Does signing a DLL invalidate a PDB file? - windbg

Analyzing a crash dump, WinDbg says my symbols (PDB file) does not match the module. The symbols are the ones generated when the DLL was compiled. The only thing that I can imagine would cause a mismatch is that the DLL was signed.
I'm using !chksym to validate symbols:
!chksym libcef.dll D:\sym\libcef.dll.pdb
libcef.dll
Timestamp: 5BB3D477
SizeOfImage: 626D000
pdb: F:\src\out\libcef.dll.pdb
pdb sig: B0065D83-113F-63BE-53BC-AEF07EC816B4
age: 1
libcef.dll.pdb
pdb sig: 9BA88A40-D168-44F2-44C1-DD2D73A38B38
age: 1
sig MISMATCH: libcef.dll.pdb and libcef.dll

Code signing an executable or DLL does not affect the debug header of the executable. Thus, it will still match the PDB.
...\SigningPdb\bin\Release>symchk signed.exe /s .
SYMCHK: FAILED files = 0
SYMCHK: PASSED + IGNORED files = 1
...\SigningPdb\bin\Release>symchk unsigned.exe /s .
SYMCHK: FAILED files = 0
SYMCHK: PASSED + IGNORED files = 1
And
...\SigningPdb\bin\Release>ChkMatch.exe -c signed.exe SigningPdb.pdb
ChkMatch - version 1.0
Copyright (C) 2004 Oleg Starodumov
http://www.debuginfo.com/
Executable: signed.exe
Debug info file: SigningPdb.pdb
Executable:
TimeDateStamp: bc78c18e
Debug info: 2 ( CodeView )
TimeStamp: a7b373e5 Characteristics: 0 MajorVer: 0 MinorVer: 0
Size: 97 RVA: 000026a0 FileOffset: 000008a0
CodeView format: RSDS
Signature: {b8ed520c-cdfc-486b-8e1a-7c0752a2a41f} Age: 1
PdbFile: ...\Release\SigningPdb.pdb
Debug info: 16 ( Unknown )
TimeStamp: 00000000 Characteristics: 0 MajorVer: 0 MinorVer: 0
Size: 0 RVA: 00000000 FileOffset: 00000000
Debug information file:
Format: PDB 7.00
Signature: {b8ed520c-cdfc-486b-8e1a-7c0752a2a41f} Age: 1
Result: Matched
The timestamp is in the COFF header. That header is only 24 bytes in size and will not change during code signing.
Most changes will happen in a new section for certificates. This section, however, will also be ignored during code signing. Otherwise, a second signature would destroy the first signature. (BTW: this section has been used to transport malicious code inside a signed executable)
Of course, the "usual" checksums, which do not consider the EXE/DLL file structure, will report a different checksum.
What may have happened to your DLL or EXE?
you have accidentally rebuilt it, so the time stamp of your DLL does not match the PDB any more
you are using aspect oriented programming (AOP) in .NET and there is some code weaving happening after a rebuild. These tools might not be able to rebuild the PDB after weaving, so the PDB mismatch is already there before the DLL is code signed.

Related

"image is too large" keeps on happening to openbmc image for Raspberrypi platform

Could someone please give me advice to make an openbmc image for Raspberrypi platform ?
Before I tried, I looked through related documents and believed an openbmc image can be worked on Raspberrypi.
Like OpenBMC with Raspberry Pi (2 or 3) and build bmcweb?
and https://kevinleeblog.github.io/project1/2019/11/25/openbmc-for-raspberry-pi-zero/.
So, I followed these instructions and tried the following steps.
#1: Git clone openbmc.git to my local PC.
tm#tm-VB1:~/Rpi4-64$ git clone https://github.com/openbmc/openbmc.git
Snip the logs but it looks no problem.
Receiving objects: 100% (182121/182121), 84.10 MiB | 5.55 MiB/s, done.
Resolving deltas: 100% (96860/96860), done.
#2: set TEMPLATECONF for raspberrypi
tm#tm-VB1:~/Rpi4-64$ export TEMPLATECONF=meta-evb/meta-evb-raspberrypi/conf
tm#tm-VB1:~/Rpi4-64$ echo $TEMPLATECONF
meta-evb/meta-evb-raspberrypi/conf
#3: set up the environment by "openbmc-env"
tm#tm-VB1:~/Rpi4-64/openbmc$ . openbmc-env
### Initializing OE build env ###
Snip the logs but it looks no problem. As you know, the script automatically creates a subdirectory,build, under openbmc.
Common targets are:
obmc-phosphor-image
tm#tm-VB1:~/Rpi4-64/openbmc/build$
#4: Change the directory and edit local.conf for my Raspberrypi platform.
tm#tm-VB1:~/Rpi4-64/openbmc/build$ cat ./conf/local.conf
Snip the log for unchanged part.
MACHINE ??= "raspberrypi4-64" <<< Change here for my platform.
DL_DIR ?= "/home/tm/Yocto/downloads" <<< Add here for build-time reduction at retry.
SSTATE_DIR ?= "/home/tm/Yocto/sstate-cache" <<< Add here for build-time reduction at retry.
#5: Change FLASH_SIZE variable based on the following sugestion. https://github.com/openbmc/openbmc/issues/3590
tm#tm-VB1:~/Rpi4-64/openbmc/meta-phosphor/classes$ cat image_types_phosphor.bbclass
Snip the log.
# Flash characteristics in KB unless otherwise noted
FLASH_SIZE ?= "131072" <<< I changed only this variable from 32768 to 131072.
#6: bitbake starts.
tm#tm-VB1:~/Rpi4-64/openbmc/bitbake obmc-phosphor-image
Then, ERROR happened.
ERROR: Logfile of failure stored in: /home/tm/Rpi/openbmc/build/tmp/work/raspberrypi-openbmc-linux-gnueabi/obmc-phosphor-image/1.0-r0/temp/log.do_generate_static.2055074
DEBUG: Executing python function do_generate_static
DEBUG: Executing shell function do_mk_static_nor_image
32768+0 records in
32768+0 records out
33554432 bytes (34 MB, 32 MiB) copied, 0.09147 s, 367 MB/s
DEBUG: Shell function do_mk_static_nor_image finished
DEBUG: Considering file size=495980 name=/home/tm/Rpi/openbmc/build/tmp/deploy/images/raspberrypi/u-boot.bin
DEBUG: Spanning start=0K end=512K
DEBUG: Compare needed=495980 available=524288 margin=28308
484+1 records in
484+1 records out
495980 bytes (496 kB, 484 KiB) copied, 0.00120141 s, 413 MB/s
DEBUG: Considering file size=8266960 name=/home/tm/Rpi/openbmc/build/tmp/deploy/images/raspberrypi/fitImage-obmc-phosphor-initramfs-raspberrypi-raspberrypi
DEBUG: Spanning start=512K end=4864K
>>>DEBUG: Compare needed=8266960 available=4456448 margin=-3810512
ERROR: Image '/home/tm/Rpi/openbmc/build/tmp/deploy/images/raspberrypi/fitImage-obmc-phosphor-initramfs-raspberrypi-raspberrypi' is too large!
DEBUG: Python function do_generate_static finished
It said margin=-3810512.
Now, my 2nd try.
I removed the whole openbmc directory and did the same steps above.
But this time, I change FLASH_SIZE from 32768 to 262144.
It is the same result like below.
ERROR: obmc-phosphor-image-1.0-r0 do_generate_static: Image '/home/tm/Rpi4/openbmc/build/tmp/deploy/images/raspberrypi4/u-boot.bin' is too large!
ERROR: Logfile of failure stored in: /home/tm/Rpi4/openbmc/build/tmp/work/raspberrypi4-openbmc-linux-gnueabi/obmc-phosphor-image/1.0-r0/temp/log.do_generate_static.2061792
ERROR: Task (/openbmc/meta-phosphor/recipes-phosphor/images/obmc-phosphor-image.bb:do_generate_static) failed with exit code '1'
NOTE: Tasks Summary: Attempted 3915 tasks of which 2633 didn't need to be rerun and 1 failed.
Summary: 1 task failed:
/openbmc/meta-phosphor/recipes-phosphor/images/obmc-phosphor-image.bb:do_generate_static
Summary: There were 2 WARNING messages shown.
Summary: There was 1 ERROR message shown, returning a non-zero exit code.
tm#tm-VB1:~/Rpi4/openbmc/build$ cat /home/tm/Rpi4/openbmc/build/tmp/work/raspberrypi4-openbmc-linux-gnueabi/obmc-phosphor-image/1.0-r0/temp/log.do_generate_static.2061792
DEBUG: Executing python function do_generate_static
DEBUG: Executing shell function do_mk_static_nor_image
32768+0 records in
32768+0 records out
33554432 bytes (34 MB, 32 MiB) copied, 0.177223 s, 189 MB/s
DEBUG: Shell function do_mk_static_nor_image finished
DEBUG: Considering file size=548224 name=/home/tm/Rpi4/openbmc/build/tmp/deploy/images/raspberrypi4/u-boot.bin
DEBUG: Spanning start=0K end=512K
>>>DEBUG: Compare needed=548224 available=524288 margin=-23936
ERROR: Image '/home/tm/Rpi4/openbmc/build/tmp/deploy/images/raspberrypi4/u-boot.bin' is too large!
DEBUG: Python function do_generate_static finished
tm#tm-VB1:~/Rpi4/openbmc/build$
It said margin=-23936.
OK. Image is too large. So,my 3rd try.
I removed the whole openbmc directory and did the same steps above.
But this time, I change FLASH_SIZE from 32768 to 9437184.
It is the same result like below.
ERROR: obmc-phosphor-image-1.0-r0 do_generate_static: Image '/home/tm/Rpi4/openbmc/build/tmp/deploy/images/raspberrypi4/u-boot.bin' is too large!
ERROR: Logfile of failure stored in: /home/tm/Rpi4/openbmc/build/tmp/work/raspberrypi4-openbmc-linux-gnueabi/obmc-phosphor-image/1.0-r0/temp/log.do_generate_static.2058361
ERROR: Task (/openbmc/meta-phosphor/recipes-phosphor/images/obmc-phosphor-image.bb:do_generate_static) failed with exit code '1'
NOTE: Tasks Summary: Attempted 3935 tasks of which 0 didn't need to be rerun and 1 failed.
Summary: 1 task failed:
/openbmc/meta-phosphor/recipes-phosphor/images/obmc-phosphor-image.bb:do_generate_static
Summary: There were 4 WARNING messages shown.
Summary: There was 1 ERROR message shown, returning a non-zero exit code.
tm#tm-VB1:~/Rpi4/openbmc$
tm#tm-VB1:~/Rpi4/openbmc$ cat /home/tm/Rpi4/openbmc/build/tmp/work/raspberrypi4-openbmc-linux-gnueabi/obmc-phosphor-image/1.0-r0/temp/log.do_generate_static.2058361
DEBUG: Executing python function do_generate_static
DEBUG: Executing shell function do_mk_static_nor_image
32768+0 records in
32768+0 records out
33554432 bytes (34 MB, 32 MiB) copied, 0.173685 s, 193 MB/s
DEBUG: Shell function do_mk_static_nor_image finished
DEBUG: Considering file size=548224 name=/home/tm/Rpi4/openbmc/build/tmp/deploy/images/raspberrypi4/u-boot.bin
DEBUG: Spanning start=0K end=512K
>>>DEBUG: Compare needed=548224 available=524288 margin=-23936
ERROR: Image '/home/tm/Rpi4/openbmc/build/tmp/deploy/images/raspberrypi4/u-boot.bin' is too large!
DEBUG: Python function do_generate_static finished
tm#tm-VB1:~/Rpi4/openbmc$
It said the same margin as 256MB case.
My 4th try.
I removed the whole openbmc directory and did the same steps above.
I changed MACHINE ??= "raspberrypi4-64" to "raspberrypi2"
But this time, I change FLASH_SIZE from 32768 to 33554432.
It is the same result before.
My 5th try.
I removed the whole openbmc directory and did the same steps above.
I used MACHINE ??= "raspberrypi2"
But this time, I change FLASH_SIZE from 32768 to 67108864.
It is the same result before.
After I tried several variations, it always said "image is too large" although I changed FLASH_SIZE to much much larger one.
So, I am wondering if I have missed some important configuration or it needs another parameter to fix this except FLASH_SIZE.
By the way, I tried romulus and made it.
My environment is ubuntu-20.04.2.0-desktop-amd64.
I really appreciate someone could kindly give me advice to make this work.
Interesting, I don't have a quick fix for you but I did notice the partition that is over sized is the uboot partition. The uboot is a smaller separate binary installed on the machine. It looks as if your uboot build is over 512k and the partition is set for 512k. Your flash size is massize
FLASH_SIZE = 9437184" that is more then a gig, (because FLASH_SIZE is in K)
If I were you I would first try to build an older version of openbmc for raspberry pi. (It used to work so you just need to find the commit before uboot grew to big). Use git to move back a month until you find it works.
If that does not work I would try to modify the partition table.
here is where you failing
this looks fine building the uboot image looks fine
increasing the kernel offset make if build, but the other targets in openbmc will not be happy with this solution. So maybe meta-raspberry-pi will have to override the partition table (if uboot can not be shrunk)
What ever you do, open an issue on the github and share you changes. Also use the discord, and gerrit.
I just replicated this issue. We should fix it

bash script calling rdiff-backup never ends

I want to run rdiff-backup and then switch of the raspberrypi it was running on.
I use the following script:
#!/bin/sh
date > /home/mik/rdiff-backup.log
echo "rsync start" >> /home/mik/rdiff-backup.log
rdiff-backup -v5 --print-statistics offlinebackup#server::/srv/backup /srv/datenserverBackup/backup >> /home/mik/rdiff-backup.log 2>&1
sync
date >> /home/mik/rdiff-backup.log
echo "rdiff-backup end" >> /home/mik/rdiff-backup.log
df -h >> /home/mik/rdiff-backup.log
sync
halt
The log file looks good (for the rdiff-backup part):
Sat 12 Aug 08:20:59 UTC 2017
rsync start
Unable to import win32security module. Windows ACLs
not supported by filesystem at /srv/backup
escape_dos_devices not required by filesystem at /srv/backup
Warning: name offlinebackup not found on system, dropping ACL entry.
Further ACL entries dropped with this name will not trigger further warnings
Using rdiff-backup version 1.2.8
Executing ssh -C offlinebackup#server rdiff-backup --server
-----------------------------------------------------------------
Detected abilities for source (read only) file system:
Access control lists On
Extended attributes On
Windows access control lists Off
Case sensitivity On
Escape DOS devices Off
Escape trailing spaces Off
Mac OS X style resource forks Off
Mac OS X Finder information Off
-----------------------------------------------------------------
Unable to import win32security module. Windows ACLs
not supported by filesystem at /srv/datenserverBackup/backup/rdiff-backup-data/rdiff-backup.tmp.0
escape_dos_devices not required by filesystem at /srv/datenserverBackup/backup/rdiff-backup-data/rdiff-backup.tmp.0
-----------------------------------------------------------------
Detected abilities for destination (read/write) file system:
Ownership changing On
Hard linking On
fsync() directories On
Directory inc permissions On
High-bit permissions On
Symlink permissions Off
Extended filenames On
Windows reserved filenames Off
Access control lists On
Extended attributes On
Windows access control lists Off
Case sensitivity On
Escape DOS devices Off
Escape trailing spaces Off
Mac OS X style resource forks Off
Mac OS X Finder information Off
-----------------------------------------------------------------
Backup: must_escape_dos_devices = 0
Starting increment operation /srv/backup to /srv/datenserverBackup/backup
Processing changed file .
Incrementing mirror file /srv/datenserverBackup/backup
Processing changed file abc
Incrementing mirror file /srv/datenserverBackup/backup/abc
Processing changed file abc/def
Incrementing mirror file /srv/datenserverBackup/backup/abc/def
Processing changed file abc/def/testfile.dxf
Incrementing mirror file /srv/datenserverBackup/backup/abc/def/testfile.dxf
--------------[ Session statistics ]--------------
StartTime 1502526061.00 (Sat Aug 12 08:21:01 2017)
EndTime 1502527913.72 (Sat Aug 12 08:51:53 2017)
ElapsedTime 1852.72 (30 minutes 52.72 seconds)
SourceFiles 151099
SourceFileSize 386321558216 (360 GB)
MirrorFiles 151097
MirrorFileSize 386321447731 (360 GB)
NewFiles 2
NewFileSize 110485 (108 KB)
DeletedFiles 0
DeletedFileSize 0 (0 bytes)
ChangedFiles 1
ChangedSourceSize 0 (0 bytes)
ChangedMirrorSize 0 (0 bytes)
IncrementFiles 4
IncrementFileSize 0 (0 bytes)
TotalDestinationSizeChange 110485 (108 KB)
Errors 0
--------------------------------------------------
The backup is working, but then the script ends right there.
rdiff-backup.log contains the full report of rdiff-backup. But neither the line "rdiff-backup end", nor the output of "df -h".
How can I make it ran to the end?
Thanks for your answers
I finally found a workaround, that solves my problem.
My sciprt which is called after booting from /etc/init.d is calling the other script which does the actual work (i.e. backup my data, and write the log file) as a background task.
/etc/init.d/CallAfterBoot.sh
#!/bin/sh
sleep 30
/home/me/DoBackup.sh & # '&' starts the script in background
/home/me/DoBackup.sh is the script I posted above which is now runing correctly.
Same script running as the same user now behaves differently. There's got to be some bug somewhere, however, it works for me now.

Process description,user and publisher information in dump

Is there a way to retrieve the process description and publisher from a Windows Kernel crash dump?
I tried !process and !dml_proc. It doesn't show that information.
Description sould be get from exe module.
Example:
0: kd> !PROCESS fffffa800482f940 2
GetPointerFromAddress: unable to read from fffff80397f65000
PROCESS fffffa800482f940
SessionId: 1 Cid: 0e3c Peb: 7f7cfefa000 ParentCid: 04bc
DirBase: 26bcc000 ObjectTable: fffff8a0028f4e80 HandleCount: <Data Not Accessible>
Image: Taskmgr.exe
0: kd> .process /p fffffa800482f940
Implicit process is now fffffa80`0482f940
0: kd> .reload /user
Loading User Symbols
..........................................................
0: kd> lmvm Taskmgr
Browse full module list
start end module name
000007f7`d08c0000 000007f7`d09da000 taskmgr (deferred)
Image path: C:\Windows\system32\taskmgr.exe
Image name: taskmgr.exe
Browse all global symbols functions data
Timestamp: Thu Jul 26 02:07:18 2012 (50107C26)
CheckSum: 00119B41
ImageSize: 0011A000
File version: 6.2.9200.16384
Product version: 6.2.9200.16384
File flags: 0 (Mask 3F)
File OS: 40004 NT Win32
File type: 1.0 App
File date: 00000000.00000000
Translations: 0409.04b0
CompanyName: Microsoft Corporation
ProductName: Microsoft® Windows® Operating System
InternalName: Taskmgr.exe
OriginalFilename: Taskmgr.exe
ProductVersion: 6.2.9200.16384
FileVersion: 6.2.9200.16384 (win8_rtm.120725-1247)
FileDescription: Task Manager
LegalCopyright: © Microsoft Corporation. All rights reserved.
i was trying to edit and post clarification to pykd-teams answer but the edit turned out ot be substantial so posted this as an answer
FileDescription from lmvm output refers to description column in Task manager Details tab
Company Name Refers refers to publisher column in startup tab
C:\Windows\system32>wmic Startup where Caption="vmware user process" get /format:list
Caption=VMware User Process
Command="C:\Program Files\VMware\VMware Tools\vmtoolsd.exe" -n vmusr
Description=VMware User Process
Location=HKLM\SOFTWARE\Microsoft\Windows\CurrentVersion\Run
Name=VMware User Process
SettingID=
User=Public
UserSID=
C:\Windows\system32>reg query hklm\software\microsoft\windows\currentversion\run
HKEY_LOCAL_MACHINE\software\microsoft\windows\currentversion\run
VMware User Process REG_SZ "C:\Program Files\VMware\VMware Tools\vmtoolsd.exe" -n vmusr
you can check by modifiying the FILE_VERSION_INFO in rsrc section of any file that is executed in startup see below an edited publisher in windbg of the same exe as posted above
how to check for validity of the above assertion
open a live kd session
run task manager in target and select a startup
look at details and locate the process name say vmtoolsd.exe
break into kd using ctrl+break
!process 0 0 vmtoolsd.exe
.process /p /r EPROCESS ADDRESS OF vmtoolsd.exe
!dh vmtoolsd find the Data directory SECURITY DIRECTORY and start searching for FILE_VERSION_INFO
loacte the string value of Company Name
use eb Address to edit the Company Name to some random string
execute using g
now execute task manager and you will see the publisher column in startup tab reflecting the random string as publisher
username is not tied to file but to process grab the token
from !process <Eproc> 1 and pass the TOKEN value to !token -n
here is sample script to retrieve user names for each running process
!for_each_process "r $t0=(##c++(((_EPROCESS*) ##Process )->Token.Object)&0xfffffff8);r? $t1=##c++(((_TOKEN*)##(#$t0))->LogonSession->AccountName);r? $t2=##c++(((_EPROCESS *) ##Process )->ImageFileName);.printf \"%mu\t\t\t%ma\\n\",##c++((wchar_t *)#$t1.Buffer),##c++((char*)#$t2)"
should get your results like this
kd> $$>a< getuname4proc.txt
xx-PC$ smss.exe
LOCAL SERVICE svchost.exe
xx taskhost.exe
You can get this information and a lot more by using !ps command from DbgKit.
Note: From a Kernel Memory Dump you can only get User Name. To get User Name, File Description and Company Name you need a Complete Memory Dump.
Copy dbgkit.dll into winext folder (For example: C:\Program Files (x86)\Windows Kits\10\Debuggers\x64\winext)
Open a dump file in WinDbg
Run .load dbgkit command
Run !ps command (to view other commands run !dbgkit.help)

Unable to verify checksum for exe

hi i have attached crash dump for an exe and symbols also.but i am getting this error:
Unable to verify checksum for abc.exe.
What would be the reason for this?
Unable to verify checksum is emitted when the checksum in the PE header isn't verifiable.
This can happen if the exe in question was compiled and linked without using /RELEASE linker option.
Normal project based compile linker sets this option. nmake or batch file based compilation can omit this switch and can lead to this output.
A simple hello world compiled and linked with and without /RELEASE linker option (PDB not generated for simpilicity and diffed to show the difference in timestamp and checksum). Loaded in WinDbg and checksum warning is generated only for the exe with no checksum in PE header.
simple hello world.cpp contents
testrelease:\>dir /b & type testrelease.cpp
testrelease.cpp
#include <stdio.h>
int main (void) {
printf("hello my relase\n");
return 0;
}
compiling without /RELEASE
testrelease:\>cl /nologo testrelease.cpp
testrelease.cpp
renaming the exe and compiling the same source with with /RELEASE
testrelease:\>ren testrelease.exe testrelease_norel.exe
testrelease:\>cl /nologo testrelease.cpp /link /release
testrelease.cpp
comparing both exes
testrelease:\>fc /b testrelease.exe testrelease_norel.exe
Comparing files testrelease.exe and TESTRELEASE_NOREL.EXE
000000E0: D6 CE
00000130: A3 00
00000131: 95 00
00000132: 01 00
analysing output of the comparison
testrelease:\>xxd -s +0x3c -l 1 testrelease.exe
000003c: d8 .
testrelease:\>xxd -s +0x3c -l 1 testrelease_norel.exe
000003c: d8 .
testrelease:\>echo d8 = NT_HEADER so e0 = TimeDateStamp and 130 = CheckSum
d8 = NT_HEADER so e0 = TimeDateStamp and 130 = CheckSum
loading both exes in windbg warning generated for only one exe without checksum
testrelease:\>cdb -c ".reload /f ; q" testrelease.exe
.*** ERROR: Module load completed but symbols could not be loaded for image00400
testrelease:\>cdb -c ".reload /f ; q" testrelease_norel.exe
.*** WARNING: Unable to verify checksum for image00400000
*** ERROR: Module load completed but symbols could not be loaded for image004000
no symbol header available error means the exe was compiled without debug information.
You can't do much about it unless you have a lot of expertise in recreating debug information from scratch.
Both the executables that are compiled above will generate the error because iIhave intentionally not created the debug information.
DBGHELP: image00400000 missing debug info. Searching for pdb anyway
DBGHELP: Can't use symbol server for image00400000.pdb - no header information available

Sometimes Windows Crash while LabVIEW operator Advantech PCI-1716 card

The system include 2 Advantech PCI-1716 card:http://www.advantech.com/products/PCI-1716/mod_86EC4C4D-F497-45C5-81DA-B8600C0EB36F.aspx
I write a program with NI LabVIEW 7.1.
The LabVIEW program can control the two PCI 1716 card.
It work very good.
But about a year later, The computer crashed and sometimes auto restart.
Nobody change the software and hardware.
I used WinDbg to analysis the Windows crash dump file and I think I find something abnormal.
For the WinDbg result, I think maybe the PCI-1716 driver was broken so I re-installed it.
But the problem also happen.
And I also re-installed my windows xp.
The also problem happen again.
I don't know how.
Maybe the PCI 1716 hardware broken.
But how to find which one, there are two PCI 1716 card.
The WinDbg result is:
Loading Dump File [F:\WinDBG\Mini012313-01.dmp] Mini Kernel Dump File: Only registers and stack trace are available
Symbol search path is: srv*C:\Documents and Settings\cky\symbols*http://msdl.microsoft.com/download/symbols Executable search path is: Windows XP Kernel Version 2600 (Service Pack 2) MP (2 procs) Free x86 compatible Product: WinNt, suite: TerminalServer SingleUserTS Built by: 2600.xpsp_sp2_rtm.040803-2158 Kernel base = 0x804d8000 PsLoadedModuleList = 0x8055d700 Debug session time: Wed Jan 23 23:09:23.250 2013 (GMT+8) System Uptime: 0 days 23:47:48.802 Loading Kernel Symbols ....................................................................................................... Loading User Symbols Loading unloaded module list ....
*******************************************************************************
* *
* Bugcheck Analysis *
* *
*******************************************************************************
Use !analyze -v to get detailed debugging information.
BugCheck 1000000A, {ffdf, 2, 1, 806e5a8e}
Unable to load image ADS1716S.sys, Win32 error 0n2
*** WARNING: Unable to verify timestamp for ADS1716S.sys
*** ERROR: Module load completed but symbols could not be loaded for ADS1716S.sys Probably caused by : ADS1716S.sys ( ADS1716S+f08 )
Followup: MachineOwner
---------
0: kd> !analyze -v
*******************************************************************************
* *
* Bugcheck Analysis *
* *
*******************************************************************************
IRQL_NOT_LESS_OR_EQUAL (a) An attempt was made to access a pageable (or completely invalid) address at an interrupt request level (IRQL) that is too high. This is usually caused by drivers using improper addresses. If a kernel debugger is available get the stack backtrace. Arguments: Arg1: 0000ffdf, memory referenced Arg2: 00000002, IRQL Arg3: 00000001, bitfield : bit 0 : value 0 = read operation, 1 = write operation bit 3 : value 0 = not an execute operation, 1 = execute operation (only on chips which support this level of status) Arg4: 806e5a8e, address which referenced memory
Debugging Details:
------------------
WRITE_ADDRESS: 0000ffdf
CURRENT_IRQL: 2
FAULTING_IP: hal!KeAcquireQueuedSpinLock+42 806e5a8e 8902 mov dword ptr [edx],eax
CUSTOMER_CRASH_COUNT: 1
DEFAULT_BUCKET_ID: DRIVER_FAULT
BUGCHECK_STR: 0xA
PROCESS_NAME: LabVIEW.exe
LAST_CONTROL_TRANSFER: from 804efbb0 to 806e5a8e
STACK_TEXT: aa286c00 804efbb0 aa286c20 804f0de4 aa286c1c hal!KeAcquireQueuedSpinLock+0x42 aa286c08 804f0de4 aa286c1c 8602dc38 806e53b8 nt!IoAcquireCancelSpinLock+0xe aa286c20 804f104a 8602dc38 00000001 aa286c58 nt!IopStartNextPacket+0x18 aa286c30 f788ef08 8602dc38 00000001 85eb0230 nt!IoStartNextPacket+0x38 WARNING: Stack unwind information not available. Following frames may be wrong. aa286c58 80575529 85eb0230 86090da0 85f32448 ADS1716S+0xf08 aa286c80 805d1cb9 85f32448 85f32448 85f32690 nt!IoCancelThreadIo+0x33 aa286d08 805d209a 00000001 85f32448 00000000 nt!PspExitThread+0x403 aa286d28 805d2275 85f32448 00000001 aa286d64 nt!PspTerminateThreadByPointer+0x52 aa286d54 8054160c 00000000 00000001 0012e810 nt!NtTerminateProcess+0x105 aa286d54 7c92eb94 00000000 00000001 0012e810 nt!KiFastCallEntry+0xfc 0012e810 00000000 00000000 00000000 00000000 0x7c92eb94
STACK_COMMAND: kb
FOLLOWUP_IP: ADS1716S+f08 f788ef08 ?? ???
SYMBOL_STACK_INDEX: 4
SYMBOL_NAME: ADS1716S+f08
FOLLOWUP_NAME: MachineOwner
MODULE_NAME: ADS1716S
IMAGE_NAME: ADS1716S.sys
DEBUG_FLR_IMAGE_TIMESTAMP: 45b96de9
FAILURE_BUCKET_ID: 0xA_W_ADS1716S+f08
BUCKET_ID: 0xA_W_ADS1716S+f08
Followup: MachineOwner
---------
0: kd> lmvm ADS1716S start end module name f788e000 f7893ac0 ADS1716S T (no symbols)
Loaded symbol image file: ADS1716S.sys
Image path: ADS1716S.sys
Image name: ADS1716S.sys
Timestamp: Fri Jan 26 10:56:41 2007 (45B96DE9)
CheckSum: 00010C75
ImageSize: 00005AC0
Translations: 0000.04b0 0000.04e0 0409.04b0 0409.04e0
I'd run Memtest86+ just to see if you have any bad RAM.
Following that maybe you can pull 1 card out and write a simple script to do some I/O (similar to what you're doing normally maybe). If it takes a while you could put it into a loop and run it for a while and see if you can force a crash. You could repeat the test on each card to see if you have any issues with just one of the cards.
Does the crash happen often or rarely?