l4-hackers

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

2827 discussions

Re: Fwd: attempted fix for asm in oskit cpuid.c
by spooq 12 Apr '06

12 Apr '06

I think this list needs to be fixing its reply-tos... I wasnt the only one caught by it... On 4/12/06, Steffen Liebergeld <s1010824(a)inf.tu-dresden.de> wrote: > On Tue, 11 Apr 2006 23:14:08 +0100, spooq <spoooq(a)gmail.com> said: > > *sigh* ---------- Forwarded message ---------- From: spooq > > <spoooq(a)gmail.com> Date: Apr 11, 2006 11:12 PM Subject: Re: > > attempted fix for asm in oskit cpuid.c To: Ronald Aigner > > <ra3(a)os.inf.tu-dresden.de> > > > > I see dietlibc's printf is not given a good review on > > http://www.and.org/vstr/printf_comparison... > > That comparison is very old, and they tested version 0.22 of > dietlibc. The current version is 0.29 (have a look at > http://www.fefe.de/dietlibc/). > > -- > Microsoft is simply one example of a proprietary software developer, a > software developer that tries to subjugate users to keep them divided > and helpless. -- Richard M. Stallman >

2 1

attempted fix for asm in oskit cpuid.c
by spooq 11 Apr '06

11 Apr '06

Hello list, I tried to fix some errors in cpuid.c ... wouldnt rely on it, my asms not very good and I havent tested it beyond seeing that it compiles. Basically it just makes the asm respect gcc using the ebx as a pic reg. Not sure about the \n\t stuff, it made it work tho. --- virgin_code/oskit/libkern/x86/cpuid.c 2006-04-02 15:41:55.000000000 +0000 +++ cpuid.c 2006-04-10 17:14:57.000000000 +0000 @@ -39,12 +39,17 @@ unsigned i; cicount++; - asm volatile("cpuid" + asm volatile( + "movl %%ebx, %%edi\n\t" + "cpuid\n\t" + "movl %1, %%ebx\n\t" + "movl %%edi, %%ebx\n\t" : "=a" (ci[0]), - "=b" (ci[1]), + "=r" (ci[1]), "=c" (ci[2]), "=d" (ci[3]) - : "a" (2)); + : "a" (2) + : "edi"); for (i = 0; i < 4; i++) { @@ -90,21 +95,29 @@ * CPUID is supported, so use it. * First get the vendor ID string. */ - asm volatile("cpuid" + asm volatile( + "movl %%ebx, %%edi\n\t" + "cpuid\n\t" + "movl %1, %%ebx\n\t" + "movl %%edi, %%ebx\n\t" : "=a" (highest_val), - "=b" (*((int*)(out_id->vendor_id+0))), + "=r" (*((int*)(out_id->vendor_id+0))), "=d" (*((int*)(out_id->vendor_id+4))), "=c" (*((int*)(out_id->vendor_id+8))) - : "a" (0)); + : "a" (0) + : "edi"); /* Now the feature information. */ if (highest_val >= 1) { - asm volatile("cpuid" - : "=a" (*((int*)out_id)), - "=d" (out_id->feature_flags) - : "a" (1) - : "ebx", "ecx"); + asm volatile( + "movl %%ebx, %%edi\n\t" + "cpuid\n\t" + "movl %%edi, %%ebx\n\t" + : "=a" (*((int*)out_id)), + "=d" (out_id->feature_flags) + : "a" (1) + : "ecx","edi"); } /* Cache and TLB information. */ -- spooq

2 1

Static Linux kernel, anyone?
by Stefan Schulz 09 Apr '06

09 Apr '06

Hi there I am currently trying to build a static linux kernel (one usable without libloader). I have read the earlier thread in this mailing list, but have run into some problems. I tried adding libraries to fix undefined references, since merely adding the static dependencies of libloader.s.so did not work. Finally, i hit upon references such as __l4sys_abs_ipc_fixup, which frankly looks like something i should not have to care about. ;) A quick scan through the libraries also seemed to indicate that it isn't even exported by any library. So, what am i doing so fundamentally wrong? Is there a way to build a static linux binary, or is it just intrinsically tied to the loader?

2 1

L4Linux and AMD64
by Zephiris 09 Apr '06

09 Apr '06

I haven't been looking around at things very long, but I was noticing, while Fiasco largely supports AMD64 processors in 64-bit mode, L4Linux appears to specifically avoid it. I couldn't find any references for this, will it ever be supported? It's something I was excited about/meaning to try, but as my system is setup with 64-bit binaries in mind, it looks like the only way to test it would be for L4Linux to support that.

2 1

Re: Building the environment fails
by Alexander Boettcher 05 Apr '06

05 Apr '06

1 0

Re: Building the environment fails
by Alexander Boettcher 05 Apr '06

05 Apr '06

1 0

Building the environment fails
by Stefan Schulz 05 Apr '06

05 Apr '06

Sorry for using this ML as my personal helpdesk, but i believe there is a problem in the build process for l4env. I checked the current version out from CVS, and used "make config" to configure. While "make" was running, the process failed with ... Compiling ARCH-all/libc/misc/internals/__uClibc_main.o LD_PRELOAD=/l4/tmp/l4/tool/gendep/libgendep.so GENDEP_TARGET=ARCH-all/libc/misc/internals/__uClibc_main.o GENDEP_BINARY=cc1 gcc -c -nostdinc -include ../../../lib/uclibc/ARCH-all/include/libc-symbols.h -DNDEBUG -D_LIBC -DSTATIC -DSYSTEM_x86_586 -DARCH_x86 -DCPUTYPE_586 -DL4API_ -DCONFIG_L4_CALL_SYSCALLS -DCONFIG_L4_ABS_SYSCALLS -I../ARCH-x86/include -I../ARCH-all/include -I../../../../../include/x86 -I../../../../../include -I/usr/lib/gcc/i686-pc-linux-gnu/3.4.5/include -fno-builtin -gstabs+ -g -O2 -fno-strict-aliasing -march=i586 -Wall -Wstrict-prototypes -fno-common /l4/tmp/l4/pkg/uclibc/lib/uclibc/ARCH-all/libc/misc/internals/__uClibc_main.c -o ARCH-all/libc/misc/internals/__uClibc_main.o /l4/tmp/l4/pkg/uclibc/lib/uclibc/ARCH-all/libc/misc/internals/__uClibc_main.c:2:49: l4/sys/consts.h: No such file or directory /l4/tmp/l4/pkg/uclibc/lib/uclibc/ARCH-all/libc/misc/internals/__uClibc_main.c:15: error: `L4_PAGESIZE' undeclared here (not in a function) make[4]: *** [ARCH-all/libc/misc/internals/__uClibc_main.o] Error 1 make[4]: Leaving directory `/l4/tmp/l4/pkg/uclibc/lib/uclibc/OBJ-x86_586' make[3]: *** [OBJ-x86_586] Error 2 make[3]: Leaving directory `/l4/tmp/l4/pkg/uclibc/lib/uclibc' make[2]: *** [virt_dep] Error 2 make[2]: Leaving directory `/l4/tmp/l4/pkg/uclibc' make[1]: *** [uclibc.headers] Error 2 make[1]: Leaving directory `/l4/tmp/l4/pkg' make: *** [pkg] Error 2 Even if i "fixed" this error (by manually defining L4_PAGESIZE", similar problems crept up all over the build, far too many to manually fix. I tried this build on an IBM T43, using current sources, gcc3.4.5 and make 3.80 Is the build process really broken, or am i doing something wrong?

1 0

Make hello stops with an error
by Ricard de Benito Panillo 05 Apr '06

05 Apr '06

Hi, the following error arises making 'hello': cc1: note: obsolete option -I- used, please use -iquote instead In file included from ../libkern/x86/base_cpu_init.c:22: ../flux/x86/base_trap.h:105: error: array type has incomplete element type make[3]: *** [base_cpu_init.o] Error 1 gcc version 4.0.2 20051125 (Red Hat 4.0.2-8) on Fedora Core 4 Any hint? it seems that the compiler versions is newer than expected. Best regards, Ricard de Benito

2 1

I have new problem!
by chenggh04＠st.lzu.edu.cn 04 Apr '06

04 Apr '06

>Can you please give more details on what you did? I cannot really say >anything helpful on a generic "it doesn't work" statement. Thank adam lackorzynski(<adam(a)os.inf.tu-dresden.de>) for your help. Now I resolve last problem and the compilation fiasco in slackware (last time is debian) is ok. The execution of hello example is also good! Now I am ready for experiencing the fiasco-ux. I compile the package but runs badly. The error information is as follows: Fiasco-UX on Linux 2.4.31 (i686) Mapped 64 MB Memory + 0 KB Framebuffer + 0 MB Input Area on FD 3 Loading Module 0x00090000-0x00094e60 [sigma0-ux] Failed to open file: roottask But according to the instructions of Fiasco-ux website the information should be like this: Fiasco-UX on Linux 2.4.31 (i686) Mapped 64 MB Memory + 0 KB Framebuffer + 0 MB Input Area on FD 3 Loading Module 0x00070000-0x000955c0 [sigma0-ux] Loading Module 0x00100000-0x001b9ce0 [rmgr-ux] ... The addresses of the sigma0-ux are different! My configuration of fiasco is as follows: Target System Options >(UX) Target Platform (IA-32) Target CPU Family (K7) Target processor >(L4V2) Kernel ABI Version [ ] TCB size of 4k [ ] Fine-grained CPU time [Y] Graphical console (requires SDL library!) [Y] Network support Kernel Debugging >[Y] Generate inline code [ ] Do not compile assertions [Y] Compile without frame pointer [ ] Measure stack depth of threads [ ] Sanity checks in low level allocator [ ] Sanity checks at syscall entry/exit [Y] Compile with gstabs+ debugging information [Y] JDB kernel debugger [ ] JDB extended logging [ ] JDB accounting [ ] Miscellaneous JDB modules (ANY) Runtime warning level Compiling and Building >gcc C compiler g++ C++ compiler [ ] Verbose build process [ ] Do additional checks at build time and then make VERBOSE=@ according to the instruction of Fiasco-ux website but I don't find the option "Assembler IPC shortcut" . Please help me! Cheng

4 3

Why don't I have /bin/ directory in L4
by chenggh04＠st.lzu.edu.cn 03 Apr '06

03 Apr '06

Hi,Everybody: I am a beginner for L4. when I compile the L4 without any errors but finally I don't have the bin directory which inlcudes hello, rmgr and sigma0. Maybe I think there is lack of some packages in my install package. But the package is downloaded from the Fiasco official WebSite. I don't know why! Please help me! Chenggh

2 1

Jump to page:

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