On Thu, Feb 06, 2014 at 09:36:23PM +0100, Eckehard Berns wrote: > On Thu, Feb 06, 2014 at 12:32:59PM +0000, sin wrote: > > As part of experimenting with a toy distro I wanted to get rid of > > busybox's init, so I hacked together sinit[1]. sinit is based on Strake's > > init[2]. > > [...] > > Let me know what you guys think, I am looking forward to use this with > > sta.li. > > I now had time to test sinit. I first had trouble getting the FIFO > to work. When booting with the root fs mounted read-only, the child > dealing with the FIFO would segfault. When booting with my root fs > mounted read-write my rcinitcmd would become a zombe since the FIFO code > never reaps children.
Thanks for testing! I think the segfault could have been caused because there was no error checking for open(2). I added that in later. Hm yes, you are right, the FIFO code never reaps children. We could probably use the double fork trick + killing the parent to force it to be reaped by the original process (the parent of the FIFO code). > So I changed spawn() to wait for its child. Now I can mount my root fs > read-write in my rcinitcmd or even setup a tmpfs, and my rcinitcmd will > have its exit status checked and not become a zombie. This means though > that you can't use the FIFO until the rcinitcmd exited (which means > you might want to chane your rc.svc to background the agetty loop). Yeah something did not feel right with having the rc script block. I presume if we use the aforementioned double fork trick, we'll be able to continue even if rc.svc doesn't exit (I may be hallucinating right now, I will check tomorrow if what I am saying makes sense). > Also, would it be worth it to deal with x86 Linux's ctrl-alt-del? It would > pull in OS specific code, and maybe people don't care for ctrl-alt-del > on the console, since everybody lives in X anyway. Hm, not sure. What you guys think?
