mtcp_peek
NAME
mtcp_peek − Peek into the bytestream from a monitoring socket
SYNOPSIS
#include <mos_api.h>
ssize_t mtcp_peek(mctx_t mctx, int sockid, int side, char *buf, size_t len);
DESCRIPTION
mtcp_peek() can be used to peek up to len number of bytes from a monitoring socket via socket descriptor sockid into the buffer (unfragmented) starting at buf. It is the responsibility of the user to allocate memory for buf array.
For cases when the total number of buffered unread bytes in the socket are less than len, mtcp_peek() copies entire data to buf and returns the actual number of bytes that can be used by the application.
An mtcp_peek() call takes two additional argument named mctx , that represents the per-core mTCP context in an application (see mtcp_create_context() for details); and side that informs the stack whether it needs to peek the client (MOS_SIDE_CLI) or server (MOS_SIDE_SVR) side.
RETURN VALUE
On success, the number of bytes read is returned. It is not an error if this number is smaller than the number of bytes requested; this may happen when the socket has fewer number of bytes buffered at that moment. On all errors except ENODATA, -1 is returned and errno is set appropriately. On ENODATA error, if the buffer outrun was occured, it would return (-1) * (the number of bytes overwritten on the buffer).
ERRORS
|
EACCES |
The mctx argument is invalid. | ||
|
EBADF |
The socket descriptor sockid is invalid. | ||
|
EINVAL |
The buffer management of the monitoring socket with sockid was already disabled. | ||
|
ENOTSOCK |
The socket referred to by sockid has invalid type | ||
|
ENOTCONN |
Th socket referred to by sockid is in a TCP state that disallows read operations. | ||
|
ENODATA |
The socket referred to by sockid does not have any available bytes in its buffer for reading. | ||
|
EPERM |
The caller is not permitted to access this function. |
NOTES
Since TCP flow control applies between end-hosts, the receive buffer managed by mOS can become full while new packets continue to arrive. Silent TCP ring buffer overwriting is undesirable since the application may not notice the buffer overflow. Instead, mOS raises an error event to explicitly notify the application about the buffer overflow. The application can either drain the buffer by reading the data (by calling mtcp_peek()) or enlarge the buffer (by calling mtcp_setsockopt()). Otherwise, the stack overwrites the buffer with the new data, and adjusts the internal peek pointer. To notify the application about the overwriting, mtcp_peek() fails right after overwriting. Subsequent function calls continue to read the data from the new position.
AUTHORS
mOS development team <mtcp-user@list.ndsl.kaist.edu>
EXAMPLES
http://mos.kaist.edu/guide/programmer/05_api_example.html#monitoring-reassembled-tcp-payload
SEE ALSO
mtcp_ppeek(), mtcp_socket(), mtcp_getsockopt()
COLOPHON
This page is part of mOS release 0.3 docs section. A description of the project, and information about reporting bugs, can be found at http://mos.kaist.edu/.
