Dear all,
I was hoping that someone might be able to clear up a query I have wrt Windows memory and how it handles memory pages (specifically information regarding the pages executable permissions). I'm assuming that PAE is in use here.
The idea is that we have some page (holding virtual address addr) within a processes address space and wish to know if that page is executable.
Within user space, we can use the VADs and obtain executable information via vad.u.VadFlags.Protection - vadinfo.PROTECT_FLAGS allows the returned value (as an int) to be converted into a string representation.
Within kernel space, we can use the PTE to determine information regarding the pages exec status - for example:
def get_page_nx_bit(addrspace, addr):
pdpte = addrspace.get_pdpte(addr)
pde = addrspace.get_pde(addr, pdpte)
pte = addrspace.get_pte(addr, pde)
return pte >> 63
gets the NX bit for the PTE associated with a given page address.
Now, in comparing bit 63 of the PTE entries against the VAD protection permissions in user space, I'm noticing the occasional difference. Naively, I'd expected the PTEs to agree with the protection information on the VADs (within user space).
Any help or pointers to information resolving the above is very much appreciated.
Many thanks,
Carl.
Thanks Michael. My whole purpose in this instance was to test the nbdserver
project for my blog post. That's why I used this particular method. I'll
follow up with Jeff and let him know what you came up with.
Thanks!
Ken
On Mon, Apr 15, 2013 at 3:51 PM, Michael Cohen <scudette(a)gmail.com> wrote:
> Ah that explains the crash then....
>
> You can not read the entire physical address space from start to finish
> because you will hit DMA regions which will cause your reads to activate
> PCI devices and blue screen the box. You must skip the regions which are
> reserved to PCI devices.
>
> The winpmem device driver allows you to read where ever you want (i.e. it
> does not protect you from shooting yourself in the foot :-). The user space
> application is supposed to figure out where its safe to read. From quickly
> looking at the code for nbdserve
>
> https://github.com/jeffbryner/NBDServer/blob/master/main.cpp
>
> it does not seem to implement the correct algorithm for skipping the
> reserved memory regions (it should make an IO control to the pmem driver
> and ask it). There seems to be some code there to do it but it does not
> seem complete to me. Maybe best ask the author?
>
> Anyway for the purpose of experimenting you could just manually write a
> short bash script with dd involving the skip and seek parameters to skip
> over the reserved regions and only image the available regions. This should
> not crash.
>
> If you goal is just to obtain the image over the network, why not use
> netcat like the example demonstrates?
>
> Thanks,
> Michael.
>
>
>
>
>
> On 15 April 2013 22:34, Ken Pryor <kdpryor(a)gmail.com> wrote:
>
>> Hi Michael,
>>
>> Yes, I ran winpmem on the subject machine and allowed it to save an image
>> file to that machine successfully. In my nbdserver test, I ran winpmem -l
>> and verified the service was running. I went to the nbd Linux client and
>> began the process of imaging pmem from the subject computer via the
>> network. Running nbd-client on the Linux workstation, I assigned the pmem
>> output coming over the network to /dev/nbd0. I used the following command
>> line then to image:
>>
>> dd if=/dev/nbd0 of=./ramoutput.dd
>>
>> This has been successful on 32 bit XP machines, but it dies on the 64 bit
>> machine. If this description doesn't make sense, I'll try to do a better
>> description later this evening.
>>
>> Ken
>>
>>
>>
>> On Mon, Apr 15, 2013 at 12:16 PM, Michael Cohen <scudette(a)gmail.com>wrote:
>>
>>> Hi Ken,
>>> I have not had a chance to play with nbdserver. Are you saying that
>>> winpmem acquisition to the local disk completed ok?
>>>
>>> Did you manage to image with winpmem over a socket and netcat? Or are
>>> you trying to image to a network share?
>>>
>>> Thanks,
>>> Michael.
>>>
>>>
>>> On 15 April 2013 18:25, Ken Pryor <kdpryor(a)gmail.com> wrote:
>>>
>>>> I recently used the latest version of winpmem in conjunction with Jeff
>>>> Bryner's nbdserver to acquire ram for a couple different systems in support
>>>> of a blog post I was writing. Acquisition of 1 gb memory from an XP 32 bit
>>>> vm via the network worked perfectly.
>>>>
>>>> However, acquiring memory from a 64 bit Win 7 physical system with 12
>>>> GB ram failed. It would start okay, but would freeze up and reboot the Win
>>>> 7 system at 3.5 GB every time when being acquired using nbdserver via the
>>>> network. Using winpmem directly on the machine works successfully, but
>>>> fails on the network.
>>>>
>>>> Any suggestions as to the problem? I can provide any data or follow up
>>>> testing if needed.
>>>>
>>>> Ken
>>>>
>>>> _______________________________________________
>>>> Vol-users mailing list
>>>> Vol-users(a)volatilityfoundation.org
>>>> http://lists.volatilityfoundation.org/mailman/listinfo/vol-users
>>>>
>>>>
>>>
>>
>
Hi Michael,
Yes, I ran winpmem on the subject machine and allowed it to save an image
file to that machine successfully. In my nbdserver test, I ran winpmem -l
and verified the service was running. I went to the nbd Linux client and
began the process of imaging pmem from the subject computer via the
network. Running nbd-client on the Linux workstation, I assigned the pmem
output coming over the network to /dev/nbd0. I used the following command
line then to image:
dd if=/dev/nbd0 of=./ramoutput.dd
This has been successful on 32 bit XP machines, but it dies on the 64 bit
machine. If this description doesn't make sense, I'll try to do a better
description later this evening.
Ken
On Mon, Apr 15, 2013 at 12:16 PM, Michael Cohen <scudette(a)gmail.com> wrote:
> Hi Ken,
> I have not had a chance to play with nbdserver. Are you saying that
> winpmem acquisition to the local disk completed ok?
>
> Did you manage to image with winpmem over a socket and netcat? Or are you
> trying to image to a network share?
>
> Thanks,
> Michael.
>
>
> On 15 April 2013 18:25, Ken Pryor <kdpryor(a)gmail.com> wrote:
>
>> I recently used the latest version of winpmem in conjunction with Jeff
>> Bryner's nbdserver to acquire ram for a couple different systems in support
>> of a blog post I was writing. Acquisition of 1 gb memory from an XP 32 bit
>> vm via the network worked perfectly.
>>
>> However, acquiring memory from a 64 bit Win 7 physical system with 12 GB
>> ram failed. It would start okay, but would freeze up and reboot the Win 7
>> system at 3.5 GB every time when being acquired using nbdserver via the
>> network. Using winpmem directly on the machine works successfully, but
>> fails on the network.
>>
>> Any suggestions as to the problem? I can provide any data or follow up
>> testing if needed.
>>
>> Ken
>>
>> _______________________________________________
>> Vol-users mailing list
>> Vol-users(a)volatilityfoundation.org
>> http://lists.volatilityfoundation.org/mailman/listinfo/vol-users
>>
>>
>
Hey all,
So far I am unable to get hivedump to work specifying an offset in
either interactive or non-interactive mode:
Offset(V) Offset(P) Name
0xe1036b60 0x4948ab60
\Device\HarddiskVolume1\Windows\system32\config\SYSTEM @ 0xe1036b60
./vol.py --profile=WinXPSP3x86 -f ../20130412-194645.raw hivedump -o
0xe1036b60
**************************************************
Hive -
Last Written Key
------------------------ ---
ERROR:root:Error: invalid literal for int() with base 10: 'x'
ERROR:root:invalid literal for int() with base 10: 'x'. Try --debug for
more information.
Not specifying an offset works however:
./vol.py --profile=WinXPSP3x86 -f ../20130412-194645.raw hivedump
**************************************************
Hive \Device\HarddiskVolume1\Documents and Settings\ACS\NTUSER.DAT @
0xe1452b60
Last Written Key
------------------------ ---
2013-04-10 18:32:35+0000
CMILoadedHive-{D6ED518C-64C9-49F9-A016-8B8A7E2C051E}/AppEvents
...
Any hints on what I could be doing wrong? Thank you.
James
I recently used the latest version of winpmem in conjunction with Jeff
Bryner's nbdserver to acquire ram for a couple different systems in support
of a blog post I was writing. Acquisition of 1 gb memory from an XP 32 bit
vm via the network worked perfectly.
However, acquiring memory from a 64 bit Win 7 physical system with 12 GB
ram failed. It would start okay, but would freeze up and reboot the Win 7
system at 3.5 GB every time when being acquired using nbdserver via the
network. Using winpmem directly on the machine works successfully, but
fails on the network.
Any suggestions as to the problem? I can provide any data or follow up
testing if needed.
Ken
On 2013-04-12 15:24, Carl Pulley wrote:
> Hi James,
> you'll be able to do this using the dumpfiles plugin (it
> reconstructs the memory view of the on disk file system using the
> shared cache). I understand that this will be released with the
> Volatility 2.3 release.
>
> All the best,
>
> Carl.
Ah...guess I am out of luck until then...thank you!
James
On 2013-04-12 15:59, Michael Cohen wrote:
> you can try to install pytz :-)
>
> for example on ubuntu the package is called python-tz:
>
> apt-get install python-tz
>
> on windows you need to get pip or easy install then
>
> pip install pytz
>
> Or you can try the windows binary from the downloads section because
> it has no required dependencies.
>
> Michael.
Heh..always so easy :) Thanks...it's flying now. Last thing I think I
need is plugins...here's what the TP responds with:
subcommands:
The following plugins can be selected.
Plugin
dis Disassemble the given address space.
dt Print a symbol.
dump Hexdump an object or memory location.
grep Search an address space for keywords.
guess_profile Guess the exact windows profile using
heuristics.
imagecopy Copies a physical address space out as a raw DD
image
info Print information about various subsystems.
load_as Load address spaces into the session if its not
already loaded.
null This plugin does absolutely nothing.
peinfo Print information about a PE binary.
is there a different spot do snag the plugins, or am I barking up the
wrong tree? Thanks again for all your help on this.
James
On 2013-04-12 15:44, Michael Cohen wrote:
> Note that this actually is already available in the tech preview
> edition using a slightly different mechanism (Its specific to
> registry
> hives instead of for generally cached files):
>
>
> https://volatility.googlecode.com/svn/branches/scudette/docs/user_manual.ht…
> [3]
>
> YMMV
> Michael.
Thanks Michael...I did give that a go but got:
ImportError: No module named pytz
Betting I'm missing something. Thanks again.
James
On 2013-04-12 15:43, david nardoni wrote:
> Here is a blog post that may suit your needs
>
> http://mikemachnik.com/2013/02/16/memdumpcracking/ [3]
>
> Dave Nardoni
> dnardoni(a)gmail.com [4]
>
> On Fri, Apr 12, 2013 at 2:18 PM, James Lay <jlay(a)slave-tothe-box.net
> [5]> wrote:
>
>> Hey all,
>>
>> Topic says it...any way to dump a full registry hive to disk? My
>> plan is to import the SYSTEM and SECURITY into Cain ;) Thank you.
>>
>> James
Thank you for the info...very useful!
James
