I'm developing a miniFilter driver and took the Microsoft's SwapBuffers
miniFilter as example.
An InstaceSetup routin by default is attaching to all volumes. But I don't want
to attach to all of them, only to some choosen...
I tried to set "NULL" instead of "InstanceSetup" in "FLT_REGISTRATION
FilterRegistration" and then to call "FltAttachVolume" in the "DriverEntry"
routin. I've done the following:
PFLT_VOLUME vol;
UNICODE_STRING vname;
....
RtlInitUnicodeString(&vname, L"E:\");
FltGetVolumeFromName(gFilterHandle, &vname, &vol);
...
FltAttachVolume(gFilterHandle, vol, NULL, NULL);
...
When i tried to call FltAttachVolume with the "NULL" 3-d parameter
(PCUNICODE_STRING InstanceName) i received a
"STATUS_FLT_INSTANCE_NAME_COLLISION" error.
If i call FltAttachVolume with a "NOT NULL" 3-d parameter, such as a
"UniqueInstaceName" it returns me "-2145452013".
I'm receiving the same errors, when i,m trying to attach a volume, using a
FilterAttach routine from my User application, like this:
...
driver.driverName = L"swapBuffers";
...
LPCWSTR vname = L"F:\";
...
FilterAttach(driver.driverName, vname, NULL, NULL, NULL);
With "NULL" 3-d parameter (LPCWSTR lpInstanceName):
"ERROR_FLT_INSTANCE_NAME_COLLISION"
With "NOT-NULL": "-2145452013".
In MiniSpy miniFilter there is a User application, and the routine FilterAttach
is used. I tried to call this routine in my application the same way - no
results.
Finally, i changed the swapBuffers inf-file:
there was no DefaultInstance parameter, i set it: "SwapBuffers - Top
Instance".
also i copied this from the MiniSpy inf-file:
[MiniFilter.AddRegistry]
HKR,"Instances","DefaultInstance",0x00000000,%DefaultInstance%
HKR,"Instances\"%Instance1.Name%,"Altitude",0x00000000,%Instance1.Altitude%
HKR,"Instances\"%Instance1.Name%,"Flags",0x00010001,%Instance1.Flags%
HKR,"Instances\"%Instance2.Name%,"Altitude",0x00000000,%Instance2.Altitude%
HKR,"Instances\"%Instance2.Name%,"Flags",0x00010001,%Instance2.Flags%
HKR,"Instances\"%Instance3.Name%,"Altitude",0x00000000,%Instance3.Altitude%
HKR,"Instances\"%Instance3.Name%,"Flags",0x00010001,%Instance3.Flags%
.............
Instance1.Name = "SwapBuffers - Middle Instance"
Instance1.Altitude = "370000"
Instance1.Flags = 0x1 ; Suppress automatic attachments
Instance2.Name = "SwapBuffers - Bottom Instance"
Instance2.Altitude = "361000"
Instance2.Flags = 0x1 ; Suppress automatic attachments
Instance3.Name = "SwapBuffers - Top Instance"
Instance3.Altitude = "385100"
Instance3.Flags = 0x1 ; Suppress automatic attachments
changing the flags to 0x1 to suppress automatic attachments.
And only installing my SwapBuffers miniFilter through this Inf file, i received
"STATUS_SUCCESS" from FltAttachVolume routine in my driver. But it isn't really
attaching to the disk...
What am i doing wrong?
Thanks.
Instance1.Flags = 0x1
That is fine.
I have somewhat similar code and that works fine.
status = FltRegisterFilter( DriverObject,
&FilterRegistration,
&gFilterHandle );
FLT_ASSERT( NT_SUCCESS( status ) );
if (NT_SUCCESS( status )) {
PSECURITY_DESCRIPTOR sd;
OBJECT_ATTRIBUTES oa;
UNICODE_STRING uniString;
status = FltBuildDefaultSecurityDescriptor(&sd,
FLT_PORT_ALL_ACCESS);
if (!NT_SUCCESS(status)) {
return status;
}
RtlInitUnicodeString(&uniString, PORT_NAME);
InitializeObjectAttributes(&oa,
&uniString,
OBJ_KERNEL_HANDLE | OBJ_CASE_INSENSITIVE,
NULL,
sd);
status = FltCreateCommunicationPort(gFilterHandle,
&gServerPort,
&oa,
NULL,
Connect,
Disconnect,
Message,
1);
FltFreeSecurityDescriptor(sd);
BREAK_HERE(); // DbgBreak() macro
//
// Start filtering i/o
//
status = FltStartFiltering(gFilterHandle);
if (!NT_SUCCESS(status)) {
FltUnregisterFilter(gFilterHandle);
}
else {
RtlInitUnicodeString(&uniString, L"\\Device\\HarddiskVolume1");
PFLT_VOLUME vol;
FltGetVolumeFromName(gFilterHandle, &uniString, &vol);
status = FltAttachVolume(gFilterHandle, vol, NULL, NULL);
// status == 0x0 at that point and the mini filter is attached to the Volume
}
}
I normally attach to all volumes but I wanted to try attaching to a single volume and it works fine.
Related
I have a windows service that creates a JobObject that i need to keep alive as long as the machine is turned on - the goal is to manage a few user session processes that can terminate/start at any time with this JobObject. I am creating it in a service to make sure the process is running at startup, and that it can't be killed by regular users.
However, i don't seem to be able to open a handle to this JobObject from the user session, I always get an access denied (5) error, despite going as far as creating it with a NULL DACL.
I have found a somewhat related question here: Open an Event object created by my service from my application, but for me, even with the NULL DACL, when asking for a JOB_OBJECT_ASSIGN_PROCESS right, i get access denied (asking for SYNCHRONIZE works for example).
The service code:
PSECURITY_DESCRIPTOR psd = (PSECURITY_DESCRIPTOR) LocalAlloc(LPTR, SECURITY_DESCRIPTOR_MIN_LENGTH);
InitializeSecurityDescriptor(psd, SECURITY_DESCRIPTOR_REVISION);
SetSecurityDescriptorDacl(psd, TRUE, NULL, FALSE);
SECURITY_ATTRIBUTES secAttr= {0};
secAttr.nLength = sizeof(secAttr);
secAttr.bInheritHandle = false;
secAttr.lpSecurityDescriptor = psd;
hJobObject = CreateJobObject(&secAttr, SCL_JOBOBJECTNAME);
LocalFree(psd);
The user session code:
hJobObject = OpenJobObject(JOB_OBJECT_ASSIGN_PROCESS, FALSE, SCL_JOBOBJECTNAME);
if (hJobObject == NULL)
{
DWORD wError = GetLastError();
printf("Error: %d\n", wError); // this always pops 5
return 1;
}
Any ideas? As a test, i tried spawning a user session process from within the service, and assign the JobObject via the service code, and that worked,.. so i'm fairly certain its related to security settings i am missing, despite the NULL DACL.
if you create Job in service - this object by default will be have WinSystemLabelSid label SID: S-1-16-16384 - System Mandatory Level. (i just check this) so you need not only set Dacl but Sacl too. for example:
ULONG cb = MAX_SID_SIZE;
PSID LowLabelSid = (PSID)alloca(MAX_SID_SIZE);
if (CreateWellKnownSid(WinLowLabelSid, 0, LowLabelSid, &cb))
{
PACL Sacl = (PACL)alloca(cb += sizeof(ACL) + sizeof(ACE_HEADER) + sizeof(ACCESS_MASK));
InitializeAcl(Sacl, cb, ACL_REVISION);
if (AddMandatoryAce(Sacl, ACL_REVISION, 0, 0, LowLabelSid))
{
SECURITY_DESCRIPTOR sd;
InitializeSecurityDescriptor(&sd, SECURITY_DESCRIPTOR_REVISION);
SetSecurityDescriptorDacl(&sd, TRUE, NULL, FALSE);
SetSecurityDescriptorSacl(&sd, TRUE, Sacl, FALSE);
SECURITY_ATTRIBUTES sa= { sizeof(sa), &sd, FALSE };
if (HANDLE hJob = CreateJobObject(&sa, L"Global\\{58BFC6DB-BE93-4cdb-919C-4C713ACB5A32}"))
{
CloseHandle(hJob);
}
}
}
I’m currently writing a driver that exposes virtual COM ports. In the driver, I send an internal IOCTL from the port FDO down the stack, which is handled from the PDO IO queue. For some reason, the output data is not written to the provided output memory.
I’ve confirmed through windbg that IoCtl_Vcp_GetPortInfo (see below) is being called, and works as intended. The request is completed with STATUS_SUCCESS. At the time that I call WdfRequestComplete, the output buffer has valid data. However, when control returns to GetPortInfo (see below), the provided buffer has not been overwritten. I confirmed this with a hardware breakpoint on access for the receiving buffer. It is not read or written to during the WdfIoTargetSendInteralIoctlSynchronously call.
The code responsible for sending the IOCTL is below:
NTSTATUS GetPortInfo(WDFDEVICE device, _Out_ PVCH_PORT_INFO port_info)
{
NTSTATUS status;
WDFIOTARGET io_target;
WDF_MEMORY_DESCRIPTOR output_descriptor;
PVOID buffer = ExAllocatePoolWithTag(NonPagedPool, sizeof(VCH_PORT_INFO), VCH_POOL_TAG);
//WDF_MEMORY_DESCRIPTOR_INIT_BUFFER(&output_descriptor, port_info, sizeof(VCH_PORT_INFO));
WDF_MEMORY_DESCRIPTOR_INIT_BUFFER(&output_descriptor, buffer, sizeof(VCH_PORT_INFO));
io_target = WdfDeviceGetIoTarget(device);
status = WdfIoTargetSendInternalIoctlSynchronously(io_target, NULL, IOCTL_VCP_INTERNAL_GET_PORT_INFO, NULL, &output_descriptor, NULL, NULL);
DbgBreakPoint();
if (!NT_SUCCESS(status))
return status;
memcpy(port_info, buffer, sizeof(VCH_PORT_INFO));
ExFreePoolWithTag(buffer, VCH_POOL_TAG);
return STATUS_SUCCESS;
}
The code that handles the IOCTL:
NTSTATUS IoCtl_Vcp_GetPortInfo(WDFDEVICE device, WDFREQUEST request)
{
NTSTATUS status;
PVCH_PORT_INFO buffer;
PPORT_PDO_DESCRIPTOR descriptor = PortPdoGetContext(device);
status = WdfRequestRetrieveOutputBuffer(request, sizeof(VCH_PORT_INFO), (PVOID*)&buffer, NULL);
if (!NT_SUCCESS(status))
return status;
buffer->Address = descriptor->Address;
buffer->ForceComIndex = FALSE; // TODO: Implement
buffer->Writeable = descriptor->Writeable;
DbgBreakPoint();
return STATUS_SUCCESS;
}
The IOCTL code definition:
#define DEVICE_TYPE_VIRTUAL_COM_PORT 0xC51
#define IOCTL_VCP_INTERNAL_GET_PORT_INFO CTL_CODE(DEVICE_TYPE_VIRTUAL_COM_PORT, 0x30, METHOD_BUFFERED, FILE_READ_DATA)
Setting the request completion information with the number of output bytes fixes the issue.
I've been banging my head on this for days, and I must have read every page on the Internet even remotely related, but I still can't find an answer. Help!
Here's the scenario: In Windows 7, I have a process running under an admin user account (not a service). It creates a global named mutex, which is later used in a child process running under a regular user account. No matter what I do, or what ACLs I put on the mutex, the child process keeps returning Access Denied when trying to get the handle.
I've distilled my code down into a test app just to experiment with the process and mutex parts, and I found something surprising: if I call OpenMutex from the user app without first creating the mutex, I would expect a Not Found error but I still get Access Denied. However, if I launch the user app from Explorer instead (shift-right-click, Run as different user...), I get the expected behavior. I also noticed that the user app has a plain blocky window border rather than the normal Windows theme when launched from the admin app.
So my guess is that there's something wrong with how I'm launching the user app, but I just can't see what I'm missing.
Here are the relevant parts:
bool CUserTest::LogInUser()
{
if ((m_hUserToken == NULL) && !LogonUser(TEST_USER_NAME, L".", TEST_USER_PASS, LOGON32_LOGON_INTERACTIVE, LOGON32_PROVIDER_DEFAULT, &m_hUserToken))
{
CloseHandle(m_hUserToken);
m_hUserToken = NULL;
}
return (m_hUserToken != NULL);
}
bool CUserTest::LaunchTestApp()
{
PROCESS_INFORMATION ProcInfo;
STARTUPINFO si;
ZeroMemory(&si, sizeof(STARTUPINFO));
si.cb = sizeof(si);
si.lpDesktop = L"winsta0\\default";
wchar_t wszCmdLine[MAX_PATH + 1] = { 0 };
wcscpy(wszCmdLine, L"UserTestClient.exe");
bool bSuccess = false;
LPVOID pEnv;
PROFILEINFO sProfileInfo;
ZeroMemory(&sProfileInfo, sizeof(PROFILEINFO));
sProfileInfo.dwSize = sizeof(PROFILEINFO);
sProfileInfo.lpUserName = TEST_USER_NAME;
if (LoadUserProfile(m_hUserToken, &sProfileInfo))
{
if (ImpersonateLoggedOnUser(m_hUserToken))
{
if (CreateEnvironmentBlock(&pEnv, m_hUserToken, FALSE))
{
bSuccess = CreateProcessAsUser(
m_hUserToken,
NULL,
wszCmdLine,
NULL, // ProcessAttributes
NULL, // ThreadAttributes
FALSE, // InheritHandles
CREATE_UNICODE_ENVIRONMENT, // CreationFlags
pEnv, // Environment
NULL, // CurrentDirectory
&si,
&ProcInfo); // ProcessInformation
DestroyEnvironmentBlock(pEnv);
}
RevertToSelf();
}
UnloadUserProfile(m_hUserToken, sProfileInfo.hProfile);
}
if (bSuccess)
{
CloseHandle(ProcInfo.hThread);
CloseHandle(ProcInfo.hProcess);
}
return bSuccess;
}
I never could get the CreateProcessAsUser call to work correctly, but I finally got it working using CreateProcessWithLogonW instead. The trick was to set si.lpDesktop to NULL rather than "winsta0\default", contrary to everything I'd read up to this point.
With UAC disabled, I need to create a process with the same characteristics as the process created with UAC enabled - basically I'm emulating process creation with UAC enabled.
My only roadblock is virtualization. The sample code below should create an instance of notedpad at medium IL with virtualization enabled. In actuality, it creates an instance of notepad at medium IL with virtualization disabled. I'm not entirely sure why the virtualization token is being ignored. Any ideas?
BOOL bRet;
HANDLE hToken;
HANDLE hNewToken;
// Notepad is used as an example
WCHAR wszProcessName[MAX_PATH] =
L"C:\\Windows\\System32\\Notepad.exe";
// Medium integrity SID
WCHAR wszIntegritySid[20] = L"S-1-16-8192";
PSID pIntegritySid = NULL;
DWORD EnableVirtualization = 1;
TOKEN_MANDATORY_LABEL TIL = {0};
PROCESS_INFORMATION ProcInfo = {0};
STARTUPINFO StartupInfo = {0};
ULONG ExitCode = 0;
if (OpenProcessToken(GetCurrentProcess(),MAXIMUM_ALLOWED, &hToken))
{
if (DuplicateTokenEx(hToken, MAXIMUM_ALLOWED, NULL,
SecurityImpersonation, TokenPrimary, &hNewToken))
{
if (ConvertStringSidToSid(wszIntegritySid, &pIntegritySid))
{
TIL.Label.Attributes = SE_GROUP_INTEGRITY;
TIL.Label.Sid = pIntegritySid;
// Set the process integrity level
if (SetTokenInformation(hNewToken, TokenIntegrityLevel, &TIL,
sizeof(TOKEN_MANDATORY_LABEL) + GetLengthSid(pIntegritySid)))
{
// Enable FS Virtualization
if (SetTokenInformation(hNewToken, TokenVirtualizationEnabled,
&EnableVirtualization, sizeof(EnableVirtualization)))
{
// Create the new process at Low integrity
bRet = CreateProcessAsUser(hNewToken, NULL,
wszProcessName, NULL, NULL, FALSE,
0, NULL, NULL, &StartupInfo, &ProcInfo);
}
}
LocalFree(pIntegritySid);
}
CloseHandle(hNewToken);
}
CloseHandle(hToken);
}
So, I was approaching this incorrectly - fs virtualization is not what I want. To emulate UAC, as described above, its necessary to create a restricted token with the administrators group disabled and use that token to create the process.
The reason that this doesn't work, is that the SetTokenInformation call to turn on virtualisation is working on the primary token created for CreateProcessAsUser. What's needed is an access token for the actual process. This can be obtained by creating the process with the CreationFlag CREATE_SUSPENDED, and calling OpenProcessToken with the process handle from ProcInfo. Use SetTokenInformation on that token to enable virtualisation, and then ResumeThread to run the process.
I am calling WsKSendTo on an opened socket (irp returns success in callback). But WskSendTo on that socket returns c0000184, what is referenced to as STATUS_INVALID_DEVICE_STATE. What kind of errors are addressed by this?
Did I miss something in the send routine?
psc->dstaddr.sin_family = AF_INET;
psc->dstaddr.sin_port = 0x6973; // big endian
psc->dstaddr.sin_addr.S_un.S_un_b.s_b1 = 0x02;
psc->dstaddr.sin_addr.S_un.S_un_b.s_b2 =
psc->dstaddr.sin_addr.S_un.S_un_b.s_b3 = 0x17;
psc->dstaddr.sin_addr.S_un.S_un_b.s_b4 = 0x0a;
// create IRP
psc->pirp = IoAllocateIrp(1, FALSE);
if(!psc->pirp){
DbgPrint("ShoutShout: IRP not allocated\n");
FreeShoutContext(psc);
return STATUS_UNSUCCESSFUL;
}
IoSetCompletionRoutine(
psc->pirp,
ShoutShoutComplete,
psc,
TRUE,
TRUE,
TRUE
);
// initiate send
const WSK_PROVIDER_DATAGRAM_DISPATCH *dispatch =
(PWSK_PROVIDER_DATAGRAM_DISPATCH)sockContext->socket->Dispatch;
status = dispatch->WskSendTo(
sockContext->socket, //PWSK_SOCKET
&psc->buf, //__in PWSK_BUF
NULL, // undocumented IN ULONG Flags,
(PSOCKADDR)&psc->dstaddr, // IN PSOCKADDR RemoteAddress OPTIONAL,
0, // IN SIZE_T ControlInfoLength,
0, // IN PCMSGHDR ControlInfo OPTIONAL,
psc->pirp); // IN PIRP Irp );
Afterwards status and pirp->IoStatus->Status contain afore mentioned error code and nothing is sent (or reaching the destination). Any suggestions?
c0000184 is returned if the socket is not fully initialised. For example if the socket is not yet bound to an address. The same problem existst in userland, always bind your socket! (Here using WskBind.)
It is also useful to synchronise the individual steps of the initialisation process: allocation of socket context, socket creation and binding using KEVENT (see KeInitializeEvent for a start).