reactos-kernel

[ros-kernel] Re: page faults in msvcrt.dll

[<we...@cw...>]

Philip Susi writes
>
>We need to fix that then, and have all demand zero allocations be >satisfied
from the zeroed pool.
>
Not doing the zero'ing is only an optimization, it doesn't affect
correctness.

====================================================
= To remove yourself from this mailing list, go to =
= http://www.reactos.com/home/mailing.html         =
====================================================

[ros-kernel] Re: page faults in msvcrt.dll

[Guido de J. <gui...@ch...>]

On Sunday 28 October 2001 10:11, we...@cw... wrote:
> Philip Susi writes
>
> >We need to fix that then, and have all demand zero allocations be
> > >satisfied
>
> from the zeroed pool.
>
> Not doing the zero'ing is only an optimization, it doesn't affect
> correctness.

OK, maybe I don't understand, but if you demand a zero'ed page and you get a 
non-zero'ed page, it seems me to that that is not correct?

Also, if someone requests and gets a non-zero'ed page, then the page might 
still contain data of another process or of the kernel (which they no longer 
need, or maybe it got swapped out). This means that someone might gain access 
to data that it would normally not be able to read. OK, interpreting the data 
is very difficult, but in theory this could be a security problem.

OTOH, zeroing all free pages is costly indeed.

====================================================
= To remove yourself from this mailing list, go to =
= http://www.reactos.com/home/mailing.html         =
====================================================

[ros-kernel] Re: page faults in msvcrt.dll

[Phillip S. <ps...@cf...>]

It most certainly does effect correctness;  it is a security hole.  MS 
zeros all pages before they are allocated to keep programs from gaining 
access to data they shouldn't and comply with C2 security 
specifications.  The optimization is to keep pages on the zeroed list that 
are already zeroed so you don't have to zero them when they are needed, you 
zero them when otherwise idle.

At 09:11 AM 10/28/2001 +0000, you wrote:
>Philip Susi writes
> >
> >We need to fix that then, and have all demand zero allocations be >satisfied
>from the zeroed pool.
> >
>Not doing the zero'ing is only an optimization, it doesn't affect
>correctness.

====================================================
= To remove yourself from this mailing list, go to =
= http://www.reactos.com/home/mailing.html         =
====================================================

[ros-kernel] Re: page faults in msvcrt.dll

[Rex J. <re...@lv...>]

At 10:48 AM 10/28/01 -0500, you wrote:
>It most certainly does effect correctness;  it is a security hole.

It effects correctness only if your requirement is to have C2 
security.  This wont be the
case for everyone using ReactOS however.  Personally I would rather see 
speed over
security in this case.  in VMS it is the allocators responsibility to mark 
the pages as
needing zeroing before reuse.  Too bad we cannot implement something like that.


Rex Jolliff
re...@lv...

====================================================
= To remove yourself from this mailing list, go to =
= http://www.reactos.com/home/mailing.html         =
====================================================

[ros-kernel] Re: page faults in msvcrt.dll

[Jakob E. <ja...@vm...>]

On Sun, Oct 28, 2001 at 10:21:50AM -0800, Rex Jolliff wrote:
> At 10:48 AM 10/28/01 -0500, you wrote:
> >It most certainly does effect correctness;  it is a security hole.
> 
> It effects correctness only if your requirement is to have C2 
> security.  This wont be the
> case for everyone using ReactOS however.  Personally I would rather see 
> speed over
> security in this case.  in VMS it is the allocators responsibility to mark 
> the pages as
> needing zeroing before reuse.  Too bad we cannot implement something like that.
> 

This will probably lead to problems running programs.
I can bet my hat on that some programs rely on that what they allocated
is zeroed out.

regards,
Jakob Eriksson


====================================================
= To remove yourself from this mailing list, go to =
= http://www.reactos.com/home/mailing.html         =
====================================================

[ros-kernel] Re: page faults in msvcrt.dll

[Phillip S. <ps...@cf...>]

What do you mean about VMS?  The kernel simply takes dirty pages and places 
them on the free list to mark them as needing zeroed.  The idle thread 
comes by and zeroes pages on the free list, moving them to the zeroed list, 
so cpu time that otherwise is wasted it put to use zeroing pages, and it 
prevents a rather large security hole.  What if, for instance, your 
password is cached by the LSA, and someone wrote an app to churn lots of 
memory around and looking at the pages the VM gives it searching for your 
cached password.  Clearly this is something we wish to prevent.  Not to 
mention that many things, such as the .bss segments of executables assume 
that pages are zeroed first.

At 10:21 AM 10/28/2001 -0800, you wrote:
>At 10:48 AM 10/28/01 -0500, you wrote:
>>It most certainly does effect correctness;  it is a security hole.
>
>It effects correctness only if your requirement is to have C2 
>security.  This wont be the
>case for everyone using ReactOS however.  Personally I would rather see 
>speed over
>security in this case.  in VMS it is the allocators responsibility to mark 
>the pages as
>needing zeroing before reuse.  Too bad we cannot implement something like 
>that.
>
>
>Rex Jolliff
>re...@lv...

====================================================
= To remove yourself from this mailing list, go to =
= http://www.reactos.com/home/mailing.html         =
====================================================

[ros-kernel] Re: page faults in msvcrt.dll

[Robert C. <rob...@it...>]

It'll also affect correctness for programs that assume zero'd pages _are
zero'd_.

Rob
----- Original Message -----
From: "Rex Jolliff" <re...@lv...>
To: <ros...@re...>
Sent: Monday, October 29, 2001 5:21 AM
Subject: [ros-kernel] Re: page faults in msvcrt.dll


> At 10:48 AM 10/28/01 -0500, you wrote:
> >It most certainly does effect correctness;  it is a security hole.
>
> It effects correctness only if your requirement is to have C2
> security.  This wont be the
> case for everyone using ReactOS however.  Personally I would rather
see
> speed over
> security in this case.  in VMS it is the allocators responsibility to
mark
> the pages as
> needing zeroing before reuse.  Too bad we cannot implement something
like that.
>
>
> Rex Jolliff
> re...@lv...
>
> ====================================================
> = To remove yourself from this mailing list, go to =
> = http://www.reactos.com/home/mailing.html         =
> ====================================================
>
>

====================================================
= To remove yourself from this mailing list, go to =
= http://www.reactos.com/home/mailing.html         =
====================================================

[ros-kernel] Re: page faults in msvcrt.dll

[Frank D. E. Jr. <fd...@ya...>]

Why not make it an option for the system administrator?

--- Robert Collins <rob...@it...> wrote:
> It'll also affect correctness for programs that assume zero'd pages
> _are
> zero'd_.
> 
> Rob
> ----- Original Message -----
> From: "Rex Jolliff" <re...@lv...>
> To: <ros...@re...>
> Sent: Monday, October 29, 2001 5:21 AM
> Subject: [ros-kernel] Re: page faults in msvcrt.dll
> 
> 
> > At 10:48 AM 10/28/01 -0500, you wrote:
> > >It most certainly does effect correctness;  it is a security hole.
> >
> > It effects correctness only if your requirement is to have C2
> > security.  This wont be the
> > case for everyone using ReactOS however.  Personally I would rather
> see
> > speed over
> > security in this case.  in VMS it is the allocators responsibility
> to
> mark
> > the pages as
> > needing zeroing before reuse.  Too bad we cannot implement
> something
> like that.
> >
> >
> > Rex Jolliff
> > re...@lv...
> >
> > ====================================================
> > = To remove yourself from this mailing list, go to =
> > = http://www.reactos.com/home/mailing.html         =
> > ====================================================
> >
> >
> 
> ====================================================
> = To remove yourself from this mailing list, go to =
> = http://www.reactos.com/home/mailing.html         =
> ====================================================
> 


=====
=======
Frank D. Engel, Jr.
Please note my new address: fd...@ya...

__________________________________________________
Do You Yahoo!?
Make a great connection at Yahoo! Personals.
http://personals.yahoo.com

====================================================
= To remove yourself from this mailing list, go to =
= http://www.reactos.com/home/mailing.html         =
====================================================