On Sun, May 4, 2008 at 6:35 PM, Anant Narayanan <anant@xxxxxx> wrote:
> Hi,
>
> While implementing a binary format loader for a foreign executable format,
What foreign executable is that?
As far as I can see, Linux kernel can handle the following types of
foreign executable:
./fs/binfmt_elf.c:
retval = setup_arg_pages(bprm, randomize_stack_top(STACK_TOP),
./fs/binfmt_elf_fdpic.c:
retval = setup_arg_pages(bprm, current->mm->start_stack,
./fs/binfmt_som.c:
setup_arg_pages(bprm, STACK_TOP, EXSTACK_DEFAULT);
./fs/binfmt_aout.c:
retval = setup_arg_pages(bprm, STACK_TOP, EXSTACK_DEFAULT);
./arch/ia64/ia32/binfmt_elf32.c:
ret = setup_arg_pages(bprm, IA32_STACK_TOP, executable_stack);
./arch/mips/kernel/irixelf.c:
setup_arg_pages(bprm, STACK_TOP, EXSTACK_DEFAULT);
./arch/x86/ia32/ia32_aout.c:
retval = setup_arg_pages(bprm, IA32_STACK_TOP, EXSTACK_DEFAULT);
Possibly u can customized based on one of these pattern.
> we have found the need to setup the user stack starting at (virtual) address
> 0xE0000000 instead of the usual 0xC0000000. However a call to
Possibly u can have small variations, which u can see via (which grep
for all the stack space virtual address of all the processes):
cat /proc/*/maps|grep stack:
bfcc6000-bfcdb000 rw-p bffeb000 00:00 0 [stack]
bf8f8000-bf90d000 rw-p bffeb000 00:00 0 [stack]
bfed1000-bfee6000 rw-p bffeb000 00:00 0 [stack]
bffa8000-bffbd000 rw-p bffeb000 00:00 0 [stack]
bfa2d000-bfa42000 rw-p bffeb000 00:00 0 [stack]
And the random address u see is in done when loading the elf
(fs/binfmt_elf.c) load_elf_binary():
/* Do this so that we can load the interpreter, if need be. We will
change some of these later */
current->mm->free_area_cache = current->mm->mmap_base;
current->mm->cached_hole_size = 0;
retval = setup_arg_pages(bprm, randomize_stack_top(STACK_TOP),
executable_stack);
if (retval < 0) {
send_sig(SIGKILL, current, 0);
goto out_free_dentry;
}
> setup_arg_pages() with that value returns -EINVAL.
>
> Is there any way to setup the stack such that it starts at 0xE0000000 and
> grows downwards?
>
Yes, u can specify upward or downwards depending on architecture:
Inside setup_arg_pages() (fs/exec.c):
#ifdef CONFIG_STACK_GROWSUP
stack_base = vma->vm_end + EXTRA_STACK_VM_PAGES * PAGE_SIZE;
#else
stack_base = vma->vm_start - EXTRA_STACK_VM_PAGES * PAGE_SIZE;
#endif
ret = expand_stack(vma, stack_base);
Finally, is the big picture. In general, it is not EASY to change
the memory layout for the different component like stack space etc, as
it may affect other component (different standards are involved, and
if they really overlapped, then kmapping mechanism is needed to share
the memory - essentially playing around with the pagetable mechanism).
Specifically, if u cat /proc/iomem:
/proc>cat iomem
00000000-0009fbff : System RAM
0009fc00-0009ffff : reserved
000a0000-000bffff : Video RAM area
000c0000-000cc7ff : Video ROM
000f0000-000fffff : System ROM
00100000-7ff8ffff : System RAM
00400000-0063df98 : Kernel code
0063df99-00765adf : Kernel data
007ab000-0086297f : Kernel bss
7ff90000-7ff9dfff : ACPI Tables
7ff9e000-7ffdffff : ACPI Non-volatile Storage
7ffe0000-7fffffff : reserved
88000000-880000ff : 0000:00:1f.3
bfe00000-dfdfffff : PCI Bus 0000:01
c0000000-cfffffff : 0000:01:00.0
dfe00000-dfefffff : PCI Bus 0000:04
e0000000-efffffff : PCI MMCONFIG 0
e0000000-efffffff : pnp 00:0e
f8700000-fe7fffff : PCI Bus 0000:01
fa000000-fbffffff : 0000:01:00.0
fd000000-fdffffff : 0000:01:00.0
And u can see that the 0xe000000 is occupied by someone else.
Correct me if wrong....but nevertheless, u can still put the stack
area there of course - using mmapping mechanism. What is this
MMCONFIG area BTW?
--
Regards,
Peter Teoh
--
To unsubscribe from this list: send an email with
"unsubscribe kernelnewbies" to ecartis@xxxxxxxxxxxx
Please read the FAQ at http://kernelnewbies.org/FAQ
