Linux 2.6.28 Porting to Oxnas

[ftcodes]

It seems that not all oxnas devices are compatible on the USB host side. The Medion NAS is not taking any USB devices on one port at least.
Update:
Seems that port connected detection is not so standard. It does not react to device connect events. However, it reacts to disconnect events and detects devices on load of usb host driver.

[WarheadsSE]

This is odd, as I have no issue with insertion/removal but then: The pogoplug has host A attached to a usb hub. Host B is attached to nothing. So, I'm likely not seeing this issue for this same reason.

Not having a Medion, I can not test this. I will see what I can of the iOmega HMND CE I have.. although I certainly wont have time this weekend.

[ftcodes]

I am checking what is inside the original kernel anyways to find out where that device connect problem comes from.
But it is good when it is checked on other hardware as well.

BTW, if a module for the LEDs and so on is useful on HMNDCE or PogoPlugPro, it is possible to derive a module from the Medion NAS specific module I did.

Update: I compared the modifications inside EHCI against vanilla EHCI. Although that stuff looks like EHCI, it is not exactly EHCI.

Remarks about HUB:

Check connect/disconnect on the Root HUB ports.

USB hubs use a different means to detect connect/disconnect. They use USB interrupt transfers.

The root hub uses registers and interrupts in the system. Root hub USB transfers are practically emulated. Those are not really using any physical USB transfers.

Update 2:

I added two LED gpio modules specific for HMNDCE and PogoPlugPro. Their LED definitions are untested and may have wrong polarity.

[ftcodes]

Did you have any issues on setuid binaries and su when not root is executing them when running kernel 3.1?

It is working when run on the 2.6 kernel. So, maybe you already ran into it before and found the reason for it.

The only thing that happens that the binary gets killed on exec call. I made a simple test program doing setreuid/setregid to see it with strace.

[ftcodes]

USB issue is fixed.
has_ppcd was set in ehci-oxnas.c. The OX820 does not seem to have per port connect detection irq.

[WarheadsSE]

I will review the system spec sheets from PLX for that when I get a chance.

[ftcodes]

Per Port Change Detection means 16 bits in USB interrupt status.

If it is not there, you still get an interrupt but you have to use Port Status/Control registers to detect what has happened. BTW that works on all EHCI USB controllers.

[WarheadsSE]

We're going to have to find a better way of defining the nand paritions:
https://github.com/ft-/OX820-3.1-Linux/ ... bef8eb707b


Also, USB: USBMPH ~ 2 port host controller, has command translation, so it always works as 2.0 even with 1.0 devices, and has a UTMI interface. It has its own DMA and SGDMA can be turned off/on. I don't know what 'AHB' means.

[ftcodes]

AHB means Advanced High-performance Bus (AHB) which is part of Advanced Microcontroller Bus Architecture (AMBA).
That is the interconnect between peripheral devices and the ARM core.

What they call command translation is actually transaction translator (TT). Which is standard for standalone USB 2.0 Hubs.

For EHCI, there are two options:
a) EHCI with companion OHCI/UHCI controllers
b) EHCI with root hub transaction translator.

The transaction translator is parsing split transactions (USB 2.0 specification) which are translated to USB1.1 transfers.

UTMI interface => USB 2.0 Transceiver Macrocell Interface (UTMI)

[WarheadsSE]

Ah..

I have merged all recent changes, and will tonight, merge the ox820 NAND support I've fixed up/updated the api calls. I have split the 2 NAND configs as well.