Me> I want to write a module that opens up a device (known to the Me> userland filesystem as "/dev/mtd/0"), runs some ioctl()s on it, Me> and writes some data to it. Greg> Do that in a userspace program, not from within the kernel. I would rather do it in a userspace program, except for one big problem: There are quite a few other userspace programs running on this embedded system, including programs that are started by the kernel to service timer interrupts, and they are all running out of a CRAMFS partition that I am trying to overwrite for a software upgrade. I'm afraid that one of those processes will get scheduled, and when CRAMFS tries to read a text segment from the partition that has been partially overwritten, it could hang the kernel before I complete the software upgrade. If I could do this in a kernel-space module, I could do a cli() to make sure I complete the upgrade with no interruptions. I can copy a userspace program to a tmpfs filesystem and run it from there, hence preventing my program from invoking a CRAMFS read. But I can't do this for every program that could run. For example, if the kernel gets a device hotswap event, it runs a userspace helper out of a specific path in the CRAMFS. I have no way to prevent that from happening. If there were a way to guarantee the CPU to my program and its device until they are finished with the upgrade, I wouldn't care if the kernel crashes as a result of the CRAMFS partition getting changed out from under it. Dave -- Kernelnewbies: Help each other learn about the Linux kernel. Archive: http://mail.nl.linux.org/kernelnewbies/ FAQ: http://kernelnewbies.org/faq/
