Get the immediate target path from symlink/reparse point - windows

Question

I am aware that GetFinalPathNameByHandle can be used to obtain the target of a symbolic link or a reparse point, but there are situations where its use is not desirable:
If the target is not available, doesn't exist or cannot be opened, CreateFile on a symlink fails, and thus the path cannot be obtained.
If I point a symlink "a" to file "b" and create a symlink "b" to file "c", the function follows the whole chain, returning "c".
The function is not useful much when I already have the handle to the actual symlink at hand.
It seems that DeviceIoControl can be used together with FSCTL_GET_REPARSE_POINT to obtain the actual reparse data of the file, but that gets me the REPARSE_DATA_BUFFER, and I would have to parse that.
I don't know how the system actually processes reparse points, but I think that the target location is a piece of information that should be available at some point. The dir command, for example, can display the target path correctly for any reparse point... well I have already seen it handle just symlinks and mount points (junctions).

Answer 1

how the system actually processes reparse points
this is done inside file system and file system filter drivers. result depend from are FILE_FLAG_OPEN_REPARSE_POINT option used in call CreateFile (or FILE_OPEN_REPARSE_POINT in NT calls).
when FILE_FLAG_OPEN_REPARSE_POINT is specified - file system bypass normal reparse point processing for the file and attempts to directly open the reparse point file as is.
If the FILE_OPEN_REPARSE_POINT flag is not specified - file-system attempts to open a file to which reparse point is point (if fs understand format of reparse point - primary only Microsoft reparse points)
the data format saved in reparse point is REPARSE_DATA_BUFFER (Microsoft reparse point format) or REPARSE_GUID_DATA_BUFFER - need look for ReparseTag at begin.
to determine whether a reparse point tag corresponds to a tag owned by Microsoft we use IsReparseTagMicrosoft macro.
code for test/print reparse point data:
volatile UCHAR guz;
ULONG TestReparsePoint(PCWSTR FileName)
{
HANDLE hFile = CreateFile(FileName, 0, FILE_SHARE_VALID_FLAGS, 0, OPEN_EXISTING,
FILE_FLAG_OPEN_REPARSE_POINT|FILE_FLAG_BACKUP_SEMANTICS, 0);
if (hFile == INVALID_HANDLE_VALUE)
{
return GetLastError();
}
union {
PVOID pv;
PULONG ReparseTag;
PREPARSE_DATA_BUFFER prdb;
PREPARSE_GUID_DATA_BUFFER prgdb;
};
PVOID stack = alloca(guz);
ULONG cb = 0, rcb = sizeof(REPARSE_DATA_BUFFER) + 0x100, BytesReturned;
ULONG dwError;
do
{
if (cb < rcb) cb = RtlPointerToOffset(pv = alloca(rcb - cb), stack);
if (DeviceIoControl(hFile, FSCTL_GET_REPARSE_POINT, 0, 0, pv, cb, &BytesReturned, 0))
{
dwError = NOERROR;
if (IsReparseTagMicrosoft(*ReparseTag))
{
char cc[16];
LPCSTR name;
switch (*ReparseTag)
{
case IO_REPARSE_TAG_SYMLINK:
name = " SYMLINK";
stack = prdb->SymbolicLinkReparseBuffer.PathBuffer;
break;
case IO_REPARSE_TAG_MOUNT_POINT:
name = " MOUNT_POINT";
stack = prdb->MountPointReparseBuffer.PathBuffer;
break;
default:
sprintf(cc, " %08x", prdb->ReparseTag);
name = cc;
}
DbgPrint(" %s->%.*S <%.*S>\n", name,
prdb->MountPointReparseBuffer.SubstituteNameLength >> 1,
RtlOffsetToPointer(stack, prdb->MountPointReparseBuffer.SubstituteNameOffset),
prdb->MountPointReparseBuffer.PrintNameLength >> 1,
RtlOffsetToPointer(stack, prdb->MountPointReparseBuffer.PrintNameOffset)
);
}
else
{
PGUID g = &prgdb->ReparseGuid;
DbgPrint(" tag=%x {%08x-%04x-%04x-%02x%02x-%02x%02x%02x%02x%02x%02x} size=%x\n", *ReparseTag,
g->Data1, g->Data2, g->Data3,
g->Data4[0],g->Data4[1],g->Data4[2],g->Data4[3],g->Data4[4],g->Data4[5],g->Data4[6],g->Data4[7],
prgdb->ReparseDataLength);
}
break;
}
rcb = IsReparseTagMicrosoft(*ReparseTag)
? REPARSE_DATA_BUFFER_HEADER_SIZE + prdb->ReparseDataLength
: REPARSE_GUID_DATA_BUFFER_HEADER_SIZE + prgdb->ReparseDataLength;
} while((dwError = GetLastError()) == ERROR_MORE_DATA);
CloseHandle(hFile);
return dwError;
}

Answer 2

Microsoft reparse points can be read with REPARSE_DATA_BUFFER instead. The MS open protocol specification might also be useful.
Parsing other GUID based tags can only be done if you know the format.