Software Diagnostics Library

WinDbg help states that NMI_HARDWARE_FAILURE (0×80) bugcheck 80 indicates a hardware fault. This description can easily lead to a conclusion that a kernel or complete crash dump you just got from your customer doesn’t worth examining. But hardware malfunction is not always the case especially if your customer mentions that their system was hanging and they forced a manual dump. Here I would advise to check whether they have a special hardware for debugging purposes, for example, a card or an integrated iLO chip (Integrated Lights-Out) for remote server administration. Both can generate NMI (Non Maskable Interrupt) on demand and therefore bugcheck the system. If this is the case then it is worth examining their dump to see why the system was hanging.

- Dmitry Vostokov -

Welcome to the new blog location at dumpanalysis.org/blog/ 

Its feed address is

http://feeds.feedburner.com/CrashDumpAnalysis

The blog has been moved from its original location at

citrite.org/blogs/dmitryv/ 

in order to bring all crash dump analysis and debugging information to one place including www.dumpanalysis.org/forum and the forthcoming online encyclopedia about assembly languages:

www.asmpedia.org

Thank you and sorry for any inconvenience this might have caused.

Merry Christmas and Happy Debugging in New Year!

- Dmitry Vostokov -

Now it’s time to ”introduce” Invalid Pointer pattern. It’s just a number saved in a register or in a memory location and when we try to interpret it as a memory address itself and follow it (dereference) to fetch memory contents (value) it points to, OS with the help of hardware tells us that the address doesn’t exist or inaccessible due to security restrictions. The following two slides from my old presentation depict the concept of a pointer:

Pointer definition
Pointers depicted

In Windows you have your process memory partitioned into two big regions: kernel space and process space. Space partition is a different concept than execution mode (kernel or user, ring 0 or ring 3) which is a processor state. Code executing in kernel mode (a driver or OS, for example) can access memory that belongs to user space.

Based on this we can make distinction between invalid pointers containing kernel space addresses (start from 0×80000000 on x86, no /3Gb switch) and invalid pointers containing user space addresses (below 0×7FFFFFFF).

On Windows x64 user space addresses are below 0×0000070000000000 and kernel space addresses start from 0xFFFF080000000000.

When you dereference invalid kernel space address you get bugcheck immediately:

UNEXPECTED_KERNEL_MODE_TRAP (7f)

PAGE_FAULT_IN_NONPAGED_AREA (50)

There is no way you can catch it in your code (by using SEH).

However when you dereference user space address the course of action depends on whether your processor is in kernel mode (ring 0) or in user mode (ring 3). In any mode you can catch the exception (by using appropriate SEH handler) or leave this to the operating system or debugger. If there was no component willing to process the exception when it happened in user mode you get your process crash and in kernel mode you get bugchecks:

SYSTEM_THREAD_EXCEPTION_NOT_HANDLED (7e) 

KERNEL_MODE_EXCEPTION_NOT_HANDLED (8e)

I summarized all of this on the following diagram: 

NULL pointer is a special class of user space pointers. Usually its value is in the range of 0×00000000 - 0×0000FFFF. You can see them used in instructions like

mov   esi, dword ptr [ecx+0×10] 

and ecx value is 0×00000000 so you try to access the value located at 0×00000010 memory address.

When you get a crash dump and you see an invalid pointer pattern the next step is to interpret the pointer value which should help in understanding possible steps that led to the crash. Pointer value interpretation is the subject of the next part.

- Dmitry Vostokov @ DumpAnalysis.org -

I consider programmers as real programmers only if they attempted to write something like an editor or a word processor. My favorite interview question is “Did you write a word processor?” This probably explains why my team is small

Why? Because writing a word processor shows your determination, persistence and if you are successful - understanding of OO principles. Even Gang of Four in their seminal book

Design Patterns: Elements of Reusable Object-Oriented Software

used word processor as a unified example. I was very satisfied when I read their book - I designed and implemented at least 2 word processors: one in pure C meant to be better than MS Write 3.x and one in Java with syntax highlighting. And I implemented design patterns without knowing they were design patterns. The Java one was better and here is the screen shot showing my editor with C/C++ syntax highlighting implemented in Java 1.0 without any 3rd-party class libraries:

The earlier one (with Windows 3.x look and feel) was written in C and has MDI interface, floating embedded objects and printing support:

- Dmitry Vostokov -

The next pattern I would like to talk about is Optimized Code. If you have such cases you should not trust your crash dump analysis tools like WinDbg. Always suspect that compiler generated code might have been optimized if you see any suspicious or strange behaviour of your tool. Let’s consider this fragment of stack:

Args to Child
77e44c24 000001ac 00000000 ntdll!KiFastSystemCallRet
000001ac 00000000 00000000 ntdll!NtFsControlFile+0xc
00000034 00000bb8 0013e3f4 kernel32!WaitNamedPipeW+0x2c3
0016fc60 00000000 67c14804 MyModule!PipeCreate+0x48

3rd-party function PipeCreate from MyModule opens a named pipe and its first parameter (0016fc60) points to a pipe name L”\\.\pipe\MyPipe”. Inside the source code it calls Win32 API function WaitNamedPipeW (to wait for the pipe to be available for connection) and passes the same pipe name. But we see that  the first parameter to WaitNamedPipeW is 00000034 which cannot be the pointer to a valid Unicode string. And the program should have been crashed if 00000034 were a pointer value.

Everything becomes clear if we look at WaitNamedPipeW disassembly (comments are mine):

0:000> uf kernel32!WaitNamedPipeW
mov     edi,edi
push    ebp
mov     ebp,esp
sub     esp,50h
push    dword ptr [ebp+8]  ; Use pipe name
lea     eax,[ebp-18h]
push    eax
call    dword ptr [kernel32!_imp__RtlCreateUnicodeString (77e411c8)]
…
…
…
…
call    dword ptr [kernel32!_imp__NtOpenFile (77e41014)]
cmp     dword ptr [ebp-4],edi
mov     esi,eax
jne     kernel32!WaitNamedPipeW+0×1d5 (77e93316)
cmp     esi,edi
jl      kernel32!WaitNamedPipeW+0×1ef (77e93331)
movzx   eax,word ptr [ebp-10h]
mov     ecx,dword ptr fs:[18h]
add     eax,0Eh
push    eax
push    dword ptr [kernel32!BaseDllTag (77ecd14c)]
mov     dword ptr [ebp+8],eax  ; reuse parameter slot

As we know [ebp+8] is the first function parameter in non-FPO calls:

Parameters and Local Variables

And we see it is reused because after we convert LPWSTR to UNICODE_STRING and call NtOpenFile to get a handle we no longer need our parameter slot and the compiler can reuse it to store other information.

There is another compiler optimization we should be aware of and it is called OMAP. It moves the code inside the code section and puts the most frequently accessed code fragments together. In that case if you type in WinDbg, for example, 

0:000> uf nt!someFunction

you get different code than if you type (assuming f4794100 is the address of the function you obtained from stack or disassembly)

0:000> uf f4794100

In conclusion the advice is to be alert and conscious during crash dump analysis and inspect any inconsistencies closer.

Happy debugging!

- Dmitry Vostokov @ DumpAnalysis.org -

I believe every Citrix user experienced clipboard breaks at least once. I remember my frustration when I coudn’t copy between Outlook and Vantive sessions and so 2.5 years ago I wrote RepairCBDChain tool to help to temporary restore clipboard functionality. Recently this feature was incorporated into ICA client. You can read about it in the client readme file (1. … [From 9.100][#112636]). However it is not enabled by default and if you experience clipboard breaks on the server side or you want to restore clipboard functionality immediately on your client without closing your session to apply changes to appsrv.ini or simply you are still using an old client then you can still benefit from this tool.

A month ago I promised to explain how my tool works. You all know that primary method for notifying windows about various events is window message mechanism. One of these notification events is clipboard notification message: WM_DRAWCLIPBOARD. Usually applications do not know whether clipboard content has changed if another program copied new data. Generally if you open Edit menu you see Paste enabled if there is data in the clipboard. This is done by application code itself by checking if clipboard is non-empty. If the application finds that clipboard is non-empty indeed it enables Paste menu item or disables it otherwise. In case of ICA client (wfica32.exe) it needs to know whether clipboard contains new data in order to send it down via ICA channel to a server session.

Windows has a mechanism to notify applications about clipboard changes. An application interested in such notifications has to register itself in the so called clipboard chain. Windows inserts it on top of that chain and that application is responsible to propagate changes down the chain:

If a 3rd-party application forgets to forward notifications down then we have a broken clipboard chain and clipboard changes are not sent via ICA protocol:

If you run RepairCBDChain.exe it tries to find the window of wfica32.exe and registers it for clipboard notifications again:

However if it finds the second instance of wfica32.exe (as on the picture above) the first instance will be still cut off from notifications and this explains why RepairCBDChain.exe doesn’t work sometimes.

On the server session side the picture is similar (the registered application is wfshell.exe):

You can see WM_DRAWCLIPBOARD messages in MessageHistory logs for wfica32.exe process:

PID.TID: c20.c0c

HWND: 0x002501D4
Class: "wMFService006600CA004"
Title: "Microsoft Outlook7718 - MetaFrame Presentation Server Client [SpeedScreen On]"

HWND: 0x003F08DC
Class: "Transparent Windows Client"
Title: "^P ^b24 of 24 - Clipboard^b^SItem collected. - \\Remote"

HWND: 0x004E0332
Class: "WFClip"
Title: "WFClip"
17:58:53:484 S WM_DRAWCLIPBOARD (0×308) wParam: 0xd0aa0 lParam: 0×0

HWND: 0x0094036E
Class: "TWI Link"
Title: ""

Hope this little excursion explained clipboard chain, how it becomes broken and how it is repaired.

- Dmitry Vostokov -

This is a follow up to my previous Q&A about crash dumps and security issues like exposing confidential information stored in memory: Crash Dumps and Security. It seems a solution exists which allows to do some sort of crash dump analysis or at least identify problem components without sending complete or kernel memory dumps.

This solution takes advantage of WinDbg ability to execute scripts of arbitrary complexity. Couple of months ago I wrote about scripts and they really help me in pulling out various information from complete memory dumps:

WinDbg scripts
Yet another Windbg script
Critical sections

Now I created the bigger script that combines together all frequent commands used for identification of potential problems in memory dumps:

• !analyze -v
• !vm 4
• lmv
• !locks
• !poolused 3
• !poolused 4
• !exqueue f
• !irpfind
• !stacks
• List of all processes’ thread stacks, loaded modues and critical sections (for complete memory dump)

Other commands can be added if necessary.

How does all this work? A customer has to install Debugging Tools for Windows from Microsoft. This can be done on any workstation and not necessarily in a production environment. Then the customer has to run WinDbg.exe with some parameters including path(s) to symbols (-y), a path to memory dump (-z) and a path to script (-c):

C:\Program Files\Debugging Tools for Windows>WinDbg.exe -y "srv*c:\mss*http://msdl.microsoft.com/download/symbols" -z MEMORY.DMP -c "$$><c:\WinDbgScripts\Dmp2Txt.txt;q" -Q -QS -QY –QSY

Once WinDbg.exe finishes (it can run for couple of hours if you have many processes in your complete memory dump) you can copy the .log file created in “C:\Program Files\Debugging Tools for Windows” folder, archive it and send it to support for analysis. Kernel and process data and cached files are not exposed in the log! And because this is a text file the customer can inspect it before sending.

Here are the contents of Dmp2Txt.txt file:

$$
$$ Dmp2Txt: Dump all necessary information from complete full memory dump into log
$$
.logopen /d
!analyze -v
!vm 4
lmv
!locks
!poolused 3
!poolused 4
!exqueue f
!irpfind
!stacks
r $t0 = nt!PsActiveProcessHead
.for (r $t1 = poi(@$t0); (@$t1 != 0) & (@$t1 != @$t0); r $t1 = poi(@$t1))
{
    r? $t2 = #CONTAINING_RECORD(@$t1, nt!_EPROCESS, ActiveProcessLinks);
    .process @$t2
    .reload
    !process @$t2
    !ntsdexts.locks
    lmv
}
.logclose
$$
$$ Dmp2Txt: End of File
$$

For kernel dumps the script is simpler: 

$$
$$ KeDmp2Txt: Dump all necessary information from kernel dump into log
$$
.logopen /d
!analyze -v
!vm 4
lmv
!locks
!poolused 3
!poolused 4
!exqueue f
!irpfind
!stacks
!process 0 7
.logclose
$$
$$ KeDmp2Txt: End of File
$$


Note: if the dump is LiveKd.exe generated then due to inconsistency scripts may run forever 

- Dmitry Vostokov -

It often happens that Citrix support advises customers to change their default postmortem debugger to NTSD. But there is no way to test new settings unless some application crashes again. And some customers come back saying dumps are not saved despite new settings and we don’t know whether dumps were not saved because a crash hadn’t yet happened or default debugger hadn’t been configured properly or something else happened.

In addition the arrival of 64-bit Windows brings another problem: there are 2 default postmortem debuggers on 64-bit Windows (for 32-bit and 64-bit applications respectively):

NTSD on x64 Windows

The new tool TestDefaultDebugger forces a crash on itself to test the presence and configuration of default postmortem debugger (Dr. Watson, NTSD or other). Then if the default postmortem debugger is configured properly OS will launch it to save a dump of TestDefaultDebugger.exe process.

If you enabled NTSD as a default postmortem debugger (CTX105888) the following console window will briefly appear:

Postmortem debuggers are explained here:

Dumps for Dummies (Part 3)

On 64-bit Windows you can run both 32-bit TestDefaultDebugger.exe and 64-bit TestDefaultDebugger64.exe applications and then open crash dumps to see whether both postmortem debuggers have been configured properly. The tool has also command line interface so you can use it remotely:

c:\>TestDefaultDebugger.exe now

You can download the tool from Citrix support web site:

TestDefaultDebugger v1.0 for 32-bit and 64-bit platforms

- Dmitry Vostokov @ DumpAnalysis.org -

The environment where Citrix software operates is so complex that some education in Systems Theory and basic understanding of “cause and effect” and impossibility of “action at a distance” is needed. In forthcoming mini-series I would try to highlight some notions of that.

- Dmitry Vostokov -

Welcome to Inside Citrix. This monthly column gives a glimpse of different aspects of Citrix through our people. Our guests have different areas of responsibility and expertise to give you an idea of what is happening behind the scenes. We discuss items of interest with people from Product Readiness, Escalation, Technical Support, and Engineering just to name a few.

In this installment of Inside Citrix, we discuss the meaning of life with Dmitry Vostokov, EMEA Development Analysis Team Lead.

Q: Hello Dmitry, how are you? I am very happy to conduct this interview as you are a creative and prolific worker. I wonder…has fame caught up to you yet, due to your creativity?

A: I’m fine, thank you! I believe there is a synergistic effect going on here. I make the company famous and the company makes me famous.

Q: So, before I get too far ahead of myself, please tell everyone a bit of your history. Where are you from? What did you do before Citrix? How long have you been with us? What kinds of things have you been doing at Citrix during your tenure?

A: I’m from Russia. I was born near Moscow and I spent 14 years there after enrolling at Moscow State University to study chemistry. In that university, I saw a computer and immediately started programming. My first program was written in FORTRAN and had almost 200 lines. My second program had commercial success: I ported 800 FORTRAN lines to about 2000 PDP-11 assembler lines and achieved a 25 percent increase in speed (the program calculated rocket fuel properties for weeks). Since then I’d been working from home for some U.S. and Russian ISV companies (mostly in speech and image processing domains) until 1999, when I went to work in an office to see a large software factory from the inside out.

In 2001 I went to Ireland to learn English. My first job in Ireland was with Ericsson in a small town as a Senior Software Designer. The title sounded great to me, but I heard rumors that the only engineers in Ericsson were hardware engineers. So that job didn’t last long because I was headhunted by a company called Programming Research and I relocated to Dublin. I spent 1.5 years there and after working briefly for a security company (that company is extinct now) I was hired by Citrix. I’ve already spent 3.16 years here. For Citrix I analyze crash dumps and provide recommendations. It’s like being a computer psychologist assessing brain damage. I also do a bit of escalation work when I have time. I like to provide full escalation and software maintenance cycles whenever I have sufficient resources to analyze the problem, contact the customer, and provide the resolution. I also have an opportunity here to apply my software design and programming skills by writing various troubleshooting tools.

Q: Most people probably didn’t know all of that. I guarantee you that Escalation knows you well. How is the blogging going? How can readers get to your blog?

A: I love blogging. I didn’t even think about blogging until I suddenly realized its potential in information sharing. When I joined the company there was no sufficient information available about crash dump analysis, so I had to learn on my own. Now I’m happy to share what I have learnt to everyone.

One topic I like to write about in my blog at the moment is crash dump analysis patterns and anti-patterns, where I summarize general solutions you can apply or should not apply in specific contexts to common recurrent dump analysis problems.

More will come…

Q: And the tools that you create, very useful! Can you take a moment to talk about each of the ones you have created? Which ones have you gotten the best feedback about? Which ones have been the most useful?

A: Thanks! I use them too. The tool I got the most complaints about is RepairCBDChain; the tool with the fewest complaints is SystemDump. I got the best feedback about PDBFinder.

All of them are useful in certain troubleshooting scenarios. I’m preparing a presentation about all these tools and I will present it to the EMEA TRM team in December. I’ll definitely publish it as soon as I get feedback about that training.

Here are brief descriptions of these tools (most of them have different versions for various platforms, and some were even ported to Windows Mobile):

• RepairCBDChain: Repairs clipboard functionality and magically you are able to copy/paste again (not always actually – I promise to write a blog post explaining why).

• ADSCleaner: Cleans Windows NT File System (NTFS) file streams created by Citrix memory optimization code if you no longer need this feature (it also frees disk space, by the way).

• ProcessHistory: Tracks processes, threads, and modules on 32-bit and 64-bit platforms. I’m going to release a Windows Mobile version soon.

• MessageHistory: Tracks window messages. It’s similar to Spy++ but much easier to use for troubleshooting and it works on 64-bit platforms too.

• WindowHistory: Tracks windows as they change their appearance, are created, and are destroyed and saves a log file. This is what Spy++ lacks and it was the primary motivation to write this tool.

• SystemDump: Forces a dump immediately or after a specified period of time. This can be done remotely too. It works on both 32-bit and 64-bit Windows! My primary motivation was that the OSR “bang” tool doesn’t work on 64-bit Windows.

• PDBFinder: Helps to find symbol files if you have zillions of them.

• DumpCheck: Verifies that you have a valid dump and even provides recommendations to avoid common mistakes before sending dumps to support.

• CtxHidEx32: Can hide any annoying windows or message boxes and reduce unnecessary support calls. It also has a peculiar feature: you can specify an action to do before hiding the window. When the Media Player window appears it can send a message to your boss.

• Dump2Wave: My most controversial tool that allows you to hear the sound of memory corruption. Some people say it’s useless but I would say it is entertaining.

Some other upcoming tools I’m working days and nights on (when I have free time) are:

• DumpDepends: Helps to automate repetitive dumping.

• DumpAlerts: Provides notification whenever new dump is saved.

• SessionHistory: Tracks session information.

• HistoryToolbar: Organizes “History” tools into one coherent super tool.

• DumpPlayer: Plays musical dumps in real-time and provides visual images based on crash dump memory contents. I coined a term—Dump Tomography—for this.

Q: They must take some upkeep, as we see a lot of improvements, updates, and so on. I also see you provide a lot of training information on escalation techniques, debugging, analysis, and more. What do you believe is the most important characteristic of a successful escalation engineer?

A: As Winston Churchill said: “Never, never, never give up!”?

Q: Any advice for Citrix administrators who might be reading this on how to avoid trouble or have their environment best situated to speed resolution, should an issue occur?

A: If you are asked to generate and/or collect crash dumps, please tell support personnel how you got that dump. And ensure that you are sending the right dump for the right issue.

I started writing Dumps for Dummies blog posts to explain dumps and I promise to continue and expand them.

Q: What do you find most challenging about your job?

A: To work with enormous amounts of information and make quick decisions at the same time.

Q: Is there anything you can share with us about new Citrix products or technologies (not giving away confidential information) that you are excited about?

A: I would tell you that with whatever new technology comes along, crash dumps will be the same! And this gives me some optimism. Whether there will be more or less crash dumps in the future is pretty confidential though…

Q: Any plans to visit Citrix headquarters in Fort Lauderdale, Florida?

A: I’m actually visiting Citrix headquarters at the end of this month! See you there.

Q: Not so much a question, make us laugh!

A: One day we got a fax from a customer where all of the blue screen information was written down by hand—hundreds of digits… How long it took to copy all that from the screen and whether or not he made any mistakes, we will never know. The copy from that fax is still hanging on my desk wall.

Q: What do you do in your free time besides analyzing dumps, debugging and programming?

A: Read books. I read lots of them and about quite diverse subjects. However, my favorite subject for the last four years has been math—the more abstract the better.

It really helps in improving the critical thinking skills required for my job.

Thanks, Dmitry. People will know to look you up online…

WindowHistory Mobile edition has been updated. It replaces the previous version of WindowHistory CE/Mobile 2.1 and now available in two separate executables: for Windows Mobile 5.0 (ARMV4I) and Windows Pocket PC 2003 (ARMV4). It has been tested under emulators, Acer n300 (480×640 screen) and mobile phone Mio A701 (240×320 screen). Here are screenshots from Windows Mobile 5.0 emulator:

The tool also includes Easter Egg (activate soft keyboard, click on and then click on About button. The following window appears with scrolling text of contributors and special thanks):

- Dmitry Vostokov -

Following the release of Dump2Wave tool some members of Citrix community have been asking me to provide some interesting sound fragments from dump files. I was also particularly interested in catching voices from the past: embedded fragments of human voice. So I recorded my “Hello” message, played it by Media Player and then saved a process dump. Then I converted the dump to CD-quality wave file and saved interesting sound fragments from it (to conserve space - the original wave file was 76Mb).

To listen to these fragments you can download wave files from the following location:

DumpSounds.zip (8Mb)

Here is the description of what I heard in these wave files:

- dump1.wav

• violin
• aliens
• train sound
• Hello

- dump2.wav

• electric guitar
• signals from cosmos

- dump3.wav

• Morse code alphabet

- dump4.wav

• helicopter

- dump5.wav

• horn
• some interesting noise and fragments of electronic music

 Enjoy

Of course, you can convert kernel memory dumps to wave files and hear voices from kernel space too…

- Dmitry Vostokov -

The tool monitors folders where dumps can be saved including Dr. Watson, a folder specified when NTSD is set as a default debugger, etc. It then alerts a user, an administrator or a software vendor whenever a new dump is saved:

• Icon in System Tray changes its color from green to red
• Popup window appears until dismissed
• E-mail is sent to a specified address
• Sound is played
• Custom action is executed, for example, automatically launching WinDbg.exe with the latest dump or copying it to an ftp server

All actions are fully configurable and can be enabled/disabled. Here is the screenshot of the main window:

I’m planning to incluide TAPI support and alerts from hung applications in the next version(s).

Later this tool will included in Dump Monitor Suite

Any comments and suggestions are welcome.

- Dmitry Vostokov -

In the previous Dumps for Dummies (Part 3) I tried to explain the nature of crashes. Another category of problems happens very often and we also need a dump for analysis: hangs. There is some confusion exists in understanding the difference between these two categories: crash and hang. Although sometimes a hang is a direct consequence of a crash most of the time hangs happen independently. They also manifest themselves differently. Let’s look at application (process) crashes and hangs first. When a crash happens an application (process) often disappears. When hang happens an application (process) is still in memory: you can see it in Task Manager, for example, but it doesn’t respond to user commands or to any other requests like pinging a TCP/IP port. If we have a crash in OS then the most visible manifestation is blue screen and/or reboot. If we have a hang then everything freezes.

Application or system hang happens because from high level view of the interaction between application or OS components (modules) is done via messages. One component sends a message to another and waits for a response. Some components are critical, for example, registry. The following hand-made picture depicts very common system hang situations when the register component stops responding. Then every running application (process) stops responding if its execution path depends on registry access.

The very common reason for hang is so called deadlock when two running applications (their execution paths, threads) are waiting for each other. Here is the analogy with a blocked road:

In order to see what’s inside the process or OS which caused a hang we need a dump. Usually this dump is called a crash dump too because in order to get it the usual method is to make some sort of a trap which causes an application or OS to crash and to save the dump. I personally prefer to call these dumps just memory dumps to avoid confusion.  

How can you get a memory dump if your application or service hangs?

• by using NTSD command line options (remember that NTSD is always present on your system)

• by using userdump.exe

• by attaching and saving the dump interactively via NTSD

• by attaching and saving the dump interactively via WinDbg

• by using ADPlus in a hang mode

How can you get a memory dump if your system hangs?

• manually via keyboad

• by using Citrix SystemDump tool remotely or via GUI if some session is still alive

For most system hangs choosing Kernel memory dump option in Control Panel\System\Advanced\Startup and Recovery applet is sufficient. Kernel memory dumps are smaller and less susceptible to corruption or truncation due to small page file size. If you discover that you need to peer inside running user applications then you can always ask for another Complete memory dump when the problem happens again.

- Dmitry Vostokov @ DumpAnalysis.org -

Suppose you work in a banking industry or for any company that has sensitive information. Is it secure to send a crash dump outside for analysis? One semi-anonymous person asked this question on www.dumpanalysis.org and here is my unedited answer based on my experience in crash dump analysis and kernel level development:

"It depends on credit card transactions software design and architecture and what type of dump is configured in Control Panel\System\Advanced\Startup and Recovery applet: Small, Kernel or Complete.

Software usually encrypts data before sending it down TCP/IP stack or other network protocol. If your credit card transactions software doesn't have any kernel space encryption drivers and doesn't rely on any MS or other 3rd-party encryption API that might send data to kernel, communicate to KSECDD or to user-space component like LSASS via LPC/RPC you can safely assume that kernel memory dumps will not have unencrypted data. If encryption is done entirely in user space Small memory dump and Kernel memory dump will only have encrypted fragments. Otherwise there is a probability that BSOD happens just before encryption or after decryption or when secure protocol is being handled. This exposure can even happen in Small memory dumps if BSOD happens in the thread that handles sensitive information in kernel mode.

The same applies if your software stores credit data on any medium. If it stores only encrypted data and decrypts entirely in user space without any transition to kernel it should be safe to enable kernel memory dump.

If your goal is ultimate security then even Small memory dump (64Kb) should not be allowed. But in reality as we consider probabilities sending small memory dump is equivalent to no more than exposing just one credit card number or one password.

What you must avoid at any cost is to enable complete memory dump option in control panel. In this case all your credit card transactions software code and data including file system cache will be exposed.

Contrary to complete memory dump kernel memory dump will not have much data even if some potion of it is being communicated during crash time. I would also be interested in hearing what other experts say. This is very interesting topic."

If you are interested too you can participate in that discussion (registration is needed to avoid spammers):

http://www.dumpanalysis.org/forum/viewtopic.php?t=56

- Dmitry Vostokov -

You probably already know or have heard about the project WINE: Windows API on top of X and Unix

winehq.com 

I first heard about it more than 10 years ago when it started. Today I rediscovered it again and was really surprised. I was looking for one NT status code I couldn’t find in MS official documentation and found it here:

dlls/ntdll/error.c

In order to run Win32 programs WINE emulates all API calls including OLE32, USER32, GDI32, KERNEL32, ADVAPI32 and of course, NTDLL:

dlls/ntdll
dlls/ole32
dlls/user32
dlls/kernel32
dlls/gdi32
dlls/advapi32

Plus hundreds of other components. All source code is located here:

http://cvs.winehq.com/cvsweb/wine/

So if want to see how particular function or protocol might have been implemented hypothetically by Windows OS designers it is a good place to start.

- Dmitry Vostokov -

We all know that macro definitions in C and C++ are evil. They cause maintenance nightmares by introducing subtle bugs. I never took that seriously until last weekend I was debugging my old code written 10 years ago which uses macros written 15 years ago :-) 

My Windows Mobile 5.0 application was crashing when I was using POOM COM interfaces (Pocket Outlook Object Model). The crash never pointed to my code. It always happened after pimstore.dll and other MS modules were loaded and COM interfaces started to return errors. I first suspected that I was using POOM incorrectly and rewrote all code several times and in different ways. No luck. Then I tried PoomMaster sample from Windows Mobile 5.0 SDK and it worked well. So I rewrote my code in exactly the same way as in that sample. No luck. My last hope was that moving code from my DLL to EXE (as in sample SDK project) would eliminate crashes but it didn’t help too. Then I slowly started to realize that the problem might have been in my old code and I also noticed that one old piece of code had never been used before. So I started debugging by elimination (commenting out less and less code) until I found a macro. I had to stare at it for couple of minutes until I realized that one pair of brackets was missing and that caused allocating less memory and worse: the returned pointer to allocated memory was multiplied by 2! So the net result was the pointer pointing to other modules and subsequent string copy was effectively overwriting their memory and eventually causing crashes inside MS dlls.  

Here is that legacy macro:

#define ALLOC(t, p, s)
((p)=(t)GlobalLock(GlobalAlloc(GHND, (s))))

It allocates memory and returns a pointer. It should have been called like this (size parameter is highlighted in blue):

if (ALLOC(LPWSTR,lpm->lpszEvents,
(lstrlen(lpszMacro)+1)*sizeof(WCHAR)))
{
lstrcpy(lpm->lpszEvents, lpszMacro);
    lpm->nEvents=lstrlen(lpm->lpszEvents)+1;
}

What I found is the missing bracket before lstrlen and last enclosing bracket (size parameter is highlighted in red):

if (ALLOC(LPWSTR,lpm->lpszEvents,
lstrlen(lpszMacro)+1)*sizeof(WCHAR))
{
lstrcpy(lpm->lpszEvents, lpszMacro);
    lpm->nEvents=lstrlen(lpm->lpszEvents)+1;
}

The resulted code after macro expansion looks like this

if (lpm->lpszEvents=(LPWSTR)GlobalLock(GlobalAlloc(GHND,
lstrlen(lpszMacro)+1))*sizeof(WCHAR))

You see that the pointer to allocated memory is multiplied by two and string copy is performed to a random place in the address space of other loaded dlls corrupting their data and causing the process to crash later.

- Dmitry Vostokov -

After looking at one dump today where all thread environment blocks were zeroed, import table corrupt and recalling some similar cases I encountered previously I came up with the next pattern: Lateral Damage.

When this problem happens you don’t have much choice and your first temptation is to apply Alien Component anti-pattern unless your module list is corrupt and you have manifestation of another common problem I will talk about next time: Corrupt Dump.

Anti-pattern is not always bad solution if complemented by subsequent verification and backed by experience. If you get damaged process and thread structures you can point to a suspicious component (supported by some evidence like raw stack analysis and educated guess) and request additional dumps in hope to get less damaged process space or see that component again. At the very end if removing it stabilizes the customer environment it proves you were right.

- Dmitry Vostokov @ DumpAnalysis.org -

In any domain of activity where patterns exist we can find anti-patterns too. They are bad solutions for recurrent problems in specific contexts. One of them I would like to introduce briefly is Alien Component. In essence, when every technique fails or you run out of WinDbg commands look at some innocent component you have never seen before or don’t have symbols for: be it some driver or hook. Of course, this component cannot be the component developed by the company you are working for.

- Dmitry Vostokov -

Another pattern I observe frequently is False Positive Dump. We get dumps pointing in a wrong direction or not useful for analysis and this usually happens when wrong tool was selected or right one was not properly configured for capturing crash dumps. Here is one example I investigated in detail.

The customer experienced frequent spooler crashes. The dump was sent for investigation to find an offending component: usually it is a printer driver. WinDbg revealed the following exception thread stack (parameters are not shown here for readability):

KERNEL32!RaiseException+0x56
KERNEL32!OutputDebugStringA+0x55
KERNEL32!OutputDebugStringW+0x39
HPZUI041!ConvertTicket+0x3c90
HPZUI041!DllGetClassObject+0x5d9b
HPZUI041!DllGetClassObject+0x11bb

The immediate response is to point to HPZUI041.DLL but if we look at parameters to KERNEL32!OutputDebugStringA we would see that the string passed to it is a valid NULL-terminated string:

0:010> da 000d0040
000d0040  ".Lower DWORD of elapsed time = 3"
000d0060  "750000."

If we disassemble OutputDebugStringA up to RaiseException call we would see:

0:010> u KERNEL32!OutputDebugStringA
KERNEL32!OutputDebugStringA+0x55
KERNEL32!OutputDebugStringA:
push    ebp
mov     ebp,esp
push    0FFFFFFFFh
push    offset KERNEL32!'string'+0x10
push    offset KERNEL32!_except_handler3
mov     eax,dword ptr fs:[00000000h]
push    eax
mov     dword ptr fs:[0],esp
push    ecx
push    ecx
sub     esp,228h
push    ebx
push    esi
push    edi
mov     dword ptr [ebp-18h],esp
and     dword ptr [ebp-4],0
mov     edx,dword ptr [ebp+8]
mov     edi,edx
or      ecx,0FFFFFFFFh
xor     eax,eax
repne scas byte ptr es:[edi]
not     ecx
mov     dword ptr [ebp-20h],ecx
mov     dword ptr [ebp-1Ch],edx
lea     eax,[ebp-20h]
push    eax
push    2
push    0
push    40010006h
call    KERNEL32!RaiseException

There is no jumps in the code prior to KERNEL32!RaiseException call and this means that raising exception was expected. Also MSDN documentation says:

“If the application has no debugger, the system debugger displays the string. If the application has no debugger and the system debugger is not active, OutputDebugString does nothing.”

So spoolsv.exe might have been monitored by a debugger which caught that exception and instead of dismissing it dumped the spooler process.

If we look at ‘analyze -v’ output we could see the following:

Comment: 'Userdump generated complete user-mode minidump
with Exception Monitor function on WS002E0O-01-MFP'ERROR_CODE: (NTSTATUS) 0x40010006 -
Debugger printed exception on control C.

Now we see that debugger was User Mode Process Dumper you can download from Microsoft web site:

How to use the Userdump.exe tool to create a dump file 

If we download it, install it and write a small console program in Visual C++ to reproduce this crash:

#include "stdafx.h"
#include
int _tmain(int argc, _TCHAR* argv[])
{
    OutputDebugString(_T("Sample string"));
    return 0;
}

and if we compile it in Release mode and configure Process Dumper applet in Control Panel to include TestOutputDebugString.exe with the following properties:

and then run our program we would see Process Dumper catching KERNEL32!RaiseException and saving the dump.

Even if we select to ignore exceptions that occur inside kernel32.dll this tool still dumps our process. Now we can see that the customer most probably enabled ‘All Exceptions’ check box too. What the customer should have done is to use default rules like on the picture below:

Or select exception codes manually. In this case no dump is generated even if we manually select all of them. Just to check that the latter configuration still catches access violations we can add a line of code dereferencing NULL pointer and Process Dumper will catch it and save the dump.

Conclusion: the customer should have used NTSD as a default postmortem debugger from the start. Then if crash happened we would have seen the real offending component or could have applied other patterns and requested additional dumps.

- Dmitry Vostokov @ DumpAnalysis.org -