From: Jungwoo H. <hab...@gm...> - 2008-11-18 23:56:05
|
Hi, As far as I remember(it was a year ago), 2.6 kernel and recent libc uses new system call that is not supported by DSS. When I looked at DSS code, it seems to be based on 2.4 kernel and libc 2.2.X. Some syscalls are emulated, and some are ignored. Many new syscalls can't be ignored + states are stored inside the kernel memory, thus difficult to implement in DSS. One of the important missing syscall I remember is sys_futex(), which is invented along with NPTL thread library. There were other syscalls related to signal handling. You can't get away this problem by linking old library to JikesRVM because classpath depends on new libc. The bottom line is it is possible to fix DSS, but you need a deep understanding of these syscalls, and requires quite an efforts for implementation. Or some experts may suggest better way. I'm going to work on Jikes RVM to run on PTLsim. I'll keep it posted when it is ready. Thanks, Jungwoo On Tue, Nov 18, 2008 at 5:29 PM, Steve Blackburn <Ste...@an... > wrote: > Hi Eliot, > > I agree that it is a limitation, not a bug. And of course I did not > mean to imply that the DSS maintainers were being slack ;-) > > However, the specific problem is not simply an accretion of new > syscalls (which would be easy enough to address one at a time, and > which we would have done), but pthreads more generally. Jungwoo Ha at > UT spent some time looking into it about a year ago, and our > conclusion was that it would be too great an effort to justify. I'm > trying to dig up email discussions with him to remind myself of the > details, but I can't find them just now. If you want to chase this > further, I'm sure Jungwoo will be happy to try to fill you in. > > Cheers, > > --Steve > > On 19/11/2008, at 7:35 AM, Eliot Moss wrote: > > > As the maintainer of DSS, let me jump in here ... > > > > The proper word is more likely "limitation" than "bug". What happens > > is > > that as Jikes RVM has evolved, and classpath, they use more and more > > Linux system calls. Every single call has to be appropriately modeled > > with code in DSS, so new calls mean new code. If I had a detailed list > > of the (new) calls being made, and where necessary/helpful, the > > arguments to those calls, I could work on it over time, but there is > > nobody to maintain this code except me. > > > > Best wishes -- Eliot Moss > > > > ------------------------------------------------------------------------- > > This SF.Net email is sponsored by the Moblin Your Move Developer's > > challenge > > Build the coolest Linux based applications with Moblin SDK & win > > great prizes > > Grand prize is a trip for two to an Open Source event anywhere in > > the world > > http://moblin-contest.org/redirect.php?banner_id=100&url=/ > > _______________________________________________ > > Jikesrvm-researchers mailing list > > Jik...@li... > > https://lists.sourceforge.net/lists/listinfo/jikesrvm-researchers > > > ------------------------------------------------------------------------- > This SF.Net email is sponsored by the Moblin Your Move Developer's > challenge > Build the coolest Linux based applications with Moblin SDK & win great > prizes > Grand prize is a trip for two to an Open Source event anywhere in the world > http://moblin-contest.org/redirect.php?banner_id=100&url=/ > _______________________________________________ > Jikesrvm-researchers mailing list > Jik...@li... > https://lists.sourceforge.net/lists/listinfo/jikesrvm-researchers > -- Jungwoo Ha |