From: Jeff D. <jd...@ka...> - 2002-11-16 17:13:01
|
mu...@ag... said: > But in UML, there's also host-side stuff that has to be restored. In > particular, memory mappings on the host have to be restored, and the > process tracing needs to be set up. So? What's the problem? > For example, when it's saving the suspend data, it writes to a swap > area. For UML, we could just write the suspend data to a file on the > host filesystem. And that file has been set up as a UML swap area. > My second idea, then, was to save the suspend data (kernel text, > kernel data, kernal bss, UML physical memory, etc) into the host > fileystem. Then to restore, I'd fire up UML, but have it map the saved > data into memory. It would then look at the process table to restore > the running UML processes. And how is this different from swsusp, except that you decided to reinvent the wheel? Jeff |