Sorry, I've read your post again and obviously you've installed software. You've described in detail what you've done with the hardware but please also describe what you've done with the software.
The hardware mod is more simple, (just USB_RREF 4.7 Kohm, and two 22 ohm impedance matching resistors are more than enough), so I think you should focus on the software and be (almost) sure that it is working as expected.
Don't bother with that small black resistor, if you look at the schematic provided by swiftik R11/R13 form some (in my opinion not important) EJTAG_SEL bootstrap jumper. Just any resistor (1 Kohm, 4.7 kOhm, etc...) should work.
By the way, about the hardware, as the 5 volt DC-DC converter is not installed on board, I've used one from Car Cigarette Powered 1000mA USB Adapter/Charger. I'm more than happy with it, it draws 1A(!) and it is much more than the USB spec (500 mA). And also, not least, it glows in the dark (not that I see it under the hood)
Also about hardware, now there are cheap routers with USB 2.0 mounted on board, like TP-LINK TL-MR3420, so next time I may get such one.
I suppose you've used trunk, I've also started with one. And I had some strange experiences with it. I can't remember which build was it, but after everything was OK with the USB, if I've tried to execute file on USB file system it crashed with SIGSEGV and SIGILL. Eventually, it turned out that it was a bug in the file system code which didn't manage correctly the data cache page flush, and, it affected only filesystems other than the embedded (squashfs and jffs were not affected, ext2/3 and vfat were!). There was no trace of this bug in the next trunk releases. I've decided not to use trunk, but a more predictable backfire release and I've compiled backfire 10.03.1-rc4 with USB support. The benefit would be that it is known stable version, and, one could use all the precompiled packages available for download, and, in this version ext_root is supported natively - no need to explicitly recompile the firmware because of it. I've done that and now the bluetooth driver and the tun driver (used by OpenVPN) are buggy - the kernel oopses I do not have much time to play with the router now so I plan to recompile and install the next stable version, when it comes out.
Now, about the USB software side. When you add the "ar71xx_add_device_usb();" line to mach-dir-615-c1.c, when the ehci module ehci-hcd.ko is inserted in the kernel the following is seen in the syslog (dmesg):
ehci_hcd: USB 2.0 'Enhanced' Host Controller (EHCI) Driver
ar71xx-ehci ar71xx-ehci: Atheros AR91xx built-in EHCI controller
ar71xx-ehci ar71xx-ehci: new USB bus registered, assigned bus number 1
ar71xx-ehci ar71xx-ehci: irq 3, io mem 0x1b000000
ar71xx-ehci ar71xx-ehci: USB 2.0 started, EHCI 1.00
If you see this you are (almost) there! The USB_PLL has to be initialized, and if it is not, you won't see anything in the syslog, but nevertheless there will still be no USB communication. This is done usually in the boot loader (u-boot) in routers in which the USB is used, but is omitted in routers like D-LINK DIR-615 C1&C2 and TP-LINK TL-WR941ND, where the USB port is there but is officially not used. So, to do this you could add another line to the kernel patch or just to try it you could initialize it by hand on the u-boot command line prompt by
Also, if you want just to play something without flashing the router you could use initramfs firmware image.
OK, that's it, focus on the software and when you are (almost) sure it should work grab the soldering iron and solder all around