On Sat, Jul 21, 2018 at 06:46:42PM +0200, cinap_lenrek@xxxxxxxxxxx wrote: > implemented support in kexec-tools to boot plan9 kernel last night > (see attachment). > > this is very basic support. it loads the kernel and provides it > a memory map (from info->memory_range) in its config area and also > has a optional parameter to pass additional plan9.ini configuration. > > but i have a couple of questions: > > who's in charge? I am the maintainer of kexec-tools. > is here interest in supporting other OS than linux such as plan9 > in the official kexec-tools? or should i write my own plan9 loader > for linux using the kexec_load() syscall directly? I think that if there is a good amount of code re-use and there is an intention to keep the plan9 code maintained then it is reasonable to include support in the current kexec-tools tree. > on uefi systems, the address of the RSDT is passed to the kernel > in the config area as theres no known region to search for it. > this is done in 9front by the bootloader (as we do not want to > call back into crappy uefi code from our kernel once virtual > memory and interrups are set up). but i'v not seen any of the > kexec modules implement such a thing. so where do i get that > information from? how does linux locate the acpi tables? > > on a uefi systems, theres no cga text mode. when linear > framebuffer was already set up. we pass this information in > the config area (done by efi loader using the GOP). but i havnt > seen any kexec module to implement passing the framebuffer > configuration to the new kernel. so where do i get that > that information in linux? this is also needed for proper > multiboot support. Its been a long time since I worked on kexec in that depth - it was on ia64. Perhaps others on the list have some more recent experience that can help guide you. > the 9front kernel supports multiboot specification, but your > multiboot loader expects all multiboot images to be ELF files. > as well. what are you THINKING? :-) I'm not sure that I was involved in the project when that design decision was made. But in general I would say that its entirely reasonable to make enhancements so long as we can be sure that existing boot-cases don't break. _______________________________________________ kexec mailing list kexec@xxxxxxxxxxxxxxxxxxx http://lists.infradead.org/mailman/listinfo/kexec
