Messages in this thread | | | Date | Tue, 19 Jan 1999 09:35:48 -0600 | From | Michael Elizabeth Chastain <> | Subject | Re: Adding checkpointing API to Linux kernel |
| |
Hi Simon,
> what needs to happen to get it to work on 2.2.x?
I need to write some more code to handle ELF execve(). ELF didn't exist when I started this project.
It doesn't handle signals at all, although I have a good idea how to do this. It also doesn't handle shared writeable memory segments or mmap'ed devices; that is inherently hard. It handles mmap'ed files just fine as long as the file doesn't change after the target mmap's it.
I have a list of every system call and every ioctl in the system. The system call list is fairly stable. Linus does not add new system calls often, and he almost never changes the binary interface of an existing system call (I think this has happened once or twice in four years; it happened to adjtimex several years ago). The ioctl list is a huge time sink because the ioctl system call does not have a parameter that describes the size of the memory region that is being smashed.
It will speed up replay a lot if writing to /proc/$pid/mem works reliably. Andi Kleen already did this but I haven't tested it.
I have a list of "it would be nice if ..." in task-kernel-lix.txt. Some of these are features that other people have asked for too, so they would be good kernel patches for the beginning of 2.3.XX:
ptrace ability to set ORIG_EAX to veto system calls clear process registers on execve some include/linux/*.h cleanup for C++ compatibility
Right now we are in the over-my-dead-body stage of 2.2.0 development, though.
Michael Elizabeth Chastain <mailto:mec@shout.net> "love without fear"
- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.rutgers.edu Please read the FAQ at http://www.tux.org/lkml/
| |