consinit() problem on -current
Izumi Tsutsui <[email protected]> Sat, 4 Jul 2009 03:59:51 +0900
| Newsgroups | gmane.os.netbsd.ports.atari,gmane.os.netbsd.ports.amiga |
|---|---|
| Message-ID | <[email protected]> |
Hi, Is there anyone who tried recent -current on atari (and amiga)? I'm afraid -current kernels won't boot (hangs after bootloader). atari (and amiga) historically uses config_console() in autoconf.c to search and initialze console devices, and that function calls autoconf(9) functions with pseudo struct device data etc. But after the following change to sys/kern/subr_autoconf.c http://mail-index.NetBSD.org/source-changes/2009/05/24/msg221624.html now autoconf(9) functions require mutex(9) functions and mutex(9) is not usable at the point where consinit() is called, so config_console() and atari_config_found() (or amiga_config_cound()) silently fails. (because there is no working console) A simple workaround is reverting the above subr_autoconf.c change, but real fix is to rewrite console initialization code not to use autoconf(9) functions but call functions which probe and initialize devices directly. NetBSD/x68k port (which was based on amiga and atari) had the similar console_config() but it has been fixed recently: http://mail-index.NetBSD.org/port-x68k/2008/12/31/msg000006.html http://mail-index.NetBSD.org/source-changes/2009/01/17/msg215673.html I guess we can put the similar fixes for atari (and amiga), but I'm afraid that atari has various possible console devices per machine types than x68k. (x68k has no variant, I guess) Is there any summary which devices (especially video) should be used/initialized as a console on each machine? Furthermore, is there any users who can test kernels on Hades and Milan? Thanks, --- Izumi Tsutsui