- l4-hackers - example.com Site

L4Env DDE questions
by Igor Shmukler 29 Jan '07

29 Jan '07

Hello, I am working on a research project where we are trying to play with information flow between device drivers, and IO stack in general. We want to be able to break disk and network IO in pieces and glue it together as our needs require. From what I understood while going over KMB slides, it seems like L4Env might be an ideal platform for our experiments. However, I am not sure whether I understand how things work correctly. I was wondering if someone could answer some of my questions, to give me a little head start. When Linux drivers are reused, does everyone of these drivers run in its' own address space? Would "reused DDE" drivers still be isolated from one another? Thank you in advance, igor

2 1

[PATCH] L4Linux Ore fix
by Alexander Warg 29 Jan '07

29 Jan '07

No comment -- Alex

1 0

Problem with syscall 243.
by Jorge Cardona 21 Jan '07

21 Jan '07

Hi. I download form the svn of NICTA all the iguana-project--devel--1.2--version-0, and when it was finished, i type: scons machine=pc99 simulate, i also try with toolsuffix=-3.3 and toolsuffix=-4.0 after a moment of compilation it start to run on qemu an image with l4 kernel, iguana and wombat at the top but get's into an infinite loop of: syscall 243 init:1:e4001 unknown syscall I was searching and found that the problem it get solved building a crosscompiler, but it was on a Red Hat system, I'm using a Debian. Some info about my machine: ii libc5 5.4.46-15 The Linux C library version 5 (run-time libr ii libc6 2.3.6-9 GNU C Library: Shared libraries ii libc6-dev 2.3.6-9 GNU C Library: Development Libraries and Hea ii libc6-i686 2.3.6-9 GNU C Library: Shared libraries [i686 optimi ii libg++2.8.1.3-glibc2.2 2.95.4-27 The GNU C++ extension library - runtime vers ii libstdc++2.10-glibc2.2 2.95.4-27 The GNU stdc++ library $ gcc -print-file-name=libc.so.6 /lib/../lib/libc.so.6 $ gcc -print-file-name=ld-linux.so.2 /lib/../lib/ld-linux.so.2 I'm using Debian on testing. The complete log is: Booting 'L4 NICTA::Iguana' kernel=/boot/grub/kickstart [Multiboot-elf, <0x800000:0x6a98:0x0>, <0x808000:0x0:0x18a8>, shtab=0x80a258 , entry=0x800000] module=/boot/grub/l4kernel [Multiboot-module @ 0x821000, 0x19f295 bytes] module=/boot/grub/ig_server [Multiboot-module @ 0x9c1000, 0x58fde bytes] module=/boot/grub/ig_init [Multiboot-module @ 0xa1a000, 0x55344 bytes] module=/boot/grub/ig_serial [Multiboot-module @ 0xa70000, 0x4dae7 bytes] module=/boot/grub/ig_naming [Multiboot-module @ 0xabe000, 0x41798 bytes] module=/boot/grub/ig_trace [Multiboot-module @ 0xb00000, 0x4104a bytes] module=/boot/grub/ig_timer [Multiboot-module @ 0xb42000, 0x4701e bytes] module=/boot/grub/vmlinux [Multiboot-module @ 0xb8a000, 0x2cbc6c bytes] module=/boot/grub/ramdisk [Multiboot-module @ 0xe56000, 0x7d0000 bytes] module=/boot/grub/startup.script [Multiboot-module @ 0x1626000, 0x1f0 bytes] KickStart 0.13.0 Detected multiboot compliant loader Found an elf32 file called "kernel" located at address 0x00821000 Program Headers: Type Offset VirtAddr PhysAddr FileSiz MemSiz Flg Align LOAD 0x0000c0 0xf0100200 0x00100200 0x0a4ec 0x0a4ec RWE 0x0010 LOAD 0x00a800 0xf010c000 0x0010c000 0x111e0 0x111e0 RWE 0x0800 LOAD 0x01c000 0x0013e000 0x0013e000 0x0b03c 0x0b03c RWE 0x1000 segment 3 is not loadable, skipping Found an elf32 file called "roottask" located at address 0x009c1000 Program Headers: Type Offset VirtAddr PhysAddr FileSiz MemSiz Flg Align LOAD 0x001000 0x00150000 0x00150000 0x10032 0x131d0 RWE 0x1000 segment 1 is not loadable, skipping Loaded normal modules.. now the rest Found an elf32 file called "/boot/grub/ig_init" located at address 0x00a1a000 Program Headers: Type Offset VirtAddr PhysAddr FileSiz MemSiz Flg Align LOAD 0x001000 0x00170000 0x00170000 0x0e032 0x0e090 RWE 0x1000 segment 1 is not loadable, skipping Found an elf32 file called "/boot/grub/ig_serial" located at address 0x00a70000 Program Headers: Type Offset VirtAddr PhysAddr FileSiz MemSiz Flg Align LOAD 0x001000 0x00180000 0x00180000 0x0c032 0x0c0a4 RWE 0x1000 segment 1 is not loadable, skipping Found an elf32 file called "/boot/grub/ig_naming" located at address 0x00abe000 Program Headers: Type Offset VirtAddr PhysAddr FileSiz MemSiz Flg Align LOAD 0x001000 0x00190000 0x00190000 0x0a032 0x0a088 RWE 0x1000 segment 1 is not loadable, skipping Found an elf32 file called "/boot/grub/ig_trace" located at address 0x00b00000 Program Headers: Type Offset VirtAddr PhysAddr FileSiz MemSiz Flg Align LOAD 0x001000 0x001a0000 0x001a0000 0x0a032 0x0a088 RWE 0x1000 segment 1 is not loadable, skipping Found an elf32 file called "/boot/grub/ig_timer" located at address 0x00b42000 Program Headers: Type Offset VirtAddr PhysAddr FileSiz MemSiz Flg Align LOAD 0x001000 0x001b0000 0x001b0000 0x0b032 0x0b0ac RWE 0x1000 segment 1 is not loadable, skipping Found an elf32 file called "/boot/grub/vmlinux" located at address 0x00b8a000 Program Headers: Type Offset VirtAddr PhysAddr FileSiz MemSiz Flg Align LOAD 0x001000 0x001c0000 0x001c0000 0x1ec06c 0x20d0b8 RWE 0x1000 segment 1 is not loadable, skipping Found an elf32 file called "/boot/grub/ramdisk" located at address 0x00e56000 Invalid elf file (-1) Magic is: 00 00 00 00 Found an elf32 file called "/boot/grub/startup.script" located at address 0x0162 6000 Invalid elf file (-1) Magic is: 23 20 49 67 Launching kernel ... NICTA::Pistachio - built on Jan 20 2007 15:57:48 by jorgee@mio using gcc version 3.3.6 (Debian 1:3.3.6-13). Initializing boot memory (f011e000 - f013e000) kmem_init (f011e000, f013e000) [128K] remap area f0100000->00100000 len=3e000, w=1, k=1, size=0 remap area 0013e000->0013e000 len=c000, w=1, k=1, size=0 remap area 00000000->00000000 len=400000, w=1, k=1, size=0 activate_gdt remap area f7000000->07000000 len=1000000, w=1, k=1, size=0 activate_gdt Registering processor 0 in KIP (0MHz, 0MHz) startup_system done domain pairs: (0, 1) Initialising scheduler... Switching to idle thread initial_switch_to:552: Initial switch to: sp = f0113ff8 system has 16 interrupts Initializing root servers root-servers: utcb_area: bf000110 (128KB), kip_area: bff000c0 (4KB) creating root server (000c0001) idle thread started ***************************************** * Iguana system, Copyright 1996-2004 * * ERTOS - National ICT Australia * ***************************************** Unknown type: 258 Iguana init starting Initialise naming: Initialise trace: Initialise timer: Initialise serial: Trace server starting no kernel tracebuffer found Backing memory section: 0x001be000 Backed memory section Driver enable Done driver enable Initial servers started Loading linux wombat is at <0x001c0000:0x003ce000> starting at 0x001c0000 Instance: 0x0500e868 -- server: dc001 objref: 1c0000 Loading eCos Couldn't find: ecos Linux version 2.6.10-i386l4_pc99 (jorgee@mio) (gcc version 4.0.4 20060507 (prere lease) (Debian 4.0.3-3)) #4 Sat Jan 20 16:00:03 COT 2007 Kernel memory ranges: 0: 0x03000000-0x04000000 (4096 pages) total 4096 pages Built 4 zonelists Kernel command line: vmlinux igms_name=ramdisk root=/dev/igms0 PID hash table entries: 128 (order: 7, 2048 bytes) Dentry cache hash table entries: 4096 (order: 2, 16384 bytes) Inode-cache hash table entries: 2048 (order: 1, 8192 bytes) Memory: 16088k/16400k available (1336k kernel code, 200k reserved, 332k data, 30 0k init) Mount-cache hash table entries: 512 (order: 0, 4096 bytes) NET: Registered protocol family 16 PCI: direct init pci_direct_init: calling request_region(0xCF8, 8, "PCI conf1") pci_direct_init: request_region returned 0x0301fe60 calling pci_check_type1() PCI: Using configuration type 1 PCI: Probing PCI hardware PCI: Probing PCI hardware (bus 00) PCI: pcibios_fixup_bus : Limiting direct PCI/PCI transfers. PCI: PIIX3: Enabling Passive Release on 0000:00:01.0 Activating ISA DMA hang workarounds. RAM (4096 kB) at 0xxe0000000, Cirrus Logic chipset on PCI bus cirrusfb: Driver for Cirrus Logic based graphic boards, v2.0-pre2 igvfb alloc = 00020000 fb1: Virtual frame buffer device, using 75K of video memory io scheduler noop registered io scheduler anticipatory registered RAMDISK driver initialized: 16 RAM disks of 4096K size 1024 blocksize loop: loaded (max 8 devices) ne2k-pci.c:v1.03 9/22/2003 D. Becker/P. Gortmaker http://www.scyld.com/network/ne2k-pci.html eth0: RealTek RTL-8029 found at 0xc100, IRQ 10, 52:54:00:12:34:56. Uniform Multi-Platform E-IDE driver Revision: 7.00alpha2 ide: Assuming 33MHz system bus speed for PIO modes; override with idebus=xx hda: QEMU HARDDISK, ATA DISK drive hda: IRQ probe failed (0xffffffff) hdb: IRQ probe failed (0xffffffff) hdb: IRQ probe failed (0xffffffff) hdc: QEMU CD-ROM, ATAPI CD/DVD-ROM drive hdc: IRQ probe failed (0xffffffff) hdd: IRQ probe failed (0xffffffff) hdd: IRQ probe failed (0xffffffff) elevator: using anticipatory as default io scheduler ide0 at 0x1f0-0x1f7,0x3f6 on irq 14 ide1 at 0x170-0x177,0x376 on irq 15 hda: max request size: 1024KiB hda: 88704 sectors (45 MB) w/256KiB Cache, CHS=88/255/63 hda: hda1 hdc: ATAPI 4X CD-ROM drive, 512kB Cache Uniform CD-ROM driver Revision: 3.20 igms0: unknown partition table NET: Registered protocol family 2 IP: routing cache hash table of 512 buckets, 4Kbytes TCP: Hash tables configured (established 1024 bind 2048) NET: Registered protocol family 1 NET: Registered protocol family 17 VFS: Mounted root (ext2 filesystem) readonly. syscall 243 init:1:e4001 unknown syscall syscall 243 init:1:e4001 unknown syscall syscall 243 init:1:e4001 unknown syscall syscall 243 init:1:e4001 unknown syscall syscall 243 init:1:e4001 unknown syscall syscall 243 init:1:e4001 unknown syscall syscall 243 init:1:e4001 unknown syscall syscall 243 init:1:e4001 unknown syscall syscall 243 init:1:e4001 unknown syscall syscall 243 init:1:e4001 unknown syscall syscall 243 init:1:e4001 unknown syscall What can i do to fix this? -- Jorge Eduardo Cardona jorgeecardona(a)gmail.com ------------------------------------------------ Linux registered user #391186 Registered machine #291871 ------------------------------------------------

1 0

GRUB and modaddr
by Manuel A. Fernandez Montecelo 17 Jan '07

17 Jan '07

Hello, If I understood it correctly (stumbling upon a memory overlapping error and searching the web), that 'modaddr' command (or similar name, don't have it around now) exists only in a fork of GRUB maintained to boot Fiasco (and maybe other Dresden/L4 projects). Is there a way to run Fiasco with standard GRUB, or any ongoing work to be able to do it? Regards. -- Manuel A. Fernandez Montecelo <manuel.montezelo(a)gmail.com>

5 6

Fiasco lazy-scheduling question
by eschierb＠sci.kun.nl 17 Jan '07

17 Jan '07

Hi, I'm Erik Schierboom and I study at the Radboud University Nijmegen. Right now, I'm writing my master thesis, under supervision of Hendrik Tews, on the formal verification of a part of Fiasco's IPC implementation. I studied the thesis by René Reusner but I don't fully grasp the lazy-scheduling details. I have the following questions. For one, the description says that the sender is not removed from the ready-list when the switch is made to the receiver (even though it is now in waiting status). However, as the sender is the active thread it is not part of the ready-list by definition (at least that is how I interpreted the system), so how could it be removed anyway? Secondly, the switching from sender to receiver is simply an execution context switch. If the receiver finishes before the sender's time is up, does the receiver switch the execution context back to that of the sender? With regards, Erik

2 1

Re: some problems of when compiling IDL with dice tools
by Ronald Aigner 17 Jan '07

17 Jan '07

-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 CC'd to l4-hackers, so the most probable solution is in the archives :) chenggh wrote on 17.01.2007 12:04 this: > On Wednesday 17 January 2007 17:04, you wrote: >> chenggh wrote on 17.01.2007 04:52 this: >>> Hi: >>> In the idl file i defined an interface including a function declaration >>> like this: >>> int foo([in] arg1, [in] void *( * start_routine)(void *), [in] arg2); >>> >>> My aim is to define a method which includes a function pointer I define >>> the function name and function parameters by myself. But when I compile >>> there is error. >> Naturally. >> >>> I checked out the dice manual there is nothing about how to deal with >>> function pointer in idl file. Is this style supported by dice IDL ? If >>> it supports how could I do? >> No, don't do it! >> >> Dice is an IDL compiler. IDL stands for Interface Definition Language. >> And IDL is *NOT* the C programming language. Even though IDL looks >> similar to C it is not C. So using function pointers as arguments in an >> interface definition won't work. > > If I want to define a method including a function as a parameter in IDL file > how can I do? You will have to find another way of expressing a call-back function, e.g., using addresses which you convert into a function pointer wherever you want to invoke the function. You could write something like: int foo([in] arg1, [in] unsigned long * addr_func, [in] arg2); (unsigned long* is used here as the type describing a memory location) and then in the component function: /* I did not test this. Maybe it compiles and works, maybe it doesn't. */ foo_component(arg1, unsigned long * addr_func, arg2) { void *(*start_routine)(void*) = addr_func; ... start_routine(...); ... } However, you usually use IDL to describe an interface for a service that runs in a *different* address space. Sending a pointer to a function to that server is useless, because it has no access to that function located in the client's address space. Then again, if the function is located in a shared memory region you could do the above, but would have to "translate" the address from the client's address space location to the server's address space location. Or client and server are located in the same address space. Then you could use the method described above. Greetings, Ron. - -- Mit freundlichen Gruessen / with regards ra3 @ inf.tu-dresden.de http://os.inf.tu-dresden.de/~ra3/ -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.4.6 (GNU/Linux) iD8DBQFFrg9m2sAx3Dtx9+YRApzmAJ92VK5ke3N1wfsLD7Aktp8cVhiR/QCggnS5 DU9EjVEN9u4A+bF9G6b7fSs= =9Noj -----END PGP SIGNATURE-----

1 0

Re: some problems of when compiling IDL with dice tools
by Bjoern Doebel 17 Jan '07

17 Jan '07

-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 chenggh wrote: > On Wednesday 17 January 2007 15:44, you wrote: >> chenggh wrote: >>> Hi: >>> In the idl file i defined an interface including a function declaration >>> like this: >>> int foo([in] arg1, [in] void *( * start_routine)(void *), [in] arg2); >>> >>> My aim is to define a method which includes a function pointer I define >>> the function name and function parameters by myself. But when I compile >>> there is error. >>> >>> I checked out the dice manual there is nothing about how to deal with >>> function pointer in idl file. Is this style supported by dice IDL ? If >>> it supports how could I do? >> What exactly are you trying to do? > I try to encapsulate the pthread function with some l4 thread functions. In > Linux the pthread definition is like this: > int pthread_create (pthread_t *__restrict __threadp, > __const pthread_attr_t *__restrict __attr, > void *(*__start_routine) (void *), > void *__restrict __arg) __THROW; > So in IDL file I define the similar style but compilation has some errors. > Maybe I must consider other solutions. Why do you need an IDL interface for starting a thread? Can't this just be implemented as a C function? You can have a look at l4/pkg/thread to see how the L4 thread library is implemented. Bjoern -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.4.5 (GNU/Linux) Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org iD8DBQFFrgu/P5ijxgQLUNkRAjUNAJ4h4VwnevgtTlxTXnYaz2s7/SwHoACfRjmj ExnO96FlWr/PL7I0fDYNRa8= =9dYs -----END PGP SIGNATURE-----

1 0

some problems of when compiling IDL with dice tools
by chenggh 17 Jan '07

17 Jan '07

Hi: In the idl file i defined an interface including a function declaration like this: int foo([in] arg1, [in] void *( * start_routine)(void *), [in] arg2); My aim is to define a method which includes a function pointer I define the function name and function parameters by myself. But when I compile there is error. I checked out the dice manual there is nothing about how to deal with function pointer in idl file. Is this style supported by dice IDL ? If it supports how could I do? Thanks. cheng

3 2

arm instruction version
by WanKeun Jo 17 Jan '07

17 Jan '07

Hi, In the l4linux-2.6, if set L4ARCH=arm, the processor type is fixed to CPU_SA1100. And the arm instruction version is armv4. I want to use arch-dependent code for pxa, so I changed CPU_SA1100 to CPU_XSCALE config and arm instruction set version to armv5te at --------------------------------------------------- $ l4linux-2.6/arch/l4/Makefile CFLAGS += -D__LINUX_ARM_ARCH__=5 -march=armv5te --------------------------------------------------- and found a code, --------------------------------------------------- $ l4linux-2.6/arch/l4/kernel/arch-arm/setup.c unsigned int processor_id = 0x0001f000; /* ARMv4 */ --------------------------------------------------- The processor_id is fixed to ARMv4 value. My question is 001. How processor_id affect the linux?? Should I also change this value to a armv5te value?? 002. If I have to change it, what is the proper value for armv5te?? I can't find the value anywhere. Thank you!!!

2 1

AW: Documentation for L4Linux/L4Env
by Christian, Martin, OPEE45 16 Jan '07

16 Jan '07

Hi Adam! > maybe the KMB lecture slides are somehow helpful. They are a good starting point to understand the principles. However the mapping to the code is still unclear. Wouldn't it be possible to get a student (HiWi) writing some documentation? > You should have some good understanding of Linux internals though Yes. And that's exactly my problem ... :-( > understand what the important code does. First understand what the important code is. > Besides that you can of course ask questions and I can try to answer them. >From your lecture I understand that a new architecture would require mainly changes to the "Lower Half" of L4Linux. But I suppose there are also changes required to the Syscall Interface. However, since I want to port it to another L4 Interface, I'll also have to rewrite a lot more. So I'm tending more to use either the L4Linux-2.4 (where there is no documentation) or Wombat, which at least builds on the L4.X2 API. Anyways, I fear I won't manage to do it within 6 weeks ... :-( Thanks for the help, Martin.

2 1