- l4-hackers - example.com Site

L4Linux and virtualization
by Robert Gubler 14 Nov '06

14 Nov '06

Hello, I'd like to run two L4Linux 2.4 based kernels and one L4Linux 2.6 based kernel. Is this something that can be achieved relatively easily or would all running kernels need to be 2.6/2.4? How do the 'guest' kernels communicate to each other in an L4 environment? In User Mode Linux (UML) and Xen on the guest kernels there is built-in support for pseudo Ethernet devices; these devices are then bridged together on the host kernel. I understand the architecture is considerably different in L4, but is there some similar mechanism already in place that allows Ethernet/IP communication between the running Linux kernels? Thanks in advance, Rob

3 2

L4 API and ABI Documentation
by Till Backhaus 12 Nov '06

12 Nov '06

Dear L4-Hackers, to get started with L4 I'd like to take a look at the L4 API- Documentation and the L4 ABI-Documentation for the PowerPC. Unfortunately I didn't find them at your site. Maybe someone can point me to the right document(s)? Thanks in advance, Till Backhaus

2 1

new system call for L4
by Suresh Iyer 12 Nov '06

12 Nov '06

how do i implement a new system call for L4 ? I am tryiing out a new real time scheduling algorithm for L4? thanks suresh

2 1

IPC from L4Linux
by Daniel Vandersee 02 Nov '06

02 Nov '06

Hi there, when I place an ipc call from a l4linux task to an l4 server, I always get a task number, which, I think, is not assigned. Is for every l4linux task a new l4 task is generated/reserved? With kind regards, Daniel

3 5

kernel bug in combination with jdb
by Johannes Richter 01 Nov '06

01 Nov '06

Hi, I found a kernel bug; while selecting the partner of my task 6 in the kernel debugger with tab in the list of present tasks i got the message: id name pr wait to state ???.?? KERNEL BUG at EIP:0001eea8 ESP:000d39e4 -- PFA:00000006 ASSERTION_FAILED (Kmem::is_tcb_page_fault((Mword) kesp, 0)) in function static void Usermode::kernel_entry(ucontext*, long unsigned int, long unsigned int, long unsigned int, long unsigned int, long unsigned int, long unsigned int, long unsigned int, long unsigned int)progr in file ../src/kern/ux/usermode.cpp:204 cu - Joe

2 1

Re: FW: XGenOS / Fiasco
by William DUCK 28 Oct '06

28 Oct '06

On Sunday 22 October 2006 21:56, you wrote: > Hello, > > I must confess your mail came as a big surprise, since nobody seemed to > care about the XGeNOS project for quite some time, but of course the > surprise is a pleasing one ;-) > > Considering your thougths and ideas below, I'll try to figure out as > much as possible for the moment ... I've been thinking over some of > them, but it will take me some time to get through it all ... > > > I would want to know if this is possible to use fiasco and a uclibc-nptl > > rewrote in cyclone for a XGenOS fork. > > http://www.xgenos.org/ > > http://os.inf.tu-dresden.de/vfiasco/ > > Well, vfiasco is an interesting project and complements to some work I'm > going to do as a part of my PhD thesis. On the other hand from the > viewpoint of the u-kernel API an implementation of X.2 seems to be a > good idea, which would rather imply to use the pistachio kernel ... > http://cyclone.thelanguage.org/ > > Interesting. I never came across that before, but it seems this could > solve some problems of mine (not only XGeNOS related ...) > http://www.realitydiluted.com/nptl-uclibc/ > > http://www.busybox.net/cgi-bin/viewcvs.cgi/branches/uClibc-nptl/ > > uhm, I'd be happy if there was an alternative to uClibc as my previous > experiences with uClibc have shown that its code is from so many > different sources that it's sometimes quite hard to keep the overview. I > guess the code is quite hard to maintain which IMHO is not the perfect > basis for secure software ... But, I cannot see any alternative to > uClibc for the moment ... I'll probably have a look at Peter Anvins > klibc in some time, maybe that could be an alternate starting point for > a suitable C-library ... > > > Could we be able to merge the best ideas from these projects ? : > > > > http://www.marcus-brinkmann.org/hurd-ng.pdf > > > > http://www.research.ibm.com/K42/papers/eurosys06.pdf > > > > http://domino.research.ibm.com/comm/research_projects.nsf/pages/prose.ind > >ex.html > > Lots to read ;-) Give me some time to check that out ... > > > Which Device Driver Framework could be the best and that will let us > > reuse the > > maximum of linux drivers code ? : > > > > http://www.jaluna.com/doc/c5/html/BSPDevGuide/c3882.html > > > > http://sourceforge.net/projects/jaluna/ > > > > http://www.circlemud.org/~jelson/software/fusd/ > > <http://www.circlemud.org/%7Ejelson/software/fusd/> > > > > http://www.gelato.unsw.edu.au/IA64wiki/UserLevelDrivers > > Well, IMHO the device driver framwork is the crucial point of the whole > system. I've got some ideas that derive partly from some of the projects > mentioned above, and others (e.g the sawmill project, etc.). I agree > that reusage of existing driver code must be on of the project goals, > but if this turns out to be of some security risk, one should think of > some other solutions ... but that's just my EUR 0.02 ... > > > Could haskell be a good alternative to C ? : > > > > http://programatica.cs.pdx.edu/House/ > > > > http://www.ertos.nicta.com.au/research/sel4/ > > > > http://repetae.net/john/computer/jhc/ > > > > http://www.cse.unsw.edu.au/~chak/haskell/c2hs/ > > <http://www.cse.unsw.edu.au/%7Echak/haskell/c2hs/> > > I must confess I haven't dealt with Haskell yet, so I can't say anything > about that ... > > > Time is not a problem, I would to know only which basis could be the most > > suitable for an innovative project ... > > Let's stay on contact, any ideas are welcome. Pls. use the above > mail-address (ta(a)xgenos.org) if possible. I'm probably going to set up a > mailinglist in some time ... > > Best regards, > > -ta

1 0

DROPS L4Env I/O client API not applicable for C++ programs
by Marcel Winandy 24 Oct '06

24 Oct '06

Hello! I have just encountered a problem with the L4Env I/O client API libio.h if you try to use this in a C++ program. You get a syntax error because there is a variable having the name "class". But that is a C++ keyword. In l4/include/x86/l4/generic_io/types.h: **************************************** struct l4io_drv: "unsigned class:8;" struct l4io_pci_dev: "unsigned long class;" In l4/include/x86/l4/generic_io/libio.h: **************************************** "int l4io_pci_find_class(unsigned long class, ...)" Although these variable names are possible in C, you run in syntax errors when trying to use these API interfaces in a C++ program. I think these variables should get other names, shouldn't they? Kind regards Marcel -- Marcel Winandy http://www.prosec.rub.de/staff/winandy.html Applied Data Security Group, Ruhr-University Bochum, Germany

2 1

PowerPC Port
by William DUCK 20 Oct '06

20 Oct '06

Hello, DId/Will someone port Fiasco to Power Architecture ? http://www.power.org/members/developers/specs/PAPR_Version_2.0_28August06.p… http://www.power.org/resources/downloads/PowerISA_203.Public.pdf Best Regards, Will

1 0

Exokernel / Device drivers
by William DUCK 20 Oct '06

20 Oct '06

Hello, The more I go forward, the more questions come to me : http://lists.tunes.org/archives/tunes/1995-June/000617.html the exokernel paper says somewhere that all the exokernel does is to securely multiplex the cpu (that's time) and memory (that's space) ressources. that's what you said you want from a good os in your `why ...' paper. and some months ago, i observed that the os may be viewed as a device driver for these two because that's what device drivers should do: securely multiplex some hardware ressource http://lists.gnu.org/archive/html/l4-hurd/2006-08/msg00005.html "In contract to our approach [L4], [Exokernel] is based on the philosophy that a kernel should /not/ provide abstractions but only a minimal set of primitives. Consequently, the Exokernel interface is architecture dependent . . . We believe that dropping the abstractional [sic] approach could only be justified by substantial performance gains. . . . It might turn out that the right abstractions are even more efficient than securely multiplexing hardware primitives or, on the other hand, that abstractions are too inflexible." http://tunes.org/wiki/Decentralized So would it be possible to build a general purpose OS without a kernel ? Isn't just a matter of language *and/or* hardware architecure ? Could we imagine a full application-level OS and link the apps directly to the drivers => so the "kernel" would be the device drivers framework ? Or at least the ressource management system ... Best Regards, Will

1 0

about L4Linux supporting for 312MHz Intel XScale Bulverde (PXA270) processor
by Wenbin Zhang 19 Oct '06

19 Oct '06

Hello, I notice in the L4Linux website, you mentioned that The ARM port currently runs on the following platforms: * PLEB The University of New South Wales' embedded hardware * Supports both PLEB1 (SA1100) and PLEB2 (XScale PXA255) * TuxScreen - StrongARM based phone platform. * Intel's IDXP450 Xscale development platform (XScale IXP450). Now I have a Motorola E680i cell phone, it is using 312MHz Intel XScale Bulverde (PXA270) processor, which is using the 5th generation of the ARM architecture. Is it supported by L4Linux currently? Thank you very much! Wenbin

2 1