On 17/12/2013 12:29 am, Sebastian Huber wrote:
Pass also zero-size buffers to the write handler since they may have
side-effects for non-regular files, e.g. sending of an empty message.
I agree regular files should follow the standard with others being
unspecified. I do not think this patch does this.
@@ -31,95 +29,31 @@ ssize_t writev(
{
ssize_t total;
int v;
- int bytes;
rtems_libio_t *iop;
- ssize_t old;
- bool all_zeros;
- rtems_libio_check_fd( fd );
- iop = rtems_libio_iop( fd );
- rtems_libio_check_is_open( iop );
- rtems_libio_check_permissions_with_error( iop, LIBIO_FLAGS_WRITE, EBADF );
-
- /*
- * Argument validation on IO vector
- */
- if ( !iov )
- rtems_set_errno_and_return_minus_one( EINVAL );
-
- if ( iovcnt <= 0 )
- rtems_set_errno_and_return_minus_one( EINVAL );
-
- if ( iovcnt > IOV_MAX )
- rtems_set_errno_and_return_minus_one( EINVAL );
-
- /*
- * OpenGroup says that you are supposed to return EINVAL if the
- * sum of the iov_len values in the iov array would overflow a
- * ssize_t.
- *
- * Also we would like to ensure that no IO is performed if there
- * are obvious errors in the iovec. So this extra loop ensures
- * that we do not do anything if there is an argument error.
- *
- * In addition,the OpenGroup specification says that if all the
- * iov_len entries are zero, then the call has no effect. So
- * this loop does that check as well and sets "all-zero" appropriately.
- * The variable "all_zero" is used as an early exit point before
- * entering the write loop.
- */
- all_zeros = true;
- for ( old=0, total=0, v=0 ; v < iovcnt ; v++ ) {
+ total = rtems_libio_iovec_eval( fd, iov, iovcnt, LIBIO_FLAGS_WRITE, &iop );
+ if ( total >= 0 ) {
I am not sure about this...
- /*
- * A writev with all zeros is supposed to have no effect per OpenGroup.
- */
- if ( all_zeros == true ) {
- return 0;
- }
... as this statement is correct if the file is a regular file. I think
a check for a regular file is needed if the total is 0.
Chris
_______________________________________________
rtems-devel mailing list
rtems-devel@rtems.org
http://www.rtems.org/mailman/listinfo/rtems-devel
