|
A classic article from the admirable genius of Spender: http://grsecurity.net/~spender/uderef.txt in the following i'll assume that you have good knowledge of i386 protected mode details and some concepts about how modern OSs make use of them and i'll only explain what linux (2.6 in particular) does and how UDEREF modifies it to achieve best-effort (read: not perfect) userland/kernel separation.
----------------------------------------------------------------------------
of the basic protected mode resources (TSS, GDT, LDT, IDT) we'll
look at the GDT only as that's what defines all the interesting
descriptors/segments needed for UDEREF (userland has some control
over the per-process LDT but whatever is there isn't relevant as
linux will load GDT selectors upon a userland->kernel transition
and will never dereference anything in the LDT).
linux 2.6 has per-CPU GDTs, each initialized from cpu_gdt_table
in arch/i386/kernel/head.S . the kernel uses 3 descriptors there,
__KERNEL_CS for %cs, __KERNEL_DS for %ss and __USER_DS for %ds/%es
(since 2.6.20 there's a per-CPU data segment stored in %gs and in
2.6.21 it'll move into %fs, but it's not relevant for UDEREF).
of these, __KERNEL_* are DPL=0 (i.e., for kernel use only), however
__USER_DS is DPL=3, the default userland data selector loaded for
any new userland process as well.
all of these descriptors define a flat segment (0-4GB), so there's
no userland/kernel separation at the segmentation level, it's
solely up to the paging logic (i guess you already know why that's
bad, but see below). the reason that the kernel uses the default
userland data selectors/segments is that presumably there's some
performance gain when the CPU reloads a segment register with the
same value it already has (coming from a typical ELF process the
data selectors will already have __USER_DS in them). i never checked
this claim and maybe i misremember it, but in any case, regardless
of which __*_DS selector the kernel uses, they're all flat anyway.
now we're getting to the actual topic ;-). the problem i set out to
solve with UDEREF was that many kernel bugs can be exploited (at all
or more reliably) due to the fact that on i386 most OSs don't separate
the userland virtual address space from that of the kernel. this in
turn means that whenever userland can make the kernel (unexpectedly)
dereference a userland controlled pointer, userland can control the
data (and sometimes, control) flow of the kernel by virtue of providing
the attack data in its own userland address range as it's fully visible
in the kernel's virtual address space as well (the two virtual address
spaces are the same because of the use of flat segments and lack of
effective address space IDs in the i386 paging logic).
since there're two stages in logical->physical address translation,
one can attack the problem at each level. one approach is to make
use of paging by realizing that %cr3 is effectively the address space
ID on i386. that is, one can switch (reload) %cr3 on every kernel
entry to point to page tables that simply don't map userland. this
approach has been tried on linux in the so-called 4:4 VM patches
(Red Hat, Ingo Molnar) and was used/offered in some RHEL series (it
wasn't for security per se, but to provide a 4GB effective userland
virtual address space for aging 32-bit machines with lots of memory).
it has one big problem: performance impact. so much that it wasn't
even on the table for me.
the other approach is to make use of segmentation, which is what i
do in UDEREF. the basic idea is very simple and very old, i first
saw it in use (don't laugh ;-) on windows 95 where this particular
feature caught NULL userland derefs in win32 processes (and prevented
them from trashing the low DOS related memory). the trick they used
was that they set up the userland segments as so-called expand-down
segments, that is, the limit field in the segment descriptor meant
the lowest valid address for the segment, so they could set it to
something low, i forget the exact number now, but it was either 4MB
or 4k.
the same trick is used in UDEREF with a few more modifications. the
new segment setup is as follows:
kernel %cs:
still __KERNEL_CS which remains flat (there's another PaX feature
which limits it properly, but let's not digress)
kernel %ss/%ds/%es:
all switched to __KERNEL_DS which is no longer flat but turned
into an expand-down segment with a lower limit set at the top of
the userland address space (__PAGE_OFFSET in linux, normally it's
at 3GB)
userland %cs: stays the same __USER_CS, flat, doesn't matter
userland %ds/%es/%ss:
stays as __USER_DS but limited (most of the time) to the userland
address space size.
this setup so far is enough to cause a GPF whenever the kernel tries
to dereference a pointer value that falls into the userland virtual
address range (say, 0-3GB, i.e., NULL pointers are automatically
included). what it's not enough for is the need to allow data transfer
between userland/kernel and kernel/kernel in certain contexts. let's
see them one by one.
as you know, syscalls often need to copy data to/from userland which
on i386 is most often implemented by the (rep) movs insn. this insn
uses two data segments, one for the source, one for the destination
(former can be overridden). this blind movs works in the usual case
because all segments are flat, so data can freely move between kernel
and userland virtual addresses.
with the above described setup however we have a problem since
neither %ds nor %es allow userland access. furthermore, even if one
could override the source segment (so that one could use one of the
remaining segment registers for copying from userland), it wouldn't
help with copying to userland. the solution is that in these special
copy functions (on linux it's all in asm) we reload the proper segment
register with __USER_DS for the duration of the copy therefore we allow
the copy to succeed *and* at the same time not allow kernel addresses
where they're not expected (e.g., when the kernel copies from userland,
this won't allow a source pointer to kernel land and vice versa).
we're still not done however, as there's a special kernel/kernel copy
problem (i don't know if it's specific to linux): certain internal
kernel functions call syscall functions more or less directly (say,
setting up a kernel thread or even just starting init makes use of
the fork interface). this among others means that any normal userland
pointer input these syscall expect will fall into the kernel address
range - and would fail input validation right at the beginning, were
it not for a trick linux uses: set_fs().
there's some history behind this function because in early days linux
actually used to be more secure in this regard as it used segmentation
logic for userland/kernel separation (yes, there's some irony of fate
here ;-) and it provided this set_fs() call to tell the kernel that
further userland/kernel memory copy operations will actually stay within
the kernel. many years later this interface was effectively switched off
because linux turned to flat segments (the baby went with the bathwater
or something like that) however one of its features stayed: the address
space limit check in the low level copying functions. this limit check
simply compared the to-be-dereferenced pointer against the current
address space limit allowed for such copying (set_fs() itself was reduced
to change only this limit as needed).
with the non-flat segment setup described above we would still pass the
address space limit check because it's not directly derived from the
GDT descriptor limit but a per-thread variable, however we would fail
the actual copy because one of the segment registers is reloaded with
__USER_DS and that won't allow kernel addresses. the assumption i broke
is that the separately maintained address space limit corresponds to
the actual segment limit therefore the fix has to be that somehow this
segment limit needs to be updated every time set_fs() is called.
this can be accomplished by either updating the limit on __USER_DS (so
that the actual copy functions don't have to change) or by loading either
__USER_DS or __KERNEL_DS in said copy functions (then each one of them
has to be patched for this).
i went with the former for two reasons. one, it's a lot less code/work to
update the (per-CPU) GDT descriptor vs. changing all the copy functions
to conditonally load either __USER_DS or __KERNEL_DS (it took some time
to find most/all such copy code, besides the obvious copy_{to/from}_user
functions linux has a special IP checksumming code used for both kernel
and userland buffers, then there were some direct userland address accesses
in reboot code (to patch BIOS variables), etc).
two, i wasn't actually sure that set_fs() is only ever called for
kernel/kernel copy situatons (since it only raises the allowed address
space limit, therefore on i386 it'd still allow userland accesses), so
doing the 'conditional load in the copy code' stuff just to find out it
was in vain wasn't worth it. i still want to try it out though because
this is the reason i called the current approach 'best-effort' separation
only: in set_fs() context a kernel bug has actually free reign over
pointers in the copy functions (yes, it's still a lot smaller attack
surface but i'm a maximalist ;-).
so, as a summary: UDEREF makes sure that (data) segments for userland and
the kernel are properly limited, either upwards (userland) or downwards
(kernel). furthermore, every userland/kernel copy code is patched to use
the proper segments for the inter-segment copy (the challenging part was
to find all such code ;-). as a linux specialty, some care is taken to
ensure that the same copy code works for kernel/kernel copying as well.
note that GDT patching can be done safely because linux 2.6 has per-CPU
GDTs (and in the context switch code i take care of the __USER_DS limit
based on the set_fs() limit), on linux 2.4 and earlier this cannot be
done (there's one shared GDT only) and there all the copy functions are
potential escape points.
----------------------------------------------------------------------------
that'd be it in a nutshell, feel free to ask me if you have questions.
cheers,
PaX Team
To: pagee...@freemail.hu
Cc: Elad Efrat
Subject: Re: PaX uderef
Date: Wed, 4 Apr 2007 15:14:45 +0900 (JST)
From: [censored]
> On 3 Apr 2007 at 20:28, Elad Efrat wrote:
>
> > [censored],
> >
> > I've cc'd the PaX author to this mail so he can elaborate on what uderef
> > is, how it works, and why it was introduced.
>
> hello guys,
>
> in the following i'll assume that you have good knowledge of i386
> protected mode details and some concepts about how modern OSs make
> use of them and i'll only explain what linux (2.6 in particular)
> does and how UDEREF modifies it to achieve best-effort (read: not
> perfect) userland/kernel separation.
thanks for explanation.
> there's some history behind this function because in early days linux
> actually used to be more secure in this regard as it used segmentation
> logic for userland/kernel separation (yes, there's some irony of fate
> here ;-) and it provided this set_fs() call to tell the kernel that
> further userland/kernel memory copy operations will actually stay within
> the kernel. many years later this interface was effectively switched off
> because linux turned to flat segments (the baby went with the bathwater
> or something like that) however one of its features stayed: the address
> space limit check in the low level copying functions. this limit check
> simply compared the to-be-dereferenced pointer against the current
> address space limit allowed for such copying (set_fs() itself was reduced
> to change only this limit as needed).
loading selectors is rather expensive operation.
well, it might be cheaper than reloading of cr3,
but it's still something i (and linux folks i guess) don't want to do
on every copying to/from userspace.
do you have some numbers of the performance penalty?
> note that GDT patching can be done safely because linux 2.6 has per-CPU
> GDTs (and in the context switch code i take care of the __USER_DS limit
> based on the set_fs() limit), on linux 2.4 and earlier this cannot be
> done (there's one shared GDT only) and there all the copy functions are
> potential escape points.
what happens when you slept at some point after set_fs(), and woke up
on another cpu?
[censored]
From: pagee...@freemail.hu
To: [censored]
Date: Wed, 04 Apr 2007 11:21:52 +0200
Subject: Re: PaX uderef
Reply-to: pagee...@freemail.hu
CC: Elad Efrat
On 4 Apr 2007 at 15:14, [censored] wrote:
> loading selectors is rather expensive operation.
> well, it might be cheaper than reloading of cr3,
the cr3 reload approach is a lot slower not only because it flushes
the TLBs on every user/kernel transition, but because it also has to
handle the very same user/kernel copy situation: due to the address
space switch, userland virtual addresses are no longer directly
accessible and require special mappings to be set up for the duration
of the copy. that means TLB entry invalidations which is one of the
slowest operations of the CPU.
> but it's still something i (and linux folks i guess) don't want to do
> on every copying to/from userspace.
> do you have some numbers of the performance penalty?
i don't have micro-benchmarks on these functions per se, only the
usual kernbench which didn't show any measurable impact (which is
to be expected as it is not heavy on user/kernel copying, most time
is spent in userland and the most used syscalls that do such copying
are already expensive enough that any impact from UDEREF would
disappear in the noise). if you have an idea for which syscall to
exercise for this (i.e., something that spends most of its time in
copying), let me know.
> > note that GDT patching can be done safely because linux 2.6 has per-CPU
> > GDTs (and in the context switch code i take care of the __USER_DS limit
> > based on the set_fs() limit), on linux 2.4 and earlier this cannot be
> > done (there's one shared GDT only) and there all the copy functions are
> > potential escape points.
>
> what happens when you slept at some point after set_fs(), and woke up
> on another cpu?
both sleeping and waking up result in a context switch on each
affected CPU (a call to __switch_to() in particular), therefore
as i stated, the changes i made to that function ensure that the
limit of __USER_DS of the given CPU's GDT will be updated to
reflect the address space limit of the incoming thread (the limit
of __USER_DS becomes part of the thread's state information as it
now faithfully follows the already existing thread_info->addr_limit
set by set_fs() - you can think of UDEREF as hardware enforcement
of set_fs(), among others).
To: pagee...@freemail.hu
Cc: Elad Efrat
Subject: Re: PaX uderef
Date: Wed, 4 Apr 2007 18:53:18 +0900 (JST)
From: [censored]
> if you have an idea for which syscall to
> exercise for this (i.e., something that spends most of its time in
> copying), let me know.
getresuid?
> > > note that GDT patching can be done safely because linux 2.6 has per-CPU
> > > GDTs (and in the context switch code i take care of the __USER_DS limit
> > > based on the set_fs() limit), on linux 2.4 and earlier this cannot be
> > > done (there's one shared GDT only) and there all the copy functions are
> > > potential escape points.
> >
> > what happens when you slept at some point after set_fs(), and woke up
> > on another cpu?
>
> both sleeping and waking up result in a context switch on each
> affected CPU (a call to __switch_to() in particular), therefore
> as i stated, the changes i made to that function ensure that the
> limit of __USER_DS of the given CPU's GDT will be updated to
> reflect the address space limit of the incoming thread (the limit
> of __USER_DS becomes part of the thread's state information as it
> now faithfully follows the already existing thread_info->addr_limit
> set by set_fs() - you can think of UDEREF as hardware enforcement
> of set_fs(), among others).
ok, i have missed "in the context switch code ..." in your
explanation. thanks.
[censored]
From: pagee...@freemail.hu
To: [censored]
Date: Sun, 08 Apr 2007 18:40:19 +0200
Subject: Re: PaX uderef
CC: Elad Efrat
On 4 Apr 2007 at 18:53, [censored] wrote:
> > if you have an idea for which syscall to
> > exercise for this (i.e., something that spends most of its time in
> > copying), let me know.
>
> getresuid?
sorry for the delay but i didn't have a native linux install and just
quickly put it on a new laptop. the box has a T7200 (core2/2GHz) with
2GB of RAM. the test was a simple looping around getresuid() with valid
pointer arguments, 10 million times (everything was compiled with gentoo's
gcc 4.1.1). i tested the following kernels:
1. 2.6.17-gentoo-r7 (from the gentoo 2006.1 livecd)
real 0m1.440s 0m1.421s 0m1.422s 0m1.415s 0m1.426s 0m1.425s 0m1.424s
0m1.421s
user 0m0.700s 0m0.680s 0m0.800s 0m0.670s 0m0.750s 0m0.740s 0m0.690s
0m0.750s
sys 0m0.720s 0m0.740s 0m0.620s 0m0.750s 0m0.670s 0m0.690s 0m0.730s
0m0.670s
2. 2.6.21-rc6 (latest on kernel.org, without PARAVIRT)
real 0m1.477s 0m1.471s 0m1.466s 0m1.466s 0m1.428s 0m1.471s 0m1.416s
0m1.466s
user 0m0.808s 0m0.876s 0m0.784s 0m0.792s 0m0.704s 0m0.960s 0m0.668s
0m0.752s
sys 0m0.648s 0m0.596s 0m0.680s 0m0.676s 0m0.724s 0m0.512s 0m0.748s
0m0.712s
3. 2.6.21-rc6 (latest on kernel.org, with PARAVIRT)
real 0m3.277s 0m3.258s 0m3.258s 0m3.268s 0m3.258s 0m3.268s 0m3.258s
0m3.258s
user 0m1.840s 0m1.884s 0m1.868s 0m2.000s 0m1.840s 0m2.040s 0m1.828s
0m1.756s
sys 0m1.416s 0m1.376s 0m1.388s 0m1.268s 0m1.416s 0m1.228s 0m1.428s
0m1.500s
4. 2.6.21-rc6-pax (not public yet, without PARAVIRT, no PaX enabled)
real 0m1.976s 0m1.949s 0m1.948s 0m1.928s 0m1.949s 0m1.948s 0m1.949s
0m1.948s
user 0m0.868s 0m0.864s 0m0.972s 0m0.716s 0m0.896s 0m0.848s 0m0.824s
0m0.780s
sys 0m1.080s 0m1.084s 0m0.976s 0m1.212s 0m1.052s 0m1.100s 0m1.124s
0m1.168s
5. 2.6.21-rc6-pax (not public yet, without PARAVIRT, with UDEREF)
real 0m2.038s 0m2.029s 0m2.030s 0m1.991s 0m2.025s 0m1.991s 0m2.030s
0m2.029s
user 0m0.880s 0m0.940s 0m0.824s 0m0.728s 0m0.784s 0m0.864s 0m0.796s
0m0.752s
sys 0m1.148s 0m1.088s 0m1.204s 0m1.264s 0m1.240s 0m1.124s 0m1.236s
0m1.276s
6. 2.6.21-rc6-pax (not public yet, with PARAVIRT, with UDEREF)
real 0m3.696s 0m3.685s 0m3.685s 0m3.686s 0m3.719s 0m3.690s 0m3.700s
0m3.696s
user 0m1.952s 0m1.936s 0m1.844s 0m1.832s 0m1.804s 0m1.632s 0m1.816s
0m1.892s
sys 0m1.732s 0m1.748s 0m1.840s 0m1.852s 0m1.912s 0m2.056s 0m1.880s
0m1.800s
some notes:
- the difference between 1. and 2. is probably due to some .config option
(e.g., i enabled syscall auditing myself)
- 2. and 3. show the effect of PARAVIRT (not that i wanted to test it
per se but i accidentally left it enabled in PaX first so i reran
everything twice ;P)
- 3. and 4. show that some PaX features are not controlled by .config,
among others the core part of UDEREF (segment reloads and overrides
on the userland accessors) is always 'on'
- 4. and 5. show that actually enabling UDEREF (changing the __USER_DS
limit) has a small impact over just 4. alone (i.e., segment register
reloads and overrides), so there seems to be some credence to the theory
that the speed of segment reloads does depend on the actual descriptor
content
To: pagee...@freemail.hu
Cc: Elad Efrat
Subject: Re: PaX uderef
Date: Tue, 10 Apr 2007 17:49:33 +0900 (JST)
From: [censored]
hi,
> On 4 Apr 2007 at 18:53, [censored] wrote:
>
> > > if you have an idea for which syscall to
> > > exercise for this (i.e., something that spends most of its time in
> > > copying), let me know.
> >
> > getresuid?
>
> sorry for the delay but i didn't have a native linux install and just
> quickly put it on a new laptop. the box has a T7200 (core2/2GHz) with
> 2GB of RAM. the test was a simple looping around getresuid() with valid
> pointer arguments, 10 million times (everything was compiled with gentoo's
> gcc 4.1.1). i tested the following kernels:
>
> 1. 2.6.17-gentoo-r7 (from the gentoo 2006.1 livecd)
> real 0m1.440s 0m1.421s 0m1.422s 0m1.415s 0m1.426s 0m1.425s 0m1.424s
> 0m1.421s
> user 0m0.700s 0m0.680s 0m0.800s 0m0.670s 0m0.750s 0m0.740s 0m0.690s
> 0m0.750s
> sys 0m0.720s 0m0.740s 0m0.620s 0m0.750s 0m0.670s 0m0.690s 0m0.730s
> 0m0.670s
does these columns mean you ran the same test 8 times?
> 2. 2.6.21-rc6 (latest on kernel.org, without PARAVIRT)
> real 0m1.477s 0m1.471s 0m1.466s 0m1.466s 0m1.428s 0m1.471s 0m1.416s
> 0m1.466s
> user 0m0.808s 0m0.876s 0m0.784s 0m0.792s 0m0.704s 0m0.960s 0m0.668s
> 0m0.752s
> sys 0m0.648s 0m0.596s 0m0.680s 0m0.676s 0m0.724s 0m0.512s 0m0.748s
> 0m0.712s
>
> 3. 2.6.21-rc6 (latest on kernel.org, with PARAVIRT)
> real 0m3.277s 0m3.258s 0m3.258s 0m3.268s 0m3.258s 0m3.268s 0m3.258s
> 0m3.258s
> user 0m1.840s 0m1.884s 0m1.868s 0m2.000s 0m1.840s 0m2.040s 0m1.828s
> 0m1.756s
> sys 0m1.416s 0m1.376s 0m1.388s 0m1.268s 0m1.416s 0m1.228s 0m1.428s
> 0m1.500s
>
> 4. 2.6.21-rc6-pax (not public yet, without PARAVIRT, no PaX enabled)
> real 0m1.976s 0m1.949s 0m1.948s 0m1.928s 0m1.949s 0m1.948s 0m1.949s
> 0m1.948s
> user 0m0.868s 0m0.864s 0m0.972s 0m0.716s 0m0.896s 0m0.848s 0m0.824s
> 0m0.780s
> sys 0m1.080s 0m1.084s 0m0.976s 0m1.212s 0m1.052s 0m1.100s 0m1.124s
> 0m1.168s
>
> 5. 2.6.21-rc6-pax (not public yet, without PARAVIRT, with UDEREF)
> real 0m2.038s 0m2.029s 0m2.030s 0m1.991s 0m2.025s 0m1.991s 0m2.030s
> 0m2.029s
> user 0m0.880s 0m0.940s 0m0.824s 0m0.728s 0m0.784s 0m0.864s 0m0.796s
> 0m0.752s
> sys 0m1.148s 0m1.088s 0m1.204s 0m1.264s 0m1.240s 0m1.124s 0m1.236s
> 0m1.276s
>
> 6. 2.6.21-rc6-pax (not public yet, with PARAVIRT, with UDEREF)
> real 0m3.696s 0m3.685s 0m3.685s 0m3.686s 0m3.719s 0m3.690s 0m3.700s
> 0m3.696s
> user 0m1.952s 0m1.936s 0m1.844s 0m1.832s 0m1.804s 0m1.632s 0m1.816s
> 0m1.892s
> sys 0m1.732s 0m1.748s 0m1.840s 0m1.852s 0m1.912s 0m2.056s 0m1.880s
> 0m1.800s
>
> some notes:
>
> - the difference between 1. and 2. is probably due to some .config option
> (e.g., i enabled syscall auditing myself)
>
> - 2. and 3. show the effect of PARAVIRT (not that i wanted to test it
> per se but i accidentally left it enabled in PaX first so i reran
> everything twice ;P)
i wonder why it affects user time..
> - 3. and 4. show that some PaX features are not controlled by .config,
> among others the core part of UDEREF (segment reloads and overrides
> on the userland accessors) is always 'on'
2. and 4. ?
> - 4. and 5. show that actually enabling UDEREF (changing the __USER_DS
> limit) has a small impact over just 4. alone (i.e., segment register
> reloads and overrides), so there seems to be some credence to the theory
> that the speed of segment reloads does depend on the actual descriptor
> content
hm, interesting.
thanks for the info.
[censored]
From: pagee...@freemail.hu
To: [censored]
Date: Tue, 10 Apr 2007 11:42:49 +0200
Subject: Re: PaX uderef
CC: Elad Efrat
On 10 Apr 2007 at 17:49, [censored] wrote:
> > 1. 2.6.17-gentoo-r7 (from the gentoo 2006.1 livecd)
> > real 0m1.440s 0m1.421s 0m1.422s 0m1.415s 0m1.426s 0m1.425s 0m1.424s
> > 0m1.421s
> > user 0m0.700s 0m0.680s 0m0.800s 0m0.670s 0m0.750s 0m0.740s 0m0.690s
> > 0m0.750s
> > sys 0m0.720s 0m0.740s 0m0.620s 0m0.750s 0m0.670s 0m0.690s 0m0.730s
> > 0m0.670s
>
> does these columns mean you ran the same test 8 times?
yes, each column is one run of 'time ./test', i thought it'd be easier
to compare the times when arranged this way.
> > - 2. and 3. show the effect of PARAVIRT (not that i wanted to test it
> > per se but i accidentally left it enabled in PaX first so i reran
> > everything twice ;P)
>
> i wonder why it affects user time..
my only guess is that PARAVIRT disables the vdso (by default, and i didn't
explicitly enable it) so that has an effect on the syscall itself (int 80h
vs. sysenter).
> > - 3. and 4. show that some PaX features are not controlled by .config,
> > among others the core part of UDEREF (segment reloads and overrides
> > on the userland accessors) is always 'on'
>
> 2. and 4. ?
yes i meant those indeed.
|
