l4-hackers December 2010

l4-hackers@os.inf.tu-dresden.de

9 participants
8 discussions

Problems booting Pistachio
by Ashford Nichols 26 Dec '10

26 Dec '10

I recently built the microkernel and tried to boot it using "qemu" but it doesn't work. qemu just seems to not even seem to recognize the image as a bootable image. I followed the instructions given on the site and everything worked, right up to the point where it doesn't boot. I must say though that I did slightly tweak the steps given on the site to create the image file since the original ones didn't work for me. I am really new to qemu so I don't have the slightest clue as to why this isn't working. If anyone experienced this problem or at least know how to resolve this please help me. The step I followed in order to create the image follows: dd if=/dev/zero of=l4-image.img bs=512 count=8640mke2fs l4-image.imgmkdir /media/mount_pointmount -o loop l4-image.img /media/mount_pointmkdir /media/mount_point/bootmkdir /media/mount_point/boot/grubcp -aR grub_stages/* /media/mount_point/boot/grubcp -aR modules/* /media/mount_pointumount /media/mount_point

1 0

L4Re: Thread local storage
by Taru Karttunen 22 Dec '10

22 Dec '10

Hello The following program crashes: #include <stdio.h> #include <pthread.h> __thread int i; int main(void) { puts("Hello World!"); printf("address of i=%p\n", &i); printf("i=%d\n", i); return 0; } with: MOE: loading 'rom/hello2' Hello World! address of i=0xb30001fc L4Re: unhandled exception: pc=0x100013c How should TLS initialization be done? This is with L4_MULTITHREADED = y Dump of assembler code for function main: 0x01000100 <+0>: push %ebp 0x01000101 <+1>: mov %esp,%ebp 0x01000103 <+3>: and $0xfffffff0,%esp 0x01000106 <+6>: sub $0x10,%esp 0x01000109 <+9>: movl $0x1017b20,(%esp) 0x01000110 <+16>: call 0x100dab0 <puts> 0x01000115 <+21>: movl $0x1,(%esp) 0x0100011c <+28>: call 0x100f9a0 <sleep> 0x01000121 <+33>: mov %gs:0x0,%eax 0x01000127 <+39>: movl $0x1017b2d,(%esp) 0x0100012e <+46>: add $0xfffffffc,%eax 0x01000133 <+51>: mov %eax,0x4(%esp) 0x01000137 <+55>: call 0x100da80 <printf> 0x0100013c <+60>: mov %gs:0xfffffffc,%eax 0x01000142 <+66>: movl $0x1017b3e,(%esp) 0x01000149 <+73>: mov %eax,0x4(%esp) 0x0100014d <+77>: call 0x100da80 <printf> 0x01000152 <+82>: xor %eax,%eax 0x01000154 <+84>: leave 0x01000155 <+85>: ret So how to get this working? ps. I have a rudimentary port of gccgo on L4Re working, and am polishing it over the holidays. - Taru Karttunen

3 2

schedule system call
by Ashford Nichols 20 Dec '10

20 Dec '10

In version 2 of the interface under the schedule system call, it says that the call can be used be schedulers. However, in the next sentence it goes on to say that the calling thread must reside in the same address space as the scheduler. Also, it goes on to say that an error code will be set if the calling thread is not the scheduler. So is the "schedule" system call made by the scheduler or by a thread that simply resides in the same address space as the scheduler?

1 0

L4 pager
by Ashford Nichols 19 Dec '10

19 Dec '10

I am new to L4 microkernels and I am reading the reference manual on version 2 of the interface. From what I've read so far the manual suggests that there's more than one pager, but if this is so what is the purpose of more than one pager?

2 2

GCC 4.5.1 and L4Re
by Taru Karttunen 17 Dec '10

17 Dec '10

Hello Has anyone gotten L4Re to work with GCC 4.5.1 on amd64? I am having issues with just a simple hello world modules.list: entry taru roottask moe --debug=all --l4re-dbg=all --init=rom/hello module l4re module hello Grub loads all the modules ok, but then the qemu just exits without anything being printed from hello. Any hints would be appreciated. - Taru Karttunen

2 3

Several questions
by Nuno Santos 10 Dec '10

10 Dec '10

Hi I've managed to build and run Fiasco with one instance of L4Linux. At this point I have several questions: - I believe that an application running in l4linux can communicate with an application sitting on fiasco. Is there a simple example I could look at? - As a follow up, how can we allow multiple l4linux instances to communicate using IPC? This post from 2005 [1] has done precisely what I need. Would it be possible to share that code? - I'd like to use and deploy a tpm driver directly on top of L4 (not directly accessible to l4linux). Is it possible to obtain that driver? How to set it up? This post [2] mentions the stpm package, but I couldn't find it in the l4re snapshot. - Is it possible to plug a TPM emulator to work directly on fiasco? It's mostly for experimentation and testing on machines without a physical TPM. - I believe that it's possible to boot fiasco with a trusted bootloader. [2] mentions oslo, but again I couldn't find the code in the snapshot. Do you know where can I find it? Do you know if trusted grub can be used as a viable alternative? I've looked extensively for online information, but I confess I'm a bit overwhelmed. Your guidance here would be fantastic. I really appreciate your help. Thanks! --Nuno [1] http://www.mail-archive.com/l4-hackers@os.inf.tu-dresden.de/msg01709.html [2] http://www.mail-archive.com/l4-hackers@os.inf.tu-dresden.de/msg03861.html

3 2

Problem with building L4Linux
by Nuno Santos 07 Dec '10

07 Dec '10

Hi I'm having trouble with building l4linux. First, I had the same problem as reported in this [1] post; then I tried to build l4linux from the latest snapshot as suggested in [2] by following the steps in the readme file ('make setup' and 'make'). If I setup the target as x86-32, at some point the build process fails with the following output (without further description of the error): " make[6]: Leaving directory `/local/l4re-snap/l4re-snapshot-2010111519/src/l4/pkg/valgrind/build' make[5]: Leaving directory `/local/l4re-snap/l4re-snapshot-2010111519/src/l4/pkg/valgrind' make[4]: Leaving directory `/local/l4re-snap/l4re-snapshot-2010111519/src/l4/pkg' make[3]: *** [pkg] Error 2 make[3]: Leaving directory `/local/l4re-snap/l4re-snapshot-2010111519/src/l4' make[2]: *** [x] Error 2 make[2]: Leaving directory `/local/l4re-snap/l4re-snapshot-2010111519/obj/l4/x86' Error building the L4Re 'obj/l4/x86' variant. " If I setup to the amd64 target, the build process fails with a compilation error: " /tmp/camlasm564f4e.s: Assembler messages: /tmp/camlasm564f4e.s:580: Error: junk `lazy_force(%rip)' after expression /tmp/camlasm564f4e.s:580: Error: suffix or operands invalid for `lea' /tmp/camlasm564f4e.s:785: Error: junk `lazy_force(%rip)' after expression /tmp/camlasm564f4e.s:785: Error: suffix or operands invalid for `lea' Assembler error, input left in file /tmp/camlasm564f4e.s make[8]: *** [stream.cmx] Error 2 ... make[4]: Leaving directory `/local/l4re-snap/l4re-snapshot-2010111519/src/l4/pkg' make[3]: *** [pkg] Error 2 make[3]: Leaving directory `/local/l4re-snap/l4re-snapshot-2010111519/src/l4' make[2]: *** [x] Error 2 make[2]: Leaving directory `/local/l4re-snap/l4re-snapshot-2010111519/obj/l4/amd64' Error building the L4Re 'obj/l4/amd64' variant. " The problematic assembly instruction is: leaq $25lazy_force(%rip), %rax I'm using the snapshot from Nov 15th, on a machine running linux 2.6.32, gcc 4.3.2. Any ideas on how to fix it? Thanks for your help! --Nuno [1] http://www.mail-archive.com/l4-hackers@os.inf.tu-dresden.de/msg04380.html [2] http://www.mail-archive.com/l4-hackers@os.inf.tu-dresden.de/msg04383.html

3 3

Stringitem
by BERTRAND Joel 04 Dec '10

04 Dec '10

Hello, I think I have misunderstood StringItem IPC but I don't know where is my mistake. My roottask spawns a thread that sends a stringitem to roottask. I have checked with debugger that this stringitem contains my string : si = L4_StringItem(str.length_trim, (void *) str.c); L4_Clear(&msg); L4_Append(&msg, si); L4_Set_Label(&msg, CALL$PRINT); L4_Load(&msg); L4_Call(L4_Pager()); str is a struct vms$string : struct vms$string { vms$pointer length; vms$pointer length_trim; unsigned char *c; }; My roottask receives this IPC in following function: void sys$loop() { int running; L4_ThreadId_t partner; L4_MsgBuffer_t buffer; L4_MsgTag_t tag; L4_Msg_t msg; running = 1; L4_Accept(L4_MapGrantItems(L4_CompleteAddressSpace) + L4_StringItemsAcceptor); tag = L4_Wait(&partner); while(running) { L4_Clear(&msg); L4_Store(tag, &msg); if ((tag.raw & L4_REQUEST_MASK) == L4_PAGEFAULT) { sys$pagefault(partner, L4_Get(&msg, 0), L4_Get(&msg, 1), tag.raw); } else { switch(L4_Label(tag)) { case CALL$PRINT: notice("Here!\n"); break; default: PANIC(1, notice(IPC_F_UNKNOWN "unknown IPC from $%lX " "with label $%lX\n", L4_ThreadNo(partner), L4_Label(tag))); } } tag = L4_ReplyWait(partner, &partner); } return; } When system boots, I can intercept on serial line: ... %SYS-F-PAGEFLT, pagefault request from $3C at $0000000000800110 %SYS-F-PAGEFLT, pagefault request from $3C at $000000000010FFF8 %SYS-F-PAGEFLT, pagefault request from $3C at $00000000008015C0 %SYS-F-PAGEFLT, pagefault request from $3C at $0000000000801C10 Here! %IPC-F-UNKNOWN, unknown IPC from $0 with label $0 Panic at sys$loop, sys/sys_loop.c line 73 Have a nice day ! Backtrace: <00> [$000000000104F5B8] -> $0000000001004ED5 (sys$loop) <01> [$000000000104F838] -> $0000000001005813 (main) Pagefault messages are written by sys$pagefault() subroutine. As sys$loop() function writes 'Here!', it receives an IPC with CALL$PRINT label. First question: I don't know how read this stringitem. I have tried with msgbuff without any result. I suppose I have misunderstood how StringItem IPC works. Second question: I don't understand last line: "%IPC-F-UNKNOWN, unknown IPC from $0 with label $0". I suppose this error comes from a bad stringitem IPC. I haven't found any stringitem IPC example. Help is welcome. Regards, JB

1 1

2024

2023

2022

2021

2020

2019

2018

2017

2016

2015

2014

2013

2012

2011

2010

2009

2008

2007

2006

2005

2004

2003

2002

2001

2000

1999

1998

l4-hackers December 2010