aboutsummaryrefslogtreecommitdiff
path: root/linux-user/syscall.c
diff options
context:
space:
mode:
authorpbrook <pbrook@c046a42c-6fe2-441c-8c8c-71466251a162>2006-03-25 19:31:22 +0000
committerpbrook <pbrook@c046a42c-6fe2-441c-8c8c-71466251a162>2006-03-25 19:31:22 +0000
commit53a5960aadd542dd27b8705ac30df154557d5ffc (patch)
treefe1b96123cd7ea7a4770ba289c239f43a0e584d0 /linux-user/syscall.c
parent26f69dc09fa3c459fca53bbfd267cbfc2682201e (diff)
Avoid accessing guest memory directly in usermode emulation.
git-svn-id: svn://svn.savannah.nongnu.org/qemu/trunk@1790 c046a42c-6fe2-441c-8c8c-71466251a162
Diffstat (limited to 'linux-user/syscall.c')
-rw-r--r--linux-user/syscall.c1118
1 files changed, 733 insertions, 385 deletions
diff --git a/linux-user/syscall.c b/linux-user/syscall.c
index ab7846d7ac..2b132477f0 100644
--- a/linux-user/syscall.c
+++ b/linux-user/syscall.c
@@ -252,44 +252,43 @@ static inline int is_error(long ret)
return (unsigned long)ret >= (unsigned long)(-4096);
}
-static char *target_brk;
-static char *target_original_brk;
+static target_ulong target_brk;
+static target_ulong target_original_brk;
-void target_set_brk(char *new_brk)
+void target_set_brk(target_ulong new_brk)
{
- target_brk = new_brk;
- target_original_brk = new_brk;
+ target_original_brk = target_brk = new_brk;
}
-long do_brk(char *new_brk)
+long do_brk(target_ulong new_brk)
{
- char *brk_page;
+ target_ulong brk_page;
long mapped_addr;
int new_alloc_size;
if (!new_brk)
- return (long)target_brk;
+ return target_brk;
if (new_brk < target_original_brk)
return -ENOMEM;
- brk_page = (char *)HOST_PAGE_ALIGN((unsigned long)target_brk);
+ brk_page = HOST_PAGE_ALIGN(target_brk);
/* If the new brk is less than this, set it and we're done... */
if (new_brk < brk_page) {
target_brk = new_brk;
- return (long)target_brk;
+ return target_brk;
}
/* We need to allocate more memory after the brk... */
new_alloc_size = HOST_PAGE_ALIGN(new_brk - brk_page + 1);
- mapped_addr = get_errno(target_mmap((unsigned long)brk_page, new_alloc_size,
+ mapped_addr = get_errno(target_mmap(brk_page, new_alloc_size,
PROT_READ|PROT_WRITE,
MAP_ANON|MAP_FIXED|MAP_PRIVATE, 0, 0));
if (is_error(mapped_addr)) {
return mapped_addr;
} else {
target_brk = new_brk;
- return (long)target_brk;
+ return target_brk;
}
}
@@ -354,9 +353,12 @@ static inline long host_to_target_clock_t(long ticks)
#endif
}
-static inline void host_to_target_rusage(struct target_rusage *target_rusage,
+static inline void host_to_target_rusage(target_ulong target_addr,
const struct rusage *rusage)
{
+ struct target_rusage *target_rusage;
+
+ lock_user_struct(target_rusage, target_addr, 0);
target_rusage->ru_utime.tv_sec = tswapl(rusage->ru_utime.tv_sec);
target_rusage->ru_utime.tv_usec = tswapl(rusage->ru_utime.tv_usec);
target_rusage->ru_stime.tv_sec = tswapl(rusage->ru_stime.tv_sec);
@@ -375,35 +377,64 @@ static inline void host_to_target_rusage(struct target_rusage *target_rusage,
target_rusage->ru_nsignals = tswapl(rusage->ru_nsignals);
target_rusage->ru_nvcsw = tswapl(rusage->ru_nvcsw);
target_rusage->ru_nivcsw = tswapl(rusage->ru_nivcsw);
+ unlock_user_struct(target_rusage, target_addr, 1);
}
-static inline void target_to_host_timeval(struct timeval *tv,
- const struct target_timeval *target_tv)
+static inline void target_to_host_timeval(struct timeval *tv,
+ target_ulong target_addr)
{
+ struct target_timeval *target_tv;
+
+ lock_user_struct(target_tv, target_addr, 1);
tv->tv_sec = tswapl(target_tv->tv_sec);
tv->tv_usec = tswapl(target_tv->tv_usec);
+ unlock_user_struct(target_tv, target_addr, 0);
}
-static inline void host_to_target_timeval(struct target_timeval *target_tv,
+static inline void host_to_target_timeval(target_ulong target_addr,
const struct timeval *tv)
{
+ struct target_timeval *target_tv;
+
+ lock_user_struct(target_tv, target_addr, 0);
target_tv->tv_sec = tswapl(tv->tv_sec);
target_tv->tv_usec = tswapl(tv->tv_usec);
+ unlock_user_struct(target_tv, target_addr, 1);
}
static long do_select(long n,
- target_long *target_rfds, target_long *target_wfds,
- target_long *target_efds, struct target_timeval *target_tv)
+ target_ulong rfd_p, target_ulong wfd_p,
+ target_ulong efd_p, target_ulong target_tv)
{
fd_set rfds, wfds, efds;
fd_set *rfds_ptr, *wfds_ptr, *efds_ptr;
+ target_long *target_rfds, *target_wfds, *target_efds;
struct timeval tv, *tv_ptr;
long ret;
+ int ok;
- rfds_ptr = target_to_host_fds(&rfds, target_rfds, n);
- wfds_ptr = target_to_host_fds(&wfds, target_wfds, n);
- efds_ptr = target_to_host_fds(&efds, target_efds, n);
+ if (rfd_p) {
+ target_rfds = lock_user(rfd_p, sizeof(target_long) * n, 1);
+ rfds_ptr = target_to_host_fds(&rfds, target_rfds, n);
+ } else {
+ target_rfds = NULL;
+ rfds_ptr = NULL;
+ }
+ if (wfd_p) {
+ target_wfds = lock_user(wfd_p, sizeof(target_long) * n, 1);
+ wfds_ptr = target_to_host_fds(&wfds, target_wfds, n);
+ } else {
+ target_wfds = NULL;
+ wfds_ptr = NULL;
+ }
+ if (efd_p) {
+ target_efds = lock_user(efd_p, sizeof(target_long) * n, 1);
+ efds_ptr = target_to_host_fds(&efds, target_efds, n);
+ } else {
+ target_efds = NULL;
+ efds_ptr = NULL;
+ }
if (target_tv) {
target_to_host_timeval(&tv, target_tv);
@@ -412,7 +443,9 @@ static long do_select(long n,
tv_ptr = NULL;
}
ret = get_errno(select(n, rfds_ptr, wfds_ptr, efds_ptr, tv_ptr));
- if (!is_error(ret)) {
+ ok = !is_error(ret);
+
+ if (ok) {
host_to_target_fds(target_rfds, rfds_ptr, n);
host_to_target_fds(target_wfds, wfds_ptr, n);
host_to_target_fds(target_efds, efds_ptr, n);
@@ -421,25 +454,41 @@ static long do_select(long n,
host_to_target_timeval(target_tv, &tv);
}
}
+ if (target_rfds)
+ unlock_user(target_rfds, rfd_p, ok ? sizeof(target_long) * n : 0);
+ if (target_wfds)
+ unlock_user(target_wfds, wfd_p, ok ? sizeof(target_long) * n : 0);
+ if (target_efds)
+ unlock_user(target_efds, efd_p, ok ? sizeof(target_long) * n : 0);
+
return ret;
}
static inline void target_to_host_sockaddr(struct sockaddr *addr,
- struct target_sockaddr *target_addr,
+ target_ulong target_addr,
socklen_t len)
{
- memcpy(addr, target_addr, len);
- addr->sa_family = tswap16(target_addr->sa_family);
+ struct target_sockaddr *target_saddr;
+
+ target_saddr = lock_user(target_addr, len, 1);
+ memcpy(addr, target_saddr, len);
+ addr->sa_family = tswap16(target_saddr->sa_family);
+ unlock_user(target_saddr, target_addr, 0);
}
-static inline void host_to_target_sockaddr(struct target_sockaddr *target_addr,
+static inline void host_to_target_sockaddr(target_ulong target_addr,
struct sockaddr *addr,
socklen_t len)
{
- memcpy(target_addr, addr, len);
- target_addr->sa_family = tswap16(addr->sa_family);
+ struct target_sockaddr *target_saddr;
+
+ target_saddr = lock_user(target_addr, len, 0);
+ memcpy(target_saddr, addr, len);
+ target_saddr->sa_family = tswap16(addr->sa_family);
+ unlock_user(target_saddr, target_addr, len);
}
+/* ??? Should this also swap msgh->name? */
static inline void target_to_host_cmsg(struct msghdr *msgh,
struct target_msghdr *target_msgh)
{
@@ -484,6 +533,7 @@ static inline void target_to_host_cmsg(struct msghdr *msgh,
msgh->msg_controllen = space;
}
+/* ??? Should this also swap msgh->name? */
static inline void host_to_target_cmsg(struct target_msghdr *target_msgh,
struct msghdr *msgh)
{
@@ -528,7 +578,7 @@ static inline void host_to_target_cmsg(struct target_msghdr *target_msgh,
}
static long do_setsockopt(int sockfd, int level, int optname,
- void *optval, socklen_t optlen)
+ target_ulong optval, socklen_t optlen)
{
int val, ret;
@@ -538,8 +588,7 @@ static long do_setsockopt(int sockfd, int level, int optname,
if (optlen < sizeof(uint32_t))
return -EINVAL;
- if (get_user(val, (uint32_t *)optval))
- return -EFAULT;
+ val = tget32(optval);
ret = get_errno(setsockopt(sockfd, level, optname, &val, sizeof(val)));
break;
case SOL_IP:
@@ -561,11 +610,9 @@ static long do_setsockopt(int sockfd, int level, int optname,
case IP_MULTICAST_LOOP:
val = 0;
if (optlen >= sizeof(uint32_t)) {
- if (get_user(val, (uint32_t *)optval))
- return -EFAULT;
+ val = tget32(optval);
} else if (optlen >= 1) {
- if (get_user(val, (uint8_t *)optval))
- return -EFAULT;
+ val = tget8(optval);
}
ret = get_errno(setsockopt(sockfd, level, optname, &val, sizeof(val)));
break;
@@ -598,8 +645,8 @@ static long do_setsockopt(int sockfd, int level, int optname,
case SO_SNDTIMEO:
if (optlen < sizeof(uint32_t))
return -EINVAL;
- if (get_user(val, (uint32_t *)optval))
- return -EFAULT;
+
+ val = tget32(optval);
ret = get_errno(setsockopt(sockfd, level, optname, &val, sizeof(val)));
break;
default:
@@ -615,7 +662,7 @@ static long do_setsockopt(int sockfd, int level, int optname,
}
static long do_getsockopt(int sockfd, int level, int optname,
- void *optval, socklen_t *optlen)
+ target_ulong optval, target_ulong optlen)
{
int len, lv, val, ret;
@@ -636,8 +683,7 @@ static long do_getsockopt(int sockfd, int level, int optname,
case SOL_TCP:
/* TCP options all take an 'int' value. */
int_case:
- if (get_user(len, optlen))
- return -EFAULT;
+ len = tget32(optlen);
if (len < 0)
return -EINVAL;
lv = sizeof(int);
@@ -647,10 +693,11 @@ static long do_getsockopt(int sockfd, int level, int optname,
val = tswap32(val);
if (len > lv)
len = lv;
- if (copy_to_user(optval, &val, len))
- return -EFAULT;
- if (put_user(len, optlen))
- return -EFAULT;
+ if (len == 4)
+ tput32(optval, val);
+ else
+ tput8(optval, val);
+ tput32(optlen, len);
break;
case SOL_IP:
switch(optname) {
@@ -669,8 +716,7 @@ static long do_getsockopt(int sockfd, int level, int optname,
#endif
case IP_MULTICAST_TTL:
case IP_MULTICAST_LOOP:
- if (get_user(len, optlen))
- return -EFAULT;
+ len = tget32(optlen);
if (len < 0)
return -EINVAL;
lv = sizeof(int);
@@ -678,20 +724,14 @@ static long do_getsockopt(int sockfd, int level, int optname,
if (ret < 0)
return ret;
if (len < sizeof(int) && len > 0 && val >= 0 && val < 255) {
- unsigned char ucval = val;
len = 1;
- if (put_user(len, optlen))
- return -EFAULT;
- if (copy_to_user(optval,&ucval,1))
- return -EFAULT;
+ tput32(optlen, len);
+ tput8(optval, val);
} else {
- val = tswap32(val);
if (len > sizeof(int))
len = sizeof(int);
- if (put_user(len, optlen))
- return -EFAULT;
- if (copy_to_user(optval, &val, len))
- return -EFAULT;
+ tput32(optlen, len);
+ tput32(optval, val);
}
break;
default:
@@ -708,25 +748,57 @@ static long do_getsockopt(int sockfd, int level, int optname,
return ret;
}
-static long do_socketcall(int num, int32_t *vptr)
+static void lock_iovec(struct iovec *vec, target_ulong target_addr,
+ int count, int copy)
+{
+ struct target_iovec *target_vec;
+ target_ulong base;
+ int i;
+
+ target_vec = lock_user(target_addr, count * sizeof(struct target_iovec), 1);
+ for(i = 0;i < count; i++) {
+ base = tswapl(target_vec[i].iov_base);
+ vec[i].iov_len = tswapl(target_vec[i].iov_len);
+ vec[i].iov_base = lock_user(base, vec[i].iov_len, copy);
+ }
+ unlock_user (target_vec, target_addr, 0);
+}
+
+static void unlock_iovec(struct iovec *vec, target_ulong target_addr,
+ int count, int copy)
+{
+ struct target_iovec *target_vec;
+ target_ulong base;
+ int i;
+
+ target_vec = lock_user(target_addr, count * sizeof(struct target_iovec), 1);
+ for(i = 0;i < count; i++) {
+ base = tswapl(target_vec[i].iov_base);
+ unlock_user(vec[i].iov_base, base, copy ? vec[i].iov_len : 0);
+ }
+ unlock_user (target_vec, target_addr, 0);
+}
+
+static long do_socketcall(int num, target_ulong vptr)
{
long ret;
+ const int n = sizeof(target_ulong);
switch(num) {
case SOCKOP_socket:
{
- int domain = tswap32(vptr[0]);
- int type = tswap32(vptr[1]);
- int protocol = tswap32(vptr[2]);
+ int domain = tgetl(vptr);
+ int type = tgetl(vptr + n);
+ int protocol = tgetl(vptr + 2 * n);
ret = get_errno(socket(domain, type, protocol));
}
break;
case SOCKOP_bind:
{
- int sockfd = tswap32(vptr[0]);
- void *target_addr = (void *)tswap32(vptr[1]);
- socklen_t addrlen = tswap32(vptr[2]);
+ int sockfd = tgetl(vptr);
+ target_ulong target_addr = tgetl(vptr + n);
+ socklen_t addrlen = tgetl(vptr + 2 * n);
void *addr = alloca(addrlen);
target_to_host_sockaddr(addr, target_addr, addrlen);
@@ -735,9 +807,9 @@ static long do_socketcall(int num, int32_t *vptr)
break;
case SOCKOP_connect:
{
- int sockfd = tswap32(vptr[0]);
- void *target_addr = (void *)tswap32(vptr[1]);
- socklen_t addrlen = tswap32(vptr[2]);
+ int sockfd = tgetl(vptr);
+ target_ulong target_addr = tgetl(vptr + n);
+ socklen_t addrlen = tgetl(vptr + 2 * n);
void *addr = alloca(addrlen);
target_to_host_sockaddr(addr, target_addr, addrlen);
@@ -746,128 +818,142 @@ static long do_socketcall(int num, int32_t *vptr)
break;
case SOCKOP_listen:
{
- int sockfd = tswap32(vptr[0]);
- int backlog = tswap32(vptr[1]);
+ int sockfd = tgetl(vptr);
+ int backlog = tgetl(vptr + n);
ret = get_errno(listen(sockfd, backlog));
}
break;
case SOCKOP_accept:
{
- int sockfd = tswap32(vptr[0]);
- void *target_addr = (void *)tswap32(vptr[1]);
- uint32_t *target_addrlen = (void *)tswap32(vptr[2]);
- socklen_t addrlen = tswap32(*target_addrlen);
+ int sockfd = tgetl(vptr);
+ target_ulong target_addr = tgetl(vptr + n);
+ target_ulong target_addrlen = tgetl(vptr + 2 * n);
+ socklen_t addrlen = tget32(target_addrlen);
void *addr = alloca(addrlen);
ret = get_errno(accept(sockfd, addr, &addrlen));
if (!is_error(ret)) {
host_to_target_sockaddr(target_addr, addr, addrlen);
- *target_addrlen = tswap32(addrlen);
+ tput32(target_addrlen, addrlen);
}
}
break;
case SOCKOP_getsockname:
{
- int sockfd = tswap32(vptr[0]);
- void *target_addr = (void *)tswap32(vptr[1]);
- uint32_t *target_addrlen = (void *)tswap32(vptr[2]);
- socklen_t addrlen = tswap32(*target_addrlen);
+ int sockfd = tgetl(vptr);
+ target_ulong target_addr = tgetl(vptr + n);
+ target_ulong target_addrlen = tgetl(vptr + 2 * n);
+ socklen_t addrlen = tget32(target_addrlen);
void *addr = alloca(addrlen);
ret = get_errno(getsockname(sockfd, addr, &addrlen));
if (!is_error(ret)) {
host_to_target_sockaddr(target_addr, addr, addrlen);
- *target_addrlen = tswap32(addrlen);
+ tput32(target_addrlen, addrlen);
}
}
break;
case SOCKOP_getpeername:
{
- int sockfd = tswap32(vptr[0]);
- void *target_addr = (void *)tswap32(vptr[1]);
- uint32_t *target_addrlen = (void *)tswap32(vptr[2]);
- socklen_t addrlen = tswap32(*target_addrlen);
+ int sockfd = tgetl(vptr);
+ target_ulong target_addr = tgetl(vptr + n);
+ target_ulong target_addrlen = tgetl(vptr + 2 * n);
+ socklen_t addrlen = tget32(target_addrlen);
void *addr = alloca(addrlen);
ret = get_errno(getpeername(sockfd, addr, &addrlen));
if (!is_error(ret)) {
host_to_target_sockaddr(target_addr, addr, addrlen);
- *target_addrlen = tswap32(addrlen);
+ tput32(target_addrlen, addrlen);
}
}
break;
case SOCKOP_socketpair:
{
- int domain = tswap32(vptr[0]);
- int type = tswap32(vptr[1]);
- int protocol = tswap32(vptr[2]);
- int32_t *target_tab = (void *)tswap32(vptr[3]);
+ int domain = tgetl(vptr);
+ int type = tgetl(vptr + n);
+ int protocol = tgetl(vptr + 2 * n);
+ target_ulong target_tab = tgetl(vptr + 3 * n);
int tab[2];
ret = get_errno(socketpair(domain, type, protocol, tab));
if (!is_error(ret)) {
- target_tab[0] = tswap32(tab[0]);
- target_tab[1] = tswap32(tab[1]);
+ tput32(target_tab, tab[0]);
+ tput32(target_tab + 4, tab[1]);
}
}
break;
case SOCKOP_send:
{
- int sockfd = tswap32(vptr[0]);
- void *msg = (void *)tswap32(vptr[1]);
- size_t len = tswap32(vptr[2]);
- int flags = tswap32(vptr[3]);
+ int sockfd = tgetl(vptr);
+ target_ulong msg = tgetl(vptr + n);
+ size_t len = tgetl(vptr + 2 * n);
+ int flags = tgetl(vptr + 3 * n);
+ void *host_msg;
- ret = get_errno(send(sockfd, msg, len, flags));
+ host_msg = lock_user(msg, len, 1);
+ ret = get_errno(send(sockfd, host_msg, len, flags));
+ unlock_user(host_msg, msg, 0);
}
break;
case SOCKOP_recv:
{
- int sockfd = tswap32(vptr[0]);
- void *msg = (void *)tswap32(vptr[1]);
- size_t len = tswap32(vptr[2]);
- int flags = tswap32(vptr[3]);
+ int sockfd = tgetl(vptr);
+ target_ulong msg = tgetl(vptr + n);
+ size_t len = tgetl(vptr + 2 * n);
+ int flags = tgetl(vptr + 3 * n);
+ void *host_msg;
- ret = get_errno(recv(sockfd, msg, len, flags));
+ host_msg = lock_user(msg, len, 0);
+ ret = get_errno(recv(sockfd, host_msg, len, flags));
+ unlock_user(host_msg, msg, ret);
}
break;
case SOCKOP_sendto:
{
- int sockfd = tswap32(vptr[0]);
- void *msg = (void *)tswap32(vptr[1]);
- size_t len = tswap32(vptr[2]);
- int flags = tswap32(vptr[3]);
- void *target_addr = (void *)tswap32(vptr[4]);
- socklen_t addrlen = tswap32(vptr[5]);
+ int sockfd = tgetl(vptr);
+ target_ulong msg = tgetl(vptr + n);
+ size_t len = tgetl(vptr + 2 * n);
+ int flags = tgetl(vptr + 3 * n);
+ target_ulong target_addr = tgetl(vptr + 4 * n);
+ socklen_t addrlen = tgetl(vptr + 5 * n);
void *addr = alloca(addrlen);
+ void *host_msg;
+ host_msg = lock_user(msg, len, 1);
target_to_host_sockaddr(addr, target_addr, addrlen);
- ret = get_errno(sendto(sockfd, msg, len, flags, addr, addrlen));
+ ret = get_errno(sendto(sockfd, host_msg, len, flags, addr, addrlen));
+ unlock_user(host_msg, msg, 0);
}
break;
case SOCKOP_recvfrom:
{
- int sockfd = tswap32(vptr[0]);
- void *msg = (void *)tswap32(vptr[1]);
- size_t len = tswap32(vptr[2]);
- int flags = tswap32(vptr[3]);
- void *target_addr = (void *)tswap32(vptr[4]);
- uint32_t *target_addrlen = (void *)tswap32(vptr[5]);
- socklen_t addrlen = tswap32(*target_addrlen);
+ int sockfd = tgetl(vptr);
+ target_ulong msg = tgetl(vptr + n);
+ size_t len = tgetl(vptr + 2 * n);
+ int flags = tgetl(vptr + 3 * n);
+ target_ulong target_addr = tgetl(vptr + 4 * n);
+ target_ulong target_addrlen = tgetl(vptr + 5 * n);
+ socklen_t addrlen = tget32(target_addrlen);
void *addr = alloca(addrlen);
+ void *host_msg;
- ret = get_errno(recvfrom(sockfd, msg, len, flags, addr, &addrlen));
+ host_msg = lock_user(msg, len, 0);
+ ret = get_errno(recvfrom(sockfd, host_msg, len, flags, addr, &addrlen));
if (!is_error(ret)) {
host_to_target_sockaddr(target_addr, addr, addrlen);
- *target_addrlen = tswap32(addrlen);
+ tput32(target_addrlen, addrlen);
+ unlock_user(host_msg, msg, len);
+ } else {
+ unlock_user(host_msg, msg, 0);
}
}
break;
case SOCKOP_shutdown:
{
- int sockfd = tswap32(vptr[0]);
- int how = tswap32(vptr[1]);
+ int sockfd = tgetl(vptr);
+ int how = tgetl(vptr + n);
ret = get_errno(shutdown(sockfd, how));
}
@@ -876,32 +962,39 @@ static long do_socketcall(int num, int32_t *vptr)
case SOCKOP_recvmsg:
{
int fd;
+ target_ulong target_msg;
struct target_msghdr *msgp;
struct msghdr msg;
- int flags, count, i;
+ int flags, count;
struct iovec *vec;
- struct target_iovec *target_vec;
-
- msgp = (void *)tswap32(vptr[1]);
- msg.msg_name = (void *)tswapl(msgp->msg_name);
- msg.msg_namelen = tswapl(msgp->msg_namelen);
+ target_ulong target_vec;
+ int send = (num == SOCKOP_sendmsg);
+
+ target_msg = tgetl(vptr + n);
+ lock_user_struct(msgp, target_msg, 1);
+ if (msgp->msg_name) {
+ msg.msg_namelen = tswap32(msgp->msg_namelen);
+ msg.msg_name = alloca(msg.msg_namelen);
+ target_to_host_sockaddr(msg.msg_name, tswapl(msgp->msg_name),
+ msg.msg_namelen);
+ } else {
+ msg.msg_name = NULL;
+ msg.msg_namelen = 0;
+ }
msg.msg_controllen = 2 * tswapl(msgp->msg_controllen);
msg.msg_control = alloca(msg.msg_controllen);
msg.msg_flags = tswap32(msgp->msg_flags);
count = tswapl(msgp->msg_iovlen);
vec = alloca(count * sizeof(struct iovec));
- target_vec = (void *)tswapl(msgp->msg_iov);
- for(i = 0;i < count; i++) {
- vec[i].iov_base = (void *)tswapl(target_vec[i].iov_base);
- vec[i].iov_len = tswapl(target_vec[i].iov_len);
- }
+ target_vec = tswapl(msgp->msg_iov);
+ lock_iovec(vec, target_vec, count, send);
msg.msg_iovlen = count;
msg.msg_iov = vec;
- fd = tswap32(vptr[0]);
- flags = tswap32(vptr[2]);
- if (num == SOCKOP_sendmsg) {
+ fd = tgetl(vptr);
+ flags = tgetl(vptr + 2 * n);
+ if (send) {
target_to_host_cmsg(&msg, msgp);
ret = get_errno(sendmsg(fd, &msg, flags));
} else {
@@ -909,26 +1002,27 @@ static long do_socketcall(int num, int32_t *vptr)
if (!is_error(ret))
host_to_target_cmsg(msgp, &msg);
}
+ unlock_iovec(vec, target_vec, count, !send);
}
break;
case SOCKOP_setsockopt:
{
- int sockfd = tswap32(vptr[0]);
- int level = tswap32(vptr[1]);
- int optname = tswap32(vptr[2]);
- void *optval = (void *)tswap32(vptr[3]);
- socklen_t optlen = tswap32(vptr[4]);
+ int sockfd = tgetl(vptr);
+ int level = tgetl(vptr + n);
+ int optname = tgetl(vptr + 2 * n);
+ target_ulong optval = tgetl(vptr + 3 * n);
+ socklen_t optlen = tgetl(vptr + 4 * n);
ret = do_setsockopt(sockfd, level, optname, optval, optlen);
}
break;
case SOCKOP_getsockopt:
{
- int sockfd = tswap32(vptr[0]);
- int level = tswap32(vptr[1]);
- int optname = tswap32(vptr[2]);
- void *optval = (void *)tswap32(vptr[3]);
- uint32_t *poptlen = (void *)tswap32(vptr[4]);
+ int sockfd = tgetl(vptr);
+ int level = tgetl(vptr + n);
+ int optname = tgetl(vptr + 2 * n);
+ target_ulong optval = tgetl(vptr + 3 * n);
+ target_ulong poptlen = tgetl(vptr + 4 * n);
ret = do_getsockopt(sockfd, level, optname, optval, poptlen);
}
@@ -949,6 +1043,7 @@ static struct shm_region {
uint32_t size;
} shm_regions[N_SHM_REGIONS];
+/* ??? This only works with linear mappings. */
static long do_ipc(long call, long first, long second, long third,
long ptr, long fifth)
{
@@ -1065,12 +1160,15 @@ IOCTLEntry ioctl_entries[] = {
{ 0, 0, },
};
+/* ??? Implement proper locking for ioctls. */
static long do_ioctl(long fd, long cmd, long arg)
{
const IOCTLEntry *ie;
const argtype *arg_type;
long ret;
uint8_t buf_temp[MAX_STRUCT_SIZE];
+ int target_size;
+ void *argptr;
ie = ioctl_entries;
for(;;) {
@@ -1098,23 +1196,32 @@ static long do_ioctl(long fd, long cmd, long arg)
break;
case TYPE_PTR:
arg_type++;
+ target_size = thunk_type_size(arg_type, 0);
switch(ie->access) {
case IOC_R:
ret = get_errno(ioctl(fd, ie->host_cmd, buf_temp));
if (!is_error(ret)) {
- thunk_convert((void *)arg, buf_temp, arg_type, THUNK_TARGET);
+ argptr = lock_user(arg, target_size, 0);
+ thunk_convert(argptr, buf_temp, arg_type, THUNK_TARGET);
+ unlock_user(argptr, arg, target_size);
}
break;
case IOC_W:
- thunk_convert(buf_temp, (void *)arg, arg_type, THUNK_HOST);
+ argptr = lock_user(arg, target_size, 1);
+ thunk_convert(buf_temp, argptr, arg_type, THUNK_HOST);
+ unlock_user(argptr, arg, 0);
ret = get_errno(ioctl(fd, ie->host_cmd, buf_temp));
break;
default:
case IOC_RW:
- thunk_convert(buf_temp, (void *)arg, arg_type, THUNK_HOST);
+ argptr = lock_user(arg, target_size, 1);
+ thunk_convert(buf_temp, argptr, arg_type, THUNK_HOST);
+ unlock_user(argptr, arg, 0);
ret = get_errno(ioctl(fd, ie->host_cmd, buf_temp));
if (!is_error(ret)) {
- thunk_convert((void *)arg, buf_temp, arg_type, THUNK_TARGET);
+ argptr = lock_user(arg, target_size, 0);
+ thunk_convert(argptr, buf_temp, arg_type, THUNK_TARGET);
+ unlock_user(argptr, arg, target_size);
}
break;
}
@@ -1338,35 +1445,41 @@ static bitmask_transtbl fcntl_flags_tbl[] = {
/* NOTE: there is really one LDT for all the threads */
uint8_t *ldt_table;
-static int read_ldt(void *ptr, unsigned long bytecount)
+static int read_ldt(target_ulong ptr, unsigned long bytecount)
{
int size;
+ void *p;
if (!ldt_table)
return 0;
size = TARGET_LDT_ENTRIES * TARGET_LDT_ENTRY_SIZE;
if (size > bytecount)
size = bytecount;
- memcpy(ptr, ldt_table, size);
+ p = lock_user(ptr, size, 0);
+ /* ??? Shoudl this by byteswapped? */
+ memcpy(p, ldt_table, size);
+ unlock_user(p, ptr, size);
return size;
}
/* XXX: add locking support */
static int write_ldt(CPUX86State *env,
- void *ptr, unsigned long bytecount, int oldmode)
+ target_ulong ptr, unsigned long bytecount, int oldmode)
{
struct target_modify_ldt_ldt_s ldt_info;
+ struct target_modify_ldt_ldt_s *target_ldt_info;
int seg_32bit, contents, read_exec_only, limit_in_pages;
int seg_not_present, useable;
uint32_t *lp, entry_1, entry_2;
if (bytecount != sizeof(ldt_info))
return -EINVAL;
- memcpy(&ldt_info, ptr, sizeof(ldt_info));
- tswap32s(&ldt_info.entry_number);
- tswapls((long *)&ldt_info.base_addr);
- tswap32s(&ldt_info.limit);
- tswap32s(&ldt_info.flags);
+ lock_user_struct(target_ldt_info, ptr, 1);
+ ldt_info.entry_number = tswap32(target_ldt_info->entry_number);
+ ldt_info.base_addr = tswapl(target_ldt_info->base_addr);
+ ldt_info.limit = tswap32(target_ldt_info->limit);
+ ldt_info.flags = tswap32(target_ldt_info->flags);
+ unlock_user_struct(target_ldt_info, ptr, 0);
if (ldt_info.entry_number >= TARGET_LDT_ENTRIES)
return -EINVAL;
@@ -1389,7 +1502,7 @@ static int write_ldt(CPUX86State *env,
if (!ldt_table)
return -ENOMEM;
memset(ldt_table, 0, TARGET_LDT_ENTRIES * TARGET_LDT_ENTRY_SIZE);
- env->ldt.base = (long)ldt_table;
+ env->ldt.base = h2g(ldt_table);
env->ldt.limit = 0xffff;
}
@@ -1432,7 +1545,7 @@ install:
}
/* specific and weird i386 syscalls */
-int do_modify_ldt(CPUX86State *env, int func, void *ptr, unsigned long bytecount)
+int do_modify_ldt(CPUX86State *env, int func, target_ulong ptr, unsigned long bytecount)
{
int ret = -ENOSYS;
@@ -1523,31 +1636,35 @@ int do_fork(CPUState *env, unsigned int flags, unsigned long newsp)
return ret;
}
-static long do_fcntl(int fd, int cmd, unsigned long arg)
+static long do_fcntl(int fd, int cmd, target_ulong arg)
{
struct flock fl;
- struct target_flock *target_fl = (void *)arg;
+ struct target_flock *target_fl;
long ret;
-
+
switch(cmd) {
case TARGET_F_GETLK:
ret = fcntl(fd, cmd, &fl);
if (ret == 0) {
+ lock_user_struct(target_fl, arg, 0);
target_fl->l_type = tswap16(fl.l_type);
target_fl->l_whence = tswap16(fl.l_whence);
target_fl->l_start = tswapl(fl.l_start);
target_fl->l_len = tswapl(fl.l_len);
target_fl->l_pid = tswapl(fl.l_pid);
+ unlock_user_struct(target_fl, arg, 1);
}
break;
case TARGET_F_SETLK:
case TARGET_F_SETLKW:
+ lock_user_struct(target_fl, arg, 1);
fl.l_type = tswap16(target_fl->l_type);
fl.l_whence = tswap16(target_fl->l_whence);
fl.l_start = tswapl(target_fl->l_start);
fl.l_len = tswapl(target_fl->l_len);
fl.l_pid = tswapl(target_fl->l_pid);
+ unlock_user_struct(target_fl, arg, 0);
ret = fcntl(fd, cmd, &fl);
break;
@@ -1690,12 +1807,35 @@ static inline long target_ftruncate64(void *cpu_env, long arg1, long arg2,
}
#endif
+static inline void target_to_host_timespec(struct timespec *host_ts,
+ target_ulong target_addr)
+{
+ struct target_timespec *target_ts;
+
+ lock_user_struct(target_ts, target_addr, 1);
+ host_ts->tv_sec = tswapl(target_ts->tv_sec);
+ host_ts->tv_nsec = tswapl(target_ts->tv_nsec);
+ unlock_user_struct(target_ts, target_addr, 0);
+}
+
+static inline void host_to_target_timespec(target_ulong target_addr,
+ struct timespec *host_ts)
+{
+ struct target_timespec *target_ts;
+
+ lock_user_struct(target_ts, target_addr, 0);
+ target_ts->tv_sec = tswapl(host_ts->tv_sec);
+ target_ts->tv_nsec = tswapl(host_ts->tv_nsec);
+ unlock_user_struct(target_ts, target_addr, 1);
+}
+
long do_syscall(void *cpu_env, int num, long arg1, long arg2, long arg3,
long arg4, long arg5, long arg6)
{
long ret;
struct stat st;
struct statfs stfs;
+ void *p;
#ifdef DEBUG
gemu_log("syscall %d", num);
@@ -1711,89 +1851,140 @@ long do_syscall(void *cpu_env, int num, long arg1, long arg2, long arg3,
ret = 0; /* avoid warning */
break;
case TARGET_NR_read:
- page_unprotect_range((void *)arg2, arg3);
- ret = get_errno(read(arg1, (void *)arg2, arg3));
+ page_unprotect_range(arg2, arg3);
+ p = lock_user(arg2, arg3, 0);
+ ret = get_errno(read(arg1, p, arg3));
+ unlock_user(p, arg2, ret);
break;
case TARGET_NR_write:
- ret = get_errno(write(arg1, (void *)arg2, arg3));
+ p = lock_user(arg2, arg3, 1);
+ ret = get_errno(write(arg1, p, arg3));
+ unlock_user(p, arg2, 0);
break;
case TARGET_NR_open:
- ret = get_errno(open(path((const char *)arg1),
+ p = lock_user_string(arg1);
+ ret = get_errno(open(path(p),
target_to_host_bitmask(arg2, fcntl_flags_tbl),
arg3));
+ unlock_user(p, arg1, 0);
break;
case TARGET_NR_close:
ret = get_errno(close(arg1));
break;
case TARGET_NR_brk:
- ret = do_brk((char *)arg1);
+ ret = do_brk(arg1);
break;
case TARGET_NR_fork:
ret = get_errno(do_fork(cpu_env, SIGCHLD, 0));
break;
case TARGET_NR_waitpid:
{
- int *status = (int *)arg2;
- ret = get_errno(waitpid(arg1, status, arg3));
- if (!is_error(ret) && status)
- tswapls((long *)&status);
+ int status;
+ ret = get_errno(waitpid(arg1, &status, arg3));
+ if (!is_error(ret) && arg2)
+ tput32(arg2, status);
}
break;
case TARGET_NR_creat:
- ret = get_errno(creat((const char *)arg1, arg2));
+ p = lock_user_string(arg1);
+ ret = get_errno(creat(p, arg2));
+ unlock_user(p, arg1, 0);
break;
case TARGET_NR_link:
- ret = get_errno(link((const char *)arg1, (const char *)arg2));
+ {
+ void * p2;
+ p = lock_user_string(arg1);
+ p2 = lock_user_string(arg2);
+ ret = get_errno(link(p, p2));
+ unlock_user(p2, arg2, 0);
+ unlock_user(p, arg1, 0);
+ }
break;
case TARGET_NR_unlink:
- ret = get_errno(unlink((const char *)arg1));
+ p = lock_user_string(arg1);
+ ret = get_errno(unlink(p));
+ unlock_user(p, arg1, 0);
break;
case TARGET_NR_execve:
{
char **argp, **envp;
int argc, envc;
- uint32_t *p;
+ target_ulong gp;
+ target_ulong guest_argp;
+ target_ulong guest_envp;
+ target_ulong addr;
char **q;
argc = 0;
- for (p = (void *)arg2; *p; p++)
+ guest_argp = arg2;
+ for (gp = guest_argp; tgetl(gp); gp++)
argc++;
envc = 0;
- for (p = (void *)arg3; *p; p++)
+ guest_envp = arg3;
+ for (gp = guest_envp; tgetl(gp); gp++)
envc++;
argp = alloca((argc + 1) * sizeof(void *));
envp = alloca((envc + 1) * sizeof(void *));
- for (p = (void *)arg2, q = argp; *p; p++, q++)
- *q = (void *)tswap32(*p);
+ for (gp = guest_argp, q = argp; ;
+ gp += sizeof(target_ulong), q++) {
+ addr = tgetl(gp);
+ if (!addr)
+ break;
+ *q = lock_user_string(addr);
+ }
*q = NULL;
- for (p = (void *)arg3, q = envp; *p; p++, q++)
- *q = (void *)tswap32(*p);
+ for (gp = guest_envp, q = envp; ;
+ gp += sizeof(target_ulong), q++) {
+ addr = tgetl(gp);
+ if (!addr)
+ break;
+ *q = lock_user_string(addr);
+ }
*q = NULL;
- ret = get_errno(execve((const char *)arg1, argp, envp));
+ p = lock_user_string(arg1);
+ ret = get_errno(execve(p, argp, envp));
+ unlock_user(p, arg1, 0);
+
+ for (gp = guest_argp, q = argp; *q;
+ gp += sizeof(target_ulong), q++) {
+ addr = tgetl(gp);
+ unlock_user(*q, addr, 0);
+ }
+ for (gp = guest_envp, q = envp; *q;
+ gp += sizeof(target_ulong), q++) {
+ addr = tgetl(gp);
+ unlock_user(*q, addr, 0);
+ }
}
break;
case TARGET_NR_chdir:
- ret = get_errno(chdir((const char *)arg1));
+ p = lock_user_string(arg1);
+ ret = get_errno(chdir(p));
+ unlock_user(p, arg1, 0);
break;
#ifdef TARGET_NR_time
case TARGET_NR_time:
{
- int *time_ptr = (int *)arg1;
- ret = get_errno(time((time_t *)time_ptr));
- if (!is_error(ret) && time_ptr)
- tswap32s(time_ptr);
+ time_t host_time;
+ ret = get_errno(time(&host_time));
+ if (!is_error(ret) && arg1)
+ tputl(arg1, host_time);
}
break;
#endif
case TARGET_NR_mknod:
- ret = get_errno(mknod((const char *)arg1, arg2, arg3));
+ p = lock_user_string(arg1);
+ ret = get_errno(mknod(p, arg2, arg3));
+ unlock_user(p, arg1, 0);
break;
case TARGET_NR_chmod:
- ret = get_errno(chmod((const char *)arg1, arg2));
+ p = lock_user_string(arg1);
+ ret = get_errno(chmod(p, arg2));
+ unlock_user(p, arg1, 0);
break;
#ifdef TARGET_NR_break
case TARGET_NR_break:
@@ -1813,14 +2004,15 @@ long do_syscall(void *cpu_env, int num, long arg1, long arg2, long arg3,
/* need to look at the data field */
goto unimplemented;
case TARGET_NR_umount:
- ret = get_errno(umount((const char *)arg1));
+ p = lock_user_string(arg1);
+ ret = get_errno(umount(p));
+ unlock_user(p, arg1, 0);
break;
case TARGET_NR_stime:
{
- int *time_ptr = (int *)arg1;
- if (time_ptr)
- tswap32s(time_ptr);
- ret = get_errno(stime((time_t *)time_ptr));
+ time_t host_time;
+ host_time = tgetl(arg1);
+ ret = get_errno(stime(&host_time));
}
break;
case TARGET_NR_ptrace:
@@ -1837,30 +2029,36 @@ long do_syscall(void *cpu_env, int num, long arg1, long arg2, long arg3,
break;
case TARGET_NR_utime:
{
- struct utimbuf tbuf, *tbuf1;
- struct target_utimbuf *target_tbuf = (void *)arg2;
- if (target_tbuf) {
- get_user(tbuf.actime, &target_tbuf->actime);
- get_user(tbuf.modtime, &target_tbuf->modtime);
- tbuf1 = &tbuf;
+ struct utimbuf tbuf, *host_tbuf;
+ struct target_utimbuf *target_tbuf;
+ if (arg2) {
+ lock_user_struct(target_tbuf, arg2, 1);
+ tbuf.actime = tswapl(target_tbuf->actime);
+ tbuf.modtime = tswapl(target_tbuf->modtime);
+ unlock_user_struct(target_tbuf, arg2, 0);
+ host_tbuf = &tbuf;
} else {
- tbuf1 = NULL;
+ host_tbuf = NULL;
}
- ret = get_errno(utime((const char *)arg1, tbuf1));
+ p = lock_user_string(arg1);
+ ret = get_errno(utime(p, host_tbuf));
+ unlock_user(p, arg1, 0);
}
break;
case TARGET_NR_utimes:
{
- struct target_timeval *target_tvp = (struct target_timeval *)arg2;
struct timeval *tvp, tv[2];
- if (target_tvp) {
- target_to_host_timeval(&tv[0], &target_tvp[0]);
- target_to_host_timeval(&tv[1], &target_tvp[1]);
+ if (arg2) {
+ target_to_host_timeval(&tv[0], arg2);
+ target_to_host_timeval(&tv[1],
+ arg2 + sizeof (struct target_timeval));
tvp = tv;
} else {
tvp = NULL;
}
- ret = get_errno(utimes((const char *)arg1, tvp));
+ p = lock_user_string(arg1);
+ ret = get_errno(utimes(p, tvp));
+ unlock_user(p, arg1, 0);
}
break;
#ifdef TARGET_NR_stty
@@ -1872,7 +2070,9 @@ long do_syscall(void *cpu_env, int num, long arg1, long arg2, long arg3,
goto unimplemented;
#endif
case TARGET_NR_access:
- ret = get_errno(access((const char *)arg1, arg2));
+ p = lock_user_string(arg1);
+ ret = get_errno(access(p, arg2));
+ unlock_user(p, arg1, 0);
break;
case TARGET_NR_nice:
ret = get_errno(nice(arg1));
@@ -1889,33 +2089,45 @@ long do_syscall(void *cpu_env, int num, long arg1, long arg2, long arg3,
ret = get_errno(kill(arg1, arg2));
break;
case TARGET_NR_rename:
- ret = get_errno(rename((const char *)arg1, (const char *)arg2));
+ {
+ void *p2;
+ p = lock_user_string(arg1);
+ p2 = lock_user_string(arg2);
+ ret = get_errno(rename(p, p2));
+ unlock_user(p2, arg2, 0);
+ unlock_user(p, arg1, 0);
+ }
break;
case TARGET_NR_mkdir:
- ret = get_errno(mkdir((const char *)arg1, arg2));
+ p = lock_user_string(arg1);
+ ret = get_errno(mkdir(p, arg2));
+ unlock_user(p, arg1, 0);
break;
case TARGET_NR_rmdir:
- ret = get_errno(rmdir((const char *)arg1));
+ p = lock_user_string(arg1);
+ ret = get_errno(rmdir(p));
+ unlock_user(p, arg1, 0);
break;
case TARGET_NR_dup:
ret = get_errno(dup(arg1));
break;
case TARGET_NR_pipe:
{
- int *pipe_ptr = (int *)arg1;
- ret = get_errno(pipe(pipe_ptr));
+ int host_pipe[2];
+ ret = get_errno(pipe(host_pipe));
if (!is_error(ret)) {
- tswap32s(&pipe_ptr[0]);
- tswap32s(&pipe_ptr[1]);
+ tput32(arg1, host_pipe[0]);
+ tput32(arg1 + 4, host_pipe[1]);
}
}
break;
case TARGET_NR_times:
{
- struct target_tms *tmsp = (void *)arg1;
+ struct target_tms *tmsp;
struct tms tms;
ret = get_errno(times(&tms));
- if (tmsp) {
+ if (arg1) {
+ tmsp = lock_user(arg1, sizeof(struct target_tms), 0);
tmsp->tms_utime = tswapl(host_to_target_clock_t(tms.tms_utime));
tmsp->tms_stime = tswapl(host_to_target_clock_t(tms.tms_stime));
tmsp->tms_cutime = tswapl(host_to_target_clock_t(tms.tms_cutime));
@@ -1935,7 +2147,9 @@ long do_syscall(void *cpu_env, int num, long arg1, long arg2, long arg3,
case TARGET_NR_acct:
goto unimplemented;
case TARGET_NR_umount2:
- ret = get_errno(umount2((const char *)arg1, arg2));
+ p = lock_user_string(arg1);
+ ret = get_errno(umount2(p, arg2));
+ unlock_user(p, arg1, 0);
break;
#ifdef TARGET_NR_lock
case TARGET_NR_lock:
@@ -1966,7 +2180,9 @@ long do_syscall(void *cpu_env, int num, long arg1, long arg2, long arg3,
ret = get_errno(umask(arg1));
break;
case TARGET_NR_chroot:
- ret = get_errno(chroot((const char *)arg1));
+ p = lock_user_string(arg1);
+ ret = get_errno(chroot(p));
+ unlock_user(p, arg1, 0);
break;
case TARGET_NR_ustat:
goto unimplemented;
@@ -1984,29 +2200,49 @@ long do_syscall(void *cpu_env, int num, long arg1, long arg2, long arg3,
break;
case TARGET_NR_sigaction:
{
- struct target_old_sigaction *old_act = (void *)arg2;
- struct target_old_sigaction *old_oact = (void *)arg3;
+ struct target_old_sigaction *old_act;
struct target_sigaction act, oact, *pact;
- if (old_act) {
+ if (arg2) {
+ lock_user_struct(old_act, arg2, 1);
act._sa_handler = old_act->_sa_handler;
target_siginitset(&act.sa_mask, old_act->sa_mask);
act.sa_flags = old_act->sa_flags;
act.sa_restorer = old_act->sa_restorer;
+ unlock_user_struct(old_act, arg2, 0);
pact = &act;
} else {
pact = NULL;
}
ret = get_errno(do_sigaction(arg1, pact, &oact));
- if (!is_error(ret) && old_oact) {
- old_oact->_sa_handler = oact._sa_handler;
- old_oact->sa_mask = oact.sa_mask.sig[0];
- old_oact->sa_flags = oact.sa_flags;
- old_oact->sa_restorer = oact.sa_restorer;
+ if (!is_error(ret) && arg3) {
+ lock_user_struct(old_act, arg3, 0);
+ old_act->_sa_handler = oact._sa_handler;
+ old_act->sa_mask = oact.sa_mask.sig[0];
+ old_act->sa_flags = oact.sa_flags;
+ old_act->sa_restorer = oact.sa_restorer;
+ unlock_user_struct(old_act, arg3, 1);
}
}
break;
case TARGET_NR_rt_sigaction:
- ret = get_errno(do_sigaction(arg1, (void *)arg2, (void *)arg3));
+ {
+ struct target_sigaction *act;
+ struct target_sigaction *oact;
+
+ if (arg2)
+ lock_user_struct(act, arg2, 1);
+ else
+ act = NULL;
+ if (arg3)
+ lock_user_struct(oact, arg3, 0);
+ else
+ oact = NULL;
+ ret = get_errno(do_sigaction(arg1, act, oact));
+ if (arg2)
+ unlock_user_struct(act, arg2, 0);
+ if (arg3)
+ unlock_user_struct(oact, arg3, 1);
+ }
break;
case TARGET_NR_sgetmask:
{
@@ -2033,9 +2269,8 @@ long do_syscall(void *cpu_env, int num, long arg1, long arg2, long arg3,
{
int how = arg1;
sigset_t set, oldset, *set_ptr;
- target_ulong *pset = (void *)arg2, *poldset = (void *)arg3;
- if (pset) {
+ if (arg2) {
switch(how) {
case TARGET_SIG_BLOCK:
how = SIG_BLOCK;
@@ -2050,15 +2285,19 @@ long do_syscall(void *cpu_env, int num, long arg1, long arg2, long arg3,
ret = -EINVAL;
goto fail;
}
- target_to_host_old_sigset(&set, pset);
+ p = lock_user(arg2, sizeof(target_sigset_t), 1);
+ target_to_host_old_sigset(&set, p);
+ unlock_user(p, arg2, 0);
set_ptr = &set;
} else {
how = 0;
set_ptr = NULL;
}
ret = get_errno(sigprocmask(arg1, set_ptr, &oldset));
- if (!is_error(ret) && poldset) {
- host_to_target_old_sigset(poldset, &oldset);
+ if (!is_error(ret) && arg3) {
+ p = lock_user(arg3, sizeof(target_sigset_t), 0);
+ host_to_target_old_sigset(p, &oldset);
+ unlock_user(p, arg3, sizeof(target_sigset_t));
}
}
break;
@@ -2066,10 +2305,8 @@ long do_syscall(void *cpu_env, int num, long arg1, long arg2, long arg3,
{
int how = arg1;
sigset_t set, oldset, *set_ptr;
- target_sigset_t *pset = (void *)arg2;
- target_sigset_t *poldset = (void *)arg3;
- if (pset) {
+ if (arg2) {
switch(how) {
case TARGET_SIG_BLOCK:
how = SIG_BLOCK;
@@ -2084,15 +2321,19 @@ long do_syscall(void *cpu_env, int num, long arg1, long arg2, long arg3,
ret = -EINVAL;
goto fail;
}
- target_to_host_sigset(&set, pset);
+ p = lock_user(arg2, sizeof(target_sigset_t), 1);
+ target_to_host_sigset(&set, p);
+ unlock_user(p, arg2, 0);
set_ptr = &set;
} else {
how = 0;
set_ptr = NULL;
}
ret = get_errno(sigprocmask(how, set_ptr, &oldset));
- if (!is_error(ret) && poldset) {
- host_to_target_sigset(poldset, &oldset);
+ if (!is_error(ret) && arg3) {
+ p = lock_user(arg3, sizeof(target_sigset_t), 0);
+ host_to_target_sigset(p, &oldset);
+ unlock_user(p, arg3, sizeof(target_sigset_t));
}
}
break;
@@ -2101,7 +2342,9 @@ long do_syscall(void *cpu_env, int num, long arg1, long arg2, long arg3,
sigset_t set;
ret = get_errno(sigpending(&set));
if (!is_error(ret)) {
- host_to_target_old_sigset((target_ulong *)arg1, &set);
+ p = lock_user(arg1, sizeof(target_sigset_t), 0);
+ host_to_target_old_sigset(p, &set);
+ unlock_user(p, arg1, sizeof(target_sigset_t));
}
}
break;
@@ -2110,51 +2353,59 @@ long do_syscall(void *cpu_env, int num, long arg1, long arg2, long arg3,
sigset_t set;
ret = get_errno(sigpending(&set));
if (!is_error(ret)) {
- host_to_target_sigset((target_sigset_t *)arg1, &set);
+ p = lock_user(arg1, sizeof(target_sigset_t), 0);
+ host_to_target_sigset(p, &set);
+ unlock_user(p, arg1, sizeof(target_sigset_t));
}
}
break;
case TARGET_NR_sigsuspend:
{
sigset_t set;
- target_to_host_old_sigset(&set, (target_ulong *)arg1);
+ p = lock_user(arg1, sizeof(target_sigset_t), 1);
+ target_to_host_old_sigset(&set, p);
+ unlock_user(p, arg1, 0);
ret = get_errno(sigsuspend(&set));
}
break;
case TARGET_NR_rt_sigsuspend:
{
sigset_t set;
- target_to_host_sigset(&set, (target_sigset_t *)arg1);
+ p = lock_user(arg1, sizeof(target_sigset_t), 1);
+ target_to_host_sigset(&set, p);
+ unlock_user(p, arg1, 0);
ret = get_errno(sigsuspend(&set));
}
break;
case TARGET_NR_rt_sigtimedwait:
{
- target_sigset_t *target_set = (void *)arg1;
- target_siginfo_t *target_uinfo = (void *)arg2;
- struct target_timespec *target_uts = (void *)arg3;
sigset_t set;
struct timespec uts, *puts;
siginfo_t uinfo;
- target_to_host_sigset(&set, target_set);
- if (target_uts) {
+ p = lock_user(arg1, sizeof(target_sigset_t), 1);
+ target_to_host_sigset(&set, p);
+ unlock_user(p, arg1, 0);
+ if (arg3) {
puts = &uts;
- puts->tv_sec = tswapl(target_uts->tv_sec);
- puts->tv_nsec = tswapl(target_uts->tv_nsec);
+ target_to_host_timespec(puts, arg3);
} else {
puts = NULL;
}
ret = get_errno(sigtimedwait(&set, &uinfo, puts));
- if (!is_error(ret) && target_uinfo) {
- host_to_target_siginfo(target_uinfo, &uinfo);
+ if (!is_error(ret) && arg2) {
+ p = lock_user(arg2, sizeof(target_sigset_t), 0);
+ host_to_target_siginfo(p, &uinfo);
+ unlock_user(p, arg2, sizeof(target_sigset_t));
}
}
break;
case TARGET_NR_rt_sigqueueinfo:
{
siginfo_t uinfo;
- target_to_host_siginfo(&uinfo, (target_siginfo_t *)arg3);
+ p = lock_user(arg3, sizeof(target_sigset_t), 1);
+ target_to_host_siginfo(&uinfo, p);
+ unlock_user(p, arg1, 0);
ret = get_errno(sys_rt_sigqueueinfo(arg1, arg2, &uinfo));
}
break;
@@ -2167,16 +2418,20 @@ long do_syscall(void *cpu_env, int num, long arg1, long arg2, long arg3,
ret = do_rt_sigreturn(cpu_env);
break;
case TARGET_NR_sethostname:
- ret = get_errno(sethostname((const char *)arg1, arg2));
+ p = lock_user_string(arg1);
+ ret = get_errno(sethostname(p, arg2));
+ unlock_user(p, arg1, 0);
break;
case TARGET_NR_setrlimit:
{
/* XXX: convert resource ? */
int resource = arg1;
- struct target_rlimit *target_rlim = (void *)arg2;
+ struct target_rlimit *target_rlim;
struct rlimit rlim;
+ lock_user_struct(target_rlim, arg2, 1);
rlim.rlim_cur = tswapl(target_rlim->rlim_cur);
rlim.rlim_max = tswapl(target_rlim->rlim_max);
+ unlock_user_struct(target_rlim, arg2, 0);
ret = get_errno(setrlimit(resource, &rlim));
}
break;
@@ -2184,72 +2439,91 @@ long do_syscall(void *cpu_env, int num, long arg1, long arg2, long arg3,
{
/* XXX: convert resource ? */
int resource = arg1;
- struct target_rlimit *target_rlim = (void *)arg2;
+ struct target_rlimit *target_rlim;
struct rlimit rlim;
ret = get_errno(getrlimit(resource, &rlim));
if (!is_error(ret)) {
- target_rlim->rlim_cur = tswapl(rlim.rlim_cur);
- target_rlim->rlim_max = tswapl(rlim.rlim_max);
+ lock_user_struct(target_rlim, arg2, 0);
+ rlim.rlim_cur = tswapl(target_rlim->rlim_cur);
+ rlim.rlim_max = tswapl(target_rlim->rlim_max);
+ unlock_user_struct(target_rlim, arg2, 1);
}
}
break;
case TARGET_NR_getrusage:
{
struct rusage rusage;
- struct target_rusage *target_rusage = (void *)arg2;
ret = get_errno(getrusage(arg1, &rusage));
if (!is_error(ret)) {
- host_to_target_rusage(target_rusage, &rusage);
+ host_to_target_rusage(arg2, &rusage);
}
}
break;
case TARGET_NR_gettimeofday:
{
- struct target_timeval *target_tv = (void *)arg1;
struct timeval tv;
ret = get_errno(gettimeofday(&tv, NULL));
if (!is_error(ret)) {
- host_to_target_timeval(target_tv, &tv);
+ host_to_target_timeval(arg1, &tv);
}
}
break;
case TARGET_NR_settimeofday:
{
- struct target_timeval *target_tv = (void *)arg1;
struct timeval tv;
- target_to_host_timeval(&tv, target_tv);
+ target_to_host_timeval(&tv, arg1);
ret = get_errno(settimeofday(&tv, NULL));
}
break;
#ifdef TARGET_NR_select
case TARGET_NR_select:
{
- struct target_sel_arg_struct *sel = (void *)arg1;
- sel->n = tswapl(sel->n);
- sel->inp = tswapl(sel->inp);
- sel->outp = tswapl(sel->outp);
- sel->exp = tswapl(sel->exp);
- sel->tvp = tswapl(sel->tvp);
- ret = do_select(sel->n, (void *)sel->inp, (void *)sel->outp,
- (void *)sel->exp, (void *)sel->tvp);
+ struct target_sel_arg_struct *sel;
+ target_ulong inp, outp, exp, tvp;
+ long nsel;
+
+ lock_user_struct(sel, arg1, 1);
+ nsel = tswapl(sel->n);
+ inp = tswapl(sel->inp);
+ outp = tswapl(sel->outp);
+ exp = tswapl(sel->exp);
+ tvp = tswapl(sel->tvp);
+ unlock_user_struct(sel, arg1, 0);
+ ret = do_select(nsel, inp, outp, exp, tvp);
}
break;
#endif
case TARGET_NR_symlink:
- ret = get_errno(symlink((const char *)arg1, (const char *)arg2));
+ {
+ void *p2;
+ p = lock_user_string(arg1);
+ p2 = lock_user_string(arg2);
+ ret = get_errno(symlink(p, p2));
+ unlock_user(p2, arg2, 0);
+ unlock_user(p, arg1, 0);
+ }
break;
#ifdef TARGET_NR_oldlstat
case TARGET_NR_oldlstat:
goto unimplemented;
#endif
case TARGET_NR_readlink:
- ret = get_errno(readlink(path((const char *)arg1), (char *)arg2, arg3));
+ {
+ void *p2;
+ p = lock_user_string(arg1);
+ p2 = lock_user(arg2, arg3, 0);
+ ret = get_errno(readlink(path(p), p2, arg3));
+ unlock_user(p2, arg2, ret);
+ unlock_user(p, arg1, 0);
+ }
break;
case TARGET_NR_uselib:
goto unimplemented;
case TARGET_NR_swapon:
- ret = get_errno(swapon((const char *)arg1, arg2));
+ p = lock_user_string(arg1);
+ ret = get_errno(swapon(p, arg2));
+ unlock_user(p, arg1, 0);
break;
case TARGET_NR_reboot:
goto unimplemented;
@@ -2258,14 +2532,16 @@ long do_syscall(void *cpu_env, int num, long arg1, long arg2, long arg3,
case TARGET_NR_mmap:
#if defined(TARGET_I386) || defined(TARGET_ARM)
{
- uint32_t v1, v2, v3, v4, v5, v6, *vptr;
- vptr = (uint32_t *)arg1;
- v1 = tswap32(vptr[0]);
- v2 = tswap32(vptr[1]);
- v3 = tswap32(vptr[2]);
- v4 = tswap32(vptr[3]);
- v5 = tswap32(vptr[4]);
- v6 = tswap32(vptr[5]);
+ target_ulong *v;
+ target_ulong v1, v2, v3, v4, v5, v6;
+ v = lock_user(arg1, 6 * sizeof(target_ulong), 1);
+ v1 = tswapl(v[0]);
+ v2 = tswapl(v[1]);
+ v3 = tswapl(v[2]);
+ v4 = tswapl(v[3]);
+ v5 = tswapl(v[4]);
+ v6 = tswapl(v[5]);
+ unlock_user(v, arg1, 0);
ret = get_errno(target_mmap(v1, v2, v3,
target_to_host_bitmask(v4, mmap_flags_tbl),
v5, v6));
@@ -2299,14 +2575,15 @@ long do_syscall(void *cpu_env, int num, long arg1, long arg2, long arg3,
case TARGET_NR_mremap:
ret = get_errno(target_mremap(arg1, arg2, arg3, arg4, arg5));
break;
+ /* ??? msync/mlock/munlock are broken for softmmu. */
case TARGET_NR_msync:
- ret = get_errno(msync((void *)arg1, arg2, arg3));
+ ret = get_errno(msync(g2h(arg1), arg2, arg3));
break;
case TARGET_NR_mlock:
- ret = get_errno(mlock((void *)arg1, arg2));
+ ret = get_errno(mlock(g2h(arg1), arg2));
break;
case TARGET_NR_munlock:
- ret = get_errno(munlock((void *)arg1, arg2));
+ ret = get_errno(munlock(g2h(arg1), arg2));
break;
case TARGET_NR_mlockall:
ret = get_errno(mlockall(arg1));
@@ -2315,7 +2592,9 @@ long do_syscall(void *cpu_env, int num, long arg1, long arg2, long arg3,
ret = get_errno(munlockall());
break;
case TARGET_NR_truncate:
- ret = get_errno(truncate((const char *)arg1, arg2));
+ p = lock_user_string(arg1);
+ ret = get_errno(truncate(p, arg2));
+ unlock_user(p, arg1, 0);
break;
case TARGET_NR_ftruncate:
ret = get_errno(ftruncate(arg1, arg2));
@@ -2334,11 +2613,15 @@ long do_syscall(void *cpu_env, int num, long arg1, long arg2, long arg3,
goto unimplemented;
#endif
case TARGET_NR_statfs:
- ret = get_errno(statfs(path((const char *)arg1), &stfs));
+ p = lock_user_string(arg1);
+ ret = get_errno(statfs(path(p), &stfs));
+ unlock_user(p, arg1, 0);
convert_statfs:
if (!is_error(ret)) {
- struct target_statfs *target_stfs = (void *)arg2;
+ struct target_statfs *target_stfs;
+ lock_user_struct(target_stfs, arg2, 0);
+ /* ??? put_user is probably wrong. */
put_user(stfs.f_type, &target_stfs->f_type);
put_user(stfs.f_bsize, &target_stfs->f_bsize);
put_user(stfs.f_blocks, &target_stfs->f_blocks);
@@ -2348,6 +2631,7 @@ long do_syscall(void *cpu_env, int num, long arg1, long arg2, long arg3,
put_user(stfs.f_ffree, &target_stfs->f_ffree);
put_user(stfs.f_fsid.__val[0], &target_stfs->f_fsid);
put_user(stfs.f_namelen, &target_stfs->f_namelen);
+ unlock_user_struct(target_stfs, arg2, 1);
}
break;
case TARGET_NR_fstatfs:
@@ -2355,11 +2639,15 @@ long do_syscall(void *cpu_env, int num, long arg1, long arg2, long arg3,
goto convert_statfs;
#ifdef TARGET_NR_statfs64
case TARGET_NR_statfs64:
- ret = get_errno(statfs(path((const char *)arg1), &stfs));
+ p = lock_user_string(arg1);
+ ret = get_errno(statfs(path(p), &stfs));
+ unlock_user(p, arg1, 0);
convert_statfs64:
if (!is_error(ret)) {
- struct target_statfs64 *target_stfs = (void *)arg3;
-
+ struct target_statfs64 *target_stfs;
+
+ lock_user_struct(target_stfs, arg3, 0);
+ /* ??? put_user is probably wrong. */
put_user(stfs.f_type, &target_stfs->f_type);
put_user(stfs.f_bsize, &target_stfs->f_bsize);
put_user(stfs.f_blocks, &target_stfs->f_blocks);
@@ -2369,6 +2657,7 @@ long do_syscall(void *cpu_env, int num, long arg1, long arg2, long arg3,
put_user(stfs.f_ffree, &target_stfs->f_ffree);
put_user(stfs.f_fsid.__val[0], &target_stfs->f_fsid);
put_user(stfs.f_namelen, &target_stfs->f_namelen);
+ unlock_user_struct(target_stfs, arg3, 0);
}
break;
case TARGET_NR_fstatfs64:
@@ -2380,60 +2669,63 @@ long do_syscall(void *cpu_env, int num, long arg1, long arg2, long arg3,
goto unimplemented;
#endif
case TARGET_NR_socketcall:
- ret = do_socketcall(arg1, (int32_t *)arg2);
+ ret = do_socketcall(arg1, arg2);
break;
case TARGET_NR_syslog:
goto unimplemented;
case TARGET_NR_setitimer:
{
- struct target_itimerval *target_value = (void *)arg2;
- struct target_itimerval *target_ovalue = (void *)arg3;
struct itimerval value, ovalue, *pvalue;
- if (target_value) {
+ if (arg2) {
pvalue = &value;
target_to_host_timeval(&pvalue->it_interval,
- &target_value->it_interval);
+ arg2);
target_to_host_timeval(&pvalue->it_value,
- &target_value->it_value);
+ arg2 + sizeof(struct target_timeval));
} else {
pvalue = NULL;
}
ret = get_errno(setitimer(arg1, pvalue, &ovalue));
- if (!is_error(ret) && target_ovalue) {
- host_to_target_timeval(&target_ovalue->it_interval,
+ if (!is_error(ret) && arg3) {
+ host_to_target_timeval(arg3,
&ovalue.it_interval);
- host_to_target_timeval(&target_ovalue->it_value,
+ host_to_target_timeval(arg3 + sizeof(struct target_timeval),
&ovalue.it_value);
}
}
break;
case TARGET_NR_getitimer:
{
- struct target_itimerval *target_value = (void *)arg2;
struct itimerval value;
ret = get_errno(getitimer(arg1, &value));
- if (!is_error(ret) && target_value) {
- host_to_target_timeval(&target_value->it_interval,
+ if (!is_error(ret) && arg2) {
+ host_to_target_timeval(arg2,
&value.it_interval);
- host_to_target_timeval(&target_value->it_value,
+ host_to_target_timeval(arg2 + sizeof(struct target_timeval),
&value.it_value);
}
}
break;
case TARGET_NR_stat:
- ret = get_errno(stat(path((const char *)arg1), &st));
+ p = lock_user_string(arg1);
+ ret = get_errno(stat(path(p), &st));
+ unlock_user(p, arg1, 0);
goto do_stat;
case TARGET_NR_lstat:
- ret = get_errno(lstat(path((const char *)arg1), &st));
+ p = lock_user_string(arg1);
+ ret = get_errno(lstat(path(p), &st));
+ unlock_user(p, arg1, 0);
goto do_stat;
case TARGET_NR_fstat:
{
ret = get_errno(fstat(arg1, &st));
do_stat:
if (!is_error(ret)) {
- struct target_stat *target_st = (void *)arg2;
+ struct target_stat *target_st;
+
+ lock_user_struct(target_st, arg2, 0);
target_st->st_dev = tswap16(st.st_dev);
target_st->st_ino = tswapl(st.st_ino);
#if defined(TARGET_PPC)
@@ -2453,6 +2745,7 @@ long do_syscall(void *cpu_env, int num, long arg1, long arg2, long arg3,
target_st->target_st_atime = tswapl(st.st_atime);
target_st->target_st_mtime = tswapl(st.st_mtime);
target_st->target_st_ctime = tswapl(st.st_ctime);
+ unlock_user_struct(target_st, arg2, 1);
}
}
break;
@@ -2479,9 +2772,9 @@ long do_syscall(void *cpu_env, int num, long arg1, long arg2, long arg3,
case TARGET_NR_wait4:
{
int status;
- target_long *status_ptr = (void *)arg2;
+ target_long status_ptr = arg2;
struct rusage rusage, *rusage_ptr;
- struct target_rusage *target_rusage = (void *)arg4;
+ target_ulong target_rusage = arg4;
if (target_rusage)
rusage_ptr = &rusage;
else
@@ -2489,7 +2782,7 @@ long do_syscall(void *cpu_env, int num, long arg1, long arg2, long arg3,
ret = get_errno(wait4(arg1, &status, arg3, rusage_ptr));
if (!is_error(ret)) {
if (status_ptr)
- *status_ptr = tswap32(status);
+ tputl(status_ptr, status);
if (target_rusage) {
host_to_target_rusage(target_rusage, &rusage);
}
@@ -2497,15 +2790,19 @@ long do_syscall(void *cpu_env, int num, long arg1, long arg2, long arg3,
}
break;
case TARGET_NR_swapoff:
- ret = get_errno(swapoff((const char *)arg1));
+ p = lock_user_string(arg1);
+ ret = get_errno(swapoff(p));
+ unlock_user(p, arg1, 0);
break;
case TARGET_NR_sysinfo:
{
- struct target_sysinfo *target_value = (void *)arg1;
+ struct target_sysinfo *target_value;
struct sysinfo value;
ret = get_errno(sysinfo(&value));
- if (!is_error(ret) && target_value)
+ if (!is_error(ret) && arg1)
{
+ /* ??? __put_user is probably wrong. */
+ lock_user_struct(target_value, arg1, 0);
__put_user(value.uptime, &target_value->uptime);
__put_user(value.loads[0], &target_value->loads[0]);
__put_user(value.loads[1], &target_value->loads[1]);
@@ -2520,6 +2817,7 @@ long do_syscall(void *cpu_env, int num, long arg1, long arg2, long arg3,
__put_user(value.totalhigh, &target_value->totalhigh);
__put_user(value.freehigh, &target_value->freehigh);
__put_user(value.mem_unit, &target_value->mem_unit);
+ unlock_user_struct(target_value, arg1, 1);
}
}
break;
@@ -2540,30 +2838,33 @@ long do_syscall(void *cpu_env, int num, long arg1, long arg2, long arg3,
break;
#endif
case TARGET_NR_setdomainname:
- ret = get_errno(setdomainname((const char *)arg1, arg2));
+ p = lock_user_string(arg1);
+ ret = get_errno(setdomainname(p, arg2));
+ unlock_user(p, arg1, 0);
break;
case TARGET_NR_uname:
/* no need to transcode because we use the linux syscall */
{
struct new_utsname * buf;
- buf = (struct new_utsname *)arg1;
+ lock_user_struct(buf, arg1, 0);
ret = get_errno(sys_uname(buf));
if (!is_error(ret)) {
/* Overrite the native machine name with whatever is being
emulated. */
strcpy (buf->machine, UNAME_MACHINE);
}
+ unlock_user_struct(buf, arg1, 1);
}
break;
#ifdef TARGET_I386
case TARGET_NR_modify_ldt:
- ret = get_errno(do_modify_ldt(cpu_env, arg1, (void *)arg2, arg3));
+ ret = get_errno(do_modify_ldt(cpu_env, arg1, arg2, arg3));
break;
case TARGET_NR_vm86old:
goto unimplemented;
case TARGET_NR_vm86:
- ret = do_vm86(cpu_env, arg1, (void *)arg2);
+ ret = do_vm86(cpu_env, arg1, arg2);
break;
#endif
case TARGET_NR_adjtimex:
@@ -2594,20 +2895,21 @@ long do_syscall(void *cpu_env, int num, long arg1, long arg2, long arg3,
{
#if defined (__x86_64__)
ret = get_errno(lseek(arg1, ((uint64_t )arg2 << 32) | arg3, arg5));
- *(int64_t *)arg4 = ret;
+ tput64(arg4, ret);
#else
int64_t res;
ret = get_errno(_llseek(arg1, arg2, arg3, &res, arg5));
- *(int64_t *)arg4 = tswap64(res);
+ tput64(arg4, res);
#endif
}
break;
case TARGET_NR_getdents:
#if TARGET_LONG_SIZE != 4
+ goto unimplemented;
#warning not supported
#elif TARGET_LONG_SIZE == 4 && HOST_LONG_SIZE == 8
{
- struct target_dirent *target_dirp = (void *)arg2;
+ struct target_dirent *target_dirp;
struct dirent *dirp;
long count = arg3;
@@ -2625,6 +2927,7 @@ long do_syscall(void *cpu_env, int num, long arg1, long arg2, long arg3,
count1 = 0;
de = dirp;
+ target_dirp = lock_user(arg2, count, 0);
tde = target_dirp;
while (len > 0) {
reclen = de->d_reclen;
@@ -2644,13 +2947,15 @@ long do_syscall(void *cpu_env, int num, long arg1, long arg2, long arg3,
}
ret = count1;
}
+ unlock_user(target_dirp, arg2, ret);
free(dirp);
}
#else
{
- struct dirent *dirp = (void *)arg2;
+ struct dirent *dirp;
long count = arg3;
+ dirp = lock_user(arg2, count, 0);
ret = get_errno(sys_getdents(arg1, dirp, count));
if (!is_error(ret)) {
struct dirent *de;
@@ -2668,14 +2973,16 @@ long do_syscall(void *cpu_env, int num, long arg1, long arg2, long arg3,
len -= reclen;
}
}
+ unlock_user(dirp, arg2, ret);
}
#endif
break;
#ifdef TARGET_NR_getdents64
case TARGET_NR_getdents64:
{
- struct dirent64 *dirp = (void *)arg2;
+ struct dirent64 *dirp;
long count = arg3;
+ dirp = lock_user(arg2, count, 0);
ret = get_errno(sys_getdents64(arg1, dirp, count));
if (!is_error(ret)) {
struct dirent64 *de;
@@ -2693,21 +3000,22 @@ long do_syscall(void *cpu_env, int num, long arg1, long arg2, long arg3,
len -= reclen;
}
}
+ unlock_user(dirp, arg2, ret);
}
break;
#endif /* TARGET_NR_getdents64 */
case TARGET_NR__newselect:
- ret = do_select(arg1, (void *)arg2, (void *)arg3, (void *)arg4,
- (void *)arg5);
+ ret = do_select(arg1, arg2, arg3, arg4, arg5);
break;
case TARGET_NR_poll:
{
- struct target_pollfd *target_pfd = (void *)arg1;
+ struct target_pollfd *target_pfd;
unsigned int nfds = arg2;
int timeout = arg3;
struct pollfd *pfd;
unsigned int i;
+ target_pfd = lock_user(arg1, sizeof(struct target_pollfd) * nfds, 1);
pfd = alloca(sizeof(struct pollfd) * nfds);
for(i = 0; i < nfds; i++) {
pfd[i].fd = tswap32(target_pfd[i].fd);
@@ -2718,7 +3026,10 @@ long do_syscall(void *cpu_env, int num, long arg1, long arg2, long arg3,
for(i = 0; i < nfds; i++) {
target_pfd[i].revents = tswap16(pfd[i].revents);
}
+ ret += nfds * (sizeof(struct target_pollfd)
+ - sizeof(struct pollfd));
}
+ unlock_user(target_pfd, arg1, ret);
}
break;
case TARGET_NR_flock:
@@ -2729,31 +3040,23 @@ long do_syscall(void *cpu_env, int num, long arg1, long arg2, long arg3,
case TARGET_NR_readv:
{
int count = arg3;
- int i;
struct iovec *vec;
- struct target_iovec *target_vec = (void *)arg2;
vec = alloca(count * sizeof(struct iovec));
- for(i = 0;i < count; i++) {
- vec[i].iov_base = (void *)tswapl(target_vec[i].iov_base);
- vec[i].iov_len = tswapl(target_vec[i].iov_len);
- }
+ lock_iovec(vec, arg2, count, 0);
ret = get_errno(readv(arg1, vec, count));
+ unlock_iovec(vec, arg2, count, 1);
}
break;
case TARGET_NR_writev:
{
int count = arg3;
- int i;
struct iovec *vec;
- struct target_iovec *target_vec = (void *)arg2;
vec = alloca(count * sizeof(struct iovec));
- for(i = 0;i < count; i++) {
- vec[i].iov_base = (void *)tswapl(target_vec[i].iov_base);
- vec[i].iov_len = tswapl(target_vec[i].iov_len);
- }
+ lock_iovec(vec, arg2, count, 1);
ret = get_errno(writev(arg1, vec, count));
+ unlock_iovec(vec, arg2, count, 0);
}
break;
case TARGET_NR_getsid:
@@ -2768,27 +3071,34 @@ long do_syscall(void *cpu_env, int num, long arg1, long arg2, long arg3,
return -ENOTDIR;
case TARGET_NR_sched_setparam:
{
- struct sched_param *target_schp = (void *)arg2;
+ struct sched_param *target_schp;
struct sched_param schp;
+
+ lock_user_struct(target_schp, arg2, 1);
schp.sched_priority = tswap32(target_schp->sched_priority);
+ unlock_user_struct(target_schp, arg2, 0);
ret = get_errno(sched_setparam(arg1, &schp));
}
break;
case TARGET_NR_sched_getparam:
{
- struct sched_param *target_schp = (void *)arg2;
+ struct sched_param *target_schp;
struct sched_param schp;
ret = get_errno(sched_getparam(arg1, &schp));
if (!is_error(ret)) {
+ lock_user_struct(target_schp, arg2, 0);
target_schp->sched_priority = tswap32(schp.sched_priority);
+ unlock_user_struct(target_schp, arg2, 1);
}
}
break;
case TARGET_NR_sched_setscheduler:
{
- struct sched_param *target_schp = (void *)arg3;
+ struct sched_param *target_schp;
struct sched_param schp;
+ lock_user_struct(target_schp, arg3, 1);
schp.sched_priority = tswap32(target_schp->sched_priority);
+ unlock_user_struct(target_schp, arg3, 0);
ret = get_errno(sched_setscheduler(arg1, arg2, &schp));
}
break;
@@ -2806,26 +3116,20 @@ long do_syscall(void *cpu_env, int num, long arg1, long arg2, long arg3,
break;
case TARGET_NR_sched_rr_get_interval:
{
- struct target_timespec *target_ts = (void *)arg2;
struct timespec ts;
ret = get_errno(sched_rr_get_interval(arg1, &ts));
if (!is_error(ret)) {
- target_ts->tv_sec = tswapl(ts.tv_sec);
- target_ts->tv_nsec = tswapl(ts.tv_nsec);
+ host_to_target_timespec(arg2, &ts);
}
}
break;
case TARGET_NR_nanosleep:
{
- struct target_timespec *target_req = (void *)arg1;
- struct target_timespec *target_rem = (void *)arg2;
struct timespec req, rem;
- req.tv_sec = tswapl(target_req->tv_sec);
- req.tv_nsec = tswapl(target_req->tv_nsec);
+ target_to_host_timespec(&req, arg1);
ret = get_errno(nanosleep(&req, &rem));
- if (is_error(ret) && target_rem) {
- target_rem->tv_sec = tswapl(rem.tv_sec);
- target_rem->tv_nsec = tswapl(rem.tv_nsec);
+ if (is_error(ret) && arg2) {
+ host_to_target_timespec(arg2, &rem);
}
}
break;
@@ -2837,15 +3141,21 @@ long do_syscall(void *cpu_env, int num, long arg1, long arg2, long arg3,
goto unimplemented;
#ifdef TARGET_NR_pread
case TARGET_NR_pread:
- page_unprotect_range((void *)arg2, arg3);
- ret = get_errno(pread(arg1, (void *)arg2, arg3, arg4));
+ page_unprotect_range(arg2, arg3);
+ p = lock_user(arg2, arg3, 0);
+ ret = get_errno(pread(arg1, p, arg3, arg4));
+ unlock_user(p, arg2, ret);
break;
case TARGET_NR_pwrite:
- ret = get_errno(pwrite(arg1, (void *)arg2, arg3, arg4));
+ p = lock_user(arg2, arg3, 1);
+ ret = get_errno(pwrite(arg1, p, arg3, arg4));
+ unlock_user(p, arg2, 0);
break;
#endif
case TARGET_NR_getcwd:
- ret = get_errno(sys_getcwd1((char *)arg1, arg2));
+ p = lock_user(arg1, arg2, 0);
+ ret = get_errno(sys_getcwd1(p, arg2));
+ unlock_user(p, arg1, ret);
break;
case TARGET_NR_capget:
goto unimplemented;
@@ -2874,16 +3184,20 @@ long do_syscall(void *cpu_env, int num, long arg1, long arg2, long arg3,
struct rlimit rlim;
ret = get_errno(getrlimit(arg1, &rlim));
if (!is_error(ret)) {
- struct target_rlimit *target_rlim = (void *)arg2;
+ struct target_rlimit *target_rlim;
+ lock_user_struct(target_rlim, arg2, 0);
target_rlim->rlim_cur = tswapl(rlim.rlim_cur);
target_rlim->rlim_max = tswapl(rlim.rlim_max);
+ unlock_user_struct(target_rlim, arg2, 1);
}
break;
}
#endif
#ifdef TARGET_NR_truncate64
case TARGET_NR_truncate64:
- ret = target_truncate64(cpu_env, (const char *)arg1, arg2, arg3, arg4);
+ p = lock_user_string(arg1);
+ ret = target_truncate64(cpu_env, p, arg2, arg3, arg4);
+ unlock_user(p, arg1, 0);
break;
#endif
#ifdef TARGET_NR_ftruncate64
@@ -2893,12 +3207,16 @@ long do_syscall(void *cpu_env, int num, long arg1, long arg2, long arg3,
#endif
#ifdef TARGET_NR_stat64
case TARGET_NR_stat64:
- ret = get_errno(stat(path((const char *)arg1), &st));
+ p = lock_user_string(arg1);
+ ret = get_errno(stat(path(p), &st));
+ unlock_user(p, arg1, 0);
goto do_stat64;
#endif
#ifdef TARGET_NR_lstat64
case TARGET_NR_lstat64:
- ret = get_errno(lstat(path((const char *)arg1), &st));
+ p = lock_user_string(arg1);
+ ret = get_errno(lstat(path(p), &st));
+ unlock_user(p, arg1, 0);
goto do_stat64;
#endif
#ifdef TARGET_NR_fstat64
@@ -2909,8 +3227,10 @@ long do_syscall(void *cpu_env, int num, long arg1, long arg2, long arg3,
if (!is_error(ret)) {
#ifdef TARGET_ARM
if (((CPUARMState *)cpu_env)->eabi) {
- struct target_eabi_stat64 *target_st = (void *)arg2;
+ struct target_eabi_stat64 *target_st;
+ lock_user_struct(target_st, arg2, 1);
memset(target_st, 0, sizeof(struct target_eabi_stat64));
+ /* put_user is probably wrong. */
put_user(st.st_dev, &target_st->st_dev);
put_user(st.st_ino, &target_st->st_ino);
#ifdef TARGET_STAT64_HAS_BROKEN_ST_INO
@@ -2928,11 +3248,14 @@ long do_syscall(void *cpu_env, int num, long arg1, long arg2, long arg3,
put_user(st.st_atime, &target_st->target_st_atime);
put_user(st.st_mtime, &target_st->target_st_mtime);
put_user(st.st_ctime, &target_st->target_st_ctime);
+ unlock_user_struct(target_st, arg2, 0);
} else
#endif
{
- struct target_stat64 *target_st = (void *)arg2;
+ struct target_stat64 *target_st;
+ lock_user_struct(target_st, arg2, 1);
memset(target_st, 0, sizeof(struct target_stat64));
+ /* ??? put_user is probably wrong. */
put_user(st.st_dev, &target_st->st_dev);
put_user(st.st_ino, &target_st->st_ino);
#ifdef TARGET_STAT64_HAS_BROKEN_ST_INO
@@ -2950,6 +3273,7 @@ long do_syscall(void *cpu_env, int num, long arg1, long arg2, long arg3,
put_user(st.st_atime, &target_st->target_st_atime);
put_user(st.st_mtime, &target_st->target_st_mtime);
put_user(st.st_ctime, &target_st->target_st_ctime);
+ unlock_user_struct(target_st, arg2, 0);
}
}
}
@@ -2957,7 +3281,9 @@ long do_syscall(void *cpu_env, int num, long arg1, long arg2, long arg3,
#endif
#ifdef USE_UID16
case TARGET_NR_lchown:
- ret = get_errno(lchown((const char *)arg1, low2highuid(arg2), low2highgid(arg3)));
+ p = lock_user_string(arg1);
+ ret = get_errno(lchown(p, low2highuid(arg2), low2highgid(arg3)));
+ unlock_user(p, arg1, 0);
break;
case TARGET_NR_getuid:
ret = get_errno(high2lowuid(getuid()));
@@ -2980,28 +3306,32 @@ long do_syscall(void *cpu_env, int num, long arg1, long arg2, long arg3,
case TARGET_NR_getgroups:
{
int gidsetsize = arg1;
- uint16_t *target_grouplist = (void *)arg2;
+ uint16_t *target_grouplist;
gid_t *grouplist;
int i;
grouplist = alloca(gidsetsize * sizeof(gid_t));
ret = get_errno(getgroups(gidsetsize, grouplist));
if (!is_error(ret)) {
+ target_grouplist = lock_user(arg2, gidsetsize * 2, 0);
for(i = 0;i < gidsetsize; i++)
target_grouplist[i] = tswap16(grouplist[i]);
+ unlock_user(target_grouplist, arg2, gidsetsize * 2);
}
}
break;
case TARGET_NR_setgroups:
{
int gidsetsize = arg1;
- uint16_t *target_grouplist = (void *)arg2;
+ uint16_t *target_grouplist;
gid_t *grouplist;
int i;
grouplist = alloca(gidsetsize * sizeof(gid_t));
+ target_grouplist = lock_user(arg2, gidsetsize * 2, 1);
for(i = 0;i < gidsetsize; i++)
grouplist[i] = tswap16(target_grouplist[i]);
+ unlock_user(target_grouplist, arg2, 0);
ret = get_errno(setgroups(gidsetsize, grouplist));
}
break;
@@ -3018,12 +3348,12 @@ long do_syscall(void *cpu_env, int num, long arg1, long arg2, long arg3,
#ifdef TARGET_NR_getresuid
case TARGET_NR_getresuid:
{
- int ruid, euid, suid;
+ uid_t ruid, euid, suid;
ret = get_errno(getresuid(&ruid, &euid, &suid));
if (!is_error(ret)) {
- *(uint16_t *)arg1 = tswap16(high2lowuid(ruid));
- *(uint16_t *)arg2 = tswap16(high2lowuid(euid));
- *(uint16_t *)arg3 = tswap16(high2lowuid(suid));
+ tput16(arg1, tswap16(high2lowuid(ruid)));
+ tput16(arg2, tswap16(high2lowuid(euid)));
+ tput16(arg3, tswap16(high2lowuid(suid)));
}
}
break;
@@ -3038,18 +3368,20 @@ long do_syscall(void *cpu_env, int num, long arg1, long arg2, long arg3,
#ifdef TARGET_NR_getresgid
case TARGET_NR_getresgid:
{
- int rgid, egid, sgid;
+ gid_t rgid, egid, sgid;
ret = get_errno(getresgid(&rgid, &egid, &sgid));
if (!is_error(ret)) {
- *(uint16_t *)arg1 = tswap16(high2lowgid(rgid));
- *(uint16_t *)arg2 = tswap16(high2lowgid(egid));
- *(uint16_t *)arg3 = tswap16(high2lowgid(sgid));
+ tput16(arg1, tswap16(high2lowgid(rgid)));
+ tput16(arg2, tswap16(high2lowgid(egid)));
+ tput16(arg3, tswap16(high2lowgid(sgid)));
}
}
break;
#endif
case TARGET_NR_chown:
- ret = get_errno(chown((const char *)arg1, low2highuid(arg2), low2highgid(arg3)));
+ p = lock_user_string(arg1);
+ ret = get_errno(chown(p, low2highuid(arg2), low2highgid(arg3)));
+ unlock_user(p, arg1, 0);
break;
case TARGET_NR_setuid:
ret = get_errno(setuid(low2highuid(arg1)));
@@ -3067,7 +3399,9 @@ long do_syscall(void *cpu_env, int num, long arg1, long arg2, long arg3,
#ifdef TARGET_NR_lchown32
case TARGET_NR_lchown32:
- ret = get_errno(lchown((const char *)arg1, arg2, arg3));
+ p = lock_user_string(arg1);
+ ret = get_errno(lchown(p, arg2, arg3));
+ unlock_user(p, arg1, 0);
break;
#endif
#ifdef TARGET_NR_getuid32
@@ -3104,15 +3438,17 @@ long do_syscall(void *cpu_env, int num, long arg1, long arg2, long arg3,
case TARGET_NR_getgroups32:
{
int gidsetsize = arg1;
- uint32_t *target_grouplist = (void *)arg2;
+ uint32_t *target_grouplist;
gid_t *grouplist;
int i;
grouplist = alloca(gidsetsize * sizeof(gid_t));
ret = get_errno(getgroups(gidsetsize, grouplist));
if (!is_error(ret)) {
+ target_grouplist = lock_user(arg2, gidsetsize * 4, 0);
for(i = 0;i < gidsetsize; i++)
- put_user(grouplist[i], &target_grouplist[i]);
+ target_grouplist[i] = tswap32(grouplist[i]);
+ unlock_user(target_grouplist, arg2, gidsetsize * 4);
}
}
break;
@@ -3121,13 +3457,15 @@ long do_syscall(void *cpu_env, int num, long arg1, long arg2, long arg3,
case TARGET_NR_setgroups32:
{
int gidsetsize = arg1;
- uint32_t *target_grouplist = (void *)arg2;
+ uint32_t *target_grouplist;
gid_t *grouplist;
int i;
grouplist = alloca(gidsetsize * sizeof(gid_t));
+ target_grouplist = lock_user(arg2, gidsetsize * 4, 1);
for(i = 0;i < gidsetsize; i++)
- get_user(grouplist[i], &target_grouplist[i]);
+ grouplist[i] = tswap32(target_grouplist[i]);
+ unlock_user(target_grouplist, arg2, 0);
ret = get_errno(setgroups(gidsetsize, grouplist));
}
break;
@@ -3145,12 +3483,12 @@ long do_syscall(void *cpu_env, int num, long arg1, long arg2, long arg3,
#ifdef TARGET_NR_getresuid32
case TARGET_NR_getresuid32:
{
- int ruid, euid, suid;
+ uid_t ruid, euid, suid;
ret = get_errno(getresuid(&ruid, &euid, &suid));
if (!is_error(ret)) {
- *(uint32_t *)arg1 = tswap32(ruid);
- *(uint32_t *)arg2 = tswap32(euid);
- *(uint32_t *)arg3 = tswap32(suid);
+ tput32(arg1, tswap32(ruid));
+ tput32(arg2, tswap32(euid));
+ tput32(arg3, tswap32(suid));
}
}
break;
@@ -3163,19 +3501,21 @@ long do_syscall(void *cpu_env, int num, long arg1, long arg2, long arg3,
#ifdef TARGET_NR_getresgid32
case TARGET_NR_getresgid32:
{
- int rgid, egid, sgid;
+ gid_t rgid, egid, sgid;
ret = get_errno(getresgid(&rgid, &egid, &sgid));
if (!is_error(ret)) {
- *(uint32_t *)arg1 = tswap32(rgid);
- *(uint32_t *)arg2 = tswap32(egid);
- *(uint32_t *)arg3 = tswap32(sgid);
+ tput32(arg1, tswap32(rgid));
+ tput32(arg2, tswap32(egid));
+ tput32(arg3, tswap32(sgid));
}
}
break;
#endif
#ifdef TARGET_NR_chown32
case TARGET_NR_chown32:
- ret = get_errno(chown((const char *)arg1, arg2, arg3));
+ p = lock_user_string(arg1);
+ ret = get_errno(chown(p, arg2, arg3));
+ unlock_user(p, arg1, 0);
break;
#endif
#ifdef TARGET_NR_setuid32
@@ -3213,9 +3553,9 @@ long do_syscall(void *cpu_env, int num, long arg1, long arg2, long arg3,
case TARGET_NR_fcntl64:
{
struct flock64 fl;
- struct target_flock64 *target_fl = (void *)arg3;
+ struct target_flock64 *target_fl;
#ifdef TARGET_ARM
- struct target_eabi_flock64 *target_efl = (void *)arg3;
+ struct target_eabi_flock64 *target_efl;
#endif
switch(arg2) {
@@ -3224,19 +3564,23 @@ long do_syscall(void *cpu_env, int num, long arg1, long arg2, long arg3,
if (ret == 0) {
#ifdef TARGET_ARM
if (((CPUARMState *)cpu_env)->eabi) {
+ lock_user_struct(target_efl, arg3, 0);
target_efl->l_type = tswap16(fl.l_type);
target_efl->l_whence = tswap16(fl.l_whence);
target_efl->l_start = tswap64(fl.l_start);
target_efl->l_len = tswap64(fl.l_len);
target_efl->l_pid = tswapl(fl.l_pid);
+ unlock_user_struct(target_efl, arg3, 1);
} else
#endif
{
+ lock_user_struct(target_fl, arg3, 0);
target_fl->l_type = tswap16(fl.l_type);
target_fl->l_whence = tswap16(fl.l_whence);
target_fl->l_start = tswap64(fl.l_start);
target_fl->l_len = tswap64(fl.l_len);
target_fl->l_pid = tswapl(fl.l_pid);
+ unlock_user_struct(target_fl, arg3, 1);
}
}
break;
@@ -3245,19 +3589,23 @@ long do_syscall(void *cpu_env, int num, long arg1, long arg2, long arg3,
case F_SETLKW64:
#ifdef TARGET_ARM
if (((CPUARMState *)cpu_env)->eabi) {
+ lock_user_struct(target_efl, arg3, 1);
fl.l_type = tswap16(target_efl->l_type);
fl.l_whence = tswap16(target_efl->l_whence);
fl.l_start = tswap64(target_efl->l_start);
fl.l_len = tswap64(target_efl->l_len);
fl.l_pid = tswapl(target_efl->l_pid);
+ unlock_user_struct(target_efl, arg3, 0);
} else
#endif
{
+ lock_user_struct(target_fl, arg3, 1);
fl.l_type = tswap16(target_fl->l_type);
fl.l_whence = tswap16(target_fl->l_whence);
fl.l_start = tswap64(target_fl->l_start);
fl.l_len = tswap64(target_fl->l_len);
fl.l_pid = tswapl(target_fl->l_pid);
+ unlock_user_struct(target_fl, arg3, 0);
}
ret = get_errno(fcntl(arg1, arg2, &fl));
break;