Write windbg command output to file, but not console - windows

In windbg, I'm looking for a mechanism to take the output of a command (specifically, a command inside of a breakpoint) and have it appended to a file, and not written to a console.
Currently I setup the process with .logappend C:\path\to\log and then enable a few breakpoints with:
bp WIN32U!{function} ".echo '===WIN32K-START==='; k; .echo '===WIN32K-END==='; g"
This works great, except the volume of output written to the console causes serious performance issues. I'm hopeful there's a way to get the same output to my log file, without the overhead of writing to the windbg console.

You want the .outmask meta-command: https://learn.microsoft.com/en-us/windows-hardware/drivers/debugger/-outmask--control-output-mask-
.outmask allows you to control which message types are sent to the output window and log file. The /l switch can allow you to change just the types that reach the output window, without affecting which types will reach the log file.
For example, this command will turn off all output to the output window while still sending normal messages to the log file:
.outmask- /l 0xffff
Although probably .outmask- /l 1 is all you need, which turns off just normal message output, but errors and warnings will still show up in the output window. Use .outmask /d to reset output settings back to the default when you're done.
In combination with the ability of .printf to output different message types, you can make it so you still have some idea what's going on, as well. Turn off normal message output to the window with .outmask- /l 1. Now you can use .printf /oe "message" in a breakpoint command somewhere to write an error message, which will still be sent to the output window so you can tell what's happening at certain points in the process.

you can patch dbgeng!g_OutputControl global to disable writing to console and only write to log file
but I don't know if you will have a performance gain or not
looking for a txt file
C:\>dir /b *.txt
File Not Found
opening a debugging session
C:\>cdb calc
Microsoft (R) Windows Debugger Version 10.0.15063.400 X86
ntdll!LdrpDoDebuggerBreak+0x2c:
774005a6 cc int 3
in the opened debugging session
spawn a parent debugger to debug the windbg running your debuggee
0:000> .dbgdbg
Debugger spawned, connect with
"-remote npipe:icfenable,pipe=cdb_pipe,server=xxxx"
in the spawned parent patch the global and detach
ed dbgeng!g_OutputControl 0
.detach
q
open a logfile in your debugging session
0:000> .logappend c:\foo.txt
Opened log file 'c:\foo.txt'
set a conditional breakpoint and start the session
0:000> bp ntdll!RtlEnterCriticalSection "kb;gc"
0:000> bl
0 e 773a7790 0001 (0001) 0:**** ntdll!RtlEnterCriticalSection "kb;gc"
0:000> g
there is no console output here
doing a ctrl+c to stop session and quitting the session
eax=7ffde000 ebx=00000000 ecx=00000000 edx=773ff1d3 esi=00000000 edi=00000000
eip=77394108 esp=016ef8a8 ebp=016ef8d4 iopl=0 nv up ei pl zr na pe nc
cs=001b ss=0023 ds=0023 es=0023 fs=003b gs=0000 efl=00000246
ntdll!DbgBreakPoint:
77394108 cc int 3
0:001> q
quit:
check for the log file and confirm if it has voluminous data written to it
RtlEnterCriticalSection Api is a very hot Api
C:\>dir /b *.txt
foo.txt
C:\>ls -l foo.txt
-rw-rw-rw- 1 0 **1754920** 2017-09-15 00:27 foo.txt
C:>head foo.txt
Opened log file 'c:\foo.txt'
0:000> bp ntdll!RtlEnterCriticalSection "kb;gc"
0:000> bl
0 e 773a7790 0001 (0001) 0:**** ntdll!RtlEnterCriticalSection "kb;gc"
0:000> g
ChildEBP RetAddr Args to Child
000cf114 77425f4b 000d0138 7724d80b 00000000 ntdll!RtlEnterCriticalSection
000cf158 773ea40a 000d0000 50180162 00000044 ntdll!RtlDebugAllocateHeap+0x9d
000cf23c 773b5ae0 00000044 00000000 00000000 ntdll!RtlpAllocateHeap+0xc4
000cf2c0 77384726 000d0000 40180060 00000044 ntdll!RtlAllocateHeap+0x23a
there are more tha 22k lines written to this file
C:\>wc -l foo.txt
22543 foo.txt
C:>tail foo.txt
000cf838 773c37be 00462d6c 7ffdb000 00000000 ntdll!__RtlUserThreadStart+0x70
000cf850 00000000 00462d6c 7ffdb000 00000000 ntdll!_RtlUserThreadStart+0x1b
(c80.8ec): Break instruction exception - code 80000003 (first chance)
eax=7ffde000 ebx=00000000 ecx=00000000 edx=773ff1d3 esi=00000000 edi=00000000
eip=77394108 esp=016ef8a8 ebp=016ef8d4 iopl=0 nv up ei pl zr na pe nc
cs=001b ss=0023 ds=0023 es=0023 fs=003b gs=0000 efl=00000246
ntdll!DbgBreakPoint:
77394108 cc int 3
0:001> q
quit:
C:\>

Related

Pending breakpoint in windbg

Remote debugging is not possible & need to debug my code for an issue not reproducible on my system.
I can set a pending breakpoint in gdb on linux. I need something similar on windbg or on windows platform as I cannot share the source file but I need to break in the file at particular line in order to debug.
what do you mean by pending breakpoint ?
does the module in which the function resides loaded dynamically and you need to set a breakpoint beforehand ?
if yes then use bu breakpoints
bu myfoomodule!myblahfunction
this will load the symbols when the module is loaded and set a breakpoint as and when needed
0:000> lm m help*
Browse full module list
start end module name
<<<<<<<<<<<< no results here
lets confirm by forcefully realoding symbols and checking for the existence of
this module
0:000> .reload /f
Reloading current modules
0:000> lm m help*
Browse full module list
start end module name
again no results we cant set a regular breakpoint
our breakpoint needs to be set using unresolved breakpoint
lets set it
0:000> bu HelpPaneProxy!DllMain
let us check
0:000> bl
0 e Disable Clear u 0001 (0001) (HelpPaneProxy!DllMain)
lets continue
0:000> g
ModLoad: 77ac0000 77adf000 C:\Windows\system32\IMM32.DLL
ModLoad: 76f90000 7705c000 C:\Windows\system32\MSCTF.dll
un needed mod load display cut off XXXXXXXXXXXXXXXXXXXX
ModLoad: 67d80000 67d91000 C:\Windows\System32\HelpPaneProxy.dll
bingo our breakpoint hit on the module load triggered after clicking help in calc
Breakpoint 0 hit
eax=00000001 ebx=00000001 ecx=00000001 edx=00000000 esi=00000000 edi=0025cfa8
eip=67d84700 esp=0025ce7c ebp=0025ced8 iopl=0 nv up ei pl nz na po nc
cs=001b ss=0023 ds=0023 es=0023 fs=003b gs=0000 efl=00000202
HelpPaneProxy!DllMain:
67d84700 8bff mov edi,edi

How to send EOF from command prompt *without newline*?

Sure, to send EOF from command prompt, Enter followed by Ctrl-Z does the trick.
C:\> type con > file.txt
line1
line2
^Z
This works, and file.txt contains line1\r\nline2\r\n. But how can you do the same without the last newline, so that file.txt contains line1\r\nline2?
In Linux, the solution is to hit Ctrl-D twice1. But what is the equivalent on Windows? Command prompt will happily print ^Zs at the end of a line without doing sending EOF. (And if you press Enter, then any ^Zs you typed get written to the file as literal escape characters!)
If there is no way to do this on Windows, then why?
1 https://askubuntu.com/questions/118548/how-do-i-end-standard-input-without-a-newline-character
The command type con > file.txt doesn't have any special handling for ^Z in the cmd shell, since the target file isn't con and the type command wasn't run in Unicode (UTF-16LE) output mode. In this case, the only ^Z handling is in the ReadFile call itself, which for a console input buffer has an undocumented behavior to return 0 bytes read if a line starts with ^Z.
Let's examine this with a debugger attached, noting that the number of bytes read (lpNumberOfBytesRead) is the 4th argument (register r9 in x64), which is returned by reference as an output parameter.
C:\Temp>type con > file.txt
Breakpoint 1 hit
KERNELBASE!ReadFile:
00007ffc`fb573cc0 48895c2410 mov qword ptr [rsp+10h],rbx
ss:00000068`c5d1dfa8=000001e3000001e7
0:000> r r9
r9=00000068c5d1dfd0
0:000> pt
line1
KERNELBASE!ReadFile+0xa9:
00007ffc`fb573d69 c3 ret
0:000> dd 68c5d1dfd0 l1
00000068`c5d1dfd0 00000007
As you see above, reading "line1\r\n" is 7 characters, as expected. Next let's enter "\x1aline2\r\n" and see how many bytes ReadFile reportedly reads:
0:000> g
Breakpoint 1 hit
KERNELBASE!ReadFile:
00007ffc`fb573cc0 48895c2410 mov qword ptr [rsp+10h],rbx
ss:00000068`c5d1dfa8=0000000000000000
0:000> r r9
r9=00000068c5d1dfd0
0:000> pt
^Zline2
KERNELBASE!ReadFile+0xa9:
00007ffc`fb573d69 c3 ret
0:000> dd 68c5d1dfd0 l1
00000068`c5d1dfd0 00000000
As you see above, this time it reads 0 bytes, i.e. EOF. Everything typed after ^Z was simply ignored.
However, what you want instead is to get this behavior in general, wherever ^Z appears in the input buffer. type will do this for you, but only if it's executed in Unicode mode, i.e. cmd /u /c type con > file.txt. In this case cmd does have special handling to scan the input for ^Z. But I bet you don't want a UTF-16LE file, especially since cmd doesn't write a BOM to allow editors to detect the UTF encoding.
You're in luck, because it happens that copy con file.txt does exactly what you want. Internally it calls cmd!ZScanA to scan each line for a ^Z character. We can see this in action back in the debugger, but this time we're in completely undocumented territory. On inspection, it appears that this function's 3rd parameter (register r8 in x64) is the number of bytes read as an in-out argument.
Let's begin again by entering the 7 character string "line1\r\n":
C:\Temp>copy con file.txt
line1
Breakpoint 0 hit
cmd!ZScanA:
00007ff7`cf4c26d0 48895c2408 mov qword ptr [rsp+8],rbx
ss:00000068`c5d1e9d0=0000000000000000
0:000> r r8; dd #r8 l1
r8=00000068c5d1ea64
00000068`c5d1ea64 00000007
On output, the scanned length remains 7 characters:
0:000> pt
cmd!ZScanA+0x4f:
00007ff7`cf4c271f c3 ret
0:000> dd 68c5d1ea64 l1
00000068`c5d1ea64 00000007
0:000> g
Next enter the 23 (0x17) character string "line2\x1a Ignore this...\r\n":
line2^Z Ignore this...
Breakpoint 0 hit
cmd!ZScanA:
00007ff7`cf4c26d0 48895c2408 mov qword ptr [rsp+8],rbx
ss:00000068`c5d1e9d0=0000000000000000
0:000> r r8; dd #r8 l1
r8=00000068c5d1ea64
00000068`c5d1ea64 00000017
This time the scanned length is only the 5 characters that precede the ^Z:
0:000> pt
cmd!ZScanA+0x4f:
00007ff7`cf4c271f c3 ret
0:000> dd 68c5d1ea64 l1
00000068`c5d1ea64 00000005
We expect file.txt to be 12 bytes, which it is:
C:\Temp>for %a in (file.txt) do #echo %~za
12
More generally, if a Windows console program wants to implement Ctrl+D handling that approximates the behavior of a Unix terminal, it can use the wide-character console function ReadConsoleW, passing a CONSOLE_READCONSOLE_CONTROL struct by reference as pInputControl. This struct's dwCtrlWakeupMask field is a bit mask that sets which control characters will immediately terminate the read. For example, bit 4 enables Ctrl+D. I wrote a simple test program that demonstrates this case:
C:\Temp>.\test
Enter some text: line1
You entered: line1\x04
You can't see this in the above example, but this read was immediately terminated by pressing Ctrl+D, without even pressing enter. The ^D control character (i.e. '\x04') remains in the input buffer, which is useful in case you want different behavior for multiple control characters.

How does Windows locate files input in the shell?

I looked for this question over the site but it looks like it's never been asked, so I thought I would share the answer to anyone looking for it.
Q: When I input executables files' name in cmd.exe, through shell-execution functions or from a Batch file, how does Windows locate the external executable file to be called?
I do not understand what "shell-execution function" is referring to, so I will ignore it.
But the rules for launching an executable from the CMD.EXE command line, or from a batch file are as follows:
If a path is specified for the executable (absolute or relative), then only that path is searched.
If only the executable name (and possibly extension) is specified, then
First search the current folder
Then search folders in PATH environment variable (in order listed)
Within each folder searched, use the provided extension. If no file extension is provided, then look for files that match extensions found in PATHEXT environment variable (in order listed).
The first matching file found is the one that gets executed.
EDIT
David Candy has pointed me to some outdated NT documentation that precisely outlines the search process, and it essentially matches what I have written above. See Launching Applications from the NT Shell
I'd love it if anyone could post links to similar CMD.EXE documentation for more recent versions of Windows, especially if it was written by Microsoft. Note that the NT doc linked above was not written by Microsoft, even though it is posted on their site.
Here's a debug of CMD using CreateProcessW to start edit.com.
Microsoft (R) Windows Debugger Version 6.2.9200.20512 X86
Copyright (c) Microsoft Corporation. All rights reserved.
CommandLine: C:\Windows\System32\cmd.exe
Symbol search path is: srv*C:\tmp*http://msdl.microsoft.com/download/symbols;c:\tmp
Executable search path is:
ModLoad: 49fe0000 4a030000 cmd.exe
ModLoad: 76dd0000 76ef7000 ntdll.dll
ModLoad: 75860000 7593b000 C:\Windows\system32\kernel32.dll
ModLoad: 76a20000 76ae6000 C:\Windows\system32\ADVAPI32.dll
ModLoad: 76d00000 76dc2000 C:\Windows\system32\RPCRT4.dll
ModLoad: 75b20000 75bca000 C:\Windows\system32\msvcrt.dll
(58c.918): Break instruction exception - code 80000003 (first chance)
*** ERROR: Symbol file could not be found. Defaulted to export symbols for ntdll.dll -
eax=00000000 ebx=00000000 ecx=0027f9a0 edx=76e29a94 esi=fffffffe edi=76e2b6f8
eip=76e17dfe esp=0027f9b8 ebp=0027f9e8 iopl=0 nv up ei pl zr na pe nc
cs=001b ss=0023 ds=0023 es=0023 fs=003b gs=0000 efl=00000246
ntdll!DbgBreakPoint:
76e17dfe cc int 3
0:000> bp kernel32!CreateProcessW
*** ERROR: Symbol file could not be found. Defaulted to export symbols for C:\Windows\system32\kernel32.dll -
breakpoint 0 redefined
0:000> g
Breakpoint 0 hit
eax=0027f590 ebx=00000000 ecx=00000000 edx=00000000 esi=4a005200 edi=00000001
eip=75861c01 esp=0027f4b4 ebp=0027f674 iopl=0 nv up ei pl zr na pe nc
cs=001b ss=0023 ds=0023 es=0023 fs=003b gs=0000 efl=00000246
kernel32!CreateProcessW:
75861c01 8bff mov edi,edi
0:000> kb
ChildEBP RetAddr Args to Child
WARNING: Stack unwind information not available. Following frames may be wrong.
0027f4b0 49fe4568 00316078 003175a8 00000000 kernel32!CreateProcessW
0027f674 49fe4315 00317508 00000000 00000000 cmd!ExecPgm+0x20a
0027f8d4 49fe43a8 00317508 00000000 00000000 cmd!ECWork+0x7f
0027f8ec 49fe43ce 00317508 e3a3ec47 00000001 cmd!ExtCom+0x47
0027fd48 49fe185f 00317508 00000002 7588372f cmd!FindFixAndRun+0xb3
0027fd98 49ff70d4 00000000 00317508 4a0041b0 cmd!Dispatch+0x14a
0027fddc 49fe985b 00000001 000a0f38 000a1840 cmd!main+0x21a
0027fe20 758a4911 7ffdb000 0027fe6c 76e0e4b6 cmd!_initterm_e+0x163
0027fe2c 76e0e4b6 7ffdb000 7871d1d4 00000000 kernel32!BaseThreadInitThunk+0x12
0027fe6c 76e0e489 49fe9797 7ffdb000 00000000 ntdll!RtlInitializeExceptionChain+0x63
0027fe84 00000000 49fe9797 7ffdb000 00000000 ntdll!RtlInitializeExceptionChain+0x36
0:000> du 316078
00316078 "C:\Windows\system32\edit.com"
See CreateProcess.
The directory from which the application loaded.
The current directory for the parent process.
The 32-bit Windows system directory. Use the GetSystemDirectory function to get the path of this directory.
Windows Me/98/95: The Windows system directory. Use the GetSystemDirectory function to get the path of this directory.
The 16-bit Windows system directory. There is no function that obtains the path of this directory, but it is searched. The name of this directory is System.
The Windows directory. Use the GetWindowsDirectory function to get the path of this directory.
The directories that are listed in the PATH environment variable. Note that this function does not search the per-application path specified by the App Paths registry key. To include this per-application path in the search sequence, use the ShellExecute function.

How is procdump -t -- dump on process termination -- used?

The question may be a bit awkward, but here's my detailed problem:
Currently I'm looking into setting up SysInternals' procdump.exe to monitor an application of ours that exhibits spurious disappearances -- that is, the user reports that the application is simply "gone" without any trace after a short visible hang of the application's window.
My first idea was to run procdump -e -x . MyApp.exe which would record a crash dump when the application encounters an unhandled exception, but then I saw that there is also a -t switch, that --
-t - Write a dump when the process terminates.
automatically generates a dump when the process terminates.
Now the problem
I have tested the -t switch with our app by inserting a ExitProcess or TerminateProcess call at a defined location where I can trigger it.
While the app behaves as expected, i.e. TerminateProcess immediately "kills" the running app and ExitProcess takes a while because global cleanup is run, the dump generated this way is useless in both cases.
The dumps I get for -t always contain only a sinlge thread (where the app was running over 20 thread at termination time) and the callstack isn't even at a useful location. (It just seems to be one random thread from the terminated app.)
Am I doing something wrong? Can I usefully use procdump -t to track down unexpected calls of process exit functions at all?
Can I usefully use procdump -t to track down unexpected calls of
process exit functions at all?
I think not and here's why:
test process calc.exe
CommandLine: "C:\Program Files\Sysinternals\procdump.exe" -t calc.exe
I try to carefully suggest that procdump is waiting on calc.exe process handle.
0:000> kb
ChildEBP RetAddr Args to Child
0017f2e0 77135e6c 75336872 00000002 0017f334 ntdll!KiFastSystemCallRet
0017f2e4 75336872 00000002 0017f334 00000001 ntdll!NtWaitForMultipleObjects+0xc
0017f380 76cbf14a 0017f334 0017f3a8 00000000 KERNELBASE!WaitForMultipleObjectsEx+0x100
0017f3c8 76cbf2c2 00000002 7ffdb000 00000000 kernel32!WaitForMultipleObjectsExImplementation+0xe0
0017f3e4 011c6135 00000002 0017f46c 00000000 kernel32!WaitForMultipleObjects+0x18
WARNING: Stack unwind information not available. Following frames may be wrong.
0017fc30 011c999e 00000003 013d1de0 013d1e78 procdump+0x6135
0017fc78 76cc1194 7ffdb000 0017fcc4 7714b495 procdump+0x999e
0017fc84 7714b495 7ffdb000 77ad79b5 00000000 kernel32!BaseThreadInitThunk+0xe
0017fcc4 7714b468 011c99f5 7ffdb000 00000000 ntdll!__RtlUserThreadStart+0x70
0017fcdc 00000000 011c99f5 7ffdb000 00000000 ntdll!_RtlUserThreadStart+0x1b
0:000> dd 17f46c
0017f46c 00000238 00000268
0:000> !handle 238 f
Handle 238
Type Process
Attributes 0
GrantedAccess 0x1fffff:
Delete,ReadControl,WriteDac,WriteOwner,Synch
Terminate,CreateThread,,VMOp,VMRead,VMWrite,DupHandle,CreateProcess,SetQuota,SetInfo,QueryInfo,SetPort
HandleCount 5
PointerCount 52
Name <none>
Object Specific Information
Process Id 1580
Parent Process 2476
Base Priority 8
In the crash dump file gets stack last complete process thread (TID 3136) just before the end of the process.
0:000> ~
. 0 Id: dc8.c40 Suspend: -1 Teb: 7ffdd000 Unfrozen
0:000> .formats c40
Evaluate expression:
Hex: 00000c40
Decimal: 3136
Crash dump file is created after the completion of the last thread, and before the end of the process.

How to break into WinDBG when ntdll.dll is mapped into the new process

I want to use the following command to break into WinDBG when ntdll.dll is mapped into the new process, and before any of ntdll's process initialization runs.
sxe ld ntdll.dll ;g
However, the trick doesn't work at all,
ModLoad: 7c900000 7c9b0000 ntdll.dll
eax=010043af ebx=7ffde000 ecx=020f18f5 edx=00000034 esi=00c2f720 edi=00c2f6f2
eip=7c810867 esp=0006fffc ebp=00000720 iopl=0 nv up ei pl nz na po nc
cs=001b ss=0023 ds=0023 es=0023 fs=0038 gs=0000 efl=00000200
7c810867 ?? ???
Processing initial command 'sxe ld ntdll.dll ;g'
0:000> sxe ld ntdll.dll ;g
(ae8.6f4): Break instruction exception - code 80000003 (first chance)
eax=00181eb4 ebx=7ffde000 ecx=00000001 edx=00000002 esi=00181f48 edi=00181eb4
eip=7c901230 esp=0006fb20 ebp=0006fc94 iopl=0 nv up ei pl nz na po nc
cs=001b ss=0023 ds=0023 es=0023 fs=003b gs=0000 efl=00000202
ntdll!DbgBreakPoint:
7c901230 cc int 3
So, How to break into WinDBG when ntdll.dll is mapped into the new process?
thanks
[UPDATE]
I exactly followed the steps as mentioned by jcopenha, but I don't know why Windbg gives a strange error(Memory access error) where Notepad.exe is running .
Please give me a hand!Thanks a lot!
0:000> .restart /f
CommandLine: C:\WINDOWS\NOTEPAD.EXE
Symbol search path is: D:\Symbols\Symbols;SRV*D:\Symbols\MySymbols*http://msdl.microsoft.com/download/symbols
Executable search path is:
ModLoad: 01000000 01014000 notepad.exe
eax=0100739d ebx=7ffd9000 ecx=020f18f5 edx=0000004e esi=00f7f73a edi=00f7f6f2
eip=7c810867 esp=0007fffc ebp=0000024c iopl=0 nv up ei pl nz na po nc
cs=001b ss=0023 ds=0023 es=0023 fs=0038 gs=0000 efl=00000200
7c810867 ?? ???
0:000> u 7c810867
7c810867 ?? ???
^ Memory access error in 'u 7c810867'
[UPDATE2]
I found an odd instruction displayed at 7c810867, but p command still can work.
Is it a bug in WinDBG?
If you go to Debug->Event Filters and change "Create process" to "enabled" then restart the application it will start before ntdll.dll shows up in the module list. If you then do sxe ld ntdll.dll;g it will stop in ntdll!RtlUserThreadStart.
0:000> .restart /f
CommandLine: C:\Windows\System32\notepad.exe
Symbol search path is: SRV*d:\symbols*http://msdl.microsoft.com/download/symbols
Executable search path is:
ModLoad: 00000000`ffe00000 00000000`ffe35000 notepad.exe
00000000`7790c500 4883ec48 sub rsp,48h
0:000> sxe ld ntdll.dll;g
ModLoad: 00000000`778e0000 00000000`77a89000 ntdll.dll
ntdll!RtlUserThreadStart:
00000000`7790c500 4883ec48 sub rsp,48h
Regarding the loading of ntdll, I recommend reading this. You can break into the process before any code runs with:
windbg -xe cpr notepad
Or
windbg -xe ld:ntdll notepad
ntdll will still be mapped into the process at this point -- you can't break in before this happens.
As for the memory access error, kernel32 is not loaded into the process yet. The initial thread is queued to run at kernel32!BaseProcessStartThunk, but since kernel32 isn't loaded into the address space yet, you aren't going to see anything at that address.
The reason the thread's start address is able to begin in unmapped memory is because before the thread begins its execution, the very first thing that happens is a user APC runs in the context of that initial thread that takes care of all the process initialization, including loading kernel32. You can see this event occur if you set a similar event, like:
sxe ld kernel32
You will need to load symbols to get the internal functions names in the stack trace.
In addition to the first link, you can read more about process initialization here. Hope this helps.
As far as I remember ntdll is not mapped by the user-mode. If it was mapped in by the user-mode, what would map the module that maps ntdll? ntdll contains a set of stubs for syscalls to enter into the kernel and also contains a few other things such as a COFF loader which is part of process initialisation. It is a bit of a magic dll and it cannot be relocated, thus is always at the same virtual address in every process.
If you want to break before ntdll is mapped I am afraid you will have to use a kernel debugger.
May I ask what you are trying to achieve by breaking before ntdll is mapped?

Resources