Rather than having every application using Wintun driver reinvent the
wheel, the Wintun device/adapter/interface management has been moved
from wireguard-go to wintun.dll deployed with Wintun itself.
Signed-off-by: Simon Rozman <[email protected]>
c85e4a410f introduced preliminary HWID
checking to speed up Wintun adapter enumeration. However, all HWID are
case insensitive by Windows convention.
Furthermore, a device might have multiple HWIDs. When DevInfo's
DeviceRegistryProperty(SPDRP_HARDWAREID) method returns []string, all
strings returned should be checked against given hardware ID.
This issue was discovered when researching Wintun and wireguard-go on
Windows 10 ARM64. The Wintun adapter was created using devcon.exe
utility with "wintun" hardware ID, causing wireguard-go fail to
enumerate the adapter properly.
Signed-off-by: Simon Rozman <[email protected]>
DeleteAllInterfaces() didn't check if SPDRP_DEVICEDESC == "WireGuard
Tunnel". It deleted _all_ Wintun adapters, not just WireGuard's.
Furthermore, the DeleteAllInterfaces() was upgraded into a new function
called DeleteMatchingInterfaces() for selectively deletion. This will
be used by WireGuard to clean stale Wintun adapters.
Signed-off-by: Simon Rozman <[email protected]>
This also introduces waiting for key to appear on initial access.
See if this resolves the issue caused by HDD power-up delay resulting in
failure to create the adapter.
Signed-off-by: Simon Rozman <[email protected]>
- Make foreign interface found error numeric to ease condition
detection.
- Update GetInterface() documentation.
- Make tun.CreateTUN() quit when foreign interface found before
attempting to create a Wintun interface with a duplicate name.
Creation is futile.
Signed-off-by: Simon Rozman <[email protected]>
- Make getStringValueRetry() reusable for reading any value type. This
merges code from GetIntegerValueWait().
- expandString() >> toString() and extend to support REG_MULTI_SZ
(to return first value of REG_MULTI_SZ). Furthermore, doing our own
UTF-16 to UTF-8 conversion works around a bug in windows/registry's
GetStringValue() non-zero terminated string handling.
- Provide toInteger() analogous to toString()
- GetStringValueWait() tolerates and reads REG_MULTI_SZ too now. It
returns REG_MULTI_SZ[0], making GetFirstStringValueWait() redundant.
Signed-off-by: Simon Rozman <[email protected]>
Go adds trailing padding to DrvInfoDetailData struct in GOARCH=386 which
confuses SetupAPI expecting exactly sizeof(SP_DRVINFO_DETAIL_DATA).
Signed-off-by: Simon Rozman <[email protected]>
DIF_INSTALLDEVICE returns almost immediately, while the device
installation continues in the background. It might take a while, before
all registry keys and values are populated.
Previously, wireguard-go waited for HKLM\SYSTEM\CurrentControlSet\
Control\Class\{4D36E972-E325-11CE-BFC1-08002BE10318}\<id> registry key
only.
Followed by a SetInterfaceName() method of Wintun struct which tried to
access HKLM\SYSTEM\CurrentControlSet\Control\Network\
{4D36E972-E325-11CE-BFC1-08002BE10318}\<id>\Connection registry key
might not be available yet.
This commit loops until both registry keys are available before
returning from CreateInterface() function.
Signed-off-by: Simon Rozman <[email protected]>
As querying non-existing registry value and reading non-existing
registry string value both return ERROR_FILE_NOT_FOUND, we can
use later only.
Signed-off-by: Simon Rozman <[email protected]>
In case reading from TUN device detected TUN device was closed, it
closed the file handle and set tunFile to nil. The tunFile is
automatically reopened on retry, but... If another packet comes in the
WireGuard calls Write() method. With tunFile set to nil, this will
cause access violation.
Therefore, locking was introduced.
Signed-off-by: Simon Rozman <[email protected]>
This allows buffer alignment and keeps it together with its meta-data.
Furthermore, the write buffer has been reduced - as long as we flush
after _every_ write, we don't need a 1MiB write buffer.
Signed-off-by: Simon Rozman <[email protected]>
Do the WaitForSingleObject() always to provide high-load responsiveness.
Reorder events so TUN_SIGNAL_CLOSE has priority over
TUN_SIGNAL_DATA_AVAIL, to provide high-load responsiveness at all.
Signed-off-by: Simon Rozman <[email protected]>
Adding functions with non-consistent order made setupapi package a mess.
While we could reorder data-types and functions by alphabet - it would
make searching easier - it would put ...Get... and ...Set... functions
quite apart.
Therefore, the SetupAPI.h order was adopted.
Signed-off-by: Simon Rozman <[email protected]>
The data returned by SetupDiGetDeviceInfoListDetail() is nil on error
which will cause the test to crash should the function fail.
Signed-off-by: Simon Rozman <[email protected]>
The SetupDiEnumDeviceInfo() now returns a SP_DEVINFO_DATA rather than
taking it on input to fill it on return.
Signed-off-by: Simon Rozman <[email protected]>
The former is only a subset of the later. To minimize future
maintenance, we'll provide support for extended version only.
Signed-off-by: Simon Rozman <[email protected]>
User should not have called or deferred the Close() method should
SetupDiGetClassDevsEx() return an error (and invalid handle). And even
if user does that, a SetupDiDestroyDeviceInfoList(INVALID_HANDLE_VALUE)
is harmless. It just returns ERROR_INVALID_HANDLE - we have a unit test
for this in TestSetupDiDestroyDeviceInfoList().
Signed-off-by: Simon Rozman <[email protected]>
Furthermore setupapi.DevInfoData has been obsoleted.
SetupDiEnumDeviceInfo() fills existing SP_DEVINFO_DATA structure now.
As other functions of SetupAPI use SP_DEVINFO_DATA, converting it to
DevInfoData and back would hurt performance.
Signed-off-by: Simon Rozman <[email protected]>
Simon Rozman