* xen and tls (fwd)
@ 2005-02-14 15:53 Ronald G. Minnich
2005-02-14 16:05 ` Anthony Liguori
2005-02-14 17:02 ` Michael A Fetterman
0 siblings, 2 replies; 6+ messages in thread
From: Ronald G. Minnich @ 2005-02-14 15:53 UTC (permalink / raw)
To: xen-devel
I forget the reason, can somebody fill me in?
thanks
ron
---------- Forwarded message ----------
Date: Mon, 14 Feb 2005 10:50:56 -0500
To: rminnich@lanl.gov
Subject: xen and tls
do you know why, when i booted xen, i was told to
mv /lib/tls /lib/tls.disabled
otherwise it would be very slow and may cause instability
(i.e. programmes would fault)?
-------------------------------------------------------
SF email is sponsored by - The IT Product Guide
Read honest & candid reviews on hundreds of IT Products from real users.
Discover which products truly live up to the hype. Start reading now.
http://ads.osdn.com/?ad_id=6595&alloc_id=14396&op=click
^ permalink raw reply [flat|nested] 6+ messages in thread
* Re: xen and tls (fwd)
2005-02-14 15:53 xen and tls (fwd) Ronald G. Minnich
@ 2005-02-14 16:05 ` Anthony Liguori
2005-02-14 16:33 ` Himanshu Raj
2005-02-14 17:02 ` Michael A Fetterman
1 sibling, 1 reply; 6+ messages in thread
From: Anthony Liguori @ 2005-02-14 16:05 UTC (permalink / raw)
To: Ronald G. Minnich; +Cc: xen-devel
Ronald G. Minnich wrote:
>I forget the reason, can somebody fill me in?
>
>
http://www.cl.cam.ac.uk/Research/SRG/netos/xen/readmes/user/user.html#SECTION02243000000000000000
It has to do with segmentation. I'm not sure exactly what the problem
is although I can imagine a few problems that could occur with
segmentation.. I'd be curious to know the exact reason myself.
Regards,
>thanks
>
>ron
>
>---------- Forwarded message ----------
>Date: Mon, 14 Feb 2005 10:50:56 -0500
>To: rminnich@lanl.gov
>Subject: xen and tls
>
>do you know why, when i booted xen, i was told to
>mv /lib/tls /lib/tls.disabled
>otherwise it would be very slow and may cause instability
>(i.e. programmes would fault)?
>
>
>-------------------------------------------------------
>SF email is sponsored by - The IT Product Guide
>Read honest & candid reviews on hundreds of IT Products from real users.
>Discover which products truly live up to the hype. Start reading now.
>http://ads.osdn.com/?ad_id=6595&alloc_id=14396&op=click
>_______________________________________________
>Xen-devel mailing list
>Xen-devel@lists.sourceforge.net
>https://lists.sourceforge.net/lists/listinfo/xen-devel
>
>
>
--
Anthony Liguori
anthony@codemonkey.ws
-------------------------------------------------------
SF email is sponsored by - The IT Product Guide
Read honest & candid reviews on hundreds of IT Products from real users.
Discover which products truly live up to the hype. Start reading now.
http://ads.osdn.com/?ad_id=6595&alloc_id=14396&op=click
^ permalink raw reply [flat|nested] 6+ messages in thread
* Re: xen and tls (fwd)
2005-02-14 16:05 ` Anthony Liguori
@ 2005-02-14 16:33 ` Himanshu Raj
2005-02-14 16:52 ` Ronald G. Minnich
0 siblings, 1 reply; 6+ messages in thread
From: Himanshu Raj @ 2005-02-14 16:33 UTC (permalink / raw)
To: xen-devel
I believe it is due to the use of both +ve and -ve offsets by TLS. It becomes
a nightmare to manage it with segmentation, and system has to emulate -ve
offsets by converting to +ve ones etc. I am still learning more on this.
Himanshu
On Mon, Feb 14, 2005 at 10:05:00AM -0600, Anthony Liguori wrote:
> Ronald G. Minnich wrote:
>
> >I forget the reason, can somebody fill me in?
> >
> >
> http://www.cl.cam.ac.uk/Research/SRG/netos/xen/readmes/user/user.html#SECTION02243000000000000000
>
> It has to do with segmentation. I'm not sure exactly what the problem
> is although I can imagine a few problems that could occur with
> segmentation.. I'd be curious to know the exact reason myself.
>
> Regards,
>
> >thanks
> >
> >ron
> >
> >---------- Forwarded message ----------
> >Date: Mon, 14 Feb 2005 10:50:56 -0500
> >To: rminnich@lanl.gov
> >Subject: xen and tls
> >
> >do you know why, when i booted xen, i was told to
> >mv /lib/tls /lib/tls.disabled
> >otherwise it would be very slow and may cause instability
> >(i.e. programmes would fault)?
> >
> >
> >-------------------------------------------------------
> >SF email is sponsored by - The IT Product Guide
> >Read honest & candid reviews on hundreds of IT Products from real users.
> >Discover which products truly live up to the hype. Start reading now.
> >http://ads.osdn.com/?ad_id=6595&alloc_id=14396&op=click
> >_______________________________________________
> >Xen-devel mailing list
> >Xen-devel@lists.sourceforge.net
> >https://lists.sourceforge.net/lists/listinfo/xen-devel
> >
> >
> >
>
>
> --
> Anthony Liguori
> anthony@codemonkey.ws
>
>
>
> -------------------------------------------------------
> SF email is sponsored by - The IT Product Guide
> Read honest & candid reviews on hundreds of IT Products from real users.
> Discover which products truly live up to the hype. Start reading now.
> http://ads.osdn.com/?ad_id=6595&alloc_id=14396&op=click
> _______________________________________________
> Xen-devel mailing list
> Xen-devel@lists.sourceforge.net
> https://lists.sourceforge.net/lists/listinfo/xen-devel
--
-------------------------------------------------------------------------
Himanshu Raj
PhD Student, GaTech (www.cc.gatech.edu/~rhim)
I prefer to receive attachments in an open, non-proprietary format.
-------------------------------------------------------------------------
-------------------------------------------------------
SF email is sponsored by - The IT Product Guide
Read honest & candid reviews on hundreds of IT Products from real users.
Discover which products truly live up to the hype. Start reading now.
http://ads.osdn.com/?ad_id=6595&alloc_id=14396&op=click
^ permalink raw reply [flat|nested] 6+ messages in thread
* Re: xen and tls (fwd)
2005-02-14 16:33 ` Himanshu Raj
@ 2005-02-14 16:52 ` Ronald G. Minnich
2005-02-14 17:07 ` Barry Silverman
0 siblings, 1 reply; 6+ messages in thread
From: Ronald G. Minnich @ 2005-02-14 16:52 UTC (permalink / raw)
To: Himanshu Raj; +Cc: xen-devel
On Mon, 14 Feb 2005, Himanshu Raj wrote:
> I believe it is due to the use of both +ve and -ve offsets by TLS. It becomes
> a nightmare to manage it with segmentation, and system has to emulate -ve
> offsets by converting to +ve ones etc. I am still learning more on this.
I vaguely remember this discussion, I just wish I could remember more.
ron
-------------------------------------------------------
SF email is sponsored by - The IT Product Guide
Read honest & candid reviews on hundreds of IT Products from real users.
Discover which products truly live up to the hype. Start reading now.
http://ads.osdn.com/?ad_id=6595&alloc_id=14396&op=click
^ permalink raw reply [flat|nested] 6+ messages in thread
* RE: xen and tls (fwd)
2005-02-14 15:53 xen and tls (fwd) Ronald G. Minnich
2005-02-14 16:05 ` Anthony Liguori
@ 2005-02-14 17:02 ` Michael A Fetterman
1 sibling, 0 replies; 6+ messages in thread
From: Michael A Fetterman @ 2005-02-14 17:02 UTC (permalink / raw)
To: 'Ronald G. Minnich', xen-devel
Normally, under linux, we run with segment limits of 4G.
That is, essentially no limit, any segment offset from 0 to 4G (unsigned)
is OK. Equivalently, any segment offset from -2G to 2G (signed) is OK.
Xen, however, uses segmentation to protect itself from guest OSes...
Xen runs in ring 0.
Guest OSes run in ring 1.
Guest user apps run in ring 3.
Paging provides a supervisor/user permission check, but there's only one
bit, and it says that rings 0-2 are all "supervisor" rings, and only ring
3 is prohibited from seeing those pages. So paging doesn't provide Xen
with any protection from guest OSes. Enter segmentation.
Xen sets up the segment registers so that the top 64MB of memory is
beyond the segment limit whenever the guest OS is running.
However, this creates a hole in memory (especially when the segment base
is non-zero). Segments are base and limit kinds of things; they don't
support "holes". Segments *do* however do a trick called "grow down",
which essentially means that you can invert the sense of the in-bounds/
out-of-bounds test for segment limits. That allows you to have segments
that are valid from segbase upwards, and segments that are valid from
just below segbase and "downward" to the limit from there. So Xen attempts
to detect accesses to the "other" side of segbase that should have been
valid, but aren't, due to not being able to express "holes" in the
segmentation model, and it flips the sense of the segment register to allow
such accesses to proceed.
The faulting and restarting associated with this are, however, relatively
expensive. The tls libraries (as currently written) exhibit approximately
worst case behavior, alternating positive and negative accesses via segment
registers with non-zero segbases. It's faster to just not use /lib/tls...
-----Original Message-----
From: xen-devel-admin@lists.sourceforge.net
[mailto:xen-devel-admin@lists.sourceforge.net] On Behalf Of Ronald G.
Minnich
Sent: Monday, February 14, 2005 3:54 PM
To: xen-devel@lists.sourceforge.net
Subject: [Xen-devel] xen and tls (fwd)
I forget the reason, can somebody fill me in?
thanks
ron
---------- Forwarded message ----------
Date: Mon, 14 Feb 2005 10:50:56 -0500
To: rminnich@lanl.gov
Subject: xen and tls
do you know why, when i booted xen, i was told to
mv /lib/tls /lib/tls.disabled
otherwise it would be very slow and may cause instability
(i.e. programmes would fault)?
-------------------------------------------------------
SF email is sponsored by - The IT Product Guide
Read honest & candid reviews on hundreds of IT Products from real users.
Discover which products truly live up to the hype. Start reading now.
http://ads.osdn.com/?ad_id=6595&alloc_id=14396&op=click
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/xen-devel
-------------------------------------------------------
SF email is sponsored by - The IT Product Guide
Read honest & candid reviews on hundreds of IT Products from real users.
Discover which products truly live up to the hype. Start reading now.
http://ads.osdn.com/?ad_id=6595&alloc_id=14396&op=click
^ permalink raw reply [flat|nested] 6+ messages in thread
* RE: xen and tls (fwd)
2005-02-14 16:52 ` Ronald G. Minnich
@ 2005-02-14 17:07 ` Barry Silverman
0 siblings, 0 replies; 6+ messages in thread
From: Barry Silverman @ 2005-02-14 17:07 UTC (permalink / raw)
To: Ronald G. Minnich, Himanshu Raj; +Cc: xen-devel
As I recall, tls maps negative offsets >0xFF000000 (from segment registers
FS, GS) into the kernel address space (IE >0xa8000000), and this conflicts
with xen's virtual address mapping.
Xen lives mapped into high virtual memory but inaccessible to the guest
kernels - however the whole guest kernel address space is accessible to xen.
Keir Fraser developed an instruction emulator/code patcher in the page fault
code that detected these large offsets from FS/GS, and then fixed them up...
I recall that the emulator ended up getting called way for often than anyone
wanted and ran slowly. The "not-to-use" TLS recommendation followed from
that.
Barry Silverman
-----Original Message-----
From: xen-devel-admin@lists.sourceforge.net
[mailto:xen-devel-admin@lists.sourceforge.net]On Behalf Of Ronald G.
Minnich
Sent: Monday, February 14, 2005 11:53 AM
To: Himanshu Raj
Cc: xen-devel@lists.sourceforge.net
Subject: Re: [Xen-devel] xen and tls (fwd)
On Mon, 14 Feb 2005, Himanshu Raj wrote:
> I believe it is due to the use of both +ve and -ve offsets by TLS. It
becomes
> a nightmare to manage it with segmentation, and system has to emulate -ve
> offsets by converting to +ve ones etc. I am still learning more on this.
I vaguely remember this discussion, I just wish I could remember more.
ron
-------------------------------------------------------
SF email is sponsored by - The IT Product Guide
Read honest & candid reviews on hundreds of IT Products from real users.
Discover which products truly live up to the hype. Start reading now.
http://ads.osdn.com/?ad_id=6595&alloc_id=14396&op=click
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/xen-devel
-------------------------------------------------------
SF email is sponsored by - The IT Product Guide
Read honest & candid reviews on hundreds of IT Products from real users.
Discover which products truly live up to the hype. Start reading now.
http://ads.osdn.com/?ad_id=6595&alloc_id=14396&op=click
^ permalink raw reply [flat|nested] 6+ messages in thread
end of thread, other threads:[~2005-02-14 17:07 UTC | newest]
Thread overview: 6+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2005-02-14 15:53 xen and tls (fwd) Ronald G. Minnich
2005-02-14 16:05 ` Anthony Liguori
2005-02-14 16:33 ` Himanshu Raj
2005-02-14 16:52 ` Ronald G. Minnich
2005-02-14 17:07 ` Barry Silverman
2005-02-14 17:02 ` Michael A Fetterman
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.