2.6.16 for linuxulator & 7.0 release

[Divacky Roman]

On Fri, Mar 16, 2007 at 12:00:38PM +0100, Alexander Leidinger wrote:
> Hi,
> 
> ATM it doesn't look like we can enable 2.6.16 by default in 7.0. The 
> reasons:
>  - java showstopper (epoll not implemented)
>  - futex/TLS for amd64 not in -current for testing
>  - futexes not completely right
>  - *at() not implemented
>  - the time to test a 2.6.16 default until 7.0 shrinks fast
> 
> Roman has some preliminary work regarding *at() and it looks good to  
> him. He wants me to test some stuff but I don't have enough time in  
> the next weeks because of work related stuff. Any volunteers to help  
> out?
 
the code looks ok and it even passes LTP tests. I hope that someone with
VFS-fu will review my linux_at() function and then I'll implement all
the *at syscalls.

as java is regarded I think its not a big problem to patch java shell script
to set emulation to 2.4 etc. but I agree that 2.6 is not ready for 7.0R

there's also problem with 2.6 emulation with flash9 (eating memory till
it dies)

> In p4 we have the futex/TLS stuff for amd64 but because of the futexes  
> not completely right part it is not committed to current yet. As we  
> already have the futex and TLS stuff for i386 on a similar level in  
> current, I would say we should go ahead and sync the amd64 stuff. It  
> is not used by default, so we don't break existing linux stuff and we  
> get the benefit of more people being able to have a look at it and  
> play with it. So what are your opinions, shall we give jkim@ the green  
> light to MFp4 the futex/TLS stuff?

futexes are broken... I just recieved a report that oracle installer
hangs on amd64/SMP waiting on a lock. The TLS at amd64 looks good though

I dont see any harm in commiting futexes/TLS for amd64 - go for it kim!

I'd like to see the linux-aio commited as well before 7.0R

> Regarding the futexes not being completely right and the epoll stuff:  
> I think it needs to be done now, not in a month or two, else we don't  
> have enough time to let people play with this before the release of  
> 7.0. Anyone with a little bit of time at hand out there? We need a  
> specification what the futexes are supposed to do (so far we didn't  
> find a good description, and the linux code is hard to read and  
> doesn't not really tell what it is _supposed_ to do) and we need  
> people which compare the current code we have with this specification.  
> Finding a regression test for futexes would also be nice.

I plan to apply for SoC this year with a plan to implement epoll/inotify
interface + "finish" the linux26 stuff... if I get elected I guess things
will move forward fast :) I definitely plan to look at the futexes (we have
a testing program so its not that hard)

roman