home *** CD-ROM | disk | FTP | other *** search
- From: Kay Roemer <roemer@informatik.uni-frankfurt.de>
- Posted-Date: Wed, 15 Dec 93 14:52:26 MEZ
- Received-Date: Wed, 15 Dec 93 14:52:26 +0100
- Message-Id: <9312151352.AA11794@hera.rbi.informatik.uni-frankfurt.de>
- Subject: extension for addtimeout()
- To: mint@atari.archive.umich.edu
- Date: Wed, 15 Dec 93 14:52:26 MEZ
- Mailer: Elm [revision: 70.85]
-
-
- The current addtimeout() lacks the possibility to specify
- a process to which the timeout should be attached.
-
- This is a problem, because all of a process' timeouts are
- deleted when the process exits. So the timeout function
- is not executed if the process exits before the timeout
- happens.
-
- This is bad news, since in MiNT-Net the IP protocol will
- be implemented as a kernel hook (ie no seperate process).
- When a packet arrives over the Network, an interrupt
- routine should add a timeout to MiNT (to ensure the
- timeout function is really executed) for the main IP
- function, which then processes the packet.
-
- This makes it also necessary to make the timeout stuff
- interrupt prove.
-
- I suggest a new function, which should be available over
- struct *kerinfo for device drivers:
-
- TIMEOUT
- *addproctimeout (long delta, void (*f) (PROC *), short pid);
-
- which adds a timeout function `f' to be called in `delta'
- milliseconds to process `pid'. addtimeout() would then be
- a special case of addproctimeout().
-
- This is only a first idea, any suggestions are welcome.
- What do you think about it, Eric ?
-
- Cheers,
- Kay.
-
-