ssize_t
__pread_chk (int fd, void *buf, size_t nbytes, off_t offset, size_t buflen)
{
if (nbytes > buflen)
__chk_fail ();
return __pread (fd, buf, nbytes, offset);
}
/* Try to get a file descriptor for the shared meory segment
containing the database. */
static struct mapped_database *
get_mapping (request_type type, const char *key,
struct mapped_database **mappedp)
{
struct mapped_database *result = NO_MAPPING;
#ifdef SCM_RIGHTS
const size_t keylen = strlen (key) + 1;
char resdata[keylen];
int saved_errno = errno;
int mapfd = -1;
/* Send the request. */
struct iovec iov[2];
request_header req;
int sock = open_socket ();
if (sock < 0)
goto out;
req.version = NSCD_VERSION;
req.type = type;
req.key_len = keylen;
iov[0].iov_base = &req;
iov[0].iov_len = sizeof (req);
iov[1].iov_base = (void *) key;
iov[1].iov_len = keylen;
if (__builtin_expect (TEMP_FAILURE_RETRY (__writev (sock, iov, 2))
!= iov[0].iov_len + iov[1].iov_len, 0))
/* We cannot even write the request. */
goto out_close2;
/* Room for the data sent along with the file descriptor. We expect
the key name back. */
iov[0].iov_base = resdata;
iov[0].iov_len = keylen;
union
{
struct cmsghdr hdr;
char bytes[CMSG_SPACE (sizeof (int))];
} buf;
struct msghdr msg = { .msg_iov = iov, .msg_iovlen = 1,
.msg_control = buf.bytes,
.msg_controllen = sizeof (buf) };
struct cmsghdr *cmsg = CMSG_FIRSTHDR (&msg);
cmsg->cmsg_level = SOL_SOCKET;
cmsg->cmsg_type = SCM_RIGHTS;
cmsg->cmsg_len = CMSG_LEN (sizeof (int));
/* This access is well-aligned since BUF is correctly aligned for an
int and CMSG_DATA preserves this alignment. */
*(int *) CMSG_DATA (cmsg) = -1;
msg.msg_controllen = cmsg->cmsg_len;
if (wait_on_socket (sock) <= 0)
goto out_close2;
# ifndef MSG_NOSIGNAL
# define MSG_NOSIGNAL 0
# endif
if (__builtin_expect (TEMP_FAILURE_RETRY (__recvmsg (sock, &msg,
MSG_NOSIGNAL))
!= keylen, 0))
goto out_close2;
mapfd = *(int *) CMSG_DATA (cmsg);
if (__builtin_expect (CMSG_FIRSTHDR (&msg)->cmsg_len
!= CMSG_LEN (sizeof (int)), 0))
goto out_close;
struct stat64 st;
if (__builtin_expect (strcmp (resdata, key) != 0, 0)
|| __builtin_expect (fstat64 (mapfd, &st) != 0, 0)
|| __builtin_expect (st.st_size < sizeof (struct database_pers_head), 0))
goto out_close;
struct database_pers_head head;
if (__builtin_expect (TEMP_FAILURE_RETRY (__pread (mapfd, &head,
sizeof (head), 0))
!= sizeof (head), 0))
goto out_close;
if (__builtin_expect (head.version != DB_VERSION, 0)
|| __builtin_expect (head.header_size != sizeof (head), 0)
/* This really should not happen but who knows, maybe the update
thread got stuck. */
|| __builtin_expect (! head.nscd_certainly_running
&& head.timestamp + MAPPING_TIMEOUT < time (NULL),
0))
goto out_close;
size_t size = (sizeof (head) + roundup (head.module * sizeof (ref_t), ALIGN)
+ head.data_size);
if (__builtin_expect (st.st_size < size, 0))
goto out_close;
/* The file is large enough, map it now. */
void *mapping = __mmap (NULL, size, PROT_READ, MAP_SHARED, mapfd, 0);
if (__builtin_expect (mapping != MAP_FAILED, 1))
{
/* Allocate a record for the mapping. */
struct mapped_database *newp = malloc (sizeof (*newp));
if (newp == NULL)
{
/* Ugh, after all we went through the memory allocation failed. */
__munmap (mapping, size);
goto out_close;
}
newp->head = mapping;
newp->data = ((char *) mapping + head.header_size
+ roundup (head.module * sizeof (ref_t), ALIGN));
newp->mapsize = size;
/* Set counter to 1 to show it is usable. */
newp->counter = 1;
result = newp;
}
out_close:
__close (mapfd);
out_close2:
__close (sock);
out:
__set_errno (saved_errno);
#endif /* SCM_RIGHTS */
struct mapped_database *oldval = *mappedp;
*mappedp = result;
if (oldval != NULL && atomic_decrement_val (&oldval->counter) == 0)
__nscd_unmap (oldval);
return result;
}
struct mapped_database *
__nscd_get_map_ref (request_type type, const char *name,
struct locked_map_ptr *mapptr, int *gc_cyclep)
{
struct mapped_database *cur = mapptr->mapped;
if (cur == NO_MAPPING)
return cur;
int cnt = 0;
while (atomic_compare_and_exchange_val_acq (&mapptr->lock, 1, 0) != 0)
{
// XXX Best number of rounds?
if (++cnt > 5)
return NO_MAPPING;
atomic_delay ();
}
cur = mapptr->mapped;
if (__builtin_expect (cur != NO_MAPPING, 1))
{
/* If not mapped or timestamp not updated, request new map. */
if (cur == NULL
|| (cur->head->nscd_certainly_running == 0
&& cur->head->timestamp + MAPPING_TIMEOUT < time (NULL)))
cur = get_mapping (type, name, &mapptr->mapped);
if (__builtin_expect (cur != NO_MAPPING, 1))
{
if (__builtin_expect (((*gc_cyclep = cur->head->gc_cycle) & 1) != 0,
0))
cur = NO_MAPPING;
else
atomic_increment (&cur->counter);
}
}
mapptr->lock = 0;
return cur;
}
const struct datahead *
__nscd_cache_search (request_type type, const char *key, size_t keylen,
const struct mapped_database *mapped)
{
unsigned long int hash = __nis_hash (key, keylen) % mapped->head->module;
ref_t work = mapped->head->array[hash];
while (work != ENDREF)
{
struct hashentry *here = (struct hashentry *) (mapped->data + work);
if (type == here->type && keylen == here->len
&& memcmp (key, mapped->data + here->key, keylen) == 0)
{
/* We found the entry. Increment the appropriate counter. */
const struct datahead *dh
= (struct datahead *) (mapped->data + here->packet);
/* See whether we must ignore the entry or whether something
is wrong because garbage collection is in progress. */
if (dh->usable && ((char *) dh + dh->allocsize
<= (char *) mapped->head + mapped->mapsize))
return dh;
}
work = here->next;
}
return NULL;
}
int
posix_fallocate (int fd, __off_t offset, __off_t len)
{
struct stat64 st;
if (offset < 0 || len < 0)
return EINVAL;
/* Perform overflow check. The outer cast relies on a GCC
extension. */
if ((__off_t) ((uint64_t) offset + (uint64_t) len) < 0)
return EFBIG;
/* pwrite below will not do the right thing in O_APPEND mode. */
{
int flags = __fcntl (fd, F_GETFL, 0);
if (flags < 0 || (flags & O_APPEND) != 0)
return EBADF;
}
/* We have to make sure that this is really a regular file. */
if (__fxstat64 (_STAT_VER, fd, &st) != 0)
return EBADF;
if (S_ISFIFO (st.st_mode))
return ESPIPE;
if (! S_ISREG (st.st_mode))
return ENODEV;
if (len == 0)
{
/* This is racy, but there is no good way to satisfy a
zero-length allocation request. */
if (st.st_size < offset)
{
int ret = __ftruncate (fd, offset);
if (ret != 0)
ret = errno;
return ret;
}
return 0;
}
/* Minimize data transfer for network file systems, by issuing
single-byte write requests spaced by the file system block size.
(Most local file systems have fallocate support, so this fallback
code is not used there.) */
unsigned increment;
{
struct statfs64 f;
if (__fstatfs64 (fd, &f) != 0)
return errno;
if (f.f_bsize == 0)
increment = 512;
else if (f.f_bsize < 4096)
increment = f.f_bsize;
else
/* NFS does not propagate the block size of the underlying
storage and may report a much larger value which would still
leave holes after the loop below, so we cap the increment at
4096. */
increment = 4096;
}
/* Write a null byte to every block. This is racy; we currently
lack a better option. Compare-and-swap against a file mapping
might additional local races, but requires interposition of a
signal handler to catch SIGBUS. */
for (offset += (len - 1) % increment; len > 0; offset += increment)
{
len -= increment;
if (offset < st.st_size)
{
unsigned char c;
ssize_t rsize = __pread (fd, &c, 1, offset);
if (rsize < 0)
return errno;
/* If there is a non-zero byte, the block must have been
allocated already. */
else if (rsize == 1 && c != 0)
continue;
}
if (__pwrite (fd, "", 1, offset) != 1)
return errno;
}
return 0;
}
size_t __libc_pread(int fd, void *buf, size_t count, off_t offset) {
return __pread(fd,buf,count,0,__LONG_LONG_PAIR(0,offset));
}
