Date: Sun, 27 Dec 1998 00:06:09 +1030 (CST) From: Mark Newton <newton@atdot.dotat.org> To: dillon@apollo.backplane.com (Matthew Dillon) Cc: cvs-all@FreeBSD.ORG, freebsd-hackers@FreeBSD.ORG Subject: Re: LDT entries - Proposed change Message-ID: <199812261336.AAA05205@atdot.dotat.org> In-Reply-To: <199812260504.VAA29295@apollo.backplane.com> from "Matthew Dillon" at Dec 25, 98 09:04:04 pm
next in thread | previous in thread | raw e-mail | index | archive | help
Matthew Dillon wrote: > :I have a patch for sys/i386/include/segments.h and sys/i386/i386/machdep.c > :which changes LUDATA_SEL from 4 to 5 (why 5? Why not?) and provides for > :system call entries via LDT entry 4 (termed LSOL26CALLS_SEL at present). > > Unless you think you'll have the solaris emulation done by the 3.0.1 > release, I don't think it'll ever be "done" any more than the Linux emulator is ever "done". It's currently in a state where it runs a decent selection of Solaris software with libraries from a Solaris 2.5.1 CD-ROM and a somewhat smaller selection of software if libraries are taken from a Solaris 2.6 CD-ROM (we've only just got executables to run with 2.6 libraries at all due to the LDT issue described above, and Sun has made enough libc/ld.so.1 changes between 2.5.1 and 2.6 to tickle the edges of the emulation's accuracy, so a bit more work will be required to provide 2.6 libraries with an environment where they can provide the same functionality as 2.5.1 libraries). I think most of those issues are actually answered by changes in the NetBSD codebase that I haven't tracked so I don't expect them to represent a major problem. I've had RealPlayer for Solaris 2.5.1 working (without audio) and most of the OpenWindows stuff from /usr/openwin on a Solaris/x86 CD. I've also done a small amount of fiddling with a SCO OpenServer CD-ROM with reasonable success. There's still a lot of work to do, but the core is there and limping along. I expect that bde will complain about some stylistic issues, but I'm keeping them in mind too :-) It's still an LKM, it compiles with too many warnings, and there are too many bits enclosed by #ifdef BOGUS or #ifdef NOTYET that I haven't had time to give attention to yet. http://www.freebsd.org/~newton/freebsd-svr4/index.html > I would recommend waiting until the source tree split > before commiting this change. I can't think of any specific problems > that might occur but I don't think a month is long enough to test > something like this and while the direct LDT references are constrained > to a few files in i386, those files are pretty convoluted and it's > a pretty scary change to make. I don't have a major problem with any of that. Whether it's committed or not I'll still be working on it, so this means it'll just be better when it is finally committed. When the time comes I'll also need to change a couple of prototypes (to un-staticize a couple of symbols in uipc_syscalls.c) and add a field to the end of struct socket in order to support emulation of icky SysV STREAMS over BSD sockets. I've been going over those with Mike Smith and discussed them briefly with Garrett earlier this year; See http://www.freebsd.org/~newton/freebsd-svr4/svr4patch for more diffs against 3.0-RELEASE. - mark -------------------------------------------------------------------- I tried an internal modem, newton@atdot.dotat.org but it hurt when I walked. Mark Newton ----- Voice: +61-4-1958-3414 ------------- Fax: +61-8-83034403 ----- To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe cvs-all" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199812261336.AAA05205>