1bootmain.c doesn't work right if the ELF sections aren't 2sector-aligned. so you can't use ld -N. and the sections may also need 3to be non-zero length, only really matters for tiny "kernels". 4 5kernel loaded at 1 megabyte. stack same place that bootasm.S left it. 6 7kinit() should find real mem size 8 and rescue useable memory below 1 meg 9 10no paging, no use of page table hardware, just segments 11 12no user area: no magic kernel stack mapping 13 so no copying of kernel stack during fork 14 though there is a kernel stack page for each process 15 16no kernel malloc(), just kalloc() for user core 17 18user pointers aren't valid in the kernel 19 20setting up first process 21 we do want a process zero, as template 22 but not runnable 23 just set up return-from-trap frame on new kernel stack 24 fake user program that calls exec 25 26map text read-only? 27shared text? 28 29what's on the stack during a trap or sys call? 30 PUSHA before scheduler switch? for callee-saved registers. 31 segment contents? 32 what does iret need to get out of the kernel? 33 how does INT know what kernel stack to use? 34 35are interrupts turned on in the kernel? probably. 36 37per-cpu curproc 38one tss per process, or one per cpu? 39one segment array per cpu, or per process? 40 41pass curproc explicitly, or implicit from cpu #? 42 e.g. argument to newproc()? 43 44test stack expansion 45test running out of memory, process slots 46 47we can't really use a separate stack segment, since stack addresses 48need to work correctly as ordinary pointers. the same may be true of 49data vs text. how can we have a gap between data and stack, so that 50both can grow, without committing 4GB of physical memory? does this 51mean we need paging? 52 53what's the simplest way to add the paging we need? 54 one page table, re-write it each time we leave the kernel? 55 page table per process? 56 probably need to use 0-0xffffffff segments, so that 57 both data and stack pointers always work 58 so is it now worth it to make a process's phys mem contiguous? 59 or could use segment limits and 4 meg pages? 60 but limits would prevent using stack pointers as data pointers 61 how to write-protect text? not important? 62 63perhaps have fixed-size stack, put it in the data segment? 64 65oops, if kernel stack is in contiguous user phys mem, then moving 66users' memory (e.g. to expand it) will wreck any pointers into the 67kernel stack. 68 69do we need to set fs and gs? so user processes can't abuse them? 70 71setupsegs() may modify current segment table, is that legal? 72 73trap() ought to lgdt on return, since currently only done in swtch() 74 75protect hardware interrupt vectors from user INT instructions? 76