table of contents
- bookworm 2.6.2-4
- bookworm-backports 2.8.6-1~bpo12+2
- testing 2.8.6-1
- unstable 2.8.6-1
gensio_set_sync(3) | Library Functions Manual | gensio_set_sync(3) |
NAME¶
gensio_set_sync, gensio_clear_sync, gensio_read_s, gensio_write_s - Synchronous I/O operations on a gensio
SYNOPSIS¶
#include <gensio/gensio.h>
- int gensio_set_sync(struct gensio *io);
- int gensio_clear_sync(struct gensio *io);
- int gensio_read_s(struct gensio *io, gensiods *count,
-
void *data, gensiods datalen,
struct gensio_time *timeout); - int gensio_read_s_intr(struct gensio *io, gensiods *count,
-
void *data, gensiods datalen,
struct gensio_time *timeout); - int gensio_write_s(struct gensio *io, gensiods *count,
-
const void *data, gensiods datalen,
struct gensio_time *timeout); - int gensio_write_s_intr(struct gensio *io, gensiods *count,
-
const void *data, gensiods datalen,
struct gensio_time *timeout);
DESCRIPTION¶
Normal gensio operation is asynchronous callback based. This serves most programs fairly well, especially if they are listening to multiple connections at the same time. You wouldn't want to write a compiler this way, but if you are writing something that is driven by external events, this event-driven type of programming works well. If you think about it, if you are using something like poll, select, etc., you almost always end up with something like:
- poll(fds...)
if (fd1 read is set)
call fd1_read_handler()
if (fd1 write is set)
call fd1_write_handler()
if (fd2 read is set)
...
The gensio handling does all this for you. Just register a handler with the gensio to get the read and write calls. It's more efficient, neater, and you end up with less code.
But occasionally you need to do something synchronous with the program execution. For instance, in gtlsshd, if the initial certificate and password verification fails, it uses PAM to handle reading the password from the remote gensio. This requires synchronous I/O, and it uses this capability.
gensio_set_sync sets up the gensio for synchronous I/O. If you do this, the event callback that is currently registered will no longer receive read and write callbacks. It *will* receive other callbacks. You must call this before doing any of the synchronous read and write operations. This function will block (while handling normal gensio events) until no callbacks are active.
gensio_clear_sync returns the gensio to asynchronous I/O. The callback will be restored to the one that was set when gensio_set_sync() was called.
gensio_read_s Waits for data from the gensio, up to datalen bytes. count (if not NULL) will be updated to the actual number of bytes read. This will wait for any read and will return whatever that read was, even if it is less than datalen. This function waits for the amount of time in timeout. timeout is updated to the amount of time left to wait. If timeout is NULL, wait forever.
gensio_write_s writes data to the gensio. count (if not NULL) will be updated to the actual number of bytes written. This function will wait until either the timeout occurs or all the data is written. This function waits for the amount of time in timeout. timeout is updated to the amount of time left to wait. If timeout is NULL, wait forever.
gensio_read_s_intr and gensio_write_s_intr are like gensio_read_s and gensio_write_s, but they return immediately if an signal interrupt occurs. On systems without signals, these are the same as gensio_read_s and gensio_write_s.
RETURN VALUES¶
Zero is returned on success, or a gensio error on failure.
SEE ALSO¶
27 Feb 2019 |