NVMe for Mac Pro 5,1

[TheRacerMaster]

How exactly are you adding new drivers to the EFI firmware? AFAIK the SMC checks the ZeroVectors' hash in the firmware modules. This can be bypassed with UEFITool or some other tool that prevents the ZeroVectors from changing, but I'm not sure how this would work when adding new drivers entirely. Again, I'm not fully sure, I've only done some small experiments with Xserve3,1 EFI to add support for Westmere CPUs (I ended up doing it, but never flashed, eventually when I get SOIC clip + other things to backup stock flash first I'll try).

 

[TheRacerMaster]

How's this going? Ive noticed Windows users can do this to get boot?

That's irrelevant. Those are instructions for a Z170 chipset motherboard. Almost all Intel 9 Series and newer motherboards have the proper EFI drivers for NVMe boot support, and older motherboards with UEFI (6 Series and up) can add NvmExpressDxe.efi and use that to boot (by modding the UEFI). Adding NVMe boot support to MacPro5,1 will also need NVMe EFI drivers, which will require a UEFI mod. Currently, you can use NVMe drives as storage drives with @jimj740's driver (great work BTW).

 

[SoyCapitanSoyCapitan]

http://barefeats.com/hard211.html

 

[jimj740]

http://barefeats.com/hard211.html

That didn't take long!

 

[ammulder]

I'm not sure this is the right problem-reporting thread, but I have a reproducible crash in the NVMeGeneric driver (with a 950 Pro). I run protractor tests against Appium which runs a Cordova app in the iPad Air simulator, and when the test finishes, my machine reboots with the following kernel panic.

Edit: I notice that of the two I captured the second one is very slightly different, and since I will also be traveling from time to time, I'm adding it below in case it should help. There are now two in the box below.

*** Panic Report ***
panic(cpu 3 caller 0xffffff801dfd155a): Kernel trap at 0xffffff801dfc3d61, type 14=page fault, registers:
CR0: 0x000000008001003b, CR2: 0x0000000000000429, CR3: 0x0000000033530000, CR4: 0x00000000003627e0
RAX: 0x0000000000000200, RBX: 0x0000000000000429, RCX: 0xffffff8041a7c490, RDX: 0x0000000000000080
RSP: 0xffffff8139ff3dc8, RBP: 0xffffff8139ff3de0, RSI: 0x00000000cb21f402, RDI: 0x0000000000000429
R8:  0xffffff8041a5c800, R9:  0xffffff8041210c08, R10: 0x0000000000000000, R11: 0x0000000000000001
R12: 0x0000000000000000, R13: 0x0000000000000001, R14: 0x0000000000000421, R15: 0x0000000000000421
RFL: 0x0000000000010002, RIP: 0xffffff801dfc3d61, CS:  0x0000000000000008, SS:  0x0000000000000010
Fault CR2: 0x0000000000000429, Error code: 0x0000000000000000, Fault CPU: 0x3, PL: 1

Backtrace (CPU 3), Frame : Return Address
0xffffff8139ff3a50 : 0xffffff801dede792
0xffffff8139ff3ad0 : 0xffffff801dfd155a
0xffffff8139ff3cb0 : 0xffffff801dfef233
0xffffff8139ff3cd0 : 0xffffff801dfc3d61
0xffffff8139ff3de0 : 0xffffff801defeb5c
0xffffff8139ff3e20 : 0xffffff801defee5e
0xffffff8139ff3e40 : 0xffffff7fa0e88abc
0xffffff8139ff3e70 : 0xffffff7fa0e89f5b
0xffffff8139ff3eb0 : 0xffffff7fa0e89d90
0xffffff8139ff3ef0 : 0xffffff801e4b6286
0xffffff8139ff3f40 : 0xffffff801e4b40f1
0xffffff8139ff3f80 : 0xffffff801e4b41e6
0xffffff8139ff3fb0 : 0xffffff801dfcbf77
      Kernel Extensions in backtrace:
         com.MinnowStor.NVMeGeneric(1.0)[27312503-CF2F-3E2B-8D8D-6E79177410C9]@0xffffff7fa0e84000->0xffffff7fa0e90fff
            dependency: com.apple.iokit.IOPCIFamily(2.9)[6ADFE830-2986-35D2-8C2F-6ABEDBC94F6A]@0xffffff7f9e72c000
            dependency: com.apple.iokit.IOStorageFamily(2.1)[956BA336-5C0A-37E9-B395-538D1BD95AC0]@0xffffff7f9e64c000

BSD process name corresponding to current thread: kernel_task

----------  SECOND REPORT ----------

*** Panic Report ***
panic(cpu 5 caller 0xffffff8013bd155a): Kernel trap at 0xffffff8013bc3d61, type 14=page fault, registers:
CR0: 0x000000008001003b, CR2: 0x0000000000000429, CR3: 0x000000002913d000, CR4: 0x00000000003627e0
RAX: 0x0000000000000200, RBX: 0x0000000000000429, RCX: 0xffffff803819cb60, RDX: 0x0000000000000080
RSP: 0xffffff91565a3dc8, RBP: 0xffffff91565a3de0, RSI: 0x00000000cb21f402, RDI: 0x0000000000000429
R8:  0xffffff80373f3000, R9:  0xffffff8036e41668, R10: 0x0000000000000000, R11: 0x0000000000000001
R12: 0x0000000000000000, R13: 0x0000000000000001, R14: 0x0000000000000421, R15: 0x0000000000000421
RFL: 0x0000000000010002, RIP: 0xffffff8013bc3d61, CS:  0x0000000000000008, SS:  0x0000000000000010
Fault CR2: 0x0000000000000429, Error code: 0x0000000000000000, Fault CPU: 0x5, PL: 1

Backtrace (CPU 5), Frame : Return Address
0xffffff91565a3a50 : 0xffffff8013ade792 
0xffffff91565a3ad0 : 0xffffff8013bd155a 
0xffffff91565a3cb0 : 0xffffff8013bef233 
0xffffff91565a3cd0 : 0xffffff8013bc3d61 
0xffffff91565a3de0 : 0xffffff8013afeb5c 
0xffffff91565a3e20 : 0xffffff8013afee5e 
0xffffff91565a3e40 : 0xffffff7f96a95abc 
0xffffff91565a3e70 : 0xffffff7f96a96f5b 
0xffffff91565a3eb0 : 0xffffff7f96a96d90 
0xffffff91565a3ef0 : 0xffffff80140b6286 
0xffffff91565a3f40 : 0xffffff80140b40f1 
0xffffff91565a3f80 : 0xffffff80140b41e6 
0xffffff91565a3fb0 : 0xffffff8013bcbf77 
      Kernel Extensions in backtrace:
         com.MinnowStor.NVMeGeneric(1.0)[27312503-CF2F-3E2B-8D8D-6E79177410C9]@0xffffff7f96a91000->0xffffff7f96a9dfff
            dependency: com.apple.iokit.IOPCIFamily(2.9)[6ADFE830-2986-35D2-8C2F-6ABEDBC94F6A]@0xffffff7f9432c000
            dependency: com.apple.iokit.IOStorageFamily(2.1)[956BA336-5C0A-37E9-B395-538D1BD95AC0]@0xffffff7f9424c000

BSD process name corresponding to current thread: kernel_task

 

[jimj740]

On travel will debug by first of year...

If possible, can you open "utilities/console" and filter on the string "NVMeGeneric" and post the error messages from the driver if there are any? By the way, both crashes are the identical issue... so being 100% reproducible this should be easy to nail...

-Thank you

 

[ammulder]

If possible, can you open "utilities/console" and filter on the string "NVMeGeneric" and post the error messages from the driver if there are any? By the way, both crashes are the identical issue... so being 100% reproducible this should be easy to nail...

Here's one more. I rebooted my machine for unrelated reasons, and on startup it said "your machine shut down because of a problem" though I didn't notice any problem...

*** Panic Report ***
panic(cpu 0 caller 0xffffff80095d155a): Kernel trap at 0xffffff80095c3d61, type 14=page fault, registers:
CR0: 0x000000008001003b, CR2: 0x0000000000000429, CR3: 0x000000001eb30000, CR4: 0x00000000003627e0
RAX: 0x0000000000000200, RBX: 0x0000000000000429, RCX: 0xffffff802d8ddb60, RDX: 0x0000000000000080
RSP: 0xffffff9126ae3dc8, RBP: 0xffffff9126ae3de0, RSI: 0x00000000cb21f402, RDI: 0x0000000000000429
R8:  0xffffff802cb34000, R9:  0x0000000000000001, R10: 0x0000000000000000, R11: 0x0000000000000001
R12: 0x0000000000000000, R13: 0x0000000000000001, R14: 0x0000000000000421, R15: 0x0000000000000421
RFL: 0x0000000000010002, RIP: 0xffffff80095c3d61, CS:  0x0000000000000008, SS:  0x0000000000000010
Fault CR2: 0x0000000000000429, Error code: 0x0000000000000000, Fault CPU: 0x0, PL: 1

Backtrace (CPU 0), Frame : Return Address
0xffffff9126ae3a50 : 0xffffff80094de792
0xffffff9126ae3ad0 : 0xffffff80095d155a
0xffffff9126ae3cb0 : 0xffffff80095ef233
0xffffff9126ae3cd0 : 0xffffff80095c3d61
0xffffff9126ae3de0 : 0xffffff80094feb5c
0xffffff9126ae3e20 : 0xffffff80094fee5e
0xffffff9126ae3e40 : 0xffffff7f8c488abc
0xffffff9126ae3e70 : 0xffffff7f8c489f5b
0xffffff9126ae3eb0 : 0xffffff7f8c489d90
0xffffff9126ae3ef0 : 0xffffff8009ab6286
0xffffff9126ae3f40 : 0xffffff8009ab40f1
0xffffff9126ae3f80 : 0xffffff8009ab41e6
0xffffff9126ae3fb0 : 0xffffff80095cbf77
      Kernel Extensions in backtrace:
         com.MinnowStor.NVMeGeneric(1.0)[27312503-CF2F-3E2B-8D8D-6E79177410C9]@0xffffff7f8c484000->0xffffff7f8c490fff
            dependency: com.apple.iokit.IOPCIFamily(2.9)[6ADFE830-2986-35D2-8C2F-6ABEDBC94F6A]@0xffffff7f89d2c000
            dependency: com.apple.iokit.IOStorageFamily(2.1)[956BA336-5C0A-37E9-B395-538D1BD95AC0]@0xffffff7f89c4c000

BSD process name corresponding to current thread: kernel_task

Meanwhile, here's what I have in the console log. It seems to only be from the most recent boot. If needed, I can cause the original crash again and see if there's anything different around when it happens.

12/20/15 3:45:30.000 PM kernel[0]: NVMeGeneric: com_MinnowStor_NVMeDisk::start
12/20/15 3:45:30.000 PM kernel[0]: NVMeGeneric: This driver must be licensed for commercial use.
12/20/15 3:45:30.000 PM kernel[0]: NVMeGeneric: This driver may not be posted, distributed, bundled, shared, or sold without the written consent of MinnowStor.com
12/20/15 3:45:30.000 PM kernel[0]: NVMeGeneric: This driver is free for personal, non-commercial use.
12/20/15 3:45:30.000 PM kernel[0]: NVMeGeneric: No guarantee of performance, suitibility, or warranty expressed or implied. Use at your own risk.
12/20/15 3:45:30.000 PM kernel[0]: NVMeGeneric: Copyright 2015 MinnowStor, all rights reserved.
12/20/15 3:45:30.000 PM kernel[0]: NVMeGeneric: devid a802144d subdev a801
12/20/15 3:45:30.000 PM kernel[0]: NVMeGeneric: desc SM951 512 GB SSD [m.2]
12/20/15 3:45:30.000 PM kernel[0]: NVMeGeneric: unable to map BAR4 resources
12/20/15 3:45:30.000 PM kernel[0]: NVMeGeneric: MSICAP: 86
12/20/15 3:45:30.000 PM kernel[0]: NVMeGeneric: DATA: 407b
12/20/15 3:45:30.000 PM kernel[0]: NVMeGeneric: Using INTX interrupts
12/20/15 3:45:30.000 PM kernel[0]: Got boot device = IOService:/AppleACPIPlatformExpert/PCI0@0/AppleACPIPCI/RP09@1D/IOPP/PXSX@0/com_MinnowStor_NVMeDisk/com_MinnowStor_NVMeDiskStorageDevice/IOBlockStorageDriver/MinnowStor Samsung SSD 950 PRO 512GB Media/IOGUIDPartitionScheme/Speedy@2

Two curious things -- in the middle there it says "SM951" which is not correct, though the correct model is on the last line (950 PRO). Also the very first time I partitioned it (GPT) I called the OS partition "Speedy". Then I reinstalled without re-partitioning it (leaving the EFI partition #1 and "erasing" the OS partition #2) and renamed the OS partition. But the last line still calls it "Speedy". When mounted in OS X the OS partition has the correct name. Not really a problem, just unexpected in the output above.

 

[Audiocream]

I was able to install OS X on the SSD 750 of Intel on a ASUS RIVE board! Only the speeds are the same as a regular SSD. But otherwise i am very happy! Any ideas on why the speeds max out on 400mb?
Thanks for the hard work good job!!!!! Be blessed in Jesus name!

 

[Pike R. Alpha]

@jimj740,

I followed your progress some time ago – I'm on holidays right now – and it is good to see that you are pushing things forward, congrats with that, but I also have three simple questions for you:

1.) Did you use anything from http://www.nvmexpress.org/drivers/ for your kext?

2.) Are you going to release/open source your kext?

3.) Is power management supported by your kext?

 

[jimj740]

@jimj740,

I followed your progress some time ago – I'm on holidays right now – and it is good to see that you are pushing things forward, congrats with that, but I also have three simple questions for you:

1.) Did you use anything from http://www.nvmexpress.org/drivers/ for your kext?

2.) Are you going to release/open source your kext?

3.) Is power management supported by your kext?

1) Sort of. Which is to say that the driver uses some FreeBSD source that derives from the Intel reference design, which happens to also be on the nvmexpress website.

2) No current plans, though possibly in the future. There is no GPL'ed code in the driver, so there is no requirement to open source.

3) No. This driver was developed for server usage. Though while idle these devices typically consume little power, implementing power management, particularly for portable devices, would yield greater power savings.

Once again, this driver was developed for internal use and reflects internal requirements - I actively encourage anyone who wants to create alternate drivers to do so.

 

[jimj740]

Two curious things -- in the middle there it says "SM951" which is not correct, though the correct model is on the last line (950 PRO). Also the very first time I partitioned it (GPT) I called the OS partition "Speedy". Then I reinstalled without re-partitioning it (leaving the EFI partition #1 and "erasing" the OS partition #2) and renamed the OS partition. But the last line still calls it "Speedy". When mounted in OS X the OS partition has the correct name. Not really a problem, just unexpected in the output above.

The description is from a lookup of the PCI ID of the controller, whereas the the model on the last line is taken from the ascii string returned by the device. The same controller may be used in multiple models of devices.

The original name of the partition is recorded in the GPT and is not affected by volume rename.

Both are normal/expected behaviors.

-JimJ

 

[Pike R. Alpha]

1) Sort of. Which is to say that the driver uses some FreeBSD source that derives from the Intel reference design, which happens to also be on the nvmexpress website.

2) No current plans, though possibly in the future. There is no GPL'ed code in the driver, so there is no requirement to open source.

3) No. This driver was developed for server usage. Though while idle these devices typically consume little power, implementing power management, particularly for portable devices, would yield greater power savings.

Once again, this driver was developed for internal use and reflects internal requirements - I actively encourage anyone who wants to create alternate drivers to do so.

Thank you. I understand that. And yeah servers don't really need PM support. About the source code; You know. I just had to ask this (as open source developer). Thanks again!

 

[Wernerempire]

I was able to install OS X on the SSD 750 of Intel on a ASUS RIVE board! Only the speeds are the same as a regular SSD. But otherwise i am very happy! Any ideas on why the speeds max out on 400mb?
Thanks for the hard work good job!!!!! Be blessed in Jesus name!

Audiocream, would you be so kind as to share what exact type and model number of "SSD 750 Intel" you are using, and in which model machine (and PCIe slot, if applicable) you were using said device?
What version OS X were you booting?
Thank you kindly.

 

[jimj740]

Here's one more. I rebooted my machine for unrelated reasons, and on startup it said "your machine shut down because of a problem" though I didn't notice any problem...

Ammulder - remote access to my development machines in painfully slow... so I have only a high level view of what is occurring with your system.

[EDIT:] Upon returning from vacation and having direct access to my machines, I discovered that the preliminary analysis I did remotely was absolutely wrong as it was based off disassembly of the wrong version of the driver. Everything stated by me below this point in the thread until post number 191 is wrong and should be disregarded. The driver was NOT in the reset code, and in fact the driver was not returning any error at all. There is no evidence of hardware failure, nor was the driver in a path that would log diagnostics. Sorry for wasting your time.
[End of EDIT]

It appears that you are hitting a device error that is serious enough that the driver is attempting a hard reset, and that there is a bug in the driver's error recovery logic under the particular circumstances of this error condition. It is unfortunate that the console log is lost during the crash; the driver is definitely logging diagnostic info on this path.

Would it be possible to keep the console window open (and filtering on NVMeGeneric) while you run the test and try to capture the output? Note that under certain circumstances the output can either not make it to the console log display, or appear relatively briefly. I have at times resorted to having my phone record the console output as a picture/video to capture the info.

I would also note that although I have done compatibility testing with the 950 series, I have not used it for extended periods. I mention this only because the barefeats article: http://barefeats.com/hard211.html indicates that they observed temperatures outside of specification using this device without a heatsink under load. I have no idea if this is happening with your usage pattern, nor do I have any independent first hand experience with this or the adapter being promoted. I only mention it because it fits the general symptoms: the device is working well for a period and then encounters errors serious enough for the driver to attempt a full reset.

-JimJ

 

[ammulder]

Would it be possible to keep the console window open (and filtering on NVMeGeneric) while you run the test and try to capture the output? Note that under certain circumstances the output can either not make it to the console log display, or appear relatively briefly. I have at times resorted to having my phone record the console output as a picture/video to capture the info.

There was no console output at all in the few seconds immediately preceding the crash -- I tried both filtered to "nvme" and not. I've now run this appium/protractor test 5 times and it's crashed at the same point every single time.

It seems a little unlikely to me that it's an overheating issue. The test in question just does a bunch of JavaScript stuff in the browser on the iOS Simulator. It's not particularly disk-intensive -- not like, say, installing OS X or restoring a backup to the drive, which went fine. Further, the crash happens just after the test has failed and is in the process of shutting down, so any strenuous activity is over. And it's so reproducible, I would have thought an overheating thing was less consistent. Finally, of the thousand things I've done with the system, it's only restarting it and this test/simulator shutdown process that have caused a panic.

I find it more likely that the iOS simulator is making some kind of unexpected call, perhaps related to virtualization or something (I have VT-d disabled, if that matters). But I don't get the crash if I manually quit the simulator, only when the test is wrapping up in its automated fashion. If you have some way to provide a build with gross quantities of debugging, I can try it, and I don't mind video-ing the screen, but at this point it may be after the holidays.

Sadly, I also can't share the app or test in question, though some time in January I can try making a clean test that you could use to reproduce the issue, if it turns out to be related to the combination of technologies I'm using rather than the specific app.

 

[jimj740]

I'm back, and with local access found that my preliminary analysis was entirely wrong as I accidentally disassembled the wrong version of the driver.

All three crashes are the same. The driver is loaded at different addresses in the three crashes, so the crash dumps look different, but the code offsets are identical. The last return address (and hence the last call made by the driver) indicates driver offset 4abc. This is the point in the driver where the driver calls the asynchronous completion routine supplied by the I/O request.

Here is the top of one of your crashes:
Backtrace (CPU 0), Frame : Return Address
0xffffff9126ae3a50 : 0xffffff80094de792
0xffffff9126ae3ad0 : 0xffffff80095d155a
0xffffff9126ae3cb0 : 0xffffff80095ef233
0xffffff9126ae3cd0 : 0xffffff80095c3d61
0xffffff9126ae3de0 : 0xffffff80094feb5c
0xffffff9126ae3e20 : 0xffffff80094fee5e
------ Everything above this is not driver code, it is the completion callback
0xffffff9126ae3e40 : 0xffffff7f8c488abc <---- This is the driver calling the completion routine (offset 4abc)

The actual page fault you are encountering is not in the driver, it is in fact 6 calls deep into the completion routine. There are two possibilities here, one that the driver somehow corrupted memory that the callback routine depends on, or that the call back routine has a logic/timing error.

If this is code that you have developed, perhaps you can instrument it to help determine if the driver is corrupting memory or, (assuming that this runs on other storage devices) whether it is a timing issue.

I will be happy to assist you.

-JimJ

 

[dragsquares]

Running OS 10.11.2. Early 2009 Mac Pro (4,1).

I'm admittedly far from a Terminal genius, so something is evading me here. I've put the .kext in /Library/Extensions, and then attempted to set ownership of the .kext as specified by the command sudo chown –R 0:0 NVMeGeneric.kext

So (again, not knowledgeable) I've tried many things to get the owner to be root:wheel, but I may be misunderstanding something here. If, say, my admin account was ADMIN, should I be entering sudo chown –R ADMIN:wheel NVMeGeneric.kext? Also, is this something that just cannot be done from the Get Info window?

Forgive what may be extremely obvious to a knowledgeable user.

 

[MacVidCards]

Why not post this in an existing thread?

Maybe ask the mods to do so?

 

[dragsquares]

I didn't see an existing thread devoted to installation issues with the NVMe driver. The driver is talked about in other threads but I didn't want to hijack them.

 

[jimj740]

Running OS 10.11.2. Early 2009 Mac Pro (4,1).

I'm admittedly far from a Terminal genius, so something is evading me here. I've put the .kext in /Library/Extensions, and then attempted to set ownership of the .kext as specified by the command sudo chown –R 0:0 NVMeGeneric.kext

So (again, not knowledgeable) I've tried many things to get the owner to be root:wheel, but I may be misunderstanding something here. If, say, my admin account was ADMIN, should I be entering sudo chown –R ADMIN:wheel NVMeGeneric.kext? Also, is this something that just cannot be done from the Get Info window?

Forgive what may be extremely obvious to a knowledgeable user.

I agree this should get moved... but will leave it up to the moderators. The danger is really that I do not see the new thread and am not notified of threads I am not following...

To be clear, it would be helpful to know exactly what is not working. There is no need to substitute the symbolic names for the admin user and wheel group; the numeric value of 0:0 is fine.

There can be several possible problems here...

1) The working directory of your terminal process must contain the NVMeGeneric.kext, or alternately you must specify the full path to the kext (such as sudo chown -R 0:0 /Library/ExtensionsNVMeGeneric.kext since you already put it there)

2) Ideally you want ownership set correctly BEFORE moving the kext to /Library/Extensions... but if you copy it there with incorrect ownership and later fix the ownership, you can recover by issuing the command sudo touch /Library/Extensions which will let the system know to scan for new drivers.

If the two steps above don't fix the issue, what does the following command line return? ls -ld /Library/Extensions/NVMeGeneric.kext

-JimJ

 

[dragsquares]

I agree this should get moved... but will leave it up to the moderators. The danger is really that I do not see the new thread and am not notified of threads I am not following...

My apologies. Thanks for finding it and replying, regardless.

To be clear, it would be helpful to know exactly what is not working. There is no need to substitute the symbolic names for the admin user and wheel group; the numeric value of 0:0 is fine.

There can be several possible problems here...

1) The working directory of your terminal process must contain the NVMeGeneric.kext, or alternately you must specify the full path to the kext (such as sudo chown -R 0:0 /Library/ExtensionsNVMeGeneric.kext since you already put it there)

2) Ideally you want ownership set correctly BEFORE moving the kext to /Library/Extensions... but if you copy it there with incorrect ownership and later fix the ownership, you can recover by issuing the command sudo touch /Library/Extensions which will let the system know to scan for new drivers.

I deleted the .kext from where it had been inside /Extensions, restarted and set ownership on a fresh copy before putting it in place. When I ran the sudo touch /Library/Extensions command I got a notification that the .kext had been improperly installed. When I looked the extensions via the system profiler, it returned the following:

Authentication Errors:

Authentication Failures:

File owner/permissions are incorrect (must be root:wheel, nonwritable by group/other): /Library/Extensions/NVMeGeneric.kext, Contents, _CodeSignature, CodeResources, Info.plist, MacOS, NVMeGeneric, Resources, en.lproj, InfoPlist.strings

Getting info on the file shows my admin account having read/write, wheel having read only and everyone having read-only.

If the two steps above don't fix the issue, what does the following command line return? ls -ld /Library/Extensions/NVMeGeneric.kext

drwxr-xr-x@ 3 (myadminacct) wheel 102 Dec 9 01:19 /Library/Extensions/NVMeGeneric.kext

Thanks again for looking into it.

 

[MacVidCards]

restarted and set ownership on a fresh copy before putting it in place.

The second you moved it you took control thus changing ownership

ie, you broke it

Just follow the instructions as written, don't improvise

 

[dragsquares]

The second you moved it you took control thus changing ownership

ie, you broke it

Just follow the instructions as written, don't improvise

You are absolutely right, I didn't do it to the letter. I had to navigate to the unzipped download and chown on the .kext without moving it first, then drop it in. Now it works. I'll test speed and post an update.

Thanks for everyone's help. This is a very cool thing.
[doublepost=1451978608,1451977007][/doublepost]So the speed I'm getting with the Lycom DT-120 card and the Samsung 950 Pro in Slot 2 (16x)of my 4,1 Mac Pro is around 750MB/s read and slightly slower write. I'm going to try it tomorrow with a BPlus card with a heatsink on it to see if matters improve.

 

[MacVidCards]

To anyone reading this.

When you move a kext to the Extensions folder by hand, YOU were in control, it changes the permissions to YOU owning it, not System.

You are absolutely right, I didn't do it to the letter. I had to navigate to the unzipped download and chown on the .kext without moving it first, then drop it in. Now it works. I'll test speed and post an update.

Thanks for everyone's help. This is a very cool thing.
[doublepost=1451978608,1451977007][/doublepost]So the speed I'm getting with the Lycom DT-120 card and the Samsung 950 Pro in Slot 2 (16x)of my 4,1 Mac Pro is around 750MB/s read and slightly slower write. I'm going to try it tomorrow with a BPlus card with a heatsink on it to see if matters improve.

Please read the existing threads.

Needs to be in slot 3 or 4

 

[dragsquares]

Needs to be in slot 3 or 4

As you said. 1450-sh reads now.