Hello!
CVSRT:/cvsroot/hurd
Module name:gnumach
Branch:gnumach-1-branch
Changes by:Samuel Thibault <sthibaul>06/11/05 20:39:25
[]
2006-11-05 Barry deFreese <bddebian (AT) comcast (DOT) net>
[task #5878 ``Backport code from GNU Mach's trunk to
gnumach-1-branch: i386_set_gdt, i386_get_gdt'']
[]
* i386/i386/user_ldt.c (i386_set_gdt, i386_get_gdt): New functions.
Since this is a back-port from SKit Mach's code, it might be interesting
to explain the following divergences from the SKit code perhaps
Roland would like to have these changes explained, as he's written the
original code, if I recall correctly:
#v+
@@ -221,16 +223,16 @@ i386_set_ldt(thread, first_selector, des
[]
- else if (thread == current_thread()) {
- struct x86_desc template = {0, 0, 0, ACC_P, 0, 0 ,0};
+ else {
+ struct real_descriptor template = {0, 0, 0, ACC_P, 0, 0 ,0};
#v-
The following differences might be caused simply by the SKit Mach code
being different from the GNU Mach one, but you can confirm that more
easily than I could do right now.
#v+
@@ -411,46 +413,45 @@ i386_set_gdt (thread_t thread, int *sele
if (*selector == -1)
{
- for (idx = sel_idx (USER_GDT); idx < sel_idx (USER_GDT) + USER_GDT_SLTS;
- ++idx)
+ for (idx = 0; idx < USER_GDT_SLTS; ++idx)
if ((thread->pcb->ims.user_gdt[idx].access & ACC_P) == 0)
{
- *selector = (idx << 3) | SEL_PL_U;
+ *selector = ((idx + sel_idx(USER_GDT)) << 3) | SEL_PL_U;
break;
}
- if (idx == sel_idx (USER_GDT) + USER_GDT_SLTS)
+ if (idx == USER_GDT_SLTS)
return KERN_NSPACE; /* ? */
}
else if ((*selector & (SEL_LDT|SEL_PL)) != SEL_PL_U
- || sel_idx (*selector) < USER_GDT
- || sel_idx (*selector) >= USER_GDT + USER_GDT_SLTS)
+ || sel_idx (*selector) < sel_idx(USER_GDT)
+ || sel_idx (*selector) >= sel_idx(USER_GDT) + USER_GDT_SLTS)
return KERN_INVALID_ARGUMENT;
else
- idx = sel_idx (*selector);
+ idx = sel_idx (*selector) - sel_idx(USER_GDT);
- if ((desc->access & ACC_P) == 0)
+ if ((desc.access & ACC_P) == 0)
memset (&thread->pcb->ims.user_gdt[idx], 0,
sizeof thread->pcb->ims.user_gdt[idx]);
- else if ((desc->access & (ACC_TYPE|ACC_PL)) != (ACC_TYPE_USER|ACC_PL_U))
+ else if ((desc.access & (ACC_TYPE_USER|ACC_PL)) != (ACC_TYPE_USER|ACC_PL_U))
return KERN_INVALID_ARGUMENT;
else
- thread->pcb->ims.user_gdt[idx] = *desc;
+ thread->pcb->ims.user_gdt[idx] = desc;
return KERN_SUCCESS;
}
[]
[i386_get_gdt]
if ((selector & (SEL_LDT|SEL_PL)) != SEL_PL_U
- || sel_idx (selector) < USER_GDT
- || sel_idx (selector) >= USER_GDT + USER_GDT_SLTS)
+ || sel_idx (selector) < sel_idx(USER_GDT)
+ || sel_idx (selector) >= sel_idx(USER_GDT) + USER_GDT_SLTS)
return KERN_INVALID_ARGUMENT;
- *desc = thread->pcb->ims.user_gdt[sel_idx (selector)];
+ *desc = thread->pcb->ims.user_gdt[sel_idx (selector) - sel_idx(USER_GDT)];
#v-
Please also add a note to the ChangeLog that this commit is a back-port,
see the Roland McGrath ChangeLog entry from 2006-05-14 for an idea how
this might look like.
* New file.
Please move that file to `linux/src/include/linux/head.h' (and then add a
proper ChangeLog entry for that file). I'm not going to follow the `src'
vs. `dev' distinction any further.
Thanks for completing this task!
Regards,
Thomas
Bug-hurd mailing list
Bug-hurd (AT) gnu (DOT) org
PGP SIGNATURE
Version: GnuPG v1.2.5 (GNU/Linux)
HHblzpt9vn59PcY3Ht2L3M0=
=Xy+I
PGP SIGNATURE

Hello again!

CVSRT:/cvsroot/hurd
Module name:gnumach
Branch:gnumach-1-branch
Changes by:Samuel Thibault <sthibaul>06/11/05 20:39:25

[]
2006-11-05 Barry deFreese <bddebian (AT) comcast (DOT) net>

[task #5878 ``Backport code from GNU Mach's trunk to
gnumach-1-branch: i386_set_gdt, i386_get_gdt'']

I saw some notes on the #hurd irc channel on freenode that you've been
writing a tiny program to test this functionality. Would things like
that be worth for being added to GNU Mach's test suite? course the
problem is that you first have to reboot to the freshly built kernel in
order to be able to test it, but it might nevertheless be useful to boot
the newly built kernel and then return to the build directory and run
`make check-SMETHING' or similar. What are other's thoughts?

Regards,
Thomas

Bug-hurd mailing list
Bug-hurd (AT) gnu (DOT) org

PGP SIGNATURE
Version: GnuPG v1.2.5 (GNU/Linux)

EvHlTk75NALkomwF+32M=
=vELK
PGP SIGNATURE

Hi,

Thomas Schwinge, le Sun 05 Nov 2006 22:34:56 +0100, a :
#v+
@@ -221,16 +223,16 @@ i386_set_ldt(thread, first_selector, des
[]
- else if (thread == current_thread()) {
- struct x86_desc template = {0, 0, 0, ACC_P, 0, 0 ,0};
+ else {
+ struct real_descriptor template = {0, 0, 0, ACC_P, 0, 0 ,0};
#v-
The following differences might be caused simply by the SKit Mach code
being different from the GNU Mach one,

Exactly. Just the names differ.
- for (idx = sel_idx (USER_GDT); idx < sel_idx (USER_GDT) + USER_GDT_SLTS;
- ++idx)
+ for (idx = 0; idx < USER_GDT_SLTS; ++idx)

There was a mixture of index / selector in this code, which eventually
wasn't working. Defining idx between 0 an USER_GDT_SLT-1 makes the
whole code easier to read, and eventually work :).

* New file.

Please move that file to `linux/src/include/linux/head.h' (and then add a
proper ChangeLog entry for that file). I'm not going to follow the `src'
vs. `dev' distinction any further.

Ah. I guess this policy change should be clearly stated somewhere.

Samuel

Bug-hurd mailing list
Bug-hurd (AT) gnu (DOT) org

Roland McGrath, le Sun 05 Nov 2006 13:41:43 -0800, a :
- else if (thread == current_thread()) {
- struct x86_desc template = {0, 0, 0, ACC_P, 0, 0 ,0};
+ else {
+ struct real_descriptor template = {0, 0, 0, ACC_P, 0, 0 ,0};

This can't be right (changing the if).

Uh, didn't notice that. Barry, do you remember why you dropped the if?

Samuel

Bug-hurd mailing list
Bug-hurd (AT) gnu (DOT) org

Roland McGrath, le Sun 05 Nov 2006 13:43:08 -0800, a :
It's worthwhile to collect such test cases, but let's not worry about
putting them into the gnumach package yet.

Where to put them then?

Samuel

Bug-hurd mailing list
Bug-hurd (AT) gnu (DOT) org

Samuel Thibault, le Sun 05 Nov 2006 22:49:59 +0100, a :
Roland McGrath, le Sun 05 Nov 2006 13:41:43 -0800, a :
- else if (thread == current_thread()) {
- struct x86_desc template = {0, 0, 0, ACC_P, 0, 0 ,0};
+ else {
+ struct real_descriptor template = {0, 0, 0, ACC_P, 0, 0 ,0};

This can't be right (changing the if).

Uh, didn't notice that. Barry, do you remember why you dropped the if?

That's actually not from Barry's patch (the code above deals with LDT,
not GDT), but from Jeroen's.

Samuel

Bug-hurd mailing list
Bug-hurd (AT) gnu (DOT) org

Hello!

Sun, Nov 05, 2006 at 10:46:38PM +0100, Samuel Thibault wrote:
Thomas Schwinge, le Sun 05 Nov 2006 22:34:56 +0100, a ?crit :
* New file.

Please move that file to `linux/src/include/linux/head.h' (and then add a
proper ChangeLog entry for that file). I'm not going to follow the `src'
vs. `dev' distinction any further.

Ah.

Thanks for changing this.

I guess this policy change should be clearly stated somewhere.

Done:

#v+
2006-11-05 Thomas Schwinge <tschwinge (AT) gnu (DOT) org>

* linux/dev/README: New file.
#v-
#v+
This hierarchy used to contain modified files, based on files from the
Linux kernel, as opposed to `/src/' containing only files that have not
been modified (or have only been modified marginally). This policy is
NT adhered to any further, so please don't change (or even add) files
below here, but instead merge the files in here back into `/src/'
(which should really be called `/linux-2.0' or similar) or even better
when adding large chunks create a more suitable hierarchy like
we've done with `/pcmcia-cs/'.
#v-

Regards,
Thomas

Bug-hurd mailing list
Bug-hurd (AT) gnu (DOT) org

PGP SIGNATURE
Version: GnuPG v1.2.5 (GNU/Linux)

ZUizTbHyqJ3UX0pY0KvdwbM=
=JI8
PGP SIGNATURE

Hello!

Sun, Nov 05, 2006 at 10:58:12PM +0100, Samuel Thibault wrote:
Samuel Thibault, le Sun 05 Nov 2006 22:49:59 +0100, a ?crit :
Roland McGrath, le Sun 05 Nov 2006 13:41:43 -0800, a ?crit :
- else if (thread == current_thread()) {
- struct x86_desc template = {0, 0, 0, ACC_P, 0, 0 ,0};
+ else {
+ struct real_descriptor template = {0, 0, 0, ACC_P, 0, 0 ,0};

This can't be right (changing the if).

Uh, didn't notice that. Barry, do you remember why you dropped the if?

That's actually not from Barry's patch (the code above deals with LDT,
not GDT), but from Jeroen's.

Ah, right.

#v+
2006-01-26 Jeroen Dekkers <jeroen (AT) dekkers (DOT) cx>

* i386/i386/locore.S (trap_push_segs): Switch fs and gs to kernel
data segment too.
(syscall_entry_2): Likewise.
* i386/i386/user_ldt.c (i386_set_ldt): Always copy the master LDT
when there is no old user LDT.
#v-
So, I guess we want to keep this installed? We installed it that that
time, because it prevented GNU Mach from crashing when using the `ld.so'
of a tls enabled glibc.

Regards,
Thomas

Bug-hurd mailing list
Bug-hurd (AT) gnu (DOT) org

PGP SIGNATURE
Version: GnuPG v1.2.5 (GNU/Linux)

D2+Te2gJykiSEF54xAIYU=
=MsVM
PGP SIGNATURE

Message
From: "Samuel Thibault" <samuel.thibault (AT) ens-lyon (DOT) org>
To: "Roland McGrath" <roland (AT) frob (DOT) com>
Cc: "Thomas Schwinge" <tschwinge (AT) gnu (DOT) org>; <bug-hurd (AT) gnu (DOT) org>; "Barry
deFreese" <bddebian (AT) comcast (DOT) net>; <commit-hurd (AT) gnu (DOT) org>
Sent: Sunday, November 05, 2006 4:49 PM
Subject: Re: gnumach ChangeLog i386/i386/gdt.h i386/i386/pcb
[gnumach-1-branch]

Roland McGrath, le Sun 05 Nov 2006 13:41:43 -0800, a :
>- else if (thread == current_thread()) {
>- struct x86_desc template = {0, 0, 0, ACC_P, 0, 0 ,0};
>+ else {
>+ struct real_descriptor template = {0, 0, 0, ACC_P, 0, 0
>,0};
>>
>This can't be right (changing the if).
>
Uh, didn't notice that. Barry, do you remember why you dropped the if?

Samuel

The only thing I intended to change there was x_86_desc -real_descriptor
so I don't know what happened to the if. Very strange.

Barry

Bug-hurd mailing list
Bug-hurd (AT) gnu (DOT) org