Project OS X Forums > Ehci Ownership Fix

Project OS X Forums > Snow Leopard Guides & Tutorials > DSDT Patching

THe KiNG

Jan 10 2010, 08:26 PM

This feature is already present in Chameleon and derivates, however I decided to fix it in DSDT b/c on some variants/implementations it dosen't allow me to boot from a USB flash stick.
So here we go, thanks to iNDi who translated the original code for me, I was able to fix it in DSDT:
Test machine - ASUS M50SV ICH8-M notebook.
Docs used - Intel ICH8 IO Controller HUB, 15.1.26 LEG_EXT_CAP�USB EHCI Legacy Support Extended Capability Register (USB EHCI�D29:F7, D26:F7), page 597.

On each EHC1 and EHC2 add this code, or complete it if is already:

CODE

                OperationRegion (EOWN, PCI_Config, 0x68, 0x04) // LEG_EXT_CAP�USB EHCI Legacy Support Extended
                Field (EOWN, ByteAcc, NoLock, Preserve)
                {
                        ,   16,
                    HCBO,    1, // HC BIOS Owned Semaphore
                        ,    7,
                    HCOO,    1, // HC OS Owned Semaphore
                }

Now in our good PINI method(I explained in other topics/hacks how/where):

CODE

    Method (PINI, 0, NotSerialized) // For PCI0/Wake INI
    {
        Store (Zero, \_SB.PCI0.EHC1.HCBO)
        Store (One, \_SB.PCI0.EHC1.HCOO)
        Store (Zero, \_SB.PCI0.EHC2.HCBO)
        Store (One, \_SB.PCI0.EHC2.HCOO)
        // ................. // If you have more stuff to add
    }

Do not forget the call of PINI in _WAK method and on PCI0 _INI method!

CODE

    Method (_WAK, 1, NotSerialized)
    {
        PINI ()
        // ....... // Rest of your stuff from _WAK method
    }

CODE

        Device (PCI0)
        {
            Method (_INI, 0, NotSerialized)
            {
                PINI ()
            }

            // ....... //
        }

Enjoy!

L.E. Fixed the code to write only on required bits...

Slice

Jan 11 2010, 07:30 PM

So, the last reason to remake IOUSBFamily is out. All fixes done by DSDT.

rob3r7o

Mar 3 2010, 05:42 AM

Thank You!!!!!!!!!

Bye bye USBF error!!!!!

Great Work

My PC: ASUS M51sn ICH8-M notebook

18seven

Jul 16 2010, 01:09 AM

Fixed my ICH10R shutdown issue. Using the configuration space is a much better method than others (like the hackie device specific method, yuck) thanks.

гык-sse2

Aug 20 2010, 10:40 AM

For ICH10 not 15 but 16 worked.

Slice

Aug 20 2010, 11:18 AM

Thanks to SunKi!
http://www.applelife.ru/topic/21865-п...post__p__204697

THe KiNG

Aug 21 2010, 07:07 AM

QUOTE (гык-sse2 @ Aug 20 2010, 01:40 PM)

For ICH10 not 15 but 16 worked.

QUOTE (Slice @ Aug 20 2010, 02:18 PM)

Thanks to SunKi!
http://www.applelife.ru/topic/21865-п...post__p__204697

Yes, my mistake when I updated the code...
Thanks for notice the bug

Always check the datasheet...
BTW for MCP79 users the offset address is 0x0A

18seven

Aug 24 2010, 03:13 AM

QUOTE (THe KiNG @ Aug 21 2010, 03:07 AM)

Always check the datasheet...

Agreed, despite the fact I have a totally different config space setup than this guide, I still would have never realized ICH8 and ICH10 were identical because I always read the docs for my hardware to implement the idea presented for my needs/understanding, however, glad it is sorted and would have posted had I seen the error.

thx againTHe KiNG for all the cs info you have posted. I may contribute something useful in return sometime...

Slice

Aug 27 2010, 06:57 AM

Hmmm. I read orByte post and found something new

QUOTE

Briefly, while Apple implements

CODE

#define EHCI_USBLEGSUP_BIOS (1 << 16) /* BIOS semaphore */
#define EHCI_USBLEGSUP_OS (1 << 24) /* OS semaphore */,

they do not IMPLEMENT

CODE

#define EHCI_USBLEGCTLSTS 4 /* legacy control/status */
#define EHCI_USBLEGCTLSTS_SOOE (1 << 13) /* SMI on ownership change */,

which is actually a critical part of the management state machine (consequently on restart/reset BIOS reclaims but OS does not let go - so when you restart - no one gets it in the end). Well, in an unquirky world it would be taken care of by the BIOS, but hey! You know how it goes...

Resetting the OS owned bit, generates the SMI event which then sorts out the rest of the registers properly after restart/reset. We need to do this check prior to writing an OS owned bit.

So... the trick in the topic is not full solution.

THe KiNG

Aug 30 2010, 05:45 AM

QUOTE (Slice @ Aug 27 2010, 09:57 AM)

Hmmm. I read orByte post and found something new

So... the trick in the topic is not full solution.

Yes it is, if you have EHCI devices right configured in DSDT.
Sleep/Wake/Restart/Shutdown works with no problem on 2 borked systems.
For me worked even on "evil" MCP79 after I found out the it has a diff offset(thanks to Signal64 and his debug booter)
Anyway a fixed driver would do a better job(if is done proper) especial if it can fix the eject after sleep bug on MCP79...
Still investigating that with no luck so far...

Slice

Sep 3 2010, 08:10 PM

Moreover, I see AppleUSBEHCI.kext set ownership bits in the same way as this DSDT patch

CODE

_device->configWrite32(eecp, data | kEHCI_USBLEGSUP_OSOwned);

The problem is more deeper that simply set the bit.

This is a "lo-fi" version of our main content. To view the full version with more information, formatting and images, please click here.