MacBook – Adapter from USB C device to USB port

Thunderbolt and USB are quite different protocols, so you can't just use a cable -- it would need to have active electronics in it to convert between the protocols. And that means it would cost a lot more than a cheap commodity USB hub and so no one would buy it -- they'd just get a hub instead. If you're concerned about the power drain, get a powered USB hub.

Patching the kernel to work around the issue

I do not recommend you do this. You're editing the binary of a very important file. If it gets messed up, well, ya. But this is the only solution.

This is likely 10.9.4 specific.

Location 0x00016942 of /System/Library/Extensions/IOUSBFamily.kext/Contents/PlugIns/AppleUSBXHCI.kext/Contents/MacOS/AppleUSBXHCI should contain 66 B9 60 00. That translates to mov cx, 60h

I edited it to 66 B9 *80* 00 and all of my devices come online now. I assume going even higher, or removing the "errata" check altogether, is also safe. If you run into trouble you can just unplug devices and reboot.

Research

It appears to be breaking inside of the function AppleUSBXHCI::TestConfiguredEpCount in AppleUSBXHCIUIM.cpp.

Which is using a max value, which is set from:

// default max number of endpoints we allow to be configured
_maxControllerEndpoints = kMaxXHCIControllerEndpoints;

// this will switch the mux for panther point
EnableXHCIPorts();

if( (_errataBits & kXHCIErrataPPT) != 0)
{   
    // Panther Point - limit endpoints to 64 <rdar://problem/10385765>
    // does this need a separate errata bit?
    _maxControllerEndpoints = 64;
}

The flags are being set by this code

static ErrataListEntry  errataList[] = {

    {0x1033, 0x0194, 0, 0xffff, kXHCIErrata_NEC},   // NEC XHCI, check firmware
    {0x1b73, 0x1000, 0, 0xffff, kXHCIErrata_NoMSI}, // Fresco Logic XHCI
    {0x8086, 0x1e31, 0, 0xffff, kXHCIErrataPPT | kXHCIErrataPPTMux | kXHCIErrata_EnableAutoCompliance | kErrataSWAssistXHCIIdle | kXHCIErrata_ParkRing},    // Intel Panther Point
    {0x1b21, 0, 0, 0xffff,  kXHCIErrata_ASMedia},   // ASMedia XHCI
    {0x1b73, 0, 0, 0xffff, kXHCIErrata_FrescoLogic},// Fresco Logic
    {0x1b73, 0x1100, 0, 16, kXHCIErrata_FL1100_Ax}, // Fresco Logic FL1100-Ax
    {0x1b6f, 0x7052, 1, 0xffff, kXHCIErrata_Etron7052}
};

0x1e31 is the Device ID of my MacBook Pro's built in USB controller!

The logging shows my max limit is set to 96 (it hits 97 and sets the err value causing it to revert). This source is old as Apple has closed the source to IOUSBFamily a while ago, so there's no real way to tell where the 96 is coming from, but the conclusion appears to be my hub is black listed by the OSX kernel for reasons unknown.

Apple increased the limit by 50% in 10.8, where it was 64, to 96 in 10.9. This shows that the number is arbitrary. The hardware did not change but there was a 50% increase allowed in the software, that's a significant change. But why not 128? Why not 95, or 96? The units of power of two are likely also meaningless here.

Why 96 end points is a very low limit

Each one of my USB hubs has 3 USB controllers inside of it. Each one supports "Full/Low Speed" and "SuperSpeed." Each one of those has an endpoint for input, and one for output. 3 hubs * 3 controllers each * 2 protocols each controller * 2 for IO = 36 end points... just on hubs!

The bluetooth built into the laptop uses 8, the HD FaceTime Camera uses 6, a Nexus 4 uses 6.

96 is a very low amount of end points.