* [patch] sys_modify_ldt extension (default_ldt)
@ 2001-06-09 23:47 Joerg Ahrens
2001-06-11 12:19 ` Christoph Hellwig
0 siblings, 1 reply; 2+ messages in thread
From: Joerg Ahrens @ 2001-06-09 23:47 UTC (permalink / raw)
To: linux-kernel
[-- Attachment #1: Type: text/plain, Size: 1678 bytes --]
Hi,
I am trying to integrate binfmt_xout.c into kernel 2.4 as part of the
linux-abi project (formerly known as iBCS). For old Xenix 286 binaries the
lcall7 gate needs to part of the LDT.
In kernels 2.0 sys_modify_ldt(0,...) used to return the default_ldt (with
lcall7 gate) if there were no segments set up. This behaviour changed in
kernels 2.2 . As a result of a discussion with Linus, David Bruce wrote a
patch for binfmt_xout.c tweaking with gdt and current->tss.ldt to get the
address of default_ldt. This patch does not work any more with kernels 2.4
as tss vanished from task_struct.
I do see 4 ways to cope with this problem:
a) extend sys_modify_ldt with a function to retrieve the default_ldt. I did
this for testing (see attached diff for arch/i386/kernel/ldt.c ).
b) do some work an Davids patch but this is kind of magic for me :-)
(see attached default_ldt patch)
c) loose the option to compile binfmt_xout (and the rest of linux-abi) as
module and simply use the symbol default_ldt. I dint't try that.
d) Forget about those old fashioned 286 binaries. This option will make some
linux users feel sad, as they run these progs for their daily business.
Joerg
--
------------------------------------------------------------------------------
Joerg Ahrens _/
Koenigsberger Strasse 32 _/_/
31226 Peine _/ _/
Tel.: 05171/57308 _/ _/_/_/_/
e-mail: joerg@hydrops.han.de _/_/_/_/ _/
------------------------------------------------------------------------------
[-- Attachment #2: ldt.c.diff --]
[-- Type: text/plain, Size: 826 bytes --]
--- linux-2.4.0/arch/i386/kernel/ldt.c Fri Dec 29 23:07:20 2000
+++ linux-2.4.0.i/arch/i386/kernel/ldt.c Sat Jun 9 22:48:46 2001
@@ -44,7 +44,24 @@
out:
return err;
}
+static int read_default_ldt(void * ptr, unsigned long bytecount)
+{
+ int err;
+ unsigned long size;
+ void *address;
+
+ err = 0;
+ address = &default_ldt[0];
+ size = sizeof(struct desc_struct);
+ if (size > bytecount)
+ size = bytecount;
+
+ err = size;
+ if (copy_to_user(ptr, address, size))
+ err = -EFAULT;
+ return err;
+}
static int write_ldt(void * ptr, unsigned long bytecount, int oldmode)
{
struct mm_struct * mm = current->mm;
@@ -156,6 +173,9 @@
break;
case 1:
ret = write_ldt(ptr, bytecount, 1);
+ break;
+ case 2:
+ ret = read_default_ldt(ptr, bytecount);
break;
case 0x11:
ret = write_ldt(ptr, bytecount, 0);
[-- Attachment #3: default_ldt_patch --]
[-- Type: text/plain, Size: 317 bytes --]
struct desc_struct def_ldt;
unsigned long *lp, *lp2;
asm volatile ( "sgdt __gdt+2" );
lp = (unsigned long *)(__gdt[1] + current->tss.ldt );
lp2 = (unsigned long *)(((*lp >> 16) & 0x0000ffff)
| (*(lp+1) & 0xff000000)
| ((*(lp+1) << 16) & 0x00ff0000));
def_ldt.a = *lp2;
def_ldt.b = *(lp2+1);
^ permalink raw reply [flat|nested] 2+ messages in thread
* Re: [patch] sys_modify_ldt extension (default_ldt)
2001-06-09 23:47 [patch] sys_modify_ldt extension (default_ldt) Joerg Ahrens
@ 2001-06-11 12:19 ` Christoph Hellwig
0 siblings, 0 replies; 2+ messages in thread
From: Christoph Hellwig @ 2001-06-11 12:19 UTC (permalink / raw)
To: Joerg Ahrens; +Cc: linux-kernel
Hi Joerg,
In article <m158sRu-0009RiC@hydrops.han.de> you wrote:
> Hi,
>
> I am trying to integrate binfmt_xout.c into kernel 2.4 as part of the
> linux-abi project (formerly known as iBCS). For old Xenix 286 binaries the
> lcall7 gate needs to part of the LDT.
>
> In kernels 2.0 sys_modify_ldt(0,...) used to return the default_ldt (with
> lcall7 gate) if there were no segments set up. This behaviour changed in
> kernels 2.2 . As a result of a discussion with Linus, David Bruce wrote a
> patch for binfmt_xout.c tweaking with gdt and current->tss.ldt to get the
> address of default_ldt. This patch does not work any more with kernels 2.4
> as tss vanished from task_struct.
>
> I do see 4 ways to cope with this problem:
>
> a) extend sys_modify_ldt with a function to retrieve the default_ldt. I did
> this for testing (see attached diff for arch/i386/kernel/ldt.c ).
Looks sane to me.
> b) do some work an Davids patch but this is kind of magic for me :-)
> (see attached default_ldt patch)
>
> c) loose the option to compile binfmt_xout (and the rest of linux-abi) as
> module and simply use the symbol default_ldt. I dint't try that.
As the linux-abi maintainer I do not think that's a good idea..
> d) Forget about those old fashioned 286 binaries. This option will make some
> linux users feel sad, as they run these progs for their daily business.
If possible at all I'll vote for having 286 support.
I'll integrate the patch into the linux-abi tree if you also send me the
other changes (mostly binfmt_xout.c changes I suppose).
Christoph
--
Of course it doesn't work. We've performed a software upgrade.
^ permalink raw reply [flat|nested] 2+ messages in thread
end of thread, other threads:[~2001-06-11 12:20 UTC | newest]
Thread overview: 2+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2001-06-09 23:47 [patch] sys_modify_ldt extension (default_ldt) Joerg Ahrens
2001-06-11 12:19 ` Christoph Hellwig
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).