Hi all and I check the host linux configuration. the page offset is CONFIG_PAGE_OFFSET=0xC0000000 which is correct as your suggestion.I guss that fiasco-ux must be failed to interpret ELF of sigma0/roottask/irq0 and execute it . I put the same "hello" bin in grub/ fiasco environment. it works fun. JDB does not help me becuase that fiasco-ux enter a stage that CRTL-C does not work, I can not bring the system to jdb stage with CRTL-C. ps can find three processes which are 'fiasco-ux,[I][irq0] and [fiasco-ux]. I can use strace for fiasco-ux and [i][irq0]. but failed to trace [fiasco-ux].
so any suggestions?
-----邮件原件----- 发件人: l4-hackers-bounces@os.inf.tu-dresden.de [mailto:l4-hackers-bounces@os.inf.tu-dresden.de] 代表 Adam Lackorzynski 发送时间: 2006年12月14日 17:51 收件人: l4-hackers@os.inf.tu-dresden.de 主题: Re: ת·¢: segment fault for fiasco-ux
On Thu Dec 14, 2006 at 17:40:52 +0800, halo.duan wrote:
Hi the linux kernel is 2.6.15-1.2054-FC5, gcc is 4.1.0 20060304. code is the latest update.
following is screen copy of gdb
Fiasco-UX on Linux 2.6.15-1.2054_FC5 (i686) Mapped 64 MB Memory + 0 KB Framebuffer + 0 MB Input Area on FD 6
Loading Module 0x00090000-0x00099380 [sigma0] Loading Module 0x00120000-0x002b09e4 [roottask]
Bootstrapping...
Welcome to Fiasco(ux)! DD-L4(v2)/x86 microkernel (C) 1998-2006 TU Dresden Rev: Fri Dec 8 18:40:04 2006 compiled with gcc 4.1.0 for Intel Pentium III
Detaching after fork from child process 2486.
Program received signal SIGSEGV, Segmentation fault. Dirq::alloc (this=0xd242d, t=0xd7b80, ack_in_kernel=-1) at /home/halo/l4/kernel/fiasco/src/drivers/ux/processor-ux.cpp:62 62 asm volatile ("cli" : : : "memory"); (gdb)
according to the fiasco-ux documents, the SIGSEGV must be handled by fiasico registered handler. but I don't know why fiasco generate such enomous number of SYSSEGV.
so what's happened
That's how Fiasco-UX works, by intercepting such instructions. All those segmentation faults are really ok.
To come back to your problem. Since you use a FC kernel... Do those kernels use a different address space layout for users? Does the user have 3G of address space available? Recent kernels (2.6.16 and up) should have CONFIG_PAGE_OFFSET=0xC0000000 in their config.
Adam