How can I set up Thread-Local Storage (TLS) callbacks on Windows without CRT?

Do this in the following way:

ULONG tlsindex;
void CALLBACK OnTls(PVOID, DWORD Reason, PVOID);
static const PIMAGETLSCALLBACK gtlscb[] = { OnTls, 0 };
#pragma constseg(".rdata$T")
EXTERNC const IMAGETLSDIRECTORY tlsused = {
    (ULONGPTR)0,           // Start of TLS data
    (ULONGPTR)0,           // End of TLS data
    (ULONGPTR)&tlsindex, // Address of tlsindex
    (ULONGPTR)gtlscb,    // Pointer to call back array
};
#pragma constseg()
#ifdef WIN64
pragma(comment(linker, "/include:tlsused"))
#else
pragma(comment(linker, "/include:tlsused"))
#endif

The Microsoft run-time library facilitates this process by defining a memory image of the TLS directory and giving it the special name "tlsused" (Intel x86 platforms) or "tls_used" (other platforms). The linker looks for this memory image and uses the data there to create the TLS directory. Other compilers that support TLS and work with the Microsoft linker must use this same technique.