DokanUnmount() always fails #57
Comments
|
Can you provide some logs ? Does Mirror.exe work ? |
|
Here's the log from the Visual Studio debug console: I took the following steps:
I'll try out Mirror.exe and enable more robust logging within the next few days. |
|
Please can you use DebugView https://technet.microsoft.com/en-us/library/bb896647.aspx so that we can see, why the mounter service failed to delete the mount point. |
|
Since we have no update from you @Corillian . |
|
I'm afraid my inactivity hardly means the problem is resolved. I have plenty of things to work on and am unable to devote all of my time to figuring out why this driver is hanging. To that end I have been unable to gather any debug output as my locally built copy of the driver fails to install with: I'm using the source tagged at v0.7.4. I followed your build/install instructions and I am running windows 7 x64 in |
|
Ok I got the install issues sorted out. It looks like my service was hung due to the inability to unmount Dokan, I hadn't explicitly killed the process, and to make a long story short when I ran the 0.7.4 uninstaller stuff got corrupted and created non-obvious issues. Unfortunately even though I now have my debug build of Dokan installed and running I am not getting any additional debug info in DebugView or my Visual Studio output window. I double checked |
|
Alrighty I got the debug output. My WinDbg kernel capture was turned off. GitHub won't let me upload the log file but you can find it at http://pastebin.com/ZPqvk5a0. The major highlights are when I call DokanUnmount() returns false and my Dokan thread continues to be hung in which still doesn't cause I hope that helps, let me know if there's anything else I can assist with. |
|
Good to have news from you :) !!! Just to be sure I understand well: If you can add some logs and see where it fail and why, it would be nice ! Sorry for taking your time but you are the only personn who has repport this issue 😢 |
|
The log obove is provided by driver and not by mounter service we are interested in. |
|
@marinkobabic I don't have any mounter service output beyond: |
|
Ok I have discovered why the call to if(!DokanUnmount(m_localDriveRoot[0])) // m_localDriveRoot[0] resolves to L'M'at which point BOOL DOKANAPI
DokanUnmount(
WCHAR DriveLetter)
{
WCHAR mountPoint[] = L"M:"; // <--- FAIL
mountPoint[0] = DriveLetter; // <--- FAIL
return DokanRemoveMountPoint(mountPoint);
}The issue is that the Dokan mounter service expects the drive name to be formatted as for (listEntry = g_MountList.Flink; listEntry != &g_MountList; listEntry = listEntry->Flink) {
mountEntry = CONTAINING_RECORD(listEntry, MOUNT_ENTRY, ListEntry);
if (useMountPoint) {
if (wcscmp(DokanControl->MountPoint, mountEntry->MountControl.MountPoint) == 0) { // <--- FAIL
found = TRUE;
break;
}
} else {
if (wcscmp(DokanControl->DeviceName, mountEntry->MountControl.DeviceName) == 0) {
found = TRUE;
break;
}
}
}
control.Type = DOKAN_CONTROL_UNMOUNT;
control.Option = DOKAN_CONTROL_OPTION_FORCE_UNMOUNT;
wcscpy_s(control.MountPoint, sizeof(control.MountPoint) / sizeof(WCHAR), MountPoint); // <--- POSSIBLE FAIL
DokanMountControl(&control);however DbgPrintW(L"DokanControl MountEntry not found. Try unmount '%s' force: %d\n", Control->MountPoint, Control->Option);
if (Control->Option == DOKAN_CONTROL_OPTION_FORCE_UNMOUNT && DokanControlUnmount(Control->MountPoint)) {
Control->Status = DOKAN_CONTROL_SUCCESS;
break;
}Now that |
|
Ok this is a bit frustrating. The documentation explicitly states that when mounting I should use a drive letter formatted as typedef struct _DOKAN_OPTIONS {
USHORT Version; // Supported Dokan Version, ex. "530" (Dokan ver 0.5.3)
ULONG ThreadCount; // number of threads to be used
ULONG Options; // combination of DOKAN_OPTIONS_*
ULONG64 GlobalContext; // FileSystem can use this variable
LPCWSTR MountPoint; // mount point "M:\" (drive letter) or
// "C:\mount\dokan" (path in NTFS)
} DOKAN_OPTIONS, *PDOKAN_OPTIONS;That's it, that's all there is for documentation right there. For unmounting the docs mention On top of that, commit a28a915 still doesn't solve the root problem: Dokan isn't normalizing mount paths. This is a major issue and if you noticed in my description above To summarize: Dokan should normalize mount paths and I suspect, at least for single letter drive mounts, it should also be case-insensitive. |
|
@Corillian, Yes I see your point ! |
|
Fixed with #79 |
I have not been able to find a way to call DokanUnmount() such that it actually succeeds. Not only that but
dokanctl.exe /u Malso fails. The only way I can unmount the drive is by executingdokanctl.exe /u M /fhowever DokanMain() still doesn't return and additional calls to DokanUnmount() also fail.I am running a console application and DokanMain() is being called on its own thread with
DOKAN_OPTION_KEEP_ALIVE | DOKAN_OPTION_DEBUG.The text was updated successfully, but these errors were encountered: