From: Jeff Dike <jdike@ka...> - 2001-04-27 22:44:55
> Unless you also are interested in my TUN/TAP userspace daemon for
> connecting a host to um_eth_serv.
> there is any interest for this I'll make a updated patch available.
Yes. I've been ignoring that stuff because I was ignoring the driver that it
was based on. Now that I'm thinking about this, I'd like whatever code anyone
has that might go into it.
Attached is all patches I currently have for UML.
* um_eth_tap, for connecting a host to um_eth_serv
* MODLIB correction for "make ARCH=um modules_install"
* config.in cleanups to more match the expected i386 layout, fix some erronous
option dependencies, and include some features which actually works fine in UML
(more should be added)
* Primitive console output flow control, modelled after how serial console ports
are managed (if busy then block for a short while waiting for the output queue to
drain). This works around the problem where data is lost on the console and
virtual terminals. Further work is required to implement real output flow control
in the driver. This approach is really only suitable for writes directly to the
console (i.e. printk and friends), writes from userspace applications should be
properly flow controlled using the normal device flow controls.
* Cleanups on forced exits (uml killed)
- Restore terminal settings
- Kill all host instances of our processes
- Kill all our helpers
* Some hostfs fixes
- Was allocating one byte to many in one of the path operations. Confusing to
anyone trying to understand these allocations..
- Was freeing some memory twice
- Reset the block size to what is expected by the block driver. Many
filesystems changes the logical block size without restoring it if mounting
* Fix some symbol export issues for MODVERSIONS and a few modules depending on
symbols not yet exported. The goal is to get MODVERSIONS to actually work some
time but a lot of work remains.
* Attempt in trying to get UBD to support partitions. Not finished yet.
Apart from these I also use the eth patch from net-0.090 I think, and the
exitcalls order patch posted yesterday. (and a lot of other generic kernel
patches not relevant to UML)