[Fwd: Re: [RFA 3/5] Prec: x86 segment register support: target]

[Fwd: Re: [RFA 3/5] Prec: x86 segment register support: target]

[Michael Snyder]

[-- Attachment #1: Type: text/plain, Size: 33 bytes --]

Sorry, hit 'reply' by mistake...

[-- Attachment #2: Re: [RFA 3/5] Prec: x86 segment register support: target.eml --]
[-- Type: message/rfc822, Size: 2537 bytes --]

From: Michael Snyder <msnyder@vmware.com>
To: Doug Evans <dje@google.com>
Subject: Re: [RFA 3/5] Prec: x86 segment register support: target
Date: Mon, 22 Mar 2010 11:23:13 -0700
Message-ID: <4BA7B591.7030308@vmware.com>

Doug Evans wrote:
> On Mon, Mar 22, 2010 at 11:04 AM, Eli Zaretskii <eliz@gnu.org> wrote:
>>> Date: Mon, 22 Mar 2010 10:53:35 -0400
>>> From: Daniel Jacobowitz <dan@codesourcery.com>
>>> Cc: teawater@gmail.com, msnyder@vmware.com, gdb-patches@sourceware.org,       hjl.tools@gmail.com
>>>
>>> On Mon, Mar 22, 2010 at 03:33:55PM +0100, Mark Kettenis wrote:
>>>> Sorry, but I don't think adding these xx_base registers is a good
>>>> idea.  They are not acrhitected registers so they don't beling in the
>>>> list of registers.
>>> What do you suggest instead?
>> Would a separate command fit the bill?  (The DJGPP build already has
>> something similar, see "info dos ldt" and its description in the
>> manual with an example of how to use it to display the DS segment base
>> address and limit.)
> 
> I'd have the regs in a separate info command (out of "info regs"),
> there's more than just base, there's the limit and flags regs too, and
> having them would really reduce the S/N ratio of "info reg", but
> having $gs_base, et.al. usable in expressions is really useful (and
> long overdue).  Either that or make "gs:<offset>" as an address work -
> you'd still want to be able to fold it to a "flat" address though.

I'd just like to point out that while all this sounds great,
it shouldn't be a prerequisite to the original task of just
getting prec to record the segments and offsets correctly.

Maybe we should split these two tasks, so that Teawater can
go ahead and accomplish his.

Michael

Re: [Fwd: Re: [RFA 3/5] Prec: x86 segment register support: target]

[Doug Evans]

On Mon, Mar 22, 2010 at 11:26 AM, Michael Snyder <msnyder@vmware.com> wrote:
> I'd just like to point out that while all this sounds great,
> it shouldn't be a prerequisite to the original task of just
> getting prec to record the segments and offsets correctly.
>
> Maybe we should split these two tasks, so that Teawater can
> go ahead and accomplish his.

To the extent that they can be split, IWBN alright.

I wonder if the interface is sufficient though (setting aside where to
put it and how it will look).
Any particular o/s might not provide sufficient hooks of course.
linux's modify_ldt, AIUI, let's one change more than just foo_base.
NativeClient http://code.google.com/p/nativeclient/ uses it, for example.

Re: [Fwd: Re: [RFA 3/5] Prec: x86 segment register support: target]

[Hui Zhu]

On Tue, Mar 23, 2010 at 02:47, Doug Evans <dje@google.com> wrote:
> On Mon, Mar 22, 2010 at 11:26 AM, Michael Snyder <msnyder@vmware.com> wrote:
>> I'd just like to point out that while all this sounds great,
>> it shouldn't be a prerequisite to the original task of just
>> getting prec to record the segments and offsets correctly.
>>
>> Maybe we should split these two tasks, so that Teawater can
>> go ahead and accomplish his.
>
> To the extent that they can be split, IWBN alright.
>
> I wonder if the interface is sufficient though (setting aside where to
> put it and how it will look).
> Any particular o/s might not provide sufficient hooks of course.
> linux's modify_ldt, AIUI, let's one change more than just foo_base.
> NativeClient http://code.google.com/p/nativeclient/ uses it, for example.
>

Thanks Doug.

I suggest we support segment base step by step.
When the OS that support it will show the xxx_base to user, the
unsupport OS will show nothing.

What do you think about it?

Best regards,
Hui

Re: [Fwd: Re: [RFA 3/5] Prec: x86 segment register support: target]

[Doug Evans]

On Mon, Mar 22, 2010 at 7:59 PM, Hui Zhu <teawater@gmail.com> wrote:
> On Tue, Mar 23, 2010 at 02:47, Doug Evans <dje@google.com> wrote:
>> On Mon, Mar 22, 2010 at 11:26 AM, Michael Snyder <msnyder@vmware.com> wrote:
>>> I'd just like to point out that while all this sounds great,
>>> it shouldn't be a prerequisite to the original task of just
>>> getting prec to record the segments and offsets correctly.
>>>
>>> Maybe we should split these two tasks, so that Teawater can
>>> go ahead and accomplish his.
>>
>> To the extent that they can be split, IWBN alright.
>>
>> I wonder if the interface is sufficient though (setting aside where to
>> put it and how it will look).
>> Any particular o/s might not provide sufficient hooks of course.
>> linux's modify_ldt, AIUI, let's one change more than just foo_base.
>> NativeClient http://code.google.com/p/nativeclient/ uses it, for example.
>>
>
> Thanks Doug.
>
> I suggest we support segment base step by step.
> When the OS that support it will show the xxx_base to user, the
> unsupport OS will show nothing.
>
> What do you think about it?

Is supporting segment base sufficient?
Or do you also need to support, e.g., segment limit and flags too?
There may be more, but they're the two that come to mind.
[That's what I was referring to regarding whether the interface was sufficient.]

Re: [Fwd: Re: [RFA 3/5] Prec: x86 segment register support: target]

[Hui Zhu]

On Thu, Mar 25, 2010 at 02:44, Doug Evans <dje@google.com> wrote:
>
> On Mon, Mar 22, 2010 at 7:59 PM, Hui Zhu <teawater@gmail.com> wrote:
> > On Tue, Mar 23, 2010 at 02:47, Doug Evans <dje@google.com> wrote:
> >> On Mon, Mar 22, 2010 at 11:26 AM, Michael Snyder <msnyder@vmware.com> wrote:
> >>> I'd just like to point out that while all this sounds great,
> >>> it shouldn't be a prerequisite to the original task of just
> >>> getting prec to record the segments and offsets correctly.
> >>>
> >>> Maybe we should split these two tasks, so that Teawater can
> >>> go ahead and accomplish his.
> >>
> >> To the extent that they can be split, IWBN alright.
> >>
> >> I wonder if the interface is sufficient though (setting aside where to
> >> put it and how it will look).
> >> Any particular o/s might not provide sufficient hooks of course.
> >> linux's modify_ldt, AIUI, let's one change more than just foo_base.
> >> NativeClient http://code.google.com/p/nativeclient/ uses it, for example.
> >>
> >
> > Thanks Doug.
> >
> > I suggest we support segment base step by step.
> > When the OS that support it will show the xxx_base to user, the
> > unsupport OS will show nothing.
> >
> > What do you think about it?
>
> Is supporting segment base sufficient?
> Or do you also need to support, e.g., segment limit and flags too?
> There may be more, but they're the two that come to mind.
> [That's what I was referring to regarding whether the interface was sufficient.]

Prec just need the base to get the insn memory operate address.  Do
you think we need other message of segment?

If need, do we need divide all message like eflags?

Thanks,
Hui

Re: [Fwd: Re: [RFA 3/5] Prec: x86 segment register support: target]

[Hui Zhu]

On Thu, Mar 25, 2010 at 10:14, Hui Zhu <teawater@gmail.com> wrote:
> On Thu, Mar 25, 2010 at 02:44, Doug Evans <dje@google.com> wrote:
>>
>> On Mon, Mar 22, 2010 at 7:59 PM, Hui Zhu <teawater@gmail.com> wrote:
>> > On Tue, Mar 23, 2010 at 02:47, Doug Evans <dje@google.com> wrote:
>> >> On Mon, Mar 22, 2010 at 11:26 AM, Michael Snyder <msnyder@vmware.com> wrote:
>> >>> I'd just like to point out that while all this sounds great,
>> >>> it shouldn't be a prerequisite to the original task of just
>> >>> getting prec to record the segments and offsets correctly.
>> >>>
>> >>> Maybe we should split these two tasks, so that Teawater can
>> >>> go ahead and accomplish his.
>> >>
>> >> To the extent that they can be split, IWBN alright.
>> >>
>> >> I wonder if the interface is sufficient though (setting aside where to
>> >> put it and how it will look).
>> >> Any particular o/s might not provide sufficient hooks of course.
>> >> linux's modify_ldt, AIUI, let's one change more than just foo_base.
>> >> NativeClient http://code.google.com/p/nativeclient/ uses it, for example.
>> >>
>> >
>> > Thanks Doug.
>> >
>> > I suggest we support segment base step by step.
>> > When the OS that support it will show the xxx_base to user, the
>> > unsupport OS will show nothing.
>> >
>> > What do you think about it?
>>
>> Is supporting segment base sufficient?
>> Or do you also need to support, e.g., segment limit and flags too?
>> There may be more, but they're the two that come to mind.
>> [That's what I was referring to regarding whether the interface was sufficient.]
>
> Prec just need the base to get the insn memory operate address.  Do
> you think we need other message of segment?
>
> If need, do we need divide all message like eflags?
>
> Thanks,
> Hui
>

Hi all,

X86 looks stab now.  Shall we wake up this patch?

Thanks,
Hui

Re: [Fwd: Re: [RFA 3/5] Prec: x86 segment register support: target]

[Mark Kettenis]

[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #1: Type: text/plain, Size: 2237 bytes --]

> From: Hui Zhu <teawater@gmail.com>
> Date: Fri, 30 Apr 2010 14:29:09 +0800
> 
> On Thu, Mar 25, 2010 at 10:14, Hui Zhu <teawater@gmail.com> wrote:
> > On Thu, Mar 25, 2010 at 02:44, Doug Evans <dje@google.com> wrote:
> >>
> >> On Mon, Mar 22, 2010 at 7:59 PM, Hui Zhu <teawater@gmail.com> wrote:
> >> > On Tue, Mar 23, 2010 at 02:47, Doug Evans <dje@google.com> wrote:
> >> >> On Mon, Mar 22, 2010 at 11:26 AM, Michael Snyder <msnyder@vmware.com> wrote:
> >> >>> I'd just like to point out that while all this sounds great,
> >> >>> it shouldn't be a prerequisite to the original task of just
> >> >>> getting prec to record the segments and offsets correctly.
> >> >>>
> >> >>> Maybe we should split these two tasks, so that Teawater can
> >> >>> go ahead and accomplish his.
> >> >>
> >> >> To the extent that they can be split, IWBN alright.
> >> >>
> >> >> I wonder if the interface is sufficient though (setting aside where to
> >> >> put it and how it will look).
> >> >> Any particular o/s might not provide sufficient hooks of course.
> >> >> linux's modify_ldt, AIUI, let's one change more than just foo_base.
> >> >> NativeClient http://code.google.com/p/nativeclient/ uses it, for example.
> >> >>
> >> >
> >> > Thanks Doug.
> >> >
> >> > I suggest we support segment base step by step.
> >> > When the OS that support it will show the xxx_base to user, the
> >> > unsupport OS will show nothing.
> >> >
> >> > What do you think about it?
> >>
> >> Is supporting segment base sufficient?
> >> Or do you also need to support, e.g., segment limit and flags too?
> >> There may be more, but they're the two that come to mind.
> >> [That's what I was referring to regarding whether the interface was sufficient.]
> >
> > Prec just need the base to get the insn memory operate address.  Do
> > you think we need other message of segment?
> >
> > If need, do we need divide all message like eflags?
> >
> > Thanks,
> > Hui
> >
> 
> Hi all,
> 
> X86 looks stab now.  Shall we wake up this patch?

Thanks for the reminder.

Let me first ask a question.  What do people expect out of this?  Do
they really want support for fully segmented code, or is it just for
small deviations like accessing per-thread storage through %fs/%gs?

Re: [Fwd: Re: [RFA 3/5] Prec: x86 segment register support: target]

[Hui Zhu]

On Fri, Apr 30, 2010 at 17:33, Mark Kettenis <mark.kettenis@xs4all.nl> wrote:
>> From: Hui Zhu <teawater@gmail.com>
>> Date: Fri, 30 Apr 2010 14:29:09 +0800
>>
>> On Thu, Mar 25, 2010 at 10:14, Hui Zhu <teawater@gmail.com> wrote:
>> > On Thu, Mar 25, 2010 at 02:44, Doug Evans <dje@google.com> wrote:
>> >>
>> >> On Mon, Mar 22, 2010 at 7:59 PM, Hui Zhu <teawater@gmail.com> wrote:
>> >> > On Tue, Mar 23, 2010 at 02:47, Doug Evans <dje@google.com> wrote:
>> >> >> On Mon, Mar 22, 2010 at 11:26 AM, Michael Snyder <msnyder@vmware.com> wrote:
>> >> >>> I'd just like to point out that while all this sounds great,
>> >> >>> it shouldn't be a prerequisite to the original task of just
>> >> >>> getting prec to record the segments and offsets correctly.
>> >> >>>
>> >> >>> Maybe we should split these two tasks, so that Teawater can
>> >> >>> go ahead and accomplish his.
>> >> >>
>> >> >> To the extent that they can be split, IWBN alright.
>> >> >>
>> >> >> I wonder if the interface is sufficient though (setting aside where to
>> >> >> put it and how it will look).
>> >> >> Any particular o/s might not provide sufficient hooks of course.
>> >> >> linux's modify_ldt, AIUI, let's one change more than just foo_base.
>> >> >> NativeClient http://code.google.com/p/nativeclient/ uses it, for example.
>> >> >>
>> >> >
>> >> > Thanks Doug.
>> >> >
>> >> > I suggest we support segment base step by step.
>> >> > When the OS that support it will show the xxx_base to user, the
>> >> > unsupport OS will show nothing.
>> >> >
>> >> > What do you think about it?
>> >>
>> >> Is supporting segment base sufficient?
>> >> Or do you also need to support, e.g., segment limit and flags too?
>> >> There may be more, but they're the two that come to mind.
>> >> [That's what I was referring to regarding whether the interface was sufficient.]
>> >
>> > Prec just need the base to get the insn memory operate address.  Do
>> > you think we need other message of segment?
>> >
>> > If need, do we need divide all message like eflags?
>> >
>> > Thanks,
>> > Hui
>> >
>>
>> Hi all,
>>
>> X86 looks stab now.  Shall we wake up this patch?
>
> Thanks for the reminder.
>
> Let me first ask a question.  What do people expect out of this?  Do
> they really want support for fully segmented code, or is it just for
> small deviations like accessing per-thread storage through %fs/%gs?
>

Prec must know the each base of segment register.

If you don't like it.  What about the old way that I use?  It doesn't
add anything to reg list.
But for the each way, we need add interface to the target part that
prec can get the value.

Thanks,
Hui

Re: [Fwd: Re: [RFA 3/5] Prec: x86 segment register support: target]

[Hui Zhu]

On Fri, Apr 30, 2010 at 19:07, Hui Zhu <teawater@gmail.com> wrote:
> On Fri, Apr 30, 2010 at 17:33, Mark Kettenis <mark.kettenis@xs4all.nl> wrote:
>>> From: Hui Zhu <teawater@gmail.com>
>>> Date: Fri, 30 Apr 2010 14:29:09 +0800
>>>
>>> On Thu, Mar 25, 2010 at 10:14, Hui Zhu <teawater@gmail.com> wrote:
>>> > On Thu, Mar 25, 2010 at 02:44, Doug Evans <dje@google.com> wrote:
>>> >>
>>> >> On Mon, Mar 22, 2010 at 7:59 PM, Hui Zhu <teawater@gmail.com> wrote:
>>> >> > On Tue, Mar 23, 2010 at 02:47, Doug Evans <dje@google.com> wrote:
>>> >> >> On Mon, Mar 22, 2010 at 11:26 AM, Michael Snyder <msnyder@vmware.com> wrote:
>>> >> >>> I'd just like to point out that while all this sounds great,
>>> >> >>> it shouldn't be a prerequisite to the original task of just
>>> >> >>> getting prec to record the segments and offsets correctly.
>>> >> >>>
>>> >> >>> Maybe we should split these two tasks, so that Teawater can
>>> >> >>> go ahead and accomplish his.
>>> >> >>
>>> >> >> To the extent that they can be split, IWBN alright.
>>> >> >>
>>> >> >> I wonder if the interface is sufficient though (setting aside where to
>>> >> >> put it and how it will look).
>>> >> >> Any particular o/s might not provide sufficient hooks of course.
>>> >> >> linux's modify_ldt, AIUI, let's one change more than just foo_base.
>>> >> >> NativeClient http://code.google.com/p/nativeclient/ uses it, for example.
>>> >> >>
>>> >> >
>>> >> > Thanks Doug.
>>> >> >
>>> >> > I suggest we support segment base step by step.
>>> >> > When the OS that support it will show the xxx_base to user, the
>>> >> > unsupport OS will show nothing.
>>> >> >
>>> >> > What do you think about it?
>>> >>
>>> >> Is supporting segment base sufficient?
>>> >> Or do you also need to support, e.g., segment limit and flags too?
>>> >> There may be more, but they're the two that come to mind.
>>> >> [That's what I was referring to regarding whether the interface was sufficient.]
>>> >
>>> > Prec just need the base to get the insn memory operate address.  Do
>>> > you think we need other message of segment?
>>> >
>>> > If need, do we need divide all message like eflags?
>>> >
>>> > Thanks,
>>> > Hui
>>> >
>>>
>>> Hi all,
>>>
>>> X86 looks stab now.  Shall we wake up this patch?
>>
>> Thanks for the reminder.
>>
>> Let me first ask a question.  What do people expect out of this?  Do
>> they really want support for fully segmented code, or is it just for
>> small deviations like accessing per-thread storage through %fs/%gs?
>>
>
> Prec must know the each base of segment register.
>
> If you don't like it.  What about the old way that I use?  It doesn't
> add anything to reg list.
> But for the each way, we need add interface to the target part that
> prec can get the value.
>
> Thanks,
> Hui
>

Ping.

Thanks,
Hui

Re: [Fwd: Re: [RFA 3/5] Prec: x86 segment register support: target]

[Hui Zhu]

On Wed, May 5, 2010 at 10:46, Hui Zhu <teawater@gmail.com> wrote:
> On Fri, Apr 30, 2010 at 19:07, Hui Zhu <teawater@gmail.com> wrote:
>> On Fri, Apr 30, 2010 at 17:33, Mark Kettenis <mark.kettenis@xs4all.nl> wrote:
>>>> From: Hui Zhu <teawater@gmail.com>
>>>> Date: Fri, 30 Apr 2010 14:29:09 +0800
>>>>
>>>> On Thu, Mar 25, 2010 at 10:14, Hui Zhu <teawater@gmail.com> wrote:
>>>> > On Thu, Mar 25, 2010 at 02:44, Doug Evans <dje@google.com> wrote:
>>>> >>
>>>> >> On Mon, Mar 22, 2010 at 7:59 PM, Hui Zhu <teawater@gmail.com> wrote:
>>>> >> > On Tue, Mar 23, 2010 at 02:47, Doug Evans <dje@google.com> wrote:
>>>> >> >> On Mon, Mar 22, 2010 at 11:26 AM, Michael Snyder <msnyder@vmware.com> wrote:
>>>> >> >>> I'd just like to point out that while all this sounds great,
>>>> >> >>> it shouldn't be a prerequisite to the original task of just
>>>> >> >>> getting prec to record the segments and offsets correctly.
>>>> >> >>>
>>>> >> >>> Maybe we should split these two tasks, so that Teawater can
>>>> >> >>> go ahead and accomplish his.
>>>> >> >>
>>>> >> >> To the extent that they can be split, IWBN alright.
>>>> >> >>
>>>> >> >> I wonder if the interface is sufficient though (setting aside where to
>>>> >> >> put it and how it will look).
>>>> >> >> Any particular o/s might not provide sufficient hooks of course.
>>>> >> >> linux's modify_ldt, AIUI, let's one change more than just foo_base.
>>>> >> >> NativeClient http://code.google.com/p/nativeclient/ uses it, for example.
>>>> >> >>
>>>> >> >
>>>> >> > Thanks Doug.
>>>> >> >
>>>> >> > I suggest we support segment base step by step.
>>>> >> > When the OS that support it will show the xxx_base to user, the
>>>> >> > unsupport OS will show nothing.
>>>> >> >
>>>> >> > What do you think about it?
>>>> >>
>>>> >> Is supporting segment base sufficient?
>>>> >> Or do you also need to support, e.g., segment limit and flags too?
>>>> >> There may be more, but they're the two that come to mind.
>>>> >> [That's what I was referring to regarding whether the interface was sufficient.]
>>>> >
>>>> > Prec just need the base to get the insn memory operate address.  Do
>>>> > you think we need other message of segment?
>>>> >
>>>> > If need, do we need divide all message like eflags?
>>>> >
>>>> > Thanks,
>>>> > Hui
>>>> >
>>>>
>>>> Hi all,
>>>>
>>>> X86 looks stab now.  Shall we wake up this patch?
>>>
>>> Thanks for the reminder.
>>>
>>> Let me first ask a question.  What do people expect out of this?  Do
>>> they really want support for fully segmented code, or is it just for
>>> small deviations like accessing per-thread storage through %fs/%gs?
>>>
>>
>> Prec must know the each base of segment register.
>>
>> If you don't like it.  What about the old way that I use?  It doesn't
>> add anything to reg list.
>> But for the each way, we need add interface to the target part that
>> prec can get the value.
>>
>> Thanks,
>> Hui
>>
>
> Ping.
>
> Thanks,
> Hui
>

Ping.

Thanks,
Hui