+ bad_close:
+ close(here);
+ bad_unmap:
+ munmap(stack, UM_KERN_PAGE_SIZE);
+ bad:
+ non_fatal("Failed - \n");
+ perror("");
+ return 0;
+}
+
+static int can_do_skas4(void)
+{
+ int ret;
+
+ non_fatal("Checking for new_mm and switch_mm support in the host:\n");
+
+ ret = check_switch_mm() && check_ptrace_switch_mm() && check_siginfo();
+ if (ret)
skas_needs_stub = 1;
+
+ return ret;
+}
+
+void can_do_skas(void)
+{
+ if (!can_do_skas4())
+ can_do_skas3();
+}
+
+int get_new_mm(void)
+{
+ int err;
+
+ err = new_mm(MM_COPY);
+ if (err < 0)
+ err = -errno;
+
+ return err;
}
int __init parse_iomem(char *str, int *add)
diff --git a/arch/um/sys-i386/stub.S b/arch/um/sys-i386/stub.S
index 890dc50..3cfb6e8 100644
--- a/arch/um/sys-i386/stub.S
+++ b/arch/um/sys-i386/stub.S
@@ -1,5 +1,8 @@
#include "uml-config.h"
#include "as-layout.h"
+#include "skas/skas.h"
+
+#define MAX_REG_NR 17
#define PROCESS_LIST \
/* load pointer to first operation */ \
@@ -43,3 +46,16 @@ batch_syscall_stub:
PROCESS_LIST
/* stop */
int3
+
+ .globl switch_mm_stub
+switch_mm_stub:
+ PROCESS_LIST
+
+ mov $__NR_switch_mm, %eax
+ mov ASM_STUB_DATA + UM_KERN_PAGE_SIZE - MAX_REG_NR * 4 - 4, %ebx
+ mov $MM_ALL_REGS, %ecx
+ mov $(ASM_STUB_DATA + UM_KERN_PAGE_SIZE - MAX_REG_NR * 4), %edx
+ xor %esi, %esi
+ int $0x80
+
+ int3
diff --git a/arch/um/sys-x86_64/stub.S b/arch/um/sys-x86_64/stub.S
index 143a16c..fb10353 100644
--- a/arch/um/sys-x86_64/stub.S
+++ b/arch/um/sys-x86_64/stub.S
@@ -1,5 +1,8 @@
#include "uml-config.h"
#include "as-layout.h"
+#include "skas/skas.h"
+
+#define MAX_REG_NR 27
#define PROCESS_LIST \
mov $(ASM_STUB_DATA >> 32), %rbx; \
@@ -46,3 +49,21 @@ batch_syscall_stub:
PROCESS_LIST
/* stop */
int3
+
+ .globl switch_mm_stub
+switch_mm_stub:
+ PROCESS_LIST
+
+ mov $__NR_switch_mm, %rax
+ mov $(ASM_STUB_DATA >> 32), %rdi
+ sal $32, %rdi
+ mov $(ASM_STUB_DATA & 0xffffffff), %rsi
+ add %rsi, %rdi
+ add $(UM_KERN_PAGE_SIZE - MAX_REG_NR * 8 - 8), %rdi
+ mov (%rdi), %rdi
+ mov $MM_ALL_REGS, %rsi
+ mov $(ASM_STUB_DATA + UM_KERN_PAGE_SIZE - MAX_REG_NR * 8), %rdx
+ xor %r10, %r10
+ syscall
+
+ int3
diff --git a/arch/um/sys-x86_64/syscalls.c b/arch/um/sys-x86_64/syscalls.c
index ff012ba..00131f9 100644
--- a/arch/um/sys-x86_64/syscalls.c
+++ b/arch/um/sys-x86_64/syscalls.c
@@ -30,7 +30,7 @@ long arch_prctl(struct task_struct *task, int code, unsigned long __user *addr)
{
unsigned long *ptr = addr, tmp;
long ret;
- int pid = task->mm->context.id.u.pid;
+ int pid = userspace_pid[0];
/*
* With ARCH_SET_FS (and ARCH_SET_GS is treated similarly to
-------------------------------------------------------------------------
Check out the new SourceForge.net Marketplace.
It's the best place to buy or sell services for
just about anything Open Source.
http://ad.doubleclick.net/clk;164216239;13503038;w?http://sf.net/marketplace
_______________________________________________
User-mode-linux-devel mailing list
User-mode-linux-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/user-mode-linux-devel
^ permalink raw reply related [flat|nested] 3+ messages in thread* [uml-devel] [RFC PATCH 10/10] SKAS4 - Guest get_mm and switch_mm support
@ 2008-01-23 17:12 Jeff Dike
0 siblings, 0 replies; 3+ messages in thread
From: Jeff Dike @ 2008-01-23 17:12 UTC (permalink / raw)
To: UML-user, uml-devel
diff --git a/arch/um/include/as-layout.h b/arch/um/include/as-layout.h
index a5cdf95..90ee798 100644
--- a/arch/um/include/as-layout.h
+++ b/arch/um/include/as-layout.h
@@ -17,6 +17,7 @@
#define ASM_STUB_CODE (UML_CONFIG_TOP_ADDR - 2 * UM_KERN_PAGE_SIZE)
#define ASM_STUB_DATA (UML_CONFIG_TOP_ADDR - UM_KERN_PAGE_SIZE)
#define ASM_STUB_START ASM_STUB_CODE
+#define ASM_STUB_END UML_CONFIG_TOP_ADDR
/*
* This file is included by the assembly stubs, which just want the
@@ -27,6 +28,7 @@
#define STUB_CODE ((unsigned long) ASM_STUB_CODE)
#define STUB_DATA ((unsigned long) ASM_STUB_DATA)
#define STUB_START ((unsigned long) ASM_STUB_START)
+#define STUB_END ((unsigned long) ASM_STUB_END)
#include "sysdep/ptrace.h"
diff --git a/arch/um/include/skas/mm_id.h b/arch/um/include/skas/mm_id.h
index 48dd098..a2e7643 100644
--- a/arch/um/include/skas/mm_id.h
+++ b/arch/um/include/skas/mm_id.h
@@ -7,7 +7,7 @@
#define __MM_ID_H
struct mm_id {
- union {
+ struct {
int mm_fd;
int pid;
} u;
diff --git a/arch/um/include/skas/skas.h b/arch/um/include/skas/skas.h
index 061a362..bd3af6c 100644
--- a/arch/um/include/skas/skas.h
+++ b/arch/um/include/skas/skas.h
@@ -6,14 +6,41 @@
#ifndef __SKAS_H
#define __SKAS_H
-#include "sysdep/ptrace.h"
+#ifndef __KERNEL__
+#include <unistd.h>
+#include <sys/syscall.h>
+#endif
+#include "uml-config.h"
-extern int have_siginfo_segv;
+#ifdef UML_CONFIG_X86_32
+#define __NR_new_mm 325
+#define __NR_switch_mm 326
+#else
+#define __NR_new_mm 286
+#define __NR_switch_mm 287
+#endif
+
+#define MM_COPY 0
+
+#define MM_ALL_REGS 0
+#define MM_SP_IP 1
+#define MM_SAME 2
+
+#define PTRACE_SWITCH_MM 33
+
+#ifndef __ASSEMBLY__
+
+#include "sysdep/ptrace.h"
extern int userspace_pid[];
extern int proc_mm, ptrace_faultinfo, ptrace_ldt;
extern int skas_needs_stub;
+extern int have_switch_mm;
+extern int have_ptrace_switch_mm;
+extern int have_siginfo_segv;
+extern int self_mm_fd;
+
extern int user_thread(unsigned long stack, int flags);
extern void new_thread_handler(void);
extern void handle_syscall(struct uml_pt_regs *regs);
@@ -22,4 +49,19 @@ extern void get_skas_faultinfo(int pid, struct faultinfo * fi);
extern long execute_syscall_skas(void *r);
extern unsigned long current_stub_stack(void);
+#ifndef __KERNEL__
+static inline long new_mm(int flags)
+{
+ return syscall(__NR_new_mm, MM_COPY, 0, 0, 0, 0, 0, 0);
+}
+
+static inline long switch_mm(int mm_fd, unsigned long flags,
+ unsigned long *new_regs, unsigned long *save_regs)
+{
+ return syscall(__NR_switch_mm, mm_fd, flags, new_regs, save_regs, 0, 0);
+}
+#endif
+
+#endif
+
#endif
diff --git a/arch/um/kernel/reboot.c b/arch/um/kernel/reboot.c
index 04cebcf..0a5468e 100644
--- a/arch/um/kernel/reboot.c
+++ b/arch/um/kernel/reboot.c
@@ -11,7 +11,7 @@ void (*pm_power_off)(void);
static void kill_off_processes(void)
{
- if(proc_mm)
+ if(proc_mm || have_switch_mm)
/*
* FIXME: need to loop over userspace_pids
*/
diff --git a/arch/um/kernel/skas/mmu.c b/arch/um/kernel/skas/mmu.c
index 3426711..3155263 100644
--- a/arch/um/kernel/skas/mmu.c
+++ b/arch/um/kernel/skas/mmu.c
@@ -65,6 +65,9 @@ static int init_stub_pte(struct mm_struct *mm, unsigned long proc,
return -ENOMEM;
}
+extern int copy_context_skas4(struct mm_id *id);
+extern int get_new_mm(void);
+
int init_new_context(struct task_struct *task, struct mm_struct *mm)
{
struct mm_context *from_mm = NULL;
@@ -109,6 +112,20 @@ int init_new_context(struct task_struct *task, struct mm_struct *mm)
}
to_mm->id.u.mm_fd = ret;
}
+ else if (have_switch_mm) {
+ to_mm->id.u.mm_fd = get_new_mm();
+ if(to_mm->id.u.mm_fd < 0) {
+ ret = to_mm->id.u.mm_fd;
+ goto out_free;
+ }
+
+ ret = copy_context_skas4(&to_mm->id);
+ if (ret < 0) {
+ os_close_file(to_mm->id.u.mm_fd);
+ to_mm->id.u.mm_fd = -1;
+ goto out_free;
+ }
+ }
else {
if (from_mm)
to_mm->id.u.pid = copy_context_skas0(stack,
@@ -136,11 +153,15 @@ void destroy_context(struct mm_struct *mm)
{
struct mm_context *mmu = &mm->context;
- if (proc_mm)
+ if (proc_mm || have_switch_mm)
os_close_file(mmu->id.u.mm_fd);
- else
+ else {
os_kill_ptraced_process(mmu->id.u.pid, 1);
+ if (have_switch_mm)
+ os_close_file(mmu->id.u.mm_fd);
+ }
+
if (!proc_mm || !ptrace_faultinfo) {
free_page(mmu->id.stack);
pte_lock_deinit(virt_to_page(mmu->last_page_table));
diff --git a/arch/um/kernel/skas/process.c b/arch/um/kernel/skas/process.c
index 2c6de0a..e5e8613 100644
--- a/arch/um/kernel/skas/process.c
+++ b/arch/um/kernel/skas/process.c
@@ -49,7 +49,7 @@ int __init start_uml(void)
{
stack_protections((unsigned long) &cpu0_irqstack);
set_sigstack(cpu0_irqstack, THREAD_SIZE);
- if (proc_mm)
+ if (proc_mm || have_switch_mm)
userspace_pid[0] = start_userspace(0);
init_new_thread_signals();
diff --git a/arch/um/kernel/um_arch.c b/arch/um/kernel/um_arch.c
index f1c7139..d92108b 100644
--- a/arch/um/kernel/um_arch.c
+++ b/arch/um/kernel/um_arch.c
@@ -271,7 +271,9 @@ int __init linux_main(int argc, char **argv)
can_do_skas();
- if (proc_mm && ptrace_faultinfo)
+ if (have_switch_mm)
+ mode = "SKAS4";
+ else if (proc_mm && ptrace_faultinfo)
mode = "SKAS3";
else
mode = "SKAS0";
diff --git a/arch/um/os-Linux/skas/mem.c b/arch/um/os-Linux/skas/mem.c
index efda5e1..cc86c0b 100644
--- a/arch/um/os-Linux/skas/mem.c
+++ b/arch/um/os-Linux/skas/mem.c
@@ -6,6 +6,7 @@
#include <stddef.h>
#include <unistd.h>
#include <errno.h>
+#include <signal.h>
#include <string.h>
#include <sys/mman.h>
#include "init.h"
@@ -22,7 +23,7 @@
#include "sysdep/stub.h"
#include "uml-config.h"
-extern unsigned long batch_syscall_stub, __syscall_stub_start;
+extern unsigned long batch_syscall_stub, switch_mm_stub, __syscall_stub_start;
extern void wait_stub_done(int pid);
@@ -43,7 +44,10 @@ static int __init init_syscall_regs(void)
unsigned long *stub_entry;
get_safe_registers(syscall_regs);
- stub_entry = &batch_syscall_stub;
+ if (have_switch_mm)
+ stub_entry = &switch_mm_stub;
+ else
+ stub_entry = &batch_syscall_stub;
syscall_regs[REGS_IP_INDEX] = STUB_CODE +
((unsigned long) stub_entry -
@@ -126,6 +130,32 @@ static long do_syscall_stub(struct mm_id *mm_idp, void **addr)
return ret;
}
+static long do_syscall_stub_skas4(struct mm_id *mm_idp, void **addr)
+{
+ long ret;
+ unsigned long *return_regs;
+ int err;
+ sigset_t sigs, old;
+
+ return_regs = (unsigned long *) (mm_idp->stack + UM_KERN_PAGE_SIZE) -
+ MAX_REG_NR;
+ *(return_regs - 1) = self_mm_fd;
+
+ return_regs = (unsigned long *) (STUB_DATA + UM_KERN_PAGE_SIZE) -
+ MAX_REG_NR;
+
+ sigfillset(&sigs);
+ sigprocmask(SIG_SETMASK, &sigs, &old);
+ err = switch_mm(mm_idp->u.mm_fd, MM_SP_IP, syscall_regs, return_regs);
+ sigprocmask(SIG_SETMASK, &old, NULL);
+
+ ret = syscall_stub_done(mm_idp->stack);
+
+ *addr = check_init_stack(mm_idp, NULL);
+
+ return ret;
+}
+
static int flush_syscalls(struct mm_id *mm_idp, void **addr, int extra)
{
unsigned long *stack = check_init_stack(mm_idp, *addr);
@@ -134,10 +164,16 @@ static int flush_syscalls(struct mm_id *mm_idp, void **addr, int extra)
current = ((unsigned long) stack) & ~UM_KERN_PAGE_MASK;
end = UM_KERN_PAGE_SIZE;
+ if(have_switch_mm)
+ end -= (MAX_REG_NR + 1) * sizeof(long);
+
if (current + (10 + extra) * sizeof(long) < end)
return 0;
- return do_syscall_stub(mm_idp, addr);
+ if (have_switch_mm)
+ return do_syscall_stub_skas4(mm_idp, addr);
+ else
+ return do_syscall_stub(mm_idp, addr);
}
long run_syscall_stub(struct mm_id *mm_idp, int syscall,
@@ -166,8 +202,12 @@ long run_syscall_stub(struct mm_id *mm_idp, int syscall,
*stack++ = expected;
*stack = 0;
- if (done)
- return do_syscall_stub(mm_idp, addr);
+ if (done) {
+ if (have_switch_mm)
+ return do_syscall_stub_skas4(mm_idp, addr);
+ else
+ return do_syscall_stub(mm_idp, addr);
+ }
*addr = stack;
diff --git a/arch/um/os-Linux/skas/process.c b/arch/um/os-Linux/skas/process.c
index 522d0f1..9c2c086 100644
--- a/arch/um/os-Linux/skas/process.c
+++ b/arch/um/os-Linux/skas/process.c
@@ -227,7 +227,7 @@ static int userspace_tramp(void *stack)
}
}
}
- if (!ptrace_faultinfo && (stack != NULL)) {
+ if (!ptrace_faultinfo) {
struct sigaction sa;
unsigned long v = STUB_CODE +
@@ -271,7 +271,7 @@ int start_userspace(unsigned long stub_stack)
sp = (unsigned long) stack + UM_KERN_PAGE_SIZE - sizeof(void *);
flags = CLONE_FILES;
- if (proc_mm)
+ if (proc_mm || have_switch_mm)
flags |= CLONE_VM;
else
flags |= SIGCHLD;
@@ -384,8 +384,14 @@ void userspace(struct uml_pt_regs *regs)
printk(UM_KERN_ERR "userspace - child stopped "
"with signal %d\n", sig);
}
- pid = userspace_pid[0];
+
+ /*
+ * userspace_pid can change in in_interrupt since
+ * PTRACE_SWITCH_MM can cause a process to change
+ * address spaces
+ */
interrupt_end();
+ pid = userspace_pid[0];
/* Avoid -ERESTARTSYS handling in host */
if (PT_SYSCALL_NR_OFFSET != PT_SYSCALL_RET_OFFSET)
@@ -473,6 +479,69 @@ int copy_context_skas0(unsigned long new_stack, int pid)
return pid;
}
+extern unsigned long switch_mm_stub;
+extern long task_size;
+
+static void unmap_new_as(void)
+{
+ void (*p)(void);
+ void *addr;
+ unsigned long stack = (unsigned long) &stack & ~(UM_KERN_PAGE_SIZE - 1);
+ unsigned long long data_offset, code_offset;
+ int fd = phys_mapping(to_phys((void *) stack), &data_offset);
+
+ addr = mmap((void *) STUB_DATA, UM_KERN_PAGE_SIZE,
+ PROT_READ | PROT_WRITE, MAP_SHARED | MAP_FIXED, fd,
+ data_offset);
+ if (addr == MAP_FAILED)
+ panic("Failed to remap stack");
+
+ fd = phys_mapping(to_phys(&__syscall_stub_start), &code_offset);
+ addr = mmap((void *) STUB_CODE, UM_KERN_PAGE_SIZE,
+ PROT_READ | PROT_WRITE, MAP_SHARED | MAP_FIXED, fd,
+ code_offset);
+ if (addr == MAP_FAILED)
+ panic("Failed to remap code");
+
+ p = (void (*)(void)) (STUB_CODE +
+ ((unsigned long) &switch_mm_stub -
+ (unsigned long) &__syscall_stub_start));
+ (*p)();
+}
+
+int copy_context_skas4(struct mm_id *id)
+{
+ void *data = NULL;
+ unsigned long *return_regs, *fd_ptr, regs[MAX_REG_NR];
+ sigset_t sigs, old;
+ int err;
+
+ err = unmap(id, 0, STUB_START, 0, &data);
+ if (err)
+ return err;
+
+ err = unmap(id, STUB_END, task_size - STUB_END, 0, &data);
+ if (err)
+ return err;
+
+ return_regs = (unsigned long *) (id->stack + UM_KERN_PAGE_SIZE -
+ MAX_REG_NR * sizeof(long));
+ fd_ptr = return_regs - 1;
+ *fd_ptr = self_mm_fd;
+
+ regs[REGS_IP_INDEX] = (unsigned long) unmap_new_as;
+ regs[REGS_SP_INDEX] = id->stack + UM_KERN_PAGE_SIZE / 2;
+
+ sigfillset(&sigs);
+ sigprocmask(SIG_SETMASK, &sigs, &old);
+
+ err = switch_mm(id->u.mm_fd, MM_SP_IP, regs, return_regs);
+
+ sigprocmask(SIG_SETMASK, &old, NULL);
+
+ return err;
+}
+
/*
* This is used only, if stub pages are needed, while proc_mm is
* available. Opening /proc/mm creates a new mm_context, which lacks
@@ -630,8 +699,15 @@ void __switch_mm(struct mm_id *mm_idp)
err = ptrace(OLD_PTRACE_SWITCH_MM, userspace_pid[0], 0,
mm_idp->u.mm_fd);
if (err)
- panic("__switch_mm - PTRACE_SWITCH_MM failed, "
+ panic("__switch_mm - OLD_PTRACE_SWITCH_MM failed, "
"errno = %d\n", errno);
}
+ else if (have_ptrace_switch_mm) {
+ err = ptrace(PTRACE_SWITCH_MM, userspace_pid[0], 0,
+ mm_idp->u.mm_fd);
+ if (err)
+ panic("__switch_mm - PTRACE_SWITCH_MM "
+ "failed, errno = %d\n", errno);
+ }
else userspace_pid[0] = mm_idp->u.pid;
}
diff --git a/arch/um/os-Linux/start_up.c b/arch/um/os-Linux/start_up.c
index 0e0f738..a0d45e7 100644
--- a/arch/um/os-Linux/start_up.c
+++ b/arch/um/os-Linux/start_up.c
@@ -161,6 +161,9 @@ static int disable_switch_mm;
int have_siginfo_segv;
static int disable_siginfo_segv;
+int have_ptrace_switch_mm;
+static int disable_ptrace_switch_mm;
+
int skas_needs_stub;
static int __init skas0_cmd_param(char *str, int* add)
@@ -168,8 +171,10 @@ static int __init skas0_cmd_param(char *str, int* add)
disable_ptrace_faultinfo = 1;
disable_ptrace_ldt = 1;
disable_proc_mm = 1;
+
disable_switch_mm = 1;
disable_siginfo_segv = 1;
+ disable_ptrace_switch_mm = 1;
return 0;
}
@@ -483,6 +488,18 @@ static inline void check_skas3_proc_mm(void)
}
}
+static void can_do_skas3(void)
+{
+ non_fatal("Checking for the skas3 patch in the host:\n");
+
+ check_skas3_proc_mm();
+ check_skas3_ptrace_faultinfo();
+ check_skas3_ptrace_ldt();
+
+ if (!proc_mm || !ptrace_faultinfo || !ptrace_ldt)
+ skas_needs_stub = 1;
+}
+
static void *fault_address;
static int check_fault_info(struct faultinfo *fi)
@@ -614,17 +631,207 @@ static int check_siginfo(void)
return ok;
}
-void can_do_skas(void)
+static char *mm_stack;
+static unsigned long return_regs[MAX_REG_NR];
+int self_mm_fd;
+
+static int switch_mm_works;
+
+static void after_switch(void)
{
- non_fatal("Checking for the skas3 patch in the host:\n");
+ /*
+ * If we are really in a new address space, setting this to
+ * zero won't affect the value of 1 already set in the old
+ * address space.
+ */
+ switch_mm_works = 0;
- check_skas3_proc_mm();
- check_skas3_ptrace_faultinfo();
- check_skas3_ptrace_ldt();
- check_siginfo();
+ switch_mm(self_mm_fd, MM_ALL_REGS, return_regs, NULL);
+}
- if (!proc_mm || !ptrace_faultinfo || !ptrace_ldt)
+static int check_switch_mm(void)
+{
+ unsigned long regs[MAX_REG_NR];
+ int err, there = -1;
+
+ non_fatal("\t/proc/self/mm ... ");
+ self_mm_fd = open("/proc/self/mm", O_RDONLY);
+ if (self_mm_fd < 0)
+ goto bad;
+ non_fatal("OK\n");
+
+ mm_stack = mmap(NULL, UM_KERN_PAGE_SIZE, PROT_READ | PROT_WRITE,
+ MAP_PRIVATE | MAP_ANONYMOUS, -1, 0);
+ if(mm_stack == MAP_FAILED)
+ goto bad;
+
+ non_fatal("\tnew_mm ... ");
+ there = new_mm(MM_COPY);
+ if(there < 0)
+ goto bad;
+ non_fatal("OK\n");
+
+ regs[REGS_IP_INDEX] = (unsigned long) after_switch;
+ regs[REGS_SP_INDEX] = ((unsigned long) &mm_stack[UM_KERN_PAGE_SIZE]) -
+ sizeof(void *);
+
+ non_fatal("\tswitching over ... ");
+ err = switch_mm(there, MM_SP_IP, regs, return_regs);
+ if (err < 0)
+ goto bad;
+ non_fatal("switched back ... ");
+ switch_mm_works = 1;
+ if(!switch_mm_works)
+ goto bad;
+ else
+ non_fatal("OK\n");
+
+ munmap(mm_stack, UM_KERN_PAGE_SIZE);
+ close(there);
+
+ if (disable_switch_mm)
+ non_fatal("switch_mm support disabled on command line\n");
+ else
+ have_switch_mm = 1;
+
+ return 1;
+ bad:
+ if (there > 0)
+ close(there);
+ munmap(mm_stack, UM_KERN_PAGE_SIZE);
+ non_fatal("Failed - \n");
+ perror("");
+ return 0;
+}
+
+static int ptrace_switch_mm_works;
+
+static int after_ptrace_switch(void)
+{
+ ptrace_switch_mm_works = 1;
+ exit(0);
+}
+
+static int check_ptrace_switch_mm(void)
+{
+ void *stack;
+ unsigned long regs[MAX_REG_NR];
+ int pid, here, err, status;
+
+ non_fatal("\tPTRACE_SWITCH_MM ... ");
+ pid = fork();
+ if(pid == 0){
+ ptrace(PTRACE_TRACEME, 0, 0, 0);
+ kill(getpid(), SIGSTOP);
+
+ exit(0);
+ }
+ else if(pid < 0)
+ goto bad;
+
+ stack = mmap(NULL, UM_KERN_PAGE_SIZE, PROT_READ | PROT_WRITE,
+ MAP_SHARED | MAP_ANONYMOUS, -1, 0);
+ if(stack == MAP_FAILED)
+ goto bad;
+
+ here = open("/proc/self/mm", O_RDONLY);
+ if(here < 0)
+ goto bad_unmap;
+
+ err = waitpid(pid, &status, WUNTRACED);
+ if (err < 0)
+ goto bad_close;
+ else if (err != pid) {
+ non_fatal("waitpid returned %d, expected %d\n", err, pid);
+ goto bad_close;
+ }
+ else if (!WIFSTOPPED(status) || (WSTOPSIG(status) != SIGSTOP)) {
+ non_fatal("waitpid returned status 0x%d\n", status);
+ goto bad_close;
+ }
+
+ err = ptrace(PTRACE_GETREGS, pid, 0, regs);
+ if (err < 0)
+ goto bad_close;
+
+ regs[REGS_IP_INDEX] = (unsigned long) after_ptrace_switch;
+ regs[REGS_SP_INDEX] = (unsigned long) stack + UM_KERN_PAGE_SIZE -
+ sizeof(void *);
+
+ if (ptrace(PTRACE_SETREGS, pid, 0, regs) < 0)
+ goto bad_close;
+
+ if (ptrace(PTRACE_SWITCH_MM, pid, NULL, here) < 0)
+ goto bad_close;
+
+ if (ptrace(PTRACE_CONT, pid, NULL, 0) < 0)
+ goto bad_close;
+
+ err = waitpid(pid, &status, WUNTRACED);
+ if (err < 0)
+ goto bad_close;
+ else if(err != pid) {
+ non_fatal("waitpid returned %d, expected %d\n", err, pid);
+ goto bad_close;
+ }
+ else if (!WIFEXITED(status) || (WEXITSTATUS(status) != 0)) {
+ non_fatal("waitpid returned status 0x%d\n", status);
+ goto bad_close;
+ }
+
+ if (!ptrace_switch_mm_works)
+ goto bad_close;
+ else non_fatal("OK\n");
+
+ if (disable_ptrace_switch_mm)
+ non_fatal("PTRACE_SWITCH_MM support disabled on command "
+ "line\n");
+ else
+ have_ptrace_switch_mm = 1;
+
+ close(here);
+ munmap(stack, UM_KERN_PAGE_SIZE);
+
+ return 1;
+
+ bad_close:
+ close(here);
+ bad_unmap:
+ munmap(stack, UM_KERN_PAGE_SIZE);
+ bad:
+ non_fatal("Failed - \n");
+ perror("");
+ return 0;
+}
+
+static int can_do_skas4(void)
+{
+ int ret;
+
+ non_fatal("Checking for new_mm and switch_mm support in the host:\n");
+
+ ret = check_switch_mm() && check_ptrace_switch_mm() && check_siginfo();
+ if (ret)
skas_needs_stub = 1;
+
+ return ret;
+}
+
+void can_do_skas(void)
+{
+ if (!can_do_skas4())
+ can_do_skas3();
+}
+
+int get_new_mm(void)
+{
+ int err;
+
+ err = new_mm(MM_COPY);
+ if (err < 0)
+ err = -errno;
+
+ return err;
}
int __init parse_iomem(char *str, int *add)
diff --git a/arch/um/sys-i386/stub.S b/arch/um/sys-i386/stub.S
index 890dc50..3cfb6e8 100644
--- a/arch/um/sys-i386/stub.S
+++ b/arch/um/sys-i386/stub.S
@@ -1,5 +1,8 @@
#include "uml-config.h"
#include "as-layout.h"
+#include "skas/skas.h"
+
+#define MAX_REG_NR 17
#define PROCESS_LIST \
/* load pointer to first operation */ \
@@ -43,3 +46,16 @@ batch_syscall_stub:
PROCESS_LIST
/* stop */
int3
+
+ .globl switch_mm_stub
+switch_mm_stub:
+ PROCESS_LIST
+
+ mov $__NR_switch_mm, %eax
+ mov ASM_STUB_DATA + UM_KERN_PAGE_SIZE - MAX_REG_NR * 4 - 4, %ebx
+ mov $MM_ALL_REGS, %ecx
+ mov $(ASM_STUB_DATA + UM_KERN_PAGE_SIZE - MAX_REG_NR * 4), %edx
+ xor %esi, %esi
+ int $0x80
+
+ int3
diff --git a/arch/um/sys-x86_64/stub.S b/arch/um/sys-x86_64/stub.S
index 143a16c..fb10353 100644
--- a/arch/um/sys-x86_64/stub.S
+++ b/arch/um/sys-x86_64/stub.S
@@ -1,5 +1,8 @@
#include "uml-config.h"
#include "as-layout.h"
+#include "skas/skas.h"
+
+#define MAX_REG_NR 27
#define PROCESS_LIST \
mov $(ASM_STUB_DATA >> 32), %rbx; \
@@ -46,3 +49,21 @@ batch_syscall_stub:
PROCESS_LIST
/* stop */
int3
+
+ .globl switch_mm_stub
+switch_mm_stub:
+ PROCESS_LIST
+
+ mov $__NR_switch_mm, %rax
+ mov $(ASM_STUB_DATA >> 32), %rdi
+ sal $32, %rdi
+ mov $(ASM_STUB_DATA & 0xffffffff), %rsi
+ add %rsi, %rdi
+ add $(UM_KERN_PAGE_SIZE - MAX_REG_NR * 8 - 8), %rdi
+ mov (%rdi), %rdi
+ mov $MM_ALL_REGS, %rsi
+ mov $(ASM_STUB_DATA + UM_KERN_PAGE_SIZE - MAX_REG_NR * 8), %rdx
+ xor %r10, %r10
+ syscall
+
+ int3
diff --git a/arch/um/sys-x86_64/syscalls.c b/arch/um/sys-x86_64/syscalls.c
index ff012ba..00131f9 100644
--- a/arch/um/sys-x86_64/syscalls.c
+++ b/arch/um/sys-x86_64/syscalls.c
@@ -30,7 +30,7 @@ long arch_prctl(struct task_struct *task, int code, unsigned long __user *addr)
{
unsigned long *ptr = addr, tmp;
long ret;
- int pid = task->mm->context.id.u.pid;
+ int pid = userspace_pid[0];
/*
* With ARCH_SET_FS (and ARCH_SET_GS is treated similarly to
-------------------------------------------------------------------------
This SF.net email is sponsored by: Microsoft
Defy all challenges. Microsoft(R) Visual Studio 2008.
http://clk.atdmt.com/MRT/go/vse0120000070mrt/direct/01/
_______________________________________________
User-mode-linux-devel mailing list
User-mode-linux-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/user-mode-linux-devel
^ permalink raw reply related [flat|nested] 3+ messages in thread* [uml-devel] [RFC PATCH 10/10] SKAS4 - Guest get_mm and switch_mm support
@ 2008-01-28 21:38 Jeff Dike
0 siblings, 0 replies; 3+ messages in thread
From: Jeff Dike @ 2008-01-28 21:38 UTC (permalink / raw)
To: uml-user, uml-devel
UML guest SKAS4 support.
diff --git a/arch/um/include/as-layout.h b/arch/um/include/as-layout.h
index a5cdf95..90ee798 100644
--- a/arch/um/include/as-layout.h
+++ b/arch/um/include/as-layout.h
@@ -17,6 +17,7 @@
#define ASM_STUB_CODE (UML_CONFIG_TOP_ADDR - 2 * UM_KERN_PAGE_SIZE)
#define ASM_STUB_DATA (UML_CONFIG_TOP_ADDR - UM_KERN_PAGE_SIZE)
#define ASM_STUB_START ASM_STUB_CODE
+#define ASM_STUB_END UML_CONFIG_TOP_ADDR
/*
* This file is included by the assembly stubs, which just want the
@@ -27,6 +28,7 @@
#define STUB_CODE ((unsigned long) ASM_STUB_CODE)
#define STUB_DATA ((unsigned long) ASM_STUB_DATA)
#define STUB_START ((unsigned long) ASM_STUB_START)
+#define STUB_END ((unsigned long) ASM_STUB_END)
#include "sysdep/ptrace.h"
diff --git a/arch/um/include/skas/mm_id.h b/arch/um/include/skas/mm_id.h
index 48dd098..a2e7643 100644
--- a/arch/um/include/skas/mm_id.h
+++ b/arch/um/include/skas/mm_id.h
@@ -7,7 +7,7 @@
#define __MM_ID_H
struct mm_id {
- union {
+ struct {
int mm_fd;
int pid;
} u;
diff --git a/arch/um/include/skas/skas.h b/arch/um/include/skas/skas.h
index 061a362..69c7470 100644
--- a/arch/um/include/skas/skas.h
+++ b/arch/um/include/skas/skas.h
@@ -6,14 +6,38 @@
#ifndef __SKAS_H
#define __SKAS_H
-#include "sysdep/ptrace.h"
+#ifndef __KERNEL__
+#include <unistd.h>
+#include <sys/syscall.h>
+#endif
+#include "uml-config.h"
-extern int have_siginfo_segv;
+#ifdef UML_CONFIG_X86_32
+#define __NR_new_mm 325
+#define __NR_switch_mm 326
+#else
+#define __NR_new_mm 286
+#define __NR_switch_mm 287
+#endif
+
+#define MM_ALL_REGS 0
+#define MM_SP_IP 1
+
+#define PTRACE_SWITCH_MM 33
+
+#ifndef __ASSEMBLY__
+
+#include "sysdep/ptrace.h"
extern int userspace_pid[];
extern int proc_mm, ptrace_faultinfo, ptrace_ldt;
extern int skas_needs_stub;
+extern int have_switch_mm;
+extern int have_ptrace_switch_mm;
+extern int have_siginfo_segv;
+extern int self_mm_fd;
+
extern int user_thread(unsigned long stack, int flags);
extern void new_thread_handler(void);
extern void handle_syscall(struct uml_pt_regs *regs);
@@ -22,4 +46,19 @@ extern void get_skas_faultinfo(int pid, struct faultinfo * fi);
extern long execute_syscall_skas(void *r);
extern unsigned long current_stub_stack(void);
+#ifndef __KERNEL__
+static inline long new_mm(void)
+{
+ return syscall(__NR_new_mm, 0, 0, 0, 0, 0, 0, 0);
+}
+
+static inline long switch_mm(int mm_fd, unsigned long flags,
+ unsigned long *new_regs, unsigned long *save_regs)
+{
+ return syscall(__NR_switch_mm, mm_fd, flags, new_regs, save_regs, 0, 0);
+}
+#endif
+
+#endif
+
#endif
diff --git a/arch/um/kernel/reboot.c b/arch/um/kernel/reboot.c
index 04cebcf..0a5468e 100644
--- a/arch/um/kernel/reboot.c
+++ b/arch/um/kernel/reboot.c
@@ -11,7 +11,7 @@ void (*pm_power_off)(void);
static void kill_off_processes(void)
{
- if(proc_mm)
+ if(proc_mm || have_switch_mm)
/*
* FIXME: need to loop over userspace_pids
*/
diff --git a/arch/um/kernel/skas/mmu.c b/arch/um/kernel/skas/mmu.c
index 7595f77..2672829 100644
--- a/arch/um/kernel/skas/mmu.c
+++ b/arch/um/kernel/skas/mmu.c
@@ -65,6 +65,9 @@ static int init_stub_pte(struct mm_struct *mm, unsigned long proc,
return -ENOMEM;
}
+extern int copy_context_skas4(struct mm_id *id);
+extern int get_new_mm(void);
+
int init_new_context(struct task_struct *task, struct mm_struct *mm)
{
struct mm_context *from_mm = NULL;
@@ -109,6 +112,20 @@ int init_new_context(struct task_struct *task, struct mm_struct *mm)
}
to_mm->id.u.mm_fd = ret;
}
+ else if (have_switch_mm) {
+ to_mm->id.u.mm_fd = get_new_mm();
+ if(to_mm->id.u.mm_fd < 0) {
+ ret = to_mm->id.u.mm_fd;
+ goto out_free;
+ }
+
+ ret = copy_context_skas4(&to_mm->id);
+ if (ret < 0) {
+ os_close_file(to_mm->id.u.mm_fd);
+ to_mm->id.u.mm_fd = -1;
+ goto out_free;
+ }
+ }
else {
if (from_mm)
to_mm->id.u.pid = copy_context_skas0(stack,
@@ -136,11 +153,15 @@ void destroy_context(struct mm_struct *mm)
{
struct mm_context *mmu = &mm->context;
- if (proc_mm)
+ if (proc_mm || have_switch_mm)
os_close_file(mmu->id.u.mm_fd);
- else
+ else {
os_kill_ptraced_process(mmu->id.u.pid, 1);
+ if (have_switch_mm)
+ os_close_file(mmu->id.u.mm_fd);
+ }
+
if (!proc_mm || !ptrace_faultinfo) {
free_page(mmu->id.stack);
pte_lock_deinit(virt_to_page(mmu->last_page_table));
diff --git a/arch/um/kernel/skas/process.c b/arch/um/kernel/skas/process.c
index 2c6de0a..e5e8613 100644
--- a/arch/um/kernel/skas/process.c
+++ b/arch/um/kernel/skas/process.c
@@ -49,7 +49,7 @@ int __init start_uml(void)
{
stack_protections((unsigned long) &cpu0_irqstack);
set_sigstack(cpu0_irqstack, THREAD_SIZE);
- if (proc_mm)
+ if (proc_mm || have_switch_mm)
userspace_pid[0] = start_userspace(0);
init_new_thread_signals();
diff --git a/arch/um/kernel/um_arch.c b/arch/um/kernel/um_arch.c
index f1c7139..d92108b 100644
--- a/arch/um/kernel/um_arch.c
+++ b/arch/um/kernel/um_arch.c
@@ -271,7 +271,9 @@ int __init linux_main(int argc, char **argv)
can_do_skas();
- if (proc_mm && ptrace_faultinfo)
+ if (have_switch_mm)
+ mode = "SKAS4";
+ else if (proc_mm && ptrace_faultinfo)
mode = "SKAS3";
else
mode = "SKAS0";
diff --git a/arch/um/os-Linux/skas/mem.c b/arch/um/os-Linux/skas/mem.c
index efda5e1..cc86c0b 100644
--- a/arch/um/os-Linux/skas/mem.c
+++ b/arch/um/os-Linux/skas/mem.c
@@ -6,6 +6,7 @@
#include <stddef.h>
#include <unistd.h>
#include <errno.h>
+#include <signal.h>
#include <string.h>
#include <sys/mman.h>
#include "init.h"
@@ -22,7 +23,7 @@
#include "sysdep/stub.h"
#include "uml-config.h"
-extern unsigned long batch_syscall_stub, __syscall_stub_start;
+extern unsigned long batch_syscall_stub, switch_mm_stub, __syscall_stub_start;
extern void wait_stub_done(int pid);
@@ -43,7 +44,10 @@ static int __init init_syscall_regs(void)
unsigned long *stub_entry;
get_safe_registers(syscall_regs);
- stub_entry = &batch_syscall_stub;
+ if (have_switch_mm)
+ stub_entry = &switch_mm_stub;
+ else
+ stub_entry = &batch_syscall_stub;
syscall_regs[REGS_IP_INDEX] = STUB_CODE +
((unsigned long) stub_entry -
@@ -126,6 +130,32 @@ static long do_syscall_stub(struct mm_id *mm_idp, void **addr)
return ret;
}
+static long do_syscall_stub_skas4(struct mm_id *mm_idp, void **addr)
+{
+ long ret;
+ unsigned long *return_regs;
+ int err;
+ sigset_t sigs, old;
+
+ return_regs = (unsigned long *) (mm_idp->stack + UM_KERN_PAGE_SIZE) -
+ MAX_REG_NR;
+ *(return_regs - 1) = self_mm_fd;
+
+ return_regs = (unsigned long *) (STUB_DATA + UM_KERN_PAGE_SIZE) -
+ MAX_REG_NR;
+
+ sigfillset(&sigs);
+ sigprocmask(SIG_SETMASK, &sigs, &old);
+ err = switch_mm(mm_idp->u.mm_fd, MM_SP_IP, syscall_regs, return_regs);
+ sigprocmask(SIG_SETMASK, &old, NULL);
+
+ ret = syscall_stub_done(mm_idp->stack);
+
+ *addr = check_init_stack(mm_idp, NULL);
+
+ return ret;
+}
+
static int flush_syscalls(struct mm_id *mm_idp, void **addr, int extra)
{
unsigned long *stack = check_init_stack(mm_idp, *addr);
@@ -134,10 +164,16 @@ static int flush_syscalls(struct mm_id *mm_idp, void **addr, int extra)
current = ((unsigned long) stack) & ~UM_KERN_PAGE_MASK;
end = UM_KERN_PAGE_SIZE;
+ if(have_switch_mm)
+ end -= (MAX_REG_NR + 1) * sizeof(long);
+
if (current + (10 + extra) * sizeof(long) < end)
return 0;
- return do_syscall_stub(mm_idp, addr);
+ if (have_switch_mm)
+ return do_syscall_stub_skas4(mm_idp, addr);
+ else
+ return do_syscall_stub(mm_idp, addr);
}
long run_syscall_stub(struct mm_id *mm_idp, int syscall,
@@ -166,8 +202,12 @@ long run_syscall_stub(struct mm_id *mm_idp, int syscall,
*stack++ = expected;
*stack = 0;
- if (done)
- return do_syscall_stub(mm_idp, addr);
+ if (done) {
+ if (have_switch_mm)
+ return do_syscall_stub_skas4(mm_idp, addr);
+ else
+ return do_syscall_stub(mm_idp, addr);
+ }
*addr = stack;
diff --git a/arch/um/os-Linux/skas/process.c b/arch/um/os-Linux/skas/process.c
index 522d0f1..9c2c086 100644
--- a/arch/um/os-Linux/skas/process.c
+++ b/arch/um/os-Linux/skas/process.c
@@ -227,7 +227,7 @@ static int userspace_tramp(void *stack)
}
}
}
- if (!ptrace_faultinfo && (stack != NULL)) {
+ if (!ptrace_faultinfo) {
struct sigaction sa;
unsigned long v = STUB_CODE +
@@ -271,7 +271,7 @@ int start_userspace(unsigned long stub_stack)
sp = (unsigned long) stack + UM_KERN_PAGE_SIZE - sizeof(void *);
flags = CLONE_FILES;
- if (proc_mm)
+ if (proc_mm || have_switch_mm)
flags |= CLONE_VM;
else
flags |= SIGCHLD;
@@ -384,8 +384,14 @@ void userspace(struct uml_pt_regs *regs)
printk(UM_KERN_ERR "userspace - child stopped "
"with signal %d\n", sig);
}
- pid = userspace_pid[0];
+
+ /*
+ * userspace_pid can change in in_interrupt since
+ * PTRACE_SWITCH_MM can cause a process to change
+ * address spaces
+ */
interrupt_end();
+ pid = userspace_pid[0];
/* Avoid -ERESTARTSYS handling in host */
if (PT_SYSCALL_NR_OFFSET != PT_SYSCALL_RET_OFFSET)
@@ -473,6 +479,69 @@ int copy_context_skas0(unsigned long new_stack, int pid)
return pid;
}
+extern unsigned long switch_mm_stub;
+extern long task_size;
+
+static void unmap_new_as(void)
+{
+ void (*p)(void);
+ void *addr;
+ unsigned long stack = (unsigned long) &stack & ~(UM_KERN_PAGE_SIZE - 1);
+ unsigned long long data_offset, code_offset;
+ int fd = phys_mapping(to_phys((void *) stack), &data_offset);
+
+ addr = mmap((void *) STUB_DATA, UM_KERN_PAGE_SIZE,
+ PROT_READ | PROT_WRITE, MAP_SHARED | MAP_FIXED, fd,
+ data_offset);
+ if (addr == MAP_FAILED)
+ panic("Failed to remap stack");
+
+ fd = phys_mapping(to_phys(&__syscall_stub_start), &code_offset);
+ addr = mmap((void *) STUB_CODE, UM_KERN_PAGE_SIZE,
+ PROT_READ | PROT_WRITE, MAP_SHARED | MAP_FIXED, fd,
+ code_offset);
+ if (addr == MAP_FAILED)
+ panic("Failed to remap code");
+
+ p = (void (*)(void)) (STUB_CODE +
+ ((unsigned long) &switch_mm_stub -
+ (unsigned long) &__syscall_stub_start));
+ (*p)();
+}
+
+int copy_context_skas4(struct mm_id *id)
+{
+ void *data = NULL;
+ unsigned long *return_regs, *fd_ptr, regs[MAX_REG_NR];
+ sigset_t sigs, old;
+ int err;
+
+ err = unmap(id, 0, STUB_START, 0, &data);
+ if (err)
+ return err;
+
+ err = unmap(id, STUB_END, task_size - STUB_END, 0, &data);
+ if (err)
+ return err;
+
+ return_regs = (unsigned long *) (id->stack + UM_KERN_PAGE_SIZE -
+ MAX_REG_NR * sizeof(long));
+ fd_ptr = return_regs - 1;
+ *fd_ptr = self_mm_fd;
+
+ regs[REGS_IP_INDEX] = (unsigned long) unmap_new_as;
+ regs[REGS_SP_INDEX] = id->stack + UM_KERN_PAGE_SIZE / 2;
+
+ sigfillset(&sigs);
+ sigprocmask(SIG_SETMASK, &sigs, &old);
+
+ err = switch_mm(id->u.mm_fd, MM_SP_IP, regs, return_regs);
+
+ sigprocmask(SIG_SETMASK, &old, NULL);
+
+ return err;
+}
+
/*
* This is used only, if stub pages are needed, while proc_mm is
* available. Opening /proc/mm creates a new mm_context, which lacks
@@ -630,8 +699,15 @@ void __switch_mm(struct mm_id *mm_idp)
err = ptrace(OLD_PTRACE_SWITCH_MM, userspace_pid[0], 0,
mm_idp->u.mm_fd);
if (err)
- panic("__switch_mm - PTRACE_SWITCH_MM failed, "
+ panic("__switch_mm - OLD_PTRACE_SWITCH_MM failed, "
"errno = %d\n", errno);
}
+ else if (have_ptrace_switch_mm) {
+ err = ptrace(PTRACE_SWITCH_MM, userspace_pid[0], 0,
+ mm_idp->u.mm_fd);
+ if (err)
+ panic("__switch_mm - PTRACE_SWITCH_MM "
+ "failed, errno = %d\n", errno);
+ }
else userspace_pid[0] = mm_idp->u.pid;
}
diff --git a/arch/um/os-Linux/start_up.c b/arch/um/os-Linux/start_up.c
index 0e0f738..a012723 100644
--- a/arch/um/os-Linux/start_up.c
+++ b/arch/um/os-Linux/start_up.c
@@ -161,6 +161,9 @@ static int disable_switch_mm;
int have_siginfo_segv;
static int disable_siginfo_segv;
+int have_ptrace_switch_mm;
+static int disable_ptrace_switch_mm;
+
int skas_needs_stub;
static int __init skas0_cmd_param(char *str, int* add)
@@ -168,8 +171,10 @@ static int __init skas0_cmd_param(char *str, int* add)
disable_ptrace_faultinfo = 1;
disable_ptrace_ldt = 1;
disable_proc_mm = 1;
+
disable_switch_mm = 1;
disable_siginfo_segv = 1;
+ disable_ptrace_switch_mm = 1;
return 0;
}
@@ -483,6 +488,18 @@ static inline void check_skas3_proc_mm(void)
}
}
+static void can_do_skas3(void)
+{
+ non_fatal("Checking for the skas3 patch in the host:\n");
+
+ check_skas3_proc_mm();
+ check_skas3_ptrace_faultinfo();
+ check_skas3_ptrace_ldt();
+
+ if (!proc_mm || !ptrace_faultinfo || !ptrace_ldt)
+ skas_needs_stub = 1;
+}
+
static void *fault_address;
static int check_fault_info(struct faultinfo *fi)
@@ -614,17 +631,207 @@ static int check_siginfo(void)
return ok;
}
-void can_do_skas(void)
+static char *mm_stack;
+static unsigned long return_regs[MAX_REG_NR];
+int self_mm_fd;
+
+static int switch_mm_works;
+
+static void after_switch(void)
{
- non_fatal("Checking for the skas3 patch in the host:\n");
+ /*
+ * If we are really in a new address space, setting this to
+ * zero won't affect the value of 1 already set in the old
+ * address space.
+ */
+ switch_mm_works = 0;
- check_skas3_proc_mm();
- check_skas3_ptrace_faultinfo();
- check_skas3_ptrace_ldt();
- check_siginfo();
+ switch_mm(self_mm_fd, MM_ALL_REGS, return_regs, NULL);
+}
- if (!proc_mm || !ptrace_faultinfo || !ptrace_ldt)
+static int check_switch_mm(void)
+{
+ unsigned long regs[MAX_REG_NR];
+ int err, there = -1;
+
+ non_fatal("\t/proc/self/mm ... ");
+ self_mm_fd = open("/proc/self/mm", O_RDONLY);
+ if (self_mm_fd < 0)
+ goto bad;
+ non_fatal("OK\n");
+
+ mm_stack = mmap(NULL, UM_KERN_PAGE_SIZE, PROT_READ | PROT_WRITE,
+ MAP_PRIVATE | MAP_ANONYMOUS, -1, 0);
+ if(mm_stack == MAP_FAILED)
+ goto bad;
+
+ non_fatal("\tnew_mm ... ");
+ there = new_mm();
+ if(there < 0)
+ goto bad;
+ non_fatal("OK\n");
+
+ regs[REGS_IP_INDEX] = (unsigned long) after_switch;
+ regs[REGS_SP_INDEX] = ((unsigned long) &mm_stack[UM_KERN_PAGE_SIZE]) -
+ sizeof(void *);
+
+ non_fatal("\tswitching over ... ");
+ err = switch_mm(there, MM_SP_IP, regs, return_regs);
+ if (err < 0)
+ goto bad;
+ non_fatal("switched back ... ");
+ switch_mm_works = 1;
+ if(!switch_mm_works)
+ goto bad;
+ else
+ non_fatal("OK\n");
+
+ munmap(mm_stack, UM_KERN_PAGE_SIZE);
+ close(there);
+
+ if (disable_switch_mm)
+ non_fatal("switch_mm support disabled on command line\n");
+ else
+ have_switch_mm = 1;
+
+ return 1;
+ bad:
+ if (there > 0)
+ close(there);
+ munmap(mm_stack, UM_KERN_PAGE_SIZE);
+ non_fatal("Failed - \n");
+ perror("");
+ return 0;
+}
+
+static int ptrace_switch_mm_works;
+
+static int after_ptrace_switch(void)
+{
+ ptrace_switch_mm_works = 1;
+ exit(0);
+}
+
+static int check_ptrace_switch_mm(void)
+{
+ void *stack;
+ unsigned long regs[MAX_REG_NR];
+ int pid, here, err, status;
+
+ non_fatal("\tPTRACE_SWITCH_MM ... ");
+ pid = fork();
+ if(pid == 0){
+ ptrace(PTRACE_TRACEME, 0, 0, 0);
+ kill(getpid(), SIGSTOP);
+
+ exit(0);
+ }
+ else if(pid < 0)
+ goto bad;
+
+ stack = mmap(NULL, UM_KERN_PAGE_SIZE, PROT_READ | PROT_WRITE,
+ MAP_SHARED | MAP_ANONYMOUS, -1, 0);
+ if(stack == MAP_FAILED)
+ goto bad;
+
+ here = open("/proc/self/mm", O_RDONLY);
+ if(here < 0)
+ goto bad_unmap;
+
+ err = waitpid(pid, &status, WUNTRACED);
+ if (err < 0)
+ goto bad_close;
+ else if (err != pid) {
+ non_fatal("waitpid returned %d, expected %d\n", err, pid);
+ goto bad_close;
+ }
+ else if (!WIFSTOPPED(status) || (WSTOPSIG(status) != SIGSTOP)) {
+ non_fatal("waitpid returned status 0x%d\n", status);
+ goto bad_close;
+ }
+
+ err = ptrace(PTRACE_GETREGS, pid, 0, regs);
+ if (err < 0)
+ goto bad_close;
+
+ regs[REGS_IP_INDEX] = (unsigned long) after_ptrace_switch;
+ regs[REGS_SP_INDEX] = (unsigned long) stack + UM_KERN_PAGE_SIZE -
+ sizeof(void *);
+
+ if (ptrace(PTRACE_SETREGS, pid, 0, regs) < 0)
+ goto bad_close;
+
+ if (ptrace(PTRACE_SWITCH_MM, pid, NULL, here) < 0)
+ goto bad_close;
+
+ if (ptrace(PTRACE_CONT, pid, NULL, 0) < 0)
+ goto bad_close;
+
+ err = waitpid(pid, &status, WUNTRACED);
+ if (err < 0)
+ goto bad_close;
+ else if(err != pid) {
+ non_fatal("waitpid returned %d, expected %d\n", err, pid);
+ goto bad_close;
+ }
+ else if (!WIFEXITED(status) || (WEXITSTATUS(status) != 0)) {
+ non_fatal("waitpid returned status 0x%d\n", status);
+ goto bad_close;
+ }
+
+ if (!ptrace_switch_mm_works)
+ goto bad_close;
+ else non_fatal("OK\n");
+
+ if (disable_ptrace_switch_mm)
+ non_fatal("PTRACE_SWITCH_MM support disabled on command "
+ "line\n");
+ else
+ have_ptrace_switch_mm = 1;
+
+ close(here);
+ munmap(stack, UM_KERN_PAGE_SIZE);
+
+ return 1;
+
+ bad_close:
+ close(here);
+ bad_unmap:
+ munmap(stack, UM_KERN_PAGE_SIZE);
+ bad:
+ non_fatal("Failed - \n");
+ perror("");
+ return 0;
+}
+
+static int can_do_skas4(void)
+{
+ int ret;
+
+ non_fatal("Checking for new_mm and switch_mm support in the host:\n");
+
+ ret = check_switch_mm() && check_ptrace_switch_mm() && check_siginfo();
+ if (ret)
skas_needs_stub = 1;
+
+ return ret;
+}
+
+void can_do_skas(void)
+{
+ if (!can_do_skas4())
+ can_do_skas3();
+}
+
+int get_new_mm(void)
+{
+ int err;
+
+ err = new_mm();
+ if (err < 0)
+ err = -errno;
+
+ return err;
}
int __init parse_iomem(char *str, int *add)
diff --git a/arch/um/sys-i386/stub.S b/arch/um/sys-i386/stub.S
index 890dc50..3cfb6e8 100644
--- a/arch/um/sys-i386/stub.S
+++ b/arch/um/sys-i386/stub.S
@@ -1,5 +1,8 @@
#include "uml-config.h"
#include "as-layout.h"
+#include "skas/skas.h"
+
+#define MAX_REG_NR 17
#define PROCESS_LIST \
/* load pointer to first operation */ \
@@ -43,3 +46,16 @@ batch_syscall_stub:
PROCESS_LIST
/* stop */
int3
+
+ .globl switch_mm_stub
+switch_mm_stub:
+ PROCESS_LIST
+
+ mov $__NR_switch_mm, %eax
+ mov ASM_STUB_DATA + UM_KERN_PAGE_SIZE - MAX_REG_NR * 4 - 4, %ebx
+ mov $MM_ALL_REGS, %ecx
+ mov $(ASM_STUB_DATA + UM_KERN_PAGE_SIZE - MAX_REG_NR * 4), %edx
+ xor %esi, %esi
+ int $0x80
+
+ int3
diff --git a/arch/um/sys-x86_64/stub.S b/arch/um/sys-x86_64/stub.S
index 143a16c..fb10353 100644
--- a/arch/um/sys-x86_64/stub.S
+++ b/arch/um/sys-x86_64/stub.S
@@ -1,5 +1,8 @@
#include "uml-config.h"
#include "as-layout.h"
+#include "skas/skas.h"
+
+#define MAX_REG_NR 27
#define PROCESS_LIST \
mov $(ASM_STUB_DATA >> 32), %rbx; \
@@ -46,3 +49,21 @@ batch_syscall_stub:
PROCESS_LIST
/* stop */
int3
+
+ .globl switch_mm_stub
+switch_mm_stub:
+ PROCESS_LIST
+
+ mov $__NR_switch_mm, %rax
+ mov $(ASM_STUB_DATA >> 32), %rdi
+ sal $32, %rdi
+ mov $(ASM_STUB_DATA & 0xffffffff), %rsi
+ add %rsi, %rdi
+ add $(UM_KERN_PAGE_SIZE - MAX_REG_NR * 8 - 8), %rdi
+ mov (%rdi), %rdi
+ mov $MM_ALL_REGS, %rsi
+ mov $(ASM_STUB_DATA + UM_KERN_PAGE_SIZE - MAX_REG_NR * 8), %rdx
+ xor %r10, %r10
+ syscall
+
+ int3
diff --git a/arch/um/sys-x86_64/syscalls.c b/arch/um/sys-x86_64/syscalls.c
index ff012ba..00131f9 100644
--- a/arch/um/sys-x86_64/syscalls.c
+++ b/arch/um/sys-x86_64/syscalls.c
@@ -30,7 +30,7 @@ long arch_prctl(struct task_struct *task, int code, unsigned long __user *addr)
{
unsigned long *ptr = addr, tmp;
long ret;
- int pid = task->mm->context.id.u.pid;
+ int pid = userspace_pid[0];
/*
* With ARCH_SET_FS (and ARCH_SET_GS is treated similarly to
-------------------------------------------------------------------------
This SF.net email is sponsored by: Microsoft
Defy all challenges. Microsoft(R) Visual Studio 2008.
http://clk.atdmt.com/MRT/go/vse0120000070mrt/direct/01/
_______________________________________________
User-mode-linux-devel mailing list
User-mode-linux-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/user-mode-linux-devel
^ permalink raw reply related [flat|nested] 3+ messages in thread
end of thread, other threads:[~2008-01-28 21:38 UTC | newest]
Thread overview: 3+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2008-01-14 21:36 [uml-devel] [RFC PATCH 10/10] SKAS4 - Guest get_mm and switch_mm support Jeff Dike
-- strict thread matches above, loose matches on Subject: below --
2008-01-23 17:12 Jeff Dike
2008-01-28 21:38 Jeff Dike
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.