Re: [Qemu-devel] [PATCH] qemu-i386 segfaults running "hello world".

[Stefan Weil]

Rob Landley schrieb:
> Ok, it's a more fundamental problem:
>
> address@hidden:/sys$ qemu-i386
> Segmentation fault (core dumped)
>
> Nothing to do with the program it's trying to run, it segfaults with no
> arguments.
>
> Is anybody else seeing this?
>
> Rob
Yes, I see this on Debian Linux since several months (libc update?).

The crash is caused by libc startup code which calls a null pointer.
QEMU provides this null pointer with the __init_array_start
workaround in linux-user/main.c.

This can be fixed with some kind of code hack - see my patch
(which is not really a solution, but one more workaround).

Nevertheless user mode emulations remains unusable even
with this patch because of TLS problems.

Regards,
Stefan

Index: linux-user/main.c
===================================================================
RCS file: /sources/qemu/qemu/linux-user/main.c,v
retrieving revision 1.116
diff -u -b -B -r1.116 main.c
--- linux-user/main.c   21 Jun 2007 22:55:02 -0000      1.116
+++ linux-user/main.c   23 Jun 2007 11:03:42 -0000
@@ -45,12 +45,16 @@
 /* for recent libc, we add these dummy symbols which are not declared
    when generating a linked object (bug in ld ?) */
 #if (__GLIBC__ > 2 || (__GLIBC__ == 2 && __GLIBC_MINOR__ >= 3)) && 
!defined(CONFIG_STATIC)
-long __preinit_array_start[0];
-long __preinit_array_end[0];
-long __init_array_start[0];
-long __init_array_end[0];
-long __fini_array_start[0];
-long __fini_array_end[0];
+typedef void (*dummy_function_t)(void);
+static void dummy_function(void)
+{
+}
+dummy_function_t __preinit_array_start = dummy_function;
+dummy_function_t __preinit_array_end = dummy_function;
+dummy_function_t __init_array_start = dummy_function;
+dummy_function_t __init_array_end = dummy_function;
+dummy_function_t __fini_array_start = dummy_function;
+dummy_function_t __fini_array_end = dummy_function;
 #endif
 
 /* XXX: on x86 MAP_GROWSDOWN only works if ESP <= address + 32, so