<< Solved >> NVMe on Mojave kills sleep/shutdown/restart

[TechieInHell]

Well thanks for trying. I doubt it makes any difference, but I went straight to /var/log/powermanagement to look for the Sleep Failure line to see if there was any more info on another line that grep would have filtered out. There's only one entry line without a new timestamp after the Sleep Failure line:

2019-06-27 02:17:32 -0700 Failure Sleep Failure [code:0x3405F9000000012]:
Sleep/Wakes since boot at 2019-06-26 23:10:26 -0700 :0 Dark Wake Count in this sleep cycle:0

Not super informative.

 

[pastrychef]

Well thanks for trying. I doubt it makes any difference, but I went straight to /var/log/powermanagement to look for the Sleep Failure line to see if there was any more info on another line that grep would have filtered out. There's only one entry line without a new timestamp after the Sleep Failure line:

2019-06-27 02:17:32 -0700 Failure Sleep Failure [code:0x3405F9000000012]:
Sleep/Wakes since boot at 2019-06-26 23:10:26 -0700 :0 Dark Wake Count in this sleep cycle:0

Not super informative.

Yeah. Not enough information...

 

[wstrohm]

Amazing, I have had the same shutdown failures as "TechieInHell" (great handle BTW) since building my latest computer (at left). I have 2 NVMe M.2 SSDs on my Z390 I AORUS PRO mobo, however they are both Samsung 970 Pros. I have had numerous problems with both High Sierra and Mojave 10.14.5 (now back to High Sierra 10.13.6, which is better). Impossible to pin down but sometimes the problems just "go away" by themselves.

At the moment I have Shutdown and Sleep working and no black screen, but it's very easy to induce either or all by playing with Whatever Green, Pike R. Alpha's AGDP Fix kernel patch ("Black Screen Patch Vega 56/64, RX580 etc. (c)Pike R. Alpha"), and the "CSM Enable/Disable" function in version F5 of Gigabyte's UEFI BIOS. If I could write a table describing the response of the system to various combinations of these parameters, I would, but nothing is consistent. If it might be of any help, my config.plist is attached. (At the moment, no WEG in Library/Extensions, and CSM is Enabled.)

 

[pastrychef]

Amazing, I have had the same shutdown failures as "TechieInHell" (great handle BTW) since building my latest computer (at left). I have 2 NVMe M.2 SSDs on my Z390 I AORUS PRO mobo, however they are both Samsung 970 Pros. I have had numerous problems with both High Sierra and Mojave 10.14.5 (now back to High Sierra 10.13.6, which is better). Impossible to pin down but sometimes the problems just "go away" by themselves.

At the moment I have Shutdown and Sleep working and no black screen, but it's very easy to induce either or all by playing with Whatever Green, Pike R. Alpha's AGDP Fix kernel patch ("Black Screen Patch Vega 56/64, RX580 etc. (c)Pike R. Alpha"), and the "CSM Enable/Disable" function in version F5 of Gigabyte's UEFI BIOS. If I could write a table describing the response of the system to various combinations of these parameters, I would, but nothing is consistent. If it might be of any help, my config.plist is attached. (At the moment, no WEG in Library/Extensions, and CSM is Enabled.)

Your config.plist is the first I've seen to have "DisableASPM" enabled. Are you sure you need this?

Which Aptio memory fix are you using?
Are you using EmuVariableUefi-64.efi?

TechieInHell has no problems when the Intel 660P is removed. Are your problems solved with the Samsung 970 Pros removed?

If the AGDP fix is the cause of your problems, you can disable it in config.plist. Very often, WhateverGreen isn't necessary for video to work. I only use it to enable DRM. Most modern/current motherboards and video card combos do better with CSM disabled.

 

[wstrohm]

Which Aptio memory fix are you using?

AptioMemoryFix-64

Are you using EmuVariableUefi-64.efi?

No. Seems to make no difference present or absent.

Are your problems solved with the Samsung 970 Pros removed?

Have no idea, since these are the only storage devices in this computer.

If the AGDP fix is the cause of your problems, you can disable it in config.plist.

Disabling it if WhateverGreen is not present seems to cause the black screen to appear, so in general I leave it enabled (and WEG uninstalled).

Most modern/current motherboards and video card combos do better with CSM disabled.

Seems to work either way, except non-UEFI boot options appear when it is enabled, so I prefer it disabled too. The less boot options appearing in the BIOS, the better I like it.

 

[pastrychef]

AptioMemoryFix-64
No. Seems to make no difference present or absent.
Have no idea, since these are the only storage devices in this computer.
Disabling it if WhateverGreen is not present seems to cause the black screen to appear, so in general I leave it enabled (and WEG uninstalled).
Seems to work either way, except non-UEFI boot options appear when it is enabled, so I prefer it disabled too. The less boot options appearing in the BIOS, the better I like it.

As far as I know, NVRAM is not working with Z390. Without NVRAM, shutdown/restart can be broken. I suggest trying with EmuVariableUefi-64.efi + RC scripts installed.

 

[wstrohm]

RC Scripts are installed (Clover version 4972 is the boot loader). I will try EmuVariable-64.efi once again.
[Edit: Nope, installing EmuVariable-64.efi causes slashed circle when exiting Clover to load the OS. (I have read that it is incompatible with AptioMemoryFix-64.)]

 

[TechieInHell]

I have fixed my issue.

On a lark this morning I decided to put the SATA card back in and remove the NVMe card and discovered it was still unable to shutdown, etc.

I had read on another build that EmuVariableUefi-64.efi was not needed with H370 so I had previously removed it. I'm sure sleep/shutdown was working without it using the SATA SSD. Despite all signs pointing to NVRAM working (nvram -p works, iMessage works, etc) it apparently was not. Putting EmuVariableUefi-64.efi back into my CLOVER folder restored sleep/shutdown/restart and works with the Intel NVMe card, so I'm a happy camper.

OK, on to the next step in this project... I've got an actual Apple Wifi+BT card from a real iMac (it was already dead, I didn't kill a perfectly good iMac) that I'm going to try sticking into one of those Mini PCIe to PCIe adapters.

Thanks again to pastrychef for helping!

 

[dragonmel]

@pastrychef

hmm ... this is compelling...

I am not using nvme... but I am having several issues moving forward with a migration of my hack from clover 4300ish and 10.11.6 which is working great.. to mojave using clover 5000+

in the past I used UEFI boot on this intel X58so but was not using native NVRAM and using RC methods..

in this latest install I noticed that OSXaptiomemoryfix3 or whatever it is that slice recommended in the latest clover builds was giving me native NVRAM.. variable and test variables are holding ..

but what was a very stable build using almost no options check in clover config with a edited DSDT was now unstable.. unreliable and a mess with the VEGA card that is replacing the GTX660 i have been using for 10.11 since its no longer compatible..

lots of black screens, and if I plug a sata card that works in other hackintoshes into pie.. it either crashes the load or black screen .. etc

I too have been told to use WEG regardless if my GFX0 dsdt edits bring the card up natively because its magic but honestly ..with my x58 and w3680 hardware .. it just doesn't seem optimized for it.. and all the documentation that is out there only describes setting it up, using hackentool, etc with options for sandy and later architectures ..so I don't thing WEG is the cleanest solution .. at least not in older hardware...

the latest updates past 10.14.5 have all but bricked me and I can't figure it out...

would be happy to upload my dsdt, ioreg and anything else if someone would care to have a look

long and short.. I might have to reinstall again from scratch and try with a RC scripts install and see if it magically makes things better..

I will say that when the native NVRAM was working.. during my attempt to upgrade the previous install. I made the mistake of re-introducing my config.plist from the el cap EFI, and without thinking of changing the SMBIOS info that specs the older firmware version for MACPRO5,1 .. I did an update and I think that apple firmware update routines in the upgrade attempted to run and successfully managed to permanently effect my motherboard...

I can no longer boot straight through to clover..

in bios, the drive order only shows the one installed hard disk

in the past, my one time boot menu in UEFI mode would show 1 entry for the UEFI and one 'legacy' bios for each disk

like this

[uefi shell volume]
ST0050050505

now the one time boot menu has

Mac OS X
[UEFI shell volume]
ST0002340340304

the UEFI shell volume is indented one character in from the Mac OS X heading/grouping whatever that is ..

I have never called anything that.. every.. it shows up nowhere .. EFI shell has no entry..

if I try and boot it manually or when the board auto boots.. it goes black screen blinking then steady cursor...

if I boot to F10 one time menu and select the UEFI volume.. it boots to clover menu ...

my board is permanently semi bricked.. can't get rid of it .. (the mac os x ) entry/header

downgrading and reloading bios doesn't fit it
cmos reset / battery out for 24 hours / nope
hell booting with no drives attached at all and Mac OS X is still there.. permanently written

very strange...

 

[pastrychef]

@pastrychef

hmm ... this is compelling...

I am not using nvme... but I am having several issues moving forward with a migration of my hack from clover 4300ish and 10.11.6 which is working great.. to mojave using clover 5000+

in the past I used UEFI boot on this intel X58so but was not using native NVRAM and using RC methods..

in this latest install I noticed that OSXaptiomemoryfix3 or whatever it is that slice recommended in the latest clover builds was giving me native NVRAM.. variable and test variables are holding ..

but what was a very stable build using almost no options check in clover config with a edited DSDT was now unstable.. unreliable and a mess with the VEGA card that is replacing the GTX660 i have been using for 10.11 since its no longer compatible..

lots of black screens, and if I plug a sata card that works in other hackintoshes into pie.. it either crashes the load or black screen .. etc

I too have been told to use WEG regardless if my GFX0 dsdt edits bring the card up natively because its magic but honestly ..with my x58 and w3680 hardware .. it just doesn't seem optimized for it.. and all the documentation that is out there only describes setting it up, using hackentool, etc with options for sandy and later architectures ..so I don't thing WEG is the cleanest solution .. at least not in older hardware...

the latest updates past 10.14.5 have all but bricked me and I can't figure it out...

would be happy to upload my dsdt, ioreg and anything else if someone would care to have a look

long and short.. I might have to reinstall again from scratch and try with a RC scripts install and see if it magically makes things better..

I will say that when the native NVRAM was working.. during my attempt to upgrade the previous install. I made the mistake of re-introducing my config.plist from the el cap EFI, and without thinking of changing the SMBIOS info that specs the older firmware version for MACPRO5,1 .. I did an update and I think that apple firmware update routines in the upgrade attempted to run and successfully managed to permanently effect my motherboard...

I can no longer boot straight through to clover..

in bios, the drive order only shows the one installed hard disk

in the past, my one time boot menu in UEFI mode would show 1 entry for the UEFI and one 'legacy' bios for each disk

like this

[uefi shell volume]
ST0050050505

now the one time boot menu has

Mac OS X
[UEFI shell volume]
ST0002340340304

the UEFI shell volume is indented one character in from the Mac OS X heading/grouping whatever that is ..

I have never called anything that.. every.. it shows up nowhere .. EFI shell has no entry..

if I try and boot it manually or when the board auto boots.. it goes black screen blinking then steady cursor...

if I boot to F10 one time menu and select the UEFI volume.. it boots to clover menu ...

my board is permanently semi bricked.. can't get rid of it .. (the mac os x ) entry/header

downgrading and reloading bios doesn't fit it
cmos reset / battery out for 24 hours / nope
hell booting with no drives attached at all and Mac OS X is still there.. permanently written

very strange...

I don't have any experience with Westmere or legacy hackintoshes.

Personally, I try to avoid DSDTs and/or SSDTs as much as possible. I just find my systems to run better without them.

If you have NVRAM working, use it and skip EmuVariableUefi.

As long as your NVMe SSD is compatible, there should be no consequences to using them.

I've almost never had to mess with UEFI shell.