similar to: What's next for klibc...

Peter, current klibc does not compile in ia64: gcc -Wp,-MT,arch/ia64/pipe.o,-MD,arch/ia64/.pipe.o.d -Os -fomit-frame-pointer -nostdinc -iwithprefix include -D__KLIBC__ -DBITSIZE=64 -I../include/arch/ia64 -I../include/bits64 -I../include -I../linux/include -I../linux/include2 -I../linux/include -DWITH_ERRLIST -W -Wall -Wpointer-arith -Wwrite-strings -Wstrict-prototypes -Winline -c -o

Peter, we found that nfs over udp will corrupt data under very extrem load, there is no way to fix it due to the way how UDP works. TCP will not have these problems. I also wonder why the package size is only 1k. Everyone who wishes a slow connection can pass the desired options via the kernel cmdline. Everyone else prefers probably the fast mount. The defaults should look more like that: ---

OK, I have just pushed out klibc-0.130; it significantly updates the run-init program (again... PLEASE TEST IN A SAFE ENVIRONMENT) based on, but not identical to, Alexander's suggestions; it also adds s390/x patches from Arnd Bergmann. -hpa

Hello, I would like to know if initrd is here to stay, now that klibc and initramfs are ready. As the multipath-tools maintainer, I'm facing the choice to 1) put the multipath configuration tool in the initrd * dynamic binary is possible * storage hba drivers as modules loaded * no klibc limitations (no mntent for libsysfs ...) 2) put the multipath configuration tool in the initramfs *

Hold on to your hats, because the world is about to break... I have been trying to figure out for a while now how to migrate klibc to a 64-bit off_t, and to handle other things like getuid32(). The final decision is that the _syscall*() macros just got to go; they aren't flexible enough, and the hope that they would make it easier to port between platforms hasn't worked out -

Marcus Meissner did the ppc64 port, I did the update for crt0.S. Now ash does not segfault, at least. diff -p -purNx linux klibc-0.92.orig/Makefile klibc-0.92/Makefile --- klibc-0.92.orig/Makefile 2003-12-01 18:50:24.000000000 +0000 +++ klibc-0.92/Makefile 2003-12-12 23:18:23.000000000 +0000 @@ -20,4 +20,4 @@ clean: spotless: @set -e; for d in $(SUBDIRS); do $(MAKE) -C $$d $@; done - rm -f

some files depend on nodes.h, but there is no dependency for that fact. olaf@mandarine:~/klibc-0.87> grep -Er 'include.*nodes.h' ash/ | sed 's-^[^/]\+/\([^.]*\).*-\1.o-' | xargs echo OBJ_NODES = OBJ_NODES = cd.o eval.o exec.o expand.o jobs.o main.o nodes.o options.o parser.o redir.o show.o trap.o var.o make spotless doesnt work reliable, the patch adds a few more targets. Is

The man page for mmap() says the size needs to be a multiple of getpagesize(). So in my code I call getpagesize() and adjust my lengths accordingly. klibc doesn't have getpagesize() implemented. Is it safe to call mmap() with arbritary lengths and assume the mmap will round up for you? Does this work on glibc too? If so, I can just eliminate my getpagesize() code. It turns out for VM86

Hi Peter, I need this small patch to get udev to compile against the latest klibc. I'm not sure if this is the right thing to do, but it fixes the compile. thanks mh -- Martin Hicks Wild Open Source Inc. mort@wildopensource.com 613-266-2296 # This is a BitKeeper generated patch for the following project: # Project Name: The kernel C library # This patch format is

I think the main bit of the puzzle that's missing for klibc integration at this point, other than the kbuild bits, is a kinit that's a true replacement for prepare_namespace() and everything south of it in init/main.c. This means that it should be able to replace all that code and any code that's removed from it, totalling about 6000 lines of kernel code. The kinit that's

I have pushed a new klibc which uses a modern NetBSD-derived ash. The initial port was done by Olaf, and I've mostly applied the chainsaw to it until it was less than 50% larger than the old one. It's still substantially larger, but should also eliminate need for some external binaries like expr and test, which are otherwise ubiquitous for any sensible shell scripts, and perhaps most

I'm not sure what the root cause is, but printf writes to fd 0: mandarine:/usr/src/packages/BUILD/udev-015/klibc# ./utils/true blah mandarine:/usr/src/packages/BUILD/udev-015/klibc# ./utils/true >/dev/null blah mandarine:/usr/src/packages/BUILD/udev-015/klibc# cat utils/true.c #include <stdio.h> int main(void) { printf("blah\n"); return 0; } -- USB is for mice,

nfsmount doesnt work anymore for me. This patch fixes at least udp nfs mounts. tcp does still hang. client and server is 2.6. --- /dev/shm/bindresvport.c 2003-12-08 02:33:24.000000000 +0100 +++ klibc/inet/bindresvport.c 2004-01-06 21:05:02.283779072 +0100 @@ -23,7 +23,7 @@ int bindresvport(int sd, struct sockaddr if (sin == NULL) { sin = &me;

Hi!! I found your ipconfig stuff in the udev rpm in SuSE 9.1. I'm about to setup a initramfs (pretty cool - much better than initrd) to boot linux diskless clients over the net. All the tools I would need, seem to be there with ipconfig/nfsmount ... I like to use the information I get from PXE/syslinux or Etherboot writing something like:

make spotless leaves 2 generated files. diff -purN klibc-0.159.orig/klibc/Makefile klibc-0.159/klibc/Makefile --- klibc-0.159.orig/klibc/Makefile 2004-08-03 23:07:05.000000000 +0200 +++ klibc-0.159/klibc/Makefile 2004-08-13 22:23:35.696699671 +0200 @@ -156,6 +156,7 @@ clean: archclean rm -f sha1hash errlist.c spotless: clean + rm -f include/klibc/havesyscall.h syscalls.nrs find . \( -name

Hi, after changing mksh to no longer use mkstemp, I decided to try and build mksh-static with klcc on most platforms, in Debian. Here are the results by Debian architecture: alpha (4) fails amd64 works armel works armhf (3) fails i386 works ia64 works m68k (1) workaround mips works mipsel works powerpc (2) fails ppc64 works (!) s390 works s390x works sparc works Architectures not listed either

This implements a sleep command. diff -p -purN klibc-0.111/utils/Makefile klibc-0.111.utils/utils/Makefile --- klibc-0.111/utils/Makefile 2004-02-10 00:18:15.000000000 +0100 +++ klibc-0.111.utils/utils/Makefile 2004-02-21 22:33:08.000000000 +0100 @@ -5,7 +5,7 @@ include ../MRULES MAKEDEPS = -Wp,-MD,.$(subst /,-,$*).d CFLAGS = $(MAKEDEPS) $(OPTFLAGS) $(REQFLAGS) -W -Wall LIBS = $(KLIBC)

this implements a ln(1) with -f and -s support. diff -p -purN klibc-0.112/utils/Makefile klibc-0.112.utils/utils/Makefile --- klibc-0.112/utils/Makefile 2004-02-21 23:49:26.000000000 +0100 +++ klibc-0.112.utils/utils/Makefile 2004-02-22 01:23:17.000000000 +0100 @@ -6,7 +6,7 @@ MAKEDEPS = -Wp,-MD,.$(subst /,-,$*). CFLAGS = $(MAKEDEPS) $(OPTFLAGS) $(REQFLAGS) -W -Wall LIBS =

O_CREAT requires a mode argument for open() This patch passes the current umask of the dd process. --- usr/klibc/klibc-0.81.orig/utils/dd.c 2002-09-03 00:49:08.000000000 +0200 +++ usr/klibc/klibc-0.81/utils/dd.c 2003-11-23 18:04:11.000000000 +0100 @@ -8,6 +8,8 @@ #include <stdio.h> #include <stdlib.h> #include <string.h> +#include <sys/types.h> +#include

2.6.4 changed asm/unistd.h, it includes now linux/types.h. This breaks compilation of fork.c and mmap.c, duplicate typedefs on ppc64 and x86_64. Moving one include fixes the compile. --- klibc/fork.c +++ klibc/fork.c 2004/03/14 15:04:51 @@ -5,8 +5,8 @@ * doesn't have sys_fork, only sys_clone... */ -#include <sys/syscall.h> #include <signal.h> +#include <sys/syscall.h>