Remove locks IRP-by-IRP

TheMSsForum.com: The Microsoft Software Forum

The MSS Forum ‹ Drivers
- Archive
  - Biz
  - MCSE
  - CRM
  - Drivers
  - Framework
  - ADO
  - ASP
  - Compact
  - Forms
  - Dotnet
  - C#
  - VB
  - FontpageGen
  - Excel
  - WorkSheet
  - Exchange
  - Setup
  - Fox
  - Fontpage
  - ASP
  - IIS
  - Entourage
  - Money
  - Messanger
  - PocketPC
  - Powerpoint
  - Project
  - Publisher
  - Excel
  - VB
  - Security
  - Portal
  - Services
  - SQLServerDev
  - SVCS
  - SQLServer
  - VB
  - VC
  - MFC
  - ExcelGen

- Previous
 - 1
 - MiniSpy source? I have downloaded the WDK 6000 from MSDN (I have a premium account) and cannot find the source MiniSpy. I'm looking for the source code to some sample MiniFilter driver. I have never developed a file system filter (either legacy or with the File System Manager archecture) and experience has taught me to find sample code before leaping into something new. Anyone know why this sample code appears to be missing? Know where I can get it? Is there better/newer MiniFilter source available? Anyone? Anyone? Bueller?? Bueller?? Tag: Remove locks IRP-by-IRP Tag: 89882
 - 2
 - What does "arp entry addition failed: 5" mean? I see the following mesg for the error of above type: Error Message: The ARP entry addition failed: parameter User Action: Check the command-line format for the Address Resolution Protocol (ARP) command. When adding an entry to the table, the Internet Protocol (IP) address must be entered in a dotted decimal format; the associated media access control (MAC) address must be entered in a 6-digit hexadecimal format with digits separated by hyphens. But in my case, I get this error only sometimes for the same syntax and arguments and for the same card and not always. The command I use: arp -s <remote IP> <remote mac> Looks like device/driver bindings got screwed up or something. If I disable my device and then reenable, the exact same command works. The above text on microsoft site does not provide any useful information on what the value "5" in error mesg means? Does anyone have info on what this error no means? Regds, Tag: Remove locks IRP-by-IRP Tag: 89879
 - 3
 - Which is the first NDIS routine registered by miniport with NDIS that needs to be called when network device is disabled? This is a multi-part message in MIME format. ------=_NextPart_000_0026_01C72391.85217160 Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: quoted-printable Which is the first NDIS routine registered by miniport with NDIS that = needs to be called when network device is disabled? When I try to disable the ethernet device, the device manager hangs and = device disable never seems to get disabled. I see lots of incoming traffic and packets getting indicated from my = driver to NDIS. \windows\inf\setuapi.app.log has the log enclosed below. Wondering why disable is not going through.. Looking at my device = context I dont have any pendings sends to be completed. Neither is my pause routine getting called. Nor is Ndis close routine = getting called. Any what the problem could be? Regds, >>> [DIF_PROPERTYCHANGE - = {1A3E09BE-1E45-494B-9174-D7385B45BBF5}\NVNET_DEV0373\4&55C493A&0&00] >>> Section start 2006/12/19 00:28:50.234 cmd: "C:\Windows\system32\mmc.exe" C:\Windows\system32\devmgmt.msc dvi: Using exported function 'NetClassInstaller' in module = 'C:\Windows\system32\NetCfgx.dll'. dvi: Class installer =3D=3D NetCfgx.dll,NetClassInstaller dvi: Using exported function 'NciDeviceInstall' in module = 'C:\Windows\system32\nci.dll'. dvi: CoInstaller 1 =3D=3D nci.dll,NciDeviceInstall dvi: Using exported function 'WlanDeviceClassCoInstaller' in module = 'C:\Windows\system32\wlaninst.dll'. dvi: CoInstaller 2 =3D=3D wlaninst.dll,WlanDeviceClassCoInstaller dvi: Using exported function 'FDCoInstaller' in module = 'C:\Windows\system32\fdco1ins.dll'. dvi: CoInstaller 3 =3D=3D fdco1ins.dll,FDCoInstaller dvi: CoInstaller 1: Enter 00:28:50.265 dvi: CoInstaller 1: Exit dvi: CoInstaller 2: Enter 00:28:50.281 dvi: CoInstaller 2: Exit dvi: CoInstaller 3: Enter 00:28:50.281 dvi: CoInstaller 3: Exit dvi: Class installer: Enter 00:28:50.281 dvi: Class installer: Exit dvi: Default installer: Enter 00:28:50.281 dvi: {Change State} dvi: Device Instance =3D = '{1A3E09BE-1E45-494B-9174-D7385B45BBF5}\NVNET_DEV0373\4&55C493A&0&00'. dvi: {DICS_DISABLE, DICS_FLAG_GLOBAL} dvi: Using exported function 'NetClassInstaller' in = module 'C:\Windows\system32\NetCfgx.dll'. dvi: Class installer =3D=3D NetCfgx.dll,NetClassInstaller dvi: Using exported function 'NciDeviceInstall' in module = 'C:\Windows\system32\nci.dll'. dvi: CoInstaller 1 =3D=3D nci.dll,NciDeviceInstall dvi: Using exported function 'WlanDeviceClassCoInstaller' = in module 'C:\Windows\system32\wlaninst.dll'. dvi: CoInstaller 2 =3D=3D = wlaninst.dll,WlanDeviceClassCoInstaller dvi: Using exported function 'FDCoInstaller' in module = 'C:\Windows\system32\fdco1ins.dll'. dvi: CoInstaller 3 =3D=3D fdco1ins.dll,FDCoInstaller dvi: CoInstaller 1: Enter 13:34:36.234 dvi: CoInstaller 1: Exit dvi: CoInstaller 2: Enter 13:34:36.234 dvi: CoInstaller 2: Exit dvi: CoInstaller 3: Enter 13:34:36.234 dvi: CoInstaller 3: Exit dvi: Class installer: Enter 13:34:36.234 dvi: Class installer: Exit dvi: Default installer: Enter 13:34:36.234 dvi: {Change State} dvi: Device Instance =3D = '{1A3E09BE-1E45-494B-9174-D7385B45BBF5}\NVNET_DEV0373\4&55C493A&0&00'. dvi: {DICS_DISABLE, DICS_FLAG_GLOBAL} dvi: Using exported function = 'NetClassInstaller' in module 'C:\Windows\system32\NetCfgx.dll'. dvi: Class installer =3D=3D = NetCfgx.dll,NetClassInstaller dvi: Using exported function 'NciDeviceInstall' = in module 'C:\Windows\system32\nci.dll'. dvi: CoInstaller 1 =3D=3D = nci.dll,NciDeviceInstall dvi: Using exported function = 'WlanDeviceClassCoInstaller' in module = 'C:\Windows\system32\wlaninst.dll'. dvi: CoInstaller 2 =3D=3D = wlaninst.dll,WlanDeviceClassCoInstaller dvi: Using exported function 'FDCoInstaller' in = module 'C:\Windows\system32\fdco1ins.dll'. dvi: CoInstaller 3 =3D=3D = fdco1ins.dll,FDCoInstaller dvi: CoInstaller 1: Enter 14:00:13.609 dvi: CoInstaller 1: Exit dvi: CoInstaller 2: Enter 14:00:13.625 dvi: CoInstaller 2: Exit dvi: CoInstaller 3: Enter 14:00:13.625 dvi: CoInstaller 3: Exit dvi: Class installer: Enter 14:00:13.625 dvi: Class installer: Exit dvi: Default installer: Enter 14:00:13.625 dvi: {Change State} dvi: Device Instance =3D = '{1A3E09BE-1E45-494B-9174-D7385B45BBF5}\NVNET_DEV0373\4&55C493A&0&00'. dvi: {DICS_DISABLE, DICS_FLAG_GLOBAL} <ins> >>> [DIF_PROPERTYCHANGE - = {1A3E09BE-1E45-494B-9174-D7385B45BBF5}\NVNET_DEV0373\4&55C493A&0&00] >>> Section start 2006/12/19 13:34:36.187 cmd: "C:\Windows\system32\mmc.exe" C:\Windows\system32\devmgmt.msc <ins> >>> [DIF_PROPERTYCHANGE - = {1A3E09BE-1E45-494B-9174-D7385B45BBF5}\NVNET_DEV0373\4&55C493A&0&00] >>> Section start 2006/12/19 14:00:13.578 cmd: "C:\Windows\system32\mmc.exe" C:\Windows\system32\devmgmt.msc <ins> ------=_NextPart_000_0026_01C72391.85217160 Content-Type: text/html; charset="iso-8859-1" Content-Transfer-Encoding: quoted-printable <!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.0 Transitional//EN"> <HTML><HEAD> <META http-equiv=3DContent-Type content=3D"text/html; = charset=3Diso-8859-1"> <META content=3D"MSHTML 6.00.2900.3020" name=3DGENERATOR> <STYLE></STYLE> </HEAD> <BODY> <DIV> Which is the first NDIS routine registered by = miniport with=20 NDIS that needs to be called when network device is disabled? When I try to disable the ethernet device, the = device=20 manager hangs and device disable never seems to get disabled.  I see lots of incoming traffic and packets = getting=20 indicated from my driver to NDIS. \windows\inf\setuapi.app.log has the log enclosed=20 below. Wondering why disable is not going through.. = Looking at my=20 device context I dont have any pendings sends to be = completed.  Neither is my pause routine getting called. = Nor is=20 Ndis close routine getting called. Any what the problem could = be?   Regds,   >>>  [DIF_PROPERTYCHANGE -=20 {1A3E09BE-1E45-494B-9174-D7385B45BBF5}\NVNET_DEV0373\4&55C493A&0&= amp;00] >>> =20 Section start 2006/12/19 00:28:50.234       = cmd:=20 "C:\Windows\system32\mmc.exe"=20 C:\Windows\system32\devmgmt.msc      dvi: Using = exported=20 function 'NetClassInstaller' in module=20 'C:\Windows\system32\NetCfgx.dll'.      dvi: = Class=20 installer =3D=3D = NetCfgx.dll,NetClassInstaller      dvi:=20 Using exported function 'NciDeviceInstall' in module=20 'C:\Windows\system32\nci.dll'.      dvi: = CoInstaller 1 =3D=3D=20 nci.dll,NciDeviceInstall      dvi: Using exported = function 'WlanDeviceClassCoInstaller' in module=20 'C:\Windows\system32\wlaninst.dll'.      dvi: = CoInstaller=20 2 =3D=3D = wlaninst.dll,WlanDeviceClassCoInstaller      dvi: = Using exported function 'FDCoInstaller' in module=20 'C:\Windows\system32\fdco1ins.dll'.      dvi: = CoInstaller=20 3 =3D=3D fdco1ins.dll,FDCoInstaller      dvi: = CoInstaller 1:=20 Enter 00:28:50.265      dvi: CoInstaller 1:=20 Exit      dvi: CoInstaller 2: Enter=20 00:28:50.281      dvi: CoInstaller 2:=20 Exit      dvi: CoInstaller 3: Enter=20 00:28:50.281      dvi: CoInstaller 3:=20 Exit      dvi: Class installer: Enter=20 00:28:50.281      dvi: Class installer:=20 Exit      dvi: Default installer: Enter=20 00:28:50.281      = dvi:     =20 {Change State}     =20 dvi:           Device = Instance=20 =3D=20 '{1A3E09BE-1E45-494B-9174-D7385B45BBF5}\NVNET_DEV0373\4&55C493A&0= &00'.     =20 dvi:           = {DICS_DISABLE,=20 DICS_FLAG_GLOBAL}     =20 dvi:           Using = exported=20 function 'NetClassInstaller' in module=20 'C:\Windows\system32\NetCfgx.dll'.     =20 dvi:           Class = installer=20 =3D=3D NetCfgx.dll,NetClassInstaller     =20 dvi:           Using = exported=20 function 'NciDeviceInstall' in module=20 'C:\Windows\system32\nci.dll'.     =20 dvi:           = CoInstaller 1=20 =3D=3D nci.dll,NciDeviceInstall     =20 dvi:           Using = exported=20 function 'WlanDeviceClassCoInstaller' in module=20 'C:\Windows\system32\wlaninst.dll'.     =20 dvi:           = CoInstaller 2=20 =3D=3D = wlaninst.dll,WlanDeviceClassCoInstaller     =20 dvi:           Using = exported=20 function 'FDCoInstaller' in module=20 'C:\Windows\system32\fdco1ins.dll'.     =20 dvi:           = CoInstaller 3=20 =3D=3D fdco1ins.dll,FDCoInstaller     =20 dvi:           = CoInstaller 1:=20 Enter 13:34:36.234     =20 dvi:           = CoInstaller 1:=20 Exit     =20 dvi:           = CoInstaller 2:=20 Enter 13:34:36.234     =20 dvi:           = CoInstaller 2:=20 Exit     =20 dvi:           = CoInstaller 3:=20 Enter 13:34:36.234     =20 dvi:           = CoInstaller 3:=20 Exit     =20 dvi:           Class=20 installer: Enter 13:34:36.234     =20 dvi:           Class=20 installer: Exit     =20 dvi:           Default = installer: Enter 13:34:36.234     =20 dvi:           &nb= sp;   =20 {Change State}     =20 dvi:           &nb= sp;        =20 Device Instance =3D=20 '{1A3E09BE-1E45-494B-9174-D7385B45BBF5}\NVNET_DEV0373\4&55C493A&0= &00'.     =20 dvi:           &nb= sp;        =20 {DICS_DISABLE, DICS_FLAG_GLOBAL}     =20 dvi:           &nb= sp;        =20 Using exported function 'NetClassInstaller' in module=20 'C:\Windows\system32\NetCfgx.dll'.     =20 dvi:           &nb= sp;        =20 Class installer =3D=3D = NetCfgx.dll,NetClassInstaller     =20 dvi:           &nb= sp;        =20 Using exported function 'NciDeviceInstall' in module=20 'C:\Windows\system32\nci.dll'.     =20 dvi:           &nb= sp;        =20 CoInstaller 1 =3D=3D = nci.dll,NciDeviceInstall     =20 dvi:           &nb= sp;        =20 Using exported function 'WlanDeviceClassCoInstaller' in module=20 'C:\Windows\system32\wlaninst.dll'.     =20 dvi:           &nb= sp;        =20 CoInstaller 2 =3D=3D=20 wlaninst.dll,WlanDeviceClassCoInstaller     =20 dvi:           &nb= sp;        =20 Using exported function 'FDCoInstaller' in module=20 'C:\Windows\system32\fdco1ins.dll'.     =20 dvi:           &nb= sp;        =20 CoInstaller 3 =3D=3D = fdco1ins.dll,FDCoInstaller     =20 dvi:           &nb= sp;        =20 CoInstaller 1: Enter 14:00:13.609     =20 dvi:           &nb= sp;        =20 CoInstaller 1: Exit     =20 dvi:           &nb= sp;        =20 CoInstaller 2: Enter 14:00:13.625     =20 dvi:           &nb= sp;        =20 CoInstaller 2: Exit     =20 dvi:           &nb= sp;        =20 CoInstaller 3: Enter 14:00:13.625     =20 dvi:           &nb= sp;        =20 CoInstaller 3: Exit     =20 dvi:           &nb= sp;        =20 Class installer: Enter 14:00:13.625     =20 dvi:           &nb= sp;        =20 Class installer: Exit     =20 dvi:           &nb= sp;        =20 Default installer: Enter 14:00:13.625     =20 dvi:           &nb= sp;           &nbs= p; =20 {Change State}     =20 dvi:           &nb= sp;           &nbs= p;      =20 Device Instance =3D=20 '{1A3E09BE-1E45-494B-9174-D7385B45BBF5}\NVNET_DEV0373\4&55C493A&0= &00'.     =20 dvi:           &nb= sp;           &nbs= p;      =20 {DICS_DISABLE, DICS_FLAG_GLOBAL} <ins> >>>  [DIF_PROPERTYCHANGE -=20 {1A3E09BE-1E45-494B-9174-D7385B45BBF5}\NVNET_DEV0373\4&55C493A&0&= amp;00] >>> =20 Section start 2006/12/19 13:34:36.187       = cmd:=20 "C:\Windows\system32\mmc.exe" = C:\Windows\system32\devmgmt.msc <ins> >>>  [DIF_PROPERTYCHANGE -=20 {1A3E09BE-1E45-494B-9174-D7385B45BBF5}\NVNET_DEV0373\4&55C493A&0&= amp;00] >>> =20 Section start 2006/12/19 14:00:13.578       = cmd:=20 "C:\Windows\system32\mmc.exe"=20 C:\Windows\system32\devmgmt.msc <ins></DIV></BODY></H= TML> ------=_NextPart_000_0026_01C72391.85217160-- Tag: Remove locks IRP-by-IRP Tag: 89873
 - 4
 - APCs Dear all, Do all APCs run at APC_LEVEL Irql? For example, user APCs that start at ntdll!KiUserDispatchApc. I always thought that they did but !pcr is reporting an IRQL of 0. Many thanks, Carl Tag: Remove locks IRP-by-IRP Tag: 89872
 - 5
 - Using i.MX21 with a monochrome display Hello, I'm using i.MX21 on the ADS eval board from freescale. I try to drive a monochrome display. The BSP is delivered to support a sharp display and I've modified initialisation code for LCD in OAL but as soon as I configure the monochrome mode (clear TFT and COLOR bits of LPCR register) the FLM and LP signal are stopped ? any clue, any experience in this domain ... If you have already a configuration that works for a monochrome display it could help. Any advises are welcome thanks in advance Nyrbyk Tag: Remove locks IRP-by-IRP Tag: 89864
 - 6
 - USB device descriptor I'm not an expert of device drivers programming on Windows so I hope my question is not stupid for you. What I'm trying to do is reading the USB descriptor of an USB hard disk (or an USB key) in order to read the product id and the serial number of the device. The problem is that I can't even read the USB descritpor. The code is this: USB_DEVICE_DESCRIPTOR data; USBSCAN_GET_DESCRIPTOR descriptor; DWORD length; BOOL ret; descriptor.DescriptorType=USB_DEVICE_DESCRIPTOR_TYPE; ret=DeviceIoControl(hPhysicalDriveIOCTL,IOCTL_GET_USB_DESCRIPTOR , &descriptor,sizeof(descriptor),&data,sizeof(data), &length,NULL); hPhysicalDriveIOCTL is an handle that I get from a CreateFile call on the device I want to get the descriptor from. I tried GetLastError() and I received the error message "The request is not supported". What do you think it could be? Thanks a lot for your help Cold Tag: Remove locks IRP-by-IRP Tag: 89862
 - 7
 - what happens when WM_DEVMODECHANGE message is broadcasted Hi.. Whenever printer settings are changed, i am passing this message. SendMessageTimeout (HWND_BROADCAST, WM_DEVMODECHANGE, 0L, (LPARAM) pszPrinterName, SMTO_NORMAL, 1000, NULL) So that applications will come to know the change in devmode. But the new changes are not reflected in opened applications. Can somebody tell me what happens when WM_DEVMODECHANGE message is broadcasted?? and why the change is not reflected in opened application?? Thanks in Advance. Tag: Remove locks IRP-by-IRP Tag: 89861
 - 8
 - Create DDK projects in Visual Studio *easily*! Hi, wanted to introduce you to my Visual Studio project creation wizard: DDKWizard. It is based on the two DDKBUILD flavors of OSR and allows you to customize many aspects of the driver project. For example you can tell DDKWizard to create a project using C++ or C source files. You can optionally create PREfast configurations, 64bit configurations (either IA-64 and/or x64). There are 3 predefined project types. "Driver" creates a generic legacy driver (i.e. includes ntddk.h, could be easily changed to WDM, though). "Native App" creates a native application such as are smss.exe, autochk.exe and so on. "Empty Driver" allows you to migrate an existing project into a Visual Studio project. DDKWizard creates one project per DDK you choose from the listbox. The solution configurations will automatically include only the project which is relevant for the current configuration. The configurations (both project and solution configurations) are named after the DDKBUILD configuration, e.g. "WLHNET checked" (i.e. Windows Vista DDK for .NET Server checked build) or "WNET checked (PREfast)" (i.e. WNET DDK checked PREfast build). Download and information about the wizard at: http://ddkwizard.assarbad.net/ Yes it's free, yes it's OpenSource - no, your created projects are not affected by any particular license, they are yours. Yes, it will be free in future, yes I will continue to maintain it ;) Give it a try. If you don't want to install before looking into the documentation, have a look at the manual first: http://ddkwizard.assarbad.net/downloads/ddkwiz_manual.pdf It supports the DDKs supported by DDKBUILD (from OSR) and is compatible with Visual Studio 2002, 2003 and 2005 (tested only with one VS installation per test-system, yet). Best regards, Oliver PS: If you like it, spread the word! Tag: Remove locks IRP-by-IRP Tag: 89857
 - 9
 - How important is IoStatus.Information? I've seen five different ways of handling IoStatus.Information in the DDK docs, but one of them doesn't make sense to me: On success IoStatus.Information is set to a count to indicate how much work was performed, which on error implies it should be 0. On success IoStatus.Information is set to the address of some data structure allocated by the driver, which on error implies it should be NULL. On success IoStatus.Information is set to the address of some data structure allocated by another driver, which on error implies it should be left alone. On success or error, IoStatus.Information is irrelevant (not mentioned in the description of the IRP). On success, IoStatus.Information is 0. The last case makes no sense to me. If IoStatus.Information is only set on success, then on error it would either be some indeterminate value or the initial value from the originator, but the originator could just as easily check IoStatus.Status, so IoStatus.Information is irrelevant. Does any code actually depend on IoStatus.Information to be 0 on success, or were the doc authors just having fun pasting "On success, a driver sets Irp->IoStatus.Information to zero." over and over? Tag: Remove locks IRP-by-IRP Tag: 89847
 - 10
 - usb filter driver - constant serial number Hi, i am new to whole driver develompent so i would like some pointers. I'm trying to make my UBS printers (same vendorid and productid, different serial number) to be detected as one ie. my idea is to set serial number to constant value. Something like IgnoreHWSerNumXXXXXXXX registry key but that method is not good enough for me because when i plug printer into another usb port, windows detects it as new device. So, I would like to know it this possible and how. Thanks in advance. Tag: Remove locks IRP-by-IRP Tag: 89827
 - 11
 - parport.sys driver installation from user-land application i am working on ddk parport sample. after make, i get the parport.sys. now, i want to replace the system32\driver\parport.sys with this new one, in order to set breakpoint in windbg. i can net stop parport, then i want to overwrite the old parport.sys with the new parport.sys together with the pdb file . is it ok? TIA Tag: Remove locks IRP-by-IRP Tag: 89826
 - 12
 - Kernel Direct-x Hi, Can i use direct x in kernel mode device like doing direct drawing to the video card??? Where can i find some examples? Thanks Tag: Remove locks IRP-by-IRP Tag: 89819
 - 13
 - Can IRP_MN_REMOVE_DEVICE race with IRP_MN_QUERY_DEVICE_RELATIONS? Is it possible for IRP_MN_QUERY_DEVICE_RELATIONS to race with IRP_MN_REMOVE_DEVICE? I get the impression from "Multiprocessor Considerations for Kernel-Mode Drivers" that it is. (If not then what documentation should I be reading that says they are serialized?) If so, then is this possible: - bus driver's FDO receives IRP_MN_QUERY_DEVICE_RELATIONS (BusRelations) and reports child - child receives IRP_MN_START_DEVICE - PnP manager sends IRP_MN_REMOVE_DEVICE, still thinking child has been reported present - bus driver calls IoInvalidateDeviceRelations - bus driver's FDO receives IRP_MN_QUERY_DEVICE_RELATIONS (BusRelations) and does not report child - child receives IRP_MN_REMOVE_DEVICE and since it was not reported calls IoDeleteDevice on itself - PnP manager sends IRP_MN_REMOVE_DEVICE, now realizing child was not reported present - child receives IRP_MN_REMOVE_DEVICE, even though IoDeleteDevice has already been called ? I think the documentation alludes to the possibility of receiving IRP_MN_REMOVE_DEVICE even after IoDeleteDevice, but the explanation is shady. Tag: Remove locks IRP-by-IRP Tag: 89795
 - 14
 - History of remove locks I don't want to discuss again when exactly remove locks are necessary. The question I still have trouble answering is why they were ever invented, why they were ever necessary even for one single IRP. It seems kind of roundabout for reference counting to always protect a device object yet not always block what is basically a device object's destructor (IRP_MN_REMOVE_DEVICE). If the OS can protect a device from being removed in the middle of certain IRPs, then why not for all IRPs? Is there some historical reason? Was it a Windows 9x philosophy? Was safe object retention not as popular on such a consumer platform? Was it awkward to merge Windows 9x plug and play into the NT stream when Windows 2000 was created? (not that I know what the NT driver model looked like) Tag: Remove locks IRP-by-IRP Tag: 89792
 - 15
 - SRB_INITIALIZE_DEVICE Hello, I have two USB attached, DirectShow cameras that when booting up, only one of the cameras is successfully enumerated. The other camera will usually show up as an "Unknown Device" (VID_0000&PID_0000) or sometimes not at all in the device manager (as if I didn't even have it plugged in). After the reboot, unplugging and plugging back in the problem camera then allows it to be successfully enumerated. After using SoftICE to poke around to see what my driver is doing on a reboot, I see that only one of the cameras is getting the SRB_INITIALIZE_DEVICE and enumerating just fine. This is behind a native UHCI, Intel USB2.0 controller and a Dell system running XPPro, SP2. DirectX9.0C has been installed, as well as all other updates. It seems as if both cameras are successfully enumerated when behind a USB1.1 hub, so I'm not sure if this is timing related or not. Suggestions? What would cause the class driver to not send another SRB_INITIALIZE device and consequently get the descriptors, etc? Am I going to have to put a USB analyzer on it to determine or? If you need more information, just let me know. Any help, suggestions, tips, links to read up on would help. Thanks in adv. Tag: Remove locks IRP-by-IRP Tag: 89788
 - 16
 - METHOD_OUT_DIRECT and ISR I am updating/writing a driver that collects a large amount of data (4MB) from an ISA card. Here is the way I want it to work: In my Dispatch routine I handle a START_SAMPLING IOCTL, which is METHOD_OUT_DIRECT. I read in some settings from the input buffer and store what I need in my device extension. It is the output buffer (at Irp->MdlAddress) that I need help with. I translate this address using MmGetSystemAddressForMdlSafe(Irp->MdlAddress,NormalPagePriority), as I should. But now I need help. After my dispatch routine finishes, the user mode code will start a timer on the ISA card, which triggers interrupts to my driver. I have an ISR that runs once for each interrupt, and reads one data point each time... which it should append to the buffer. The interrupts keep occuring until the ISR sees that it has collected the all the data, and turns off the interrupts. So what my ISR needs to do is access and write to that output buffer that was given to the driver back in the dispatch routine? But how? I gather that in my dispatch routine I need to do something like.. lock the buffer and/or mark the irp as pending and/or queue the irp, etc. Then in my ISR I might need to do more of the same. Can someone tell me what I need to do to use this METHOD_OUT_DIRECT buffer in my ISR? Buzz Tag: Remove locks IRP-by-IRP Tag: 89784
 - 17
 - IoAttachDevice oddities I'm curious to understand a couple things about IoAttachDevice. Why does IoAttachDevice (not IoAttachDeviceToDeviceStack) require the AttachedDevice out parameter to point to a global memory location? Why can't it be on the stack? The stack will be around until IoAttachDevice returns, and why would IoAttachDevice have to access the pointer after that? I also wonder, why does IoAttachDevice open a file on the target device and then perform the attach _before_ it closes the file object instead of after? (so that the source device being attached on top receives spurious IRP_MJ_CLEANUP and IRP_MJ_CLOSE IRPs that have no matching IRP_MJ_CREATE) Was this an accident or is there some architectural reason for it? (which I'd like to understand if there is) Tag: Remove locks IRP-by-IRP Tag: 89783
 - 18
 - METHOD_OUT_DIRECT and ISR I am updating/writing a driver that collects a large amount of data (4MB) from an ISA card. Here is the way I want it to work: In my Dispatch routine I handle a START_SAMPLING IOCTL, which is METHOD_OUT_DIRECT. I read in some settings from the input buffer and store what I need in my device extension. It is the output buffer (at Irp->MdlAddress) that I need help with. I translate this address using MmGetSystemAddressForMdlSafe(Irp->MdlAddress,NormalPagePriority), as I should. But now I need help. After my dispatch routine finishes, the user mode code will start a timer on the ISA card, which triggers interrupts to my driver. I have an ISR that runs once for each interrupt, and reads one data point each time... which it should append to the buffer. The interrupts keep occuring until the ISR sees that it has collected the all the data, and turns off the interrupts. So what my ISR needs to do is access and write to that output buffer that was given to the driver back in the dispatch routine? But how? I gather that in my dispatch routine I need to do something like.. lock the buffer and/or mark the irp as pending and/or queue the irp, etc. Then in my ISR I might need to do more of the same. Can someone tell me what I need to do to use this METHOD_OUT_DIRECT buffer in my ISR? Buzz Tag: Remove locks IRP-by-IRP Tag: 89780
 - 19
 - METHOD_OUT_DIRECT and ISR I am updating/writing a driver that collects a large amount of data (4MB) from an ISA card. Here is the way I want it to work: In my Dispatch routine I handle a START_SAMPLING IOCTL, which is METHOD_OUT_DIRECT. I read in some settings from the input buffer and store what I need in my device extension. It is the output buffer (at Irp->MdlAddress) that I need help with. I translate this address using MmGetSystemAddressForMdlSafe(Irp->MdlAddress,NormalPagePriority), as I should. But now I need help. After my dispatch routine finishes, the user mode code will start a timer on the ISA card, which triggers interrupts to my driver. I have an ISR that runs once for each interrupt, and reads one data point each time... which it should append to the buffer. The interrupts keep occuring until the ISR sees that it has collected the all the data, and turns off the interrupts. So what my ISR needs to do is access and write to that output buffer that was given to the driver back in the dispatch routine? But how? I gather that in my dispatch routine I need to do something like.. lock the buffer and/or mark the irp as pending and/or queue the irp, etc. Then in my ISR I might need to do more of the same. Can someone tell me what I need to do to use this METHOD_OUT_DIRECT buffer in my ISR? Buzz Tag: Remove locks IRP-by-IRP Tag: 89779
 - 20
 - METHOD_OUT_DIRECT and ISR I am updating/writing a driver that collects a large amount of data (4MB) from an ISA card. Here is the way I want it to work: In my Dispatch routine I handle a START_SAMPLING IOCTL, which is METHOD_OUT_DIRECT. I read in some settings from the input buffer and store what I need in my device extension. It is the output buffer (at Irp->MdlAddress) that I need help with. I translate this address using MmGetSystemAddressForMdlSafe(Irp->MdlAddress,NormalPagePriority), as I should. But now I need help. After my dispatch routine finishes, the user mode code will start a timer on the ISA card, which triggers interrupts to my driver. I have an ISR that runs once for each interrupt, and reads one data point each time... which it should append to the buffer. The interrupts keep occuring until the ISR sees that it has collected the all the data, and turns off the interrupts. So what my ISR needs to do is access and write to that output buffer that was given to the driver back in the dispatch routine? But how? I gather that in my dispatch routine I need to do something like.. lock the buffer and/or mark the irp as pending and/or queue the irp, etc. Then in my ISR I might need to do more of the same. Can someone tell me what I need to do to use this METHOD_OUT_DIRECT buffer in my ISR? Buzz Tag: Remove locks IRP-by-IRP Tag: 89775
 - 21
 - BUGCODE_NDIS_DRIVER 7C, Duplicate packet descriptor on Vista RTM 32 bit Hi all Got this bug once on Visdt which cant be repeated, and has never been seen on XP, either 32 or 64 bit. Four packet descriptors are allocated and put in a list at startup. Four 2k buffers are also allocated. The header size is set for the packet too at start up. When there is data to send up the driver calls: NdisInterlockedRemoveHeadList() to get a packet descriptor off the queue. NdisReinitializePacket() NdisAllocateBuffer() using one of the presupplied 2k buffers. NdisAdjustBufferLength() NdisChainBufferAtBack() NDIS_SET_PACKET_STATUS() to SUCCESS or RESOUCES depending if this was the last descriptor on the queue NdisMIndicateReceivePacket() <----- this is where the BSOD was. The descriptor is returned to the miniports return packets handler whicjh calls: NdisUnchainBufferAtFront() NdisFreeBuffer() NdisInterlockedInsertTailList() So, does anyone know of a bug in Vista that might cause it to think it still has the oacket after it has been given back to the miniport driver? Thanks in advance Tag: Remove locks IRP-by-IRP Tag: 89769
 - 22
 - USB<->Serial Virtual Driver Hi All, i have my own firmware( USB hardware) and i developed a driver for that which communicate to it. Now i want to develop USB<->Serial driver for it. So from where can i start? I have DDK2003 and also check serial.sys and serenum.sys. Currently i changed my inf such that which loads serenum.sys and my own driver and displays in device manager as a virtual "com12" or any. But now i want to communicate with it. I know without doing nothing i can not grab the data but here is the my problem.. From where pass the data from driver. Meaand IRP to URB or whatever is.... I did not find any help on the net. SO if possible then provide me a direction or link. I have a hardware with 2 bulk and 1 control end point. If i need then i will use interrupt end point for serial settings communication but for simplicity i am using control pipe with some vendor specific API and pass serial settings to the hardware. Now please help me on that. Thanks in advance. Regards, Tejas -- slsnios ------------------------------------------------------------------------ Posted via http://www.codecomments.com ------------------------------------------------------------------------ Tag: Remove locks IRP-by-IRP Tag: 89767
 - 23
 - Loading data using isochronous USB Driver crashes the computer When I connect my USB device to computers (hp dc71007 and hp d530) with USB controller Int 82801FB or more recent versions, while sending data from the PC the computer will crash frequently. I couldn't see the issue when using old pcs (hp d53c and hp Brio) that have USB controller 82801AA and 82801EB. Are there any known issues with the USB driver for these devices in Windows XP SP2? Does anyone know of a way of avoiding the crash? Tag: Remove locks IRP-by-IRP Tag: 89738
 - 24
 - How to get a drive letter for a virtual device Hello, I'm new in the device driver world. So I appreciate all the help that I get. I have a written a device driver for a virtual device. Is there anyway that I can assign a drive letter ( lets say f: or g: ) to that virtual drive, such that when a user tries to do a file drag and drop to that drive letter, my driver can intercept the data? Thanks pada Tag: Remove locks IRP-by-IRP Tag: 89735
 - 25
 - Dynamically loading export drivers Hello, I want to know how an export driver can be dynamically loaded. Article http://www.microsoft.com/whdc/driver/tips/KmDLL.mspx states at the end "In the export driver, define an I/O control code (IOCTL) that instantiates a class and returns an interface pointer that other drivers can use to call methods on the interface object in the usual way. The class should provide a method for freeing the object, and the header file should contain only pure abstract methods." I don't know how this is done and have not found any reference or code which explains this. Another way of dynamically loaded an export driver is by using a Device interface and I have found some discussions on this but again no reference code to explainn how it is done. Can anyone please help me? Any reference code explaining the process would be a great help. Thanks, ..Ram Tag: Remove locks IRP-by-IRP Tag: 89718
- Next
 - 1
 - How to deal with selective suspend/wait wake/stand by ? Hi, I have an issue on my USB driver that features both selective suspend, wait wake when a request for stand by occurs. This is on XP, but I guess it will be thr same on Vista. Here is the scenario: Note that this device is a smart card reader, capable of waking up the system after a stand by occured 1. After some period of inactivity, the driver issues a selective suspend request to the device 2. The device successfully enters selective suspend 3. The user issues a stand by mode 4. The driver receives an IRP QUERY POWER request 5. Then the Wait Wake IRP is completed, so as the WDK says, the driver issues a PowerD0 request and a new wait wake IRP 6. Then comes the boiler plate of QUERY/SET POWER IRPs, that finally ends with a D2 request to the device 7. Problem is that the device is no longer capable of waking up the system when inserting a smart card. This does NOT occur if the device is not suspended before the system stands by. I guess there is some problem because there are some mixing D0 (because the Wait Wake IRP is completed) and D2 (because of stand by request) messages. But as issuing a D0 request generates QUERY/SET POWER IRPs as the stand by request does, I don't really see how to hold the stand by request while the D0 request is being completed. I had a look at the bulkusb sample from WDK 6000, it seems that there is no particular mecanism to manage that. Any suggestions ? Thanks, Roger Tag: Remove locks IRP-by-IRP Tag: 89710
 - 2
 - How to install VHIDMINI driver program on VISTA 64BIT OS Hi, My device need setup 2 drivers, one driver manage USB communication with device, the other is a virtual HID class driver which enumerate some HID devices such as keyboard, mouse and so on. 2 drivers can be installed and work normanlly on windows MCE OS, WINXP 64 bit OS and Vista RC2 32 bit OS. Now, I need setup this 2 drivers on Vista 64 bit OS. I known that every driver on Vista 64 bit OS require to pass digital signatures. Through pressing F8 key after OS started, I can disable the digital signatures for a while for testing my driver. unfortunately, I only can setup the USB driver, the HID driver always return failure. My HID class driver is based on the VHIDMINI driver, it locate in these directory {system disk}\WINDDK\5744\src\hid\vhidmini\sys. I have tried to setup the VHIDMINI driver by running ''devcon'' tool, It also return failure. Can someone try to install it on VISTA 64 bit OS? I check the setupapi.dev.log, it report "Installing NULL driver" information and other information. I don't understand it. I think that some infomation in inf file bring out. I attached these inf and sys files for reference. Any additional links or information will be greatly appreciated! Lejianz, john at cwtek.com.tw THE inf file's content: [Version] Signature="$CHICAGO$" Class=HIDClass ClassGuid={745a17a0-74d3-11d0-b6fe-00a0c90f57da} Provider=%VENDOR% CatalogFile=vhidmini.cat ;LayoutFile=layout.inf DriverVer=09/21/2006,6.0.5736.1 ; ; Layout.inf (etc.) list all files shipped with the operating system so the ; source description sections are only needed if other files are needed. ; ; In order to use IHV drivers, the SourceDisksNames section must list the ; disk(s) containing the drivers and the SourceDisksFiles section must list ; which disk number each file to be copied is found on. ; Disk number 99 is used to avoid a conflict with the disk numbers defined ; in layout.inf ; ; Files used in a driver installation need to be digitally signed otherwise ; installation may fail. See documentation elsewhere in the DDK regarding ; driver signing. [SourceDisksFiles] vhidmini.sys = 99 [SourceDisksNames.x86] 99 = %DISK_NAME%,,,\i386 ;[SourceDisksNames.ia64] ;99 = %DISK_NAME%,,,\ia64 [SourceDisksNames.amd64] 99 = %DISK_NAME%,,,\amd64 [DestinationDirs] CopyFilesSYS = 12 ; should it be 10 to take care of 98 stuff?????? CopyFilesDLL = 11 ; %SystemRoot%\system or system32 - 98 or Win2000 CopyFilesSYS_98 = 10, system32/drivers [Manufacturer] %VENDOR%=Vendor, NTx86, NTia64, NTamd64 ; For Win2K [Vendor] %VHidMini% = VHidMini.Inst, {D49F883C-6486-400a-8C22-1A9EF48577E4}\HID_DEVICE %VHidDevice% = VHidDevice.Inst, HID\MyVirtualHidDevice ; For XP and later [Vendor.NTx86] %VHidMini% = VHidMini.Inst, {D49F883C-6486-400a-8C22-1A9EF48577E4}\HID_DEVICE %VHidDevice% = VHidDevice.Inst, HID\MyVirtualHidDevice [Vendor.NTia64] %VHidMini% = VHidMini.Inst, {D49F883C-6486-400a-8C22-1A9EF48577E4}\HID_DEVICE %VHidDevice% = VHidDevice.Inst, HID\MyVirtualHidDevice [Vendor.NTamd64] %VHidMini% = VHidMini.Inst, {D49F883C-6486-400a-8C22-1A9EF48577E4}\HID_DEVICE %VHidDevice% = VHidDevice.Inst, HID\MyVirtualHidDevice [ControlFlags] ExcludeFromSelect = HID\MyVirtualHidDevice ;=============================================================== ; VHidMini (Root Enumerated Hid Device) install section ;=============================================================== [VHidMini.Inst] CopyFiles = CopyFilesSYS_98, CopyFilesDLL AddReg = VHidMini_98me.AddReg [VHidMini.Inst.NT] CopyFiles = CopyFilesSYS, CopyFilesDLL [VHidMini.Inst.NT.HW] AddReg = VHidMini_Parameters.AddReg [VHidMini.Inst.NT.Services] AddService = vhidmini,0x00000002,vhidmini_Service_Inst, [CopyFilesSYS] vhidmini.sys ;HidClass.sys ;HidParse.sys [CopyFilesSYS_98] vhidmini.sys HidClass.sys HidParse.sys [CopyFilesDLL] Hid.dll [VHidMini_Parameters.AddReg] HKR,,"ReadFromRegistry",%REG_DWORD%, 0x00000001 HKR,,"MyReportDescriptor",%REG_BINARY%, 06, 00, FF, 09, 01, A1, 01, 85, 01, 09, 01, 15, 00, 26, ff, 00, 75, 08, 95, 01, B1, 00, 09, 01, 15, 00, 26, FF, 00, 75, 08, 95, 01, B1, 00, 09, 01, 75, 08, 95, 01, 81,00, C0 [VHidMini_98me.AddReg] HKR,,DevLoader,,*ntkern HKR,,NTMPDriver,,"vhidmini.sys" [vhidmini_Service_Inst] DisplayName = %VHidMini% ServiceType = %SERVICE_KERNEL_DRIVER% StartType = %SERVICE_DEMAND_START% ErrorControl = %SERVICE_ERROR_IGNORE% ServiceBinary = %12%\vhidmini.sys ;=============================================================== ; VHidDevice (Virtual Hid Device) install section ; - Only a Null service is installed. ;=============================================================== [VHidDevice.Inst] [VHidDevice.Inst.NT] [VHidDevice.Inst.NT.Services] AddService = ,0x00000002, ; NULL Service ;================================================================ [Strings] ; *******Localizable Strings******* VENDOR = "Microsoft Corp" VHidMini = "Root Enumerated Hid Device (sample)" VHidDevice = "Virtual Hid Device (sample)" DISK_NAME = "Virtual Hid Device Sample Install Disk" ; *******Non Localizable Strings******* SERVICE_BOOT_START = 0x0 SERVICE_SYSTEM_START = 0x1 SERVICE_AUTO_START = 0x2 SERVICE_DEMAND_START = 0x3 SERVICE_DISABLED = 0x4 SERVICE_KERNEL_DRIVER = 0x1 SERVICE_ERROR_IGNORE = 0x0 SERVICE_ERROR_NORMAL = 0x1 SERVICE_ERROR_SEVERE = 0x2 SERVICE_ERROR_CRITICAL = 0x3 REG_EXPAND_SZ = 0x00020000 REG_DWORD = 0x00010001 REG_MULTI_SZ = 0x00010000 REG_BINARY = 0x00000001 REG_SZ = 0x00000000 Tag: Remove locks IRP-by-IRP Tag: 89708
 - 3
 - dynamically allocated memory I am trying to find out how the memory is distributed among kernel components and the drivers. I could find this for the static memory. Is there any way to find this information for the dynamic memory, i.e. memory allocated using malloc etc. Basically, I have a driver for which I do not have the source code. I want to figure out if it does mallocs etc and what are the memory ranges it uses. Thanks Tag: Remove locks IRP-by-IRP Tag: 89706
 - 4
 - DDK with Cygwin Hi Group, I want to know whether DDK can work with Cygwin and gcc compiler. Or do I compulsorily need MS Visual Studio. Also can someone give me an idea about what is the role of driver studio in developing windows device drivers. thanks Sai K Tag: Remove locks IRP-by-IRP Tag: 89694
 - 5
 - HidP_TranslateUsagesToI8042ScanCodes API Crash? Hi, I'm new to driver programming and I'm writing a keyboard driver that translates Usage to Scan Code, then to a Unicode string. I modified a similar sample code, but I keep getting a crash from the API (according to the call stack, it's from hid.dll), and the debugger doesn't show any bad pointer. It says "Unhandled exception at <address>. Access violation writing location <address>". I've been searching for more sample code, but so far I can't find any, not in Google search results... Below are the two functions I have. I call UnpackReport. The InsertScanCode callback function is called and able to convert the scan code to unicode string successfully, but it will crash after that at the line of HidP_TranslateUsagesToI8042ScanCodes. Am I missing something? Thanks in advanced. BOOLEAN InsertScanCode(PVOID context, PCHAR newScanCodes, ULONG length) { HKL activeKeyboardLayout = GetKeyboardLayout(0); UINT vkCode = MapVirtualKeyEx(*newScanCodes, 1, activeKeyboardLayout); BYTE keyStateArray[256]; BOOL bKeyState = GetKeyboardState(keyStateArray); if (vkCode != -1 && bKeyState) { TCHAR buffer[16] = {0}; int nConvert = ToUnicode(vkCode, 0, keyStateArray, buffer, 16, 0); if (nConvert > 0) { int haha = 1; } } return true; } BOOLEAN UnpackReport ( IN PCHAR ReportBuffer, IN USHORT ReportBufferLength, IN HIDP_REPORT_TYPE ReportType, IN OUT PHID_DATA Data, IN ULONG DataLength, IN PHIDP_PREPARSED_DATA Ppd ) { HIDP_KEYBOARD_MODIFIER_STATE modState; memset(&modState, 0, sizeof(HIDP_KEYBOARD_MODIFIER_STATE)); ULONG maxUsageLen = 14; maxUsageLen = HidP_MaxUsageListLength(HidP_Input, HID_USAGE_PAGE_KEYBOARD, Ppd); if (maxUsageLen == 0) { Dbg(_T("Invalid HidP_MaxUsageListLength returned.\n")); return FALSE; } ULONG numUsages; // Number of usages returned from GetUsages. ULONG i; UCHAR reportID; ULONG Index; ULONG nextUsage; reportID = ReportBuffer[0]; for (i = 0; i < DataLength; i++, Data++) { if (reportID == Data->ReportID) { if (Data->IsButtonData) { numUsages = Data->ButtonData.MaxUsageLength; Data->Status = HidP_GetUsages (ReportType, Data->UsagePage, 0, // All collections Data->ButtonData.Usages, &numUsages, Ppd, ReportBuffer, ReportBufferLength); for (Index = 0, nextUsage = 0; Index < numUsages; Index++) { if (Data -> ButtonData.UsageMin <= Data -> ButtonData.Usages[Index] && Data -> ButtonData.Usages[Index] <= Data -> ButtonData.UsageMax) { Data -> ButtonData.Usages[nextUsage++] = Data -> ButtonData.Usages[Index]; } } if (nextUsage < Data -> ButtonData.MaxUsageLength) { Data->ButtonData.Usages[nextUsage] = 0; } } Data -> IsDataSet = TRUE; } } Data--; USAGE curUsageList[maxPreviousUsages] = {0}; USAGE breakUsageList[maxPreviousUsages] = {0}; USAGE makeUsageList[maxPreviousUsages] = {0}; memset(curUsageList, 0, sizeof(curUsageList)); for (i = 0; i < numUsages; i++) { curUsageList[i] = Data->ButtonData.Usages[i]; } HidP_UsageListDifference(prevUsageList, curUsageList, breakUsageList, makeUsageList, maxUsageLen); if (makeUsageList[0] != 0) { int context; ULONG tstatus = HidP_TranslateUsagesToI8042ScanCodes(makeUsageList, maxUsageLen, HidP_Keyboard_Make, &modState, InsertScanCode, &context); if (tstatus = HIDP_STATUS_I8042_TRANS_UNKNOWN) { int haha = 1; } } memcpy(prevUsageList, curUsageList, maxUsageLen * sizeof(USAGE)); Data++; return (TRUE); } Tag: Remove locks IRP-by-IRP Tag: 89687
 - 6
 - Accessing scsi read/write buffer in storport I am writing a storport miniport and my device supports scatter gather DMA. In my PORT_CONFIGURATION_INFORMATION structure I will be setting the MapBuffers variable to STOR_MAP_NO_BUFFERS. There are certain error conditions for which my device requires the driver to "help out" and this requires mapping the Srb->Buffer into kernel address space so the driver can access it. In iSCSI there was a function called iSCSIPortGetOriginalMdl(Srb) that could get me to the MDL. Once I have the MDL I can map the buffer to system address space. For storport, there is no equivalent function so how can I get to the MDL from the Srb? thanks David Tag: Remove locks IRP-by-IRP Tag: 89685
 - 7
 - AddDevice after IRP_MN_START_DEVICE? Is it possible for a filter DO to attach to the top of my FDO after my FDO has completed IRP_MN_START_DEVICE? What if I get stopped? Could it then attach before the next start? (I'm wondering if it's safe to store away the result of IoGetAttachedDeviceReference once as a part of handling IRP_MN_START_DEVICE without worrying about other drivers attaching to the top of the stack later) Tag: Remove locks IRP-by-IRP Tag: 89684
 - 8
 - Extra reference/remove-lock and completion routines I'm not sure how many people believe in the extra reference or extra remove lock needed for IRPs sent asynchronously with a completion routine (or how rare this situation is), but I'm still waiting for the documentation on IoReleaseRemoveLock to be updated, and in the meantime I thought of a simple way to summarize it (maybe it's been said before): If a driver releases the last known reference to itself then unpredictable behavior will result. Right? There would be no reference to maintain the return instructions at the end of the routine that does the dereference. ----- In more detail, if a lower driver calls IoCompleteRequest from its own dispatch routine then it will call the completion routine of the higher driver, and if the completion routine releases the higher driver's remove lock then the higher driver's IRP_MN_REMOVE_DEVICE handler would be allowed to proceed and dereference the lower driver. The driver has just indirectly released the last known reference to itself. If the lower driver wanted to rely on the assumption that a remove lock and IRP_MN_REMOVE_DEVICE was involved (as opposed to a direct dereference), then the PnP manager still has a reference from before the IRP_MN_REMOVE_DEVICE was sent, but when IoCompleteRequest returns, the lower driver will indirectly release that. It will release its own remove lock, which would allow its own IRP_MN_REMOVE_DEVICE handler to proceed, delete its own device object, and when the IRP_MN_REMOVE_DEVICE is completed the PnP manager will release its reference to all the device objects in the stack and may unload their drivers. Unfortunately, the lower driver is one of them, and it may not have yet returned from the routine that called IoCompleteRequest, and when it tries to return it will touch memory that may have already been deallocated. (If the completion routine release a reference to the lower driver instead of indirectly allowing the IRP_MN_REMOVE_DEVICE handler to release the reference, then it's an even more direct problem.) It's a narrow window, and I've tried to think about other possibilities over the last few months, but I still see it. I see a need for two remove lock acquires in this situation, but I only read one remove lock acquire in the description of IoReleaseRemoveLock: "For I/O operations (including power and PnP IRPs) that set an IoCompletion routine, a driver should call IoReleaseRemoveLock in the IoCompletion routine, after calling IoCompleteRequest. For I/O operations that do not set an IoCompletion routine, a driver should call IoReleaseRemoveLock after passing the current IRP to the next-lower driver, but before exiting the dispatch routine." I think that second sentence should read "... do _or_ do not set an IoCompletion routine ...", with the implication that setting an IoCompletion routine implies calling IoAcquireRemoveLock twice and calling IoReleaseRemoveLock in two separate locations. Tag: Remove locks IRP-by-IRP Tag: 89680
 - 9
 - Spin lock vs. fast mutex This is an old issue I've seen mentioned before, but I still can't decide whether I like spin locks or fast mutexes better. Here's a situation where my driver is non-pageable and has these two pieces of code that are entered at passive level and must be synchronized: thread_a pseudo-code: <acquire some sort of lock> if(State != UNFREEZABLE_STATE) { StateFrozen = TRUE; } <release some sort of lock> thread_b pseudo-code: <acquire some sort of lock> if(StateFrozen) { Failure = TRUE; } else { OldState = State; State = UNFREEZABLE_STATE; } <release some sort of lock> What's the preference for the sort of lock, fast mutex or spin lock? (I think a spin lock because each piece of code is so short) Tag: Remove locks IRP-by-IRP Tag: 89675
 - 10
 - ... while holding a ... I know not to sleep ("wait") while holding a spin lock on both Linux and Windows for the same reason: at best another CPU may spin for a long time and hurt performance and at worst the current CPU may spin forever because scheduling is disabled. What about calling routines outside a driver (instead of sleeping) or with mutexes (instead of spin locks)? Isn't it safe to call outside a driver (E.g. to queue-manipulation routines) while holding a spinlock as long as the called routine expects to run at dispatch level? Is it safe to call outside a driver (E.g. to memory allocation routines) while holding a fast mutex? Is it simply undesirable to limit the use of such techniques because the time taken by the called routine to run will impact performance by increasing the amount of time other threads trying to acquiring the lock or mutex will be blocked? Tag: Remove locks IRP-by-IRP Tag: 89663
 - 11
 - NDIS IM & NdisMDeregisterDevice As I know, i should call NdisMDeregisterDevice in MiniportHalt routine when last miniport is halted ( http://groups.google.ru/group/comp.os.ms-windows.programmer.nt.kernel-mode/browse_thread/thread/1d8ee27725dd24c2/b8e565e360364c61?lnk=st&q=NdisMDeregisterDevice+Unload&rnum=1&hl=ru#b8e565e360364c61) Then my driver will be unloaded. But if client app has opened handle on my device, driver is still in memory. I can understand why. When app close the last handle i suggest calling driver Unload routine, but driver is still referenced. I can notify app about stop, but I think when app handle such notify it will be late. What sould I do? Tag: Remove locks IRP-by-IRP Tag: 89662
 - 12
 - Sharing memory between user and kernel applications Hello all, Could any one please tell me what are different possible ways to share memory between a user level application and a kernel level driver? I am interested in transferring data from kernel driver to user level application and vice versa. In my current implementation, the user-level application uses ReadFile(), WriteFile() to read and write data from/to kernel driver. However I am looking on different ways to improve this transfer of data. So is it possible to share a memory block between user application and kernel driver to transfer data between them in fastest possible way? Any hint or idea will be appreciated. Thanks, Arsalan Tag: Remove locks IRP-by-IRP Tag: 89644
 - 13
 - Driver signing and WHQL There are four kind of Plug and Play Kernel-Mode Driver Signings: Unsigned, Self-signed, Authenticode-signed and WHQL-signed. Do I need Self-signed or Authenticode-signed drivers when go to WHQL testing? Sape Tag: Remove locks IRP-by-IRP Tag: 89643
 - 14
 - driver signing hi all, i have a drm virtual audio driver, and I want to test it with drm content. How do i get a test signature or something like that?(since drmk.sys forwards audio only to digitally signed drivers) Also what is the process to get the actual signature from ms? thanks! Tag: Remove locks IRP-by-IRP Tag: 89639
 - 15
 - error C2106 and error LNK2019: i build a ddk program using vc6.0,some code as follow: __declspec(dllimport) ServiceDescriptorTableEntry_t KeServiceDescriptorTable; #define SYSTEMSERVICE(_function) KeServiceDescriptorTable.ServiceTableBase[ *(PULONG)((PUCHAR)_function+1)] NTSYSAPI NTSTATUS NTAPI ZwQueryDirectoryFile( IN HANDLE hFile, IN HANDLE hEvent OPTIONAL, IN PIO_APC_ROUTINE IoApcRoutine OPTIONAL, IN PVOID IoApcContext OPTIONAL, OUT PIO_STATUS_BLOCK pIoStatusBlock, OUT PVOID FileInformationBuffer, IN ULONG FileInformationBufferLength, IN FILE_INFORMATION_CLASS FileInfoClass, IN BOOLEAN bReturnOnlyOneEntry, IN PUNICODE_STRING PathMask OPTIONAL, IN BOOLEAN bRestartQuery ); ........... NTSTATUS DriverEntry( IN PDRIVER_OBJECT theDriverObject, IN PUNICODE_STRING theRegistryPath ) { ..................... (ZWQUERYDIRECTORYFILE) (SYSTEMSERVICE(ZwQueryDirectoryFile))= NewZwQueryDirectoryFile; ...................... } some wrong rise like: error C2106: '=' : left operand must be l-value\ it means the follow sentense is wrong: (ZWQUERYDIRECTORYFILE) (SYSTEMSERVICE(ZwQueryDirectoryFile))= NewZwQueryDirectoryFile; when i modify this sentence as follow: (KeServiceDescriptorTable.ServiceTableBase[ *(PULONG)((PUCHAR)ZwQueryDirectoryFile+1)]) = (ULONG)OldZwQueryDirectoryFile; another wrong rise: error LNK2019: unresolved external symbol "__declspec(dllimport) long __stdcall ZwQueryDirectoryFile(void *,void *,void (__stdcall*)(void *,struct _IO_STATUS_BLOCK *,unsigned long),void *,struct _IO_STATUS_BLOCK *,void *,unsig ned long,enum _FILE_INFORMATION_CLASS,unsigned char,struct _UNICODE_STRING *,unsigned char)" (__imp_?ZwQueryDirectoryFile@@YGJPAX0P6GX0PAU_IO_STATUS_BLOCK@@K@Z010KW4_FILE_INFORMATION_CLASS@@EPAU_UNICODE_STRING@@E@Z) referenced in function "void __st dcall OnDriUnload(struct _DRIVER_OBJECT *)" (?OnDriUnload@@YGXPAU_DRIVER_OBJECT@@@Z) what can i to do? thank you! Tag: Remove locks IRP-by-IRP Tag: 89632
 - 16
 - ExAllocatePoolWithTag cause BSOD Dear All, I am developing a USB device driver in Vista, Sometimes I call ExAllocatePoolWithTag may cause BSOD. STOP code: 0x000000C5 (4, 2, 1, 0x8046A992) But it's ok in 2k/XP. Are there some tricks to call ExAllocatePoolWithTag in Vista ? Tag: Remove locks IRP-by-IRP Tag: 89629
 - 17
 - iSCSI Target FAILBACK I have a dual port EVA iSCSI array controller connected through MS iSCSI Initiator 2.02. When one of the controller restarts, the MS iSCSI Initiator marks the connection as "INACTIVE" and doesnt seem to FAILBACK the connection when the PATH comes back online ( means when the controller is online). I am guessing this is not the behaviour it is suppose to be. Most of the times when the controller comes back online, MS iSCSI initator should bring the PATH back online right? Can anyone please throw some light on this issue. PS: I am talking about already logged in path which goes INACTIVE when the controller goes down and needs "login" when the PATH comes online. Thanks KRA Tag: Remove locks IRP-by-IRP Tag: 89625
 - 18
 - Mutex at IRQL = DISPATCH_LEVEL? What?! I found this old post from comp.os.ms-windows.programmer.nt.kernel-mode back in 1997: "You are allowed to wait for a mutex at IRQL = DISPATCH_LEVEL, but unless you know what your thread context is, and you know you can make it block (i.e. it's either the caller of your dispatch routine or your own worker thread), don't do it. You'll get a blue screen I should think." I'm thrown back by the first phrase. I know you can check whether a mutex is free with a zero-length timeout, but that's not "waiting". I would thinking waiting on a mutex at DISPATCH_LEVEL would definitely deadlock a uniprocessor system (because the scheduler wouldn't be able to run and switch to the thread that's currently holding the mutex so the mutex would never be released) and likely deadlock a multiprocessor system (because the thread that would release the mutex might be in the scheduling queue of the same CPU that's doing the wait). Was NT designed this much differently back in 1997? Tag: Remove locks IRP-by-IRP Tag: 89622
 - 19
 - USB Newbie Question I'll show some (well, a lot...) of ignorance here. But I have been asked a question in an area I'm not familiar with. So, here goes. The question concerns testing USB-Ethernet devices by a house that simply tests these devices as a service. Basically, a hardware QA house. The company does not actually manufacture the devices - justs tests them. The USB-Ethernet devices may come from a variety of hardware vendors and use different driver designs. Some are RNDIS and some are NDIS-WDM widgets. Test house wants to test multiple USB-Ethernet devices by plugging in 8 at a time into a USB hub. Their test software generally works, with one exception... Problem is that in some cases Windows XP recognizes and installs NDIS miniport instances for only _some_ of the 8 USB devices that are plugged into the hub. USBView can see all of the davices, but the NDIS wrapper does _not_ load all of them as a miniport that can be seen in SetupDi or network control panel. I think that this is related to whether or not the USB device has a serial number. Some do, some don't. This can't be controlled by the company asking the question... Can the lack of a serial number (or having a duplicate serial number for all devices) prevent multiple instances of otherwise identical USB devices from being recognized and installed on Windows XP? (My research so far seeme to suggest this can cause this type of problem). If so, could a WDF USB filter driver be cooked up to "fake" a serial number for each USB port, possibily eliminating the non-recognition problem for this test station? If so, what would be some pointers? TIA, Thomas F. Divine Tag: Remove locks IRP-by-IRP Tag: 89619
 - 20
 - Where to find Setupapi.log error codes I am looking to find out where the error codes in the setupapi.log are defined. Are they on the MSDN website? In a header? I am specifically interested in error code 0xe0000228. Dan Tag: Remove locks IRP-by-IRP Tag: 89602
 - 21
 - DriverPackagePreinstall returns ERROR_ACCESS_DENIED error on Vista Hello, I have developed MSI installer which uses DLL for driver installation/preinstallation/uninstallation. Drivers preinstallation is implemented using DIFx functionality (DriverPackagePreinstall function), but DriverPackagePreinstall always returns ERROR_ACCESS_DENIED error, when I run installation on Vista RC2. WDK tells that user has to have Administrator rights in order to call DriverPackagePreinstall function, user has such rights. Have somebody ever seen such problem? Thanks. Tag: Remove locks IRP-by-IRP Tag: 89598
 - 22
 - Found new hardware wizard reappears for mouse driver in win 2k Hi all. I've written two HID drivers for a custom joystick that is not connected through a USB port. The first driver registers the device as a normal joystick that can be used with DirectX. It works great in both XP and 2k. The second driver registers the device as a mouse. The mouse driver works great in XP. In 2k it works fine but every 5 minutes or so Windows pops up the "Found new hardware" wizard. If you click Next and then Finish it works fine again for 5 minutes before Windows displays the wizard again and so forth. I use the 2003 DDK to compile the drivers. The drivers are based on the "vhidmini" sample from the DDK and the "HidFake" sample from Walter Oneys "Programming the Windows Driver Model". DevicesArePolled is set to TRUE in both drivers. The drivers are not signed. Does anybody know why Windows 2000 keeps displaying the "Found new hardware" wizard, and what I can do to fix it? /Frasse Tag: Remove locks IRP-by-IRP Tag: 89595
 - 23
 - question regarding load-balancing intermediate driver Hello, I am working on an load-balancing/fail-over intermediate driver on win2k3, NDIS 5.1. The driver can load balance the traffic across many ports. I am facing one critical issue. Suppose, I have formed a team of two drivers. Whenever I disable one of the physical adapter which is part of team, the IM driver stops receiving packets. The physical device is receiving the packets (this I verified using ethereal capture) but none of the receive function(s) [ProtocolReceive()/ProtocolReceivePacket()] of the IM driver are not getting called. Thanks. Tag: Remove locks IRP-by-IRP Tag: 89594
 - 24
 - Sample WDM driver for SeaLevel DIO-24 card Hi All, I have seen some WDF sample for this SeaLevel PIO-24 card in OSROnline website. Is there any WDM based sample for this card?. Thanks in advance, Subbu Tag: Remove locks IRP-by-IRP Tag: 89593
 - 25
 - ICMP: not resolved. hello All, i am writing a network driver.. i am able to configure it properly and also resolving the arp entry. i had tried much to ping using the device, but could not finished. i am tracing the problem using ethreal. i found that i am able to receive the ICMP request but my driver is not sending its reply,,, so it is like this,, My driver other device <------ARP request ARP reply ---->> <------ICMP request there is no ICMP reply,, can anybody please tell me what could be the problem?? is it a driver problem or a configuration problem,,, thanks from hitesh ughreja. Tag: Remove locks IRP-by-IRP Tag: 89591

Okay, despite earlier doubts I suppose I can believe the file-based
IRPs cannot be concurrent with IRP_MN_REMOVE_DEVICE, but I still have
questions about the earliest OS version where this applies and about
exactly which IRPs are file-based.

Assuming Windows 2000 or later, here's what I interpret for every IRP
major code, but I have questions about some cases, and some are not
even documented (just found in the .h file so maybe don't matter?)

IRP_MJ_CREATE
- cannot be concurrent with IRP_MN_REMOVE_DEVICE
IRP_MJ_READ
- cannot be concurrent with IRP_MN_REMOVE_DEVICE
IRP_MJ_WRITE
- cannot be concurrent with IRP_MN_REMOVE_DEVICE
IRP_MJ_DEVICE_CONTROL
- cannot be concurrent with IRP_MN_REMOVE_DEVICE
IRP_MJ_INTERNAL_DEVICE_CONTROL (IRP_MJ_SCSI)
- cannot be concurrent with IRP_MN_REMOVE_DEVICE
IRP_MJ_FLUSH_BUFFERS
- cannot be concurrent with IRP_MN_REMOVE_DEVICE
IRP_MJ_CLEANUP
- cannot be concurrent with IRP_MN_REMOVE_DEVICE
IRP_MJ_CLOSE
- cannot be concurrent with IRP_MN_REMOVE_DEVICE
IRP_MJ_QUERY_INFORMATION
- cannot be concurrent with IRP_MN_REMOVE_DEVICE
IRP_MJ_SET_INFORMATION
- cannot be concurrent with IRP_MN_REMOVE_DEVICE
IRP_MJ_PNP
IRP_MN_CANCEL_REMOVE_DEVICE
- cannot be concurrent with IRP_MN_REMOVE_DEVICE
IRP_MN_CANCEL_STOP_DEVICE
- cannot be concurrent with IRP_MN_REMOVE_DEVICE
IRP_MN_QUERY_REMOVE_DEVICE
- cannot be concurrent with IRP_MN_REMOVE_DEVICE
IRP_MN_QUERY_STOP_DEVICE
- cannot be concurrent with IRP_MN_REMOVE_DEVICE
IRP_MN_START_DEVICE
- cannot be concurrent with IRP_MN_REMOVE_DEVICE
IRP_MN_STOP_DEVICE
- cannot be concurrent with IRP_MN_REMOVE_DEVICE
IRP_MN_SURPRISE_REMOVAL
- cannot be concurrent with IRP_MN_REMOVE_DEVICE
<other IRP_MJ_PNP minor IRPs>
- can be concurrent with IRP_MN_REMOVE_DEVICE
IRP_MJ_SYSTEM_CONTROL
- can be concurrent with IRP_MN_REMOVE_DEVICE, but
IoWMIRegistrationControl(..., WMIREG_ACTION_DEREGISTER) can
be used in place of a remove lock
IRP_MJ_POWER
- can be concurrent with IRP_MN_REMOVE_DEVICE
IRP_MJ_SHUTDOWN
- ?
IRP_MJ_CREATE_NAMED_PIPE
- ?
IRP_MJ_QUERY_EA
- ?
IRP_MJ_SET_EA
- ?
IRP_MJ_QUERY_VOLUME_INFORMATION
- ?
IRP_MJ_SET_VOLUME_INFORMATION
- ?
IRP_MJ_DIRECTORY_CONTROL
- ?
IRP_MJ_FILE_SYSTEM_CONTROL
- ?
IRP_MJ_LOCK_CONTROL
- ?
IRP_MJ_CREATE_MAILSLOT
- ?
IRP_MJ_QUERY_SECURITY
- ?
IRP_MJ_SET_SECURITY
- ?
IRP_MJ_DEVICE_CHANGE
- ?
IRP_MJ_QUERY_QUOTA
- ?
IRP_MJ_SET_QUOTA
- ?

Are any cases different in Windows 98 or Me?

Top

Re: Remove locks IRP-by-IRP by BubbaGump

BubbaGump
Tue Dec 19 12:54:49 CST 2006

Instead of "be concurrent", I should have phrased the question "happen
during or after".

On Tue, 19 Dec 2006 13:16:49 -0500, BubbaGump <> wrote:

>Okay, despite earlier doubts I suppose I can believe the file-based
>IRPs cannot be concurrent with IRP_MN_REMOVE_DEVICE, but I still have
>questions about the earliest OS version where this applies and about
>exactly which IRPs are file-based.
>
>...

Top