netns xfrm: fix "ip xfrm state|policy count" misreport
[safe/jmp/linux-2.6] / kernel / compat.c
index d4898aa..f6c204f 100644 (file)
 #include <linux/timex.h>
 #include <linux/migrate.h>
 #include <linux/posix-timers.h>
+#include <linux/times.h>
+#include <linux/ptrace.h>
 
 #include <asm/uaccess.h>
 
+/*
+ * Note that the native side is already converted to a timespec, because
+ * that's what we want anyway.
+ */
+static int compat_get_timeval(struct timespec *o,
+               struct compat_timeval __user *i)
+{
+       long usec;
+
+       if (get_user(o->tv_sec, &i->tv_sec) ||
+           get_user(usec, &i->tv_usec))
+               return -EFAULT;
+       o->tv_nsec = usec * 1000;
+       return 0;
+}
+
+static int compat_put_timeval(struct compat_timeval __user *o,
+               struct timeval *i)
+{
+       return (put_user(i->tv_sec, &o->tv_sec) ||
+               put_user(i->tv_usec, &o->tv_usec)) ? -EFAULT : 0;
+}
+
+asmlinkage long compat_sys_gettimeofday(struct compat_timeval __user *tv,
+               struct timezone __user *tz)
+{
+       if (tv) {
+               struct timeval ktv;
+               do_gettimeofday(&ktv);
+               if (compat_put_timeval(tv, &ktv))
+                       return -EFAULT;
+       }
+       if (tz) {
+               if (copy_to_user(tz, &sys_tz, sizeof(sys_tz)))
+                       return -EFAULT;
+       }
+
+       return 0;
+}
+
+asmlinkage long compat_sys_settimeofday(struct compat_timeval __user *tv,
+               struct timezone __user *tz)
+{
+       struct timespec kts;
+       struct timezone ktz;
+
+       if (tv) {
+               if (compat_get_timeval(&kts, tv))
+                       return -EFAULT;
+       }
+       if (tz) {
+               if (copy_from_user(&ktz, tz, sizeof(ktz)))
+                       return -EFAULT;
+       }
+
+       return do_sys_settimeofday(tv ? &kts : NULL, tz ? &ktz : NULL);
+}
+
 int get_compat_timespec(struct timespec *ts, const struct compat_timespec __user *cts)
 {
        return (!access_ok(VERIFY_READ, cts, sizeof(*cts)) ||
@@ -42,60 +102,59 @@ int put_compat_timespec(const struct timespec *ts, struct compat_timespec __user
 
 static long compat_nanosleep_restart(struct restart_block *restart)
 {
-       unsigned long expire = restart->arg0, now = jiffies;
        struct compat_timespec __user *rmtp;
+       struct timespec rmt;
+       mm_segment_t oldfs;
+       long ret;
 
-       /* Did it expire while we handled signals? */
-       if (!time_after(expire, now))
-               return 0;
-
-       expire = schedule_timeout_interruptible(expire - now);
-       if (expire == 0)
-               return 0;
+       restart->nanosleep.rmtp = (struct timespec __user *) &rmt;
+       oldfs = get_fs();
+       set_fs(KERNEL_DS);
+       ret = hrtimer_nanosleep_restart(restart);
+       set_fs(oldfs);
 
-       rmtp = (struct compat_timespec __user *)restart->arg1;
-       if (rmtp) {
-               struct compat_timespec ct;
-               struct timespec t;
+       if (ret) {
+               rmtp = restart->nanosleep.compat_rmtp;
 
-               jiffies_to_timespec(expire, &t);
-               ct.tv_sec = t.tv_sec;
-               ct.tv_nsec = t.tv_nsec;
-               if (copy_to_user(rmtp, &ct, sizeof(ct)))
+               if (rmtp && put_compat_timespec(&rmt, rmtp))
                        return -EFAULT;
        }
-       /* The 'restart' block is already filled in */
-       return -ERESTART_RESTARTBLOCK;
+
+       return ret;
 }
 
 asmlinkage long compat_sys_nanosleep(struct compat_timespec __user *rqtp,
-               struct compat_timespec __user *rmtp)
+                                    struct compat_timespec __user *rmtp)
 {
-       struct timespec t;
-       struct restart_block *restart;
-       unsigned long expire;
+       struct timespec tu, rmt;
+       mm_segment_t oldfs;
+       long ret;
 
-       if (get_compat_timespec(&t, rqtp))
+       if (get_compat_timespec(&tu, rqtp))
                return -EFAULT;
 
-       if ((t.tv_nsec >= 1000000000L) || (t.tv_nsec < 0) || (t.tv_sec < 0))
+       if (!timespec_valid(&tu))
                return -EINVAL;
 
-       expire = timespec_to_jiffies(&t) + (t.tv_sec || t.tv_nsec);
-       expire = schedule_timeout_interruptible(expire);
-       if (expire == 0)
-               return 0;
+       oldfs = get_fs();
+       set_fs(KERNEL_DS);
+       ret = hrtimer_nanosleep(&tu,
+                               rmtp ? (struct timespec __user *)&rmt : NULL,
+                               HRTIMER_MODE_REL, CLOCK_MONOTONIC);
+       set_fs(oldfs);
+
+       if (ret) {
+               struct restart_block *restart
+                       = &current_thread_info()->restart_block;
 
-       if (rmtp) {
-               jiffies_to_timespec(expire, &t);
-               if (put_compat_timespec(&t, rmtp))
+               restart->fn = compat_nanosleep_restart;
+               restart->nanosleep.compat_rmtp = rmtp;
+
+               if (rmtp && put_compat_timespec(&rmt, rmtp))
                        return -EFAULT;
        }
-       restart = &current_thread_info()->restart_block;
-       restart->fn = compat_nanosleep_restart;
-       restart->arg0 = jiffies + expire;
-       restart->arg1 = (unsigned long) rmtp;
-       return -ERESTART_RESTARTBLOCK;
+
+       return ret;
 }
 
 static inline long get_compat_itimerval(struct itimerval *o,
@@ -151,52 +210,27 @@ asmlinkage long compat_sys_setitimer(int which,
        return 0;
 }
 
+static compat_clock_t clock_t_to_compat_clock_t(clock_t x)
+{
+       return compat_jiffies_to_clock_t(clock_t_to_jiffies(x));
+}
+
 asmlinkage long compat_sys_times(struct compat_tms __user *tbuf)
 {
-       /*
-        *      In the SMP world we might just be unlucky and have one of
-        *      the times increment as we use it. Since the value is an
-        *      atomically safe type this is just fine. Conceptually its
-        *      as if the syscall took an instant longer to occur.
-        */
        if (tbuf) {
+               struct tms tms;
                struct compat_tms tmp;
-               struct task_struct *tsk = current;
-               struct task_struct *t;
-               cputime_t utime, stime, cutime, cstime;
-
-               read_lock(&tasklist_lock);
-               utime = tsk->signal->utime;
-               stime = tsk->signal->stime;
-               t = tsk;
-               do {
-                       utime = cputime_add(utime, t->utime);
-                       stime = cputime_add(stime, t->stime);
-                       t = next_thread(t);
-               } while (t != tsk);
-
-               /*
-                * While we have tasklist_lock read-locked, no dying thread
-                * can be updating current->signal->[us]time.  Instead,
-                * we got their counts included in the live thread loop.
-                * However, another thread can come in right now and
-                * do a wait call that updates current->signal->c[us]time.
-                * To make sure we always see that pair updated atomically,
-                * we take the siglock around fetching them.
-                */
-               spin_lock_irq(&tsk->sighand->siglock);
-               cutime = tsk->signal->cutime;
-               cstime = tsk->signal->cstime;
-               spin_unlock_irq(&tsk->sighand->siglock);
-               read_unlock(&tasklist_lock);
-
-               tmp.tms_utime = compat_jiffies_to_clock_t(cputime_to_jiffies(utime));
-               tmp.tms_stime = compat_jiffies_to_clock_t(cputime_to_jiffies(stime));
-               tmp.tms_cutime = compat_jiffies_to_clock_t(cputime_to_jiffies(cutime));
-               tmp.tms_cstime = compat_jiffies_to_clock_t(cputime_to_jiffies(cstime));
+
+               do_sys_times(&tms);
+               /* Convert our struct tms to the compat version. */
+               tmp.tms_utime = clock_t_to_compat_clock_t(tms.tms_utime);
+               tmp.tms_stime = clock_t_to_compat_clock_t(tms.tms_stime);
+               tmp.tms_cutime = clock_t_to_compat_clock_t(tms.tms_cutime);
+               tmp.tms_cstime = clock_t_to_compat_clock_t(tms.tms_cstime);
                if (copy_to_user(tbuf, &tmp, sizeof(tmp)))
                        return -EFAULT;
        }
+       force_successful_syscall_return();
        return compat_jiffies_to_clock_t(jiffies);
 }
 
@@ -247,8 +281,8 @@ asmlinkage long compat_sys_setrlimit(unsigned int resource,
        int ret;
        mm_segment_t old_fs = get_fs ();
 
-       if (resource >= RLIM_NLIMITS) 
-               return -EINVAL; 
+       if (resource >= RLIM_NLIMITS)
+               return -EINVAL;
 
        if (!access_ok(VERIFY_READ, rlim, sizeof(*rlim)) ||
            __get_user(r.rlim_cur, &rlim->rlim_cur) ||
@@ -422,16 +456,16 @@ asmlinkage long compat_sys_waitid(int which, compat_pid_t pid,
 }
 
 static int compat_get_user_cpu_mask(compat_ulong_t __user *user_mask_ptr,
-                                   unsigned len, cpumask_t *new_mask)
+                                   unsigned len, struct cpumask *new_mask)
 {
        unsigned long *k;
 
-       if (len < sizeof(cpumask_t))
-               memset(new_mask, 0, sizeof(cpumask_t));
-       else if (len > sizeof(cpumask_t))
-               len = sizeof(cpumask_t);
+       if (len < cpumask_size())
+               memset(new_mask, 0, cpumask_size());
+       else if (len > cpumask_size())
+               len = cpumask_size();
 
-       k = cpus_addr(*new_mask);
+       k = cpumask_bits(new_mask);
        return compat_get_bitmap(k, user_mask_ptr, len * 8);
 }
 
@@ -439,59 +473,70 @@ asmlinkage long compat_sys_sched_setaffinity(compat_pid_t pid,
                                             unsigned int len,
                                             compat_ulong_t __user *user_mask_ptr)
 {
-       cpumask_t new_mask;
+       cpumask_var_t new_mask;
        int retval;
 
-       retval = compat_get_user_cpu_mask(user_mask_ptr, len, &new_mask);
+       if (!alloc_cpumask_var(&new_mask, GFP_KERNEL))
+               return -ENOMEM;
+
+       retval = compat_get_user_cpu_mask(user_mask_ptr, len, new_mask);
        if (retval)
-               return retval;
+               goto out;
 
-       return sched_setaffinity(pid, new_mask);
+       retval = sched_setaffinity(pid, new_mask);
+out:
+       free_cpumask_var(new_mask);
+       return retval;
 }
 
 asmlinkage long compat_sys_sched_getaffinity(compat_pid_t pid, unsigned int len,
                                             compat_ulong_t __user *user_mask_ptr)
 {
        int ret;
-       cpumask_t mask;
+       cpumask_var_t mask;
        unsigned long *k;
-       unsigned int min_length = sizeof(cpumask_t);
+       unsigned int min_length = cpumask_size();
 
-       if (NR_CPUS <= BITS_PER_COMPAT_LONG)
+       if (nr_cpu_ids <= BITS_PER_COMPAT_LONG)
                min_length = sizeof(compat_ulong_t);
 
        if (len < min_length)
                return -EINVAL;
 
-       ret = sched_getaffinity(pid, &mask);
+       if (!alloc_cpumask_var(&mask, GFP_KERNEL))
+               return -ENOMEM;
+
+       ret = sched_getaffinity(pid, mask);
        if (ret < 0)
-               return ret;
+               goto out;
 
-       k = cpus_addr(mask);
+       k = cpumask_bits(mask);
        ret = compat_put_bitmap(user_mask_ptr, k, min_length * 8);
-       if (ret)
-               return ret;
+       if (ret == 0)
+               ret = min_length;
 
-       return min_length;
+out:
+       free_cpumask_var(mask);
+       return ret;
 }
 
-static int get_compat_itimerspec(struct itimerspec *dst, 
-                                struct compat_itimerspec __user *src)
-{ 
+int get_compat_itimerspec(struct itimerspec *dst,
+                         const struct compat_itimerspec __user *src)
+{
        if (get_compat_timespec(&dst->it_interval, &src->it_interval) ||
            get_compat_timespec(&dst->it_value, &src->it_value))
                return -EFAULT;
        return 0;
-} 
+}
 
-static int put_compat_itimerspec(struct compat_itimerspec __user *dst, 
-                                struct itimerspec *src)
-{ 
+int put_compat_itimerspec(struct compat_itimerspec __user *dst,
+                         const struct itimerspec *src)
+{
        if (put_compat_timespec(&src->it_interval, &dst->it_interval) ||
            put_compat_timespec(&src->it_value, &dst->it_value))
                return -EFAULT;
        return 0;
-} 
+}
 
 long compat_sys_timer_create(clockid_t which_clock,
                        struct compat_sigevent __user *timer_event_spec,
@@ -512,9 +557,9 @@ long compat_sys_timer_create(clockid_t which_clock,
 }
 
 long compat_sys_timer_settime(timer_t timer_id, int flags,
-                         struct compat_itimerspec __user *new, 
+                         struct compat_itimerspec __user *new,
                          struct compat_itimerspec __user *old)
-{ 
+{
        long err;
        mm_segment_t oldfs;
        struct itimerspec newts, oldts;
@@ -522,58 +567,58 @@ long compat_sys_timer_settime(timer_t timer_id, int flags,
        if (!new)
                return -EINVAL;
        if (get_compat_itimerspec(&newts, new))
-               return -EFAULT; 
+               return -EFAULT;
        oldfs = get_fs();
        set_fs(KERNEL_DS);
        err = sys_timer_settime(timer_id, flags,
                                (struct itimerspec __user *) &newts,
                                (struct itimerspec __user *) &oldts);
-       set_fs(oldfs); 
+       set_fs(oldfs);
        if (!err && old && put_compat_itimerspec(old, &oldts))
                return -EFAULT;
        return err;
-} 
+}
 
 long compat_sys_timer_gettime(timer_t timer_id,
                struct compat_itimerspec __user *setting)
-{ 
+{
        long err;
        mm_segment_t oldfs;
-       struct itimerspec ts; 
+       struct itimerspec ts;
 
        oldfs = get_fs();
        set_fs(KERNEL_DS);
        err = sys_timer_gettime(timer_id,
-                               (struct itimerspec __user *) &ts); 
-       set_fs(oldfs); 
+                               (struct itimerspec __user *) &ts);
+       set_fs(oldfs);
        if (!err && put_compat_itimerspec(setting, &ts))
                return -EFAULT;
        return err;
-} 
+}
 
 long compat_sys_clock_settime(clockid_t which_clock,
                struct compat_timespec __user *tp)
 {
        long err;
        mm_segment_t oldfs;
-       struct timespec ts; 
+       struct timespec ts;
 
        if (get_compat_timespec(&ts, tp))
-               return -EFAULT; 
+               return -EFAULT;
        oldfs = get_fs();
-       set_fs(KERNEL_DS);      
+       set_fs(KERNEL_DS);
        err = sys_clock_settime(which_clock,
                                (struct timespec __user *) &ts);
        set_fs(oldfs);
        return err;
-} 
+}
 
 long compat_sys_clock_gettime(clockid_t which_clock,
                struct compat_timespec __user *tp)
 {
        long err;
        mm_segment_t oldfs;
-       struct timespec ts; 
+       struct timespec ts;
 
        oldfs = get_fs();
        set_fs(KERNEL_DS);
@@ -581,16 +626,16 @@ long compat_sys_clock_gettime(clockid_t which_clock,
                                (struct timespec __user *) &ts);
        set_fs(oldfs);
        if (!err && put_compat_timespec(&ts, tp))
-               return -EFAULT; 
+               return -EFAULT;
        return err;
-} 
+}
 
 long compat_sys_clock_getres(clockid_t which_clock,
                struct compat_timespec __user *tp)
 {
        long err;
        mm_segment_t oldfs;
-       struct timespec ts; 
+       struct timespec ts;
 
        oldfs = get_fs();
        set_fs(KERNEL_DS);
@@ -598,18 +643,18 @@ long compat_sys_clock_getres(clockid_t which_clock,
                               (struct timespec __user *) &ts);
        set_fs(oldfs);
        if (!err && tp && put_compat_timespec(&ts, tp))
-               return -EFAULT; 
+               return -EFAULT;
        return err;
-} 
+}
 
 static long compat_clock_nanosleep_restart(struct restart_block *restart)
 {
        long err;
        mm_segment_t oldfs;
        struct timespec tu;
-       struct compat_timespec *rmtp = (struct compat_timespec *)(restart->arg1);
+       struct compat_timespec *rmtp = restart->nanosleep.compat_rmtp;
 
-       restart->arg1 = (unsigned long) &tu;
+       restart->nanosleep.rmtp = (struct timespec __user *) &tu;
        oldfs = get_fs();
        set_fs(KERNEL_DS);
        err = clock_nanosleep_restart(restart);
@@ -621,7 +666,7 @@ static long compat_clock_nanosleep_restart(struct restart_block *restart)
 
        if (err == -ERESTART_RESTARTBLOCK) {
                restart->fn = compat_clock_nanosleep_restart;
-               restart->arg1 = (unsigned long) rmtp;
+               restart->nanosleep.compat_rmtp = rmtp;
        }
        return err;
 }
@@ -632,10 +677,10 @@ long compat_sys_clock_nanosleep(clockid_t which_clock, int flags,
 {
        long err;
        mm_segment_t oldfs;
-       struct timespec in, out; 
+       struct timespec in, out;
        struct restart_block *restart;
 
-       if (get_compat_timespec(&in, rqtp)) 
+       if (get_compat_timespec(&in, rqtp))
                return -EFAULT;
 
        oldfs = get_fs();
@@ -652,10 +697,10 @@ long compat_sys_clock_nanosleep(clockid_t which_clock, int flags,
        if (err == -ERESTART_RESTARTBLOCK) {
                restart = &current_thread_info()->restart_block;
                restart->fn = compat_clock_nanosleep_restart;
-               restart->arg1 = (unsigned long) rmtp;
+               restart->nanosleep.compat_rmtp = rmtp;
        }
-       return err;     
-} 
+       return err;
+}
 
 /*
  * We currently only need the following fields from the sigevent
@@ -837,6 +882,17 @@ compat_sys_rt_sigtimedwait (compat_sigset_t __user *uthese,
 
 }
 
+asmlinkage long
+compat_sys_rt_tgsigqueueinfo(compat_pid_t tgid, compat_pid_t pid, int sig,
+                            struct compat_siginfo __user *uinfo)
+{
+       siginfo_t info;
+
+       if (copy_siginfo_from_user32(&info, uinfo))
+               return -EFAULT;
+       return do_rt_tgsigqueueinfo(tgid, pid, sig, &info);
+}
+
 #ifdef __ARCH_WANT_COMPAT_SYS_TIME
 
 /* compat_time_t is a 32 bit "long" and needs to get converted. */
@@ -851,8 +907,9 @@ asmlinkage long compat_sys_time(compat_time_t __user * tloc)
 
        if (tloc) {
                if (put_user(i,tloc))
-                       i = -EFAULT;
+                       return -EFAULT;
        }
+       force_successful_syscall_return();
        return i;
 }
 
@@ -899,7 +956,7 @@ asmlinkage long compat_sys_rt_sigsuspend(compat_sigset_t __user *unewset, compat
 
        current->state = TASK_INTERRUPTIBLE;
        schedule();
-       set_thread_flag(TIF_RESTORE_SIGMASK);
+       set_restore_sigmask();
        return -ERESTARTNOHAND;
 }
 #endif /* __ARCH_WANT_COMPAT_SYS_RT_SIGSUSPEND */
@@ -956,7 +1013,8 @@ asmlinkage long compat_sys_adjtimex(struct compat_timex __user *utp)
                        __put_user(txc.jitcnt, &utp->jitcnt) ||
                        __put_user(txc.calcnt, &utp->calcnt) ||
                        __put_user(txc.errcnt, &utp->errcnt) ||
-                       __put_user(txc.stbcnt, &utp->stbcnt))
+                       __put_user(txc.stbcnt, &utp->stbcnt) ||
+                       __put_user(txc.tai, &utp->tai))
                ret = -EFAULT;
 
        return ret;
@@ -982,4 +1040,102 @@ asmlinkage long compat_sys_move_pages(pid_t pid, unsigned long nr_pages,
        }
        return sys_move_pages(pid, nr_pages, pages, nodes, status, flags);
 }
+
+asmlinkage long compat_sys_migrate_pages(compat_pid_t pid,
+                       compat_ulong_t maxnode,
+                       const compat_ulong_t __user *old_nodes,
+                       const compat_ulong_t __user *new_nodes)
+{
+       unsigned long __user *old = NULL;
+       unsigned long __user *new = NULL;
+       nodemask_t tmp_mask;
+       unsigned long nr_bits;
+       unsigned long size;
+
+       nr_bits = min_t(unsigned long, maxnode - 1, MAX_NUMNODES);
+       size = ALIGN(nr_bits, BITS_PER_LONG) / 8;
+       if (old_nodes) {
+               if (compat_get_bitmap(nodes_addr(tmp_mask), old_nodes, nr_bits))
+                       return -EFAULT;
+               old = compat_alloc_user_space(new_nodes ? size * 2 : size);
+               if (new_nodes)
+                       new = old + size / sizeof(unsigned long);
+               if (copy_to_user(old, nodes_addr(tmp_mask), size))
+                       return -EFAULT;
+       }
+       if (new_nodes) {
+               if (compat_get_bitmap(nodes_addr(tmp_mask), new_nodes, nr_bits))
+                       return -EFAULT;
+               if (new == NULL)
+                       new = compat_alloc_user_space(size);
+               if (copy_to_user(new, nodes_addr(tmp_mask), size))
+                       return -EFAULT;
+       }
+       return sys_migrate_pages(pid, nr_bits + 1, old, new);
+}
 #endif
+
+struct compat_sysinfo {
+       s32 uptime;
+       u32 loads[3];
+       u32 totalram;
+       u32 freeram;
+       u32 sharedram;
+       u32 bufferram;
+       u32 totalswap;
+       u32 freeswap;
+       u16 procs;
+       u16 pad;
+       u32 totalhigh;
+       u32 freehigh;
+       u32 mem_unit;
+       char _f[20-2*sizeof(u32)-sizeof(int)];
+};
+
+asmlinkage long
+compat_sys_sysinfo(struct compat_sysinfo __user *info)
+{
+       struct sysinfo s;
+
+       do_sysinfo(&s);
+
+       /* Check to see if any memory value is too large for 32-bit and scale
+        *  down if needed
+        */
+       if ((s.totalram >> 32) || (s.totalswap >> 32)) {
+               int bitcount = 0;
+
+               while (s.mem_unit < PAGE_SIZE) {
+                       s.mem_unit <<= 1;
+                       bitcount++;
+               }
+
+               s.totalram >>= bitcount;
+               s.freeram >>= bitcount;
+               s.sharedram >>= bitcount;
+               s.bufferram >>= bitcount;
+               s.totalswap >>= bitcount;
+               s.freeswap >>= bitcount;
+               s.totalhigh >>= bitcount;
+               s.freehigh >>= bitcount;
+       }
+
+       if (!access_ok(VERIFY_WRITE, info, sizeof(struct compat_sysinfo)) ||
+           __put_user (s.uptime, &info->uptime) ||
+           __put_user (s.loads[0], &info->loads[0]) ||
+           __put_user (s.loads[1], &info->loads[1]) ||
+           __put_user (s.loads[2], &info->loads[2]) ||
+           __put_user (s.totalram, &info->totalram) ||
+           __put_user (s.freeram, &info->freeram) ||
+           __put_user (s.sharedram, &info->sharedram) ||
+           __put_user (s.bufferram, &info->bufferram) ||
+           __put_user (s.totalswap, &info->totalswap) ||
+           __put_user (s.freeswap, &info->freeswap) ||
+           __put_user (s.procs, &info->procs) ||
+           __put_user (s.totalhigh, &info->totalhigh) ||
+           __put_user (s.freehigh, &info->freehigh) ||
+           __put_user (s.mem_unit, &info->mem_unit))
+               return -EFAULT;
+
+       return 0;
+}