Workaround/Semi-Fix for Mountroot Issues with 9.3!

Status
Not open for further replies.

9C1 Newbee

Patron
Joined
Oct 9, 2012
Messages
485
I think I see what Big is getting at.

If it is a timing issue, are the 2.0 drivers being loaded sooner than the 3.0 drivers? If so, would a delay of 90 sec...120 sec...180 sec make them work? If not, then would it rule out the "timing" hypothesis (for the 3.0 side of things at least).

Has anyone booted 9.3 via 3.0 USB port successfully?
 

Rich N

Cadet
Joined
Feb 10, 2014
Messages
6
I dont have usb 3.0 but still require the turntable to boot.
 

eire1274

Dabbler
Joined
Jan 1, 2015
Messages
13
This workaround is completely useless for my issue regarding the new GPT format.

I am using an Asus AM1M-A motherboard with an AMD Kabini 5350 SOC, booting from a SSD on the built-in SATA controller via the AMD SOC (all other drives are hosted on an eSATA controller card). The UEFI absolutely disregards the SSD (not seeing a MBR boot partition?) and dumps me back into UEFI setup. I have tried the default (Legacy and UEFI auto), Legacy only, and UEFI only, and in every case the drive is never addressed to boot.

I am incredibly frustrated because this is a BRAND NEW server on BRAND NEW parts. If I would have known that FreeNAS was intent on making this change, I would have chosen other parts.

I am trying to write a custom ISOLinux boot CD as a workaround, as a simple "boot this SSD only" method, and will address if I can come up with a way to properly boot. This server is in a cabinet in an isolated area, and running down to it for every reboot would kill me. I'm not having great luck at the moment, as I'm looking at 17 drive interrupts and if my brain doesn't settle down I may just burn 17 discs and see which one boots.
 

cyberjock

Inactive Account
Joined
Mar 25, 2012
Messages
19,526
I can understand your frustration, but...

1. We've made it pretty clear that AMD builds aren't exactly something to hang your hat on. Regardless of what you think, the CTO of iX has made it 100% clear that this is NOT recommended because it may or may not work today and tomorrow.
2. Your hardware manufacturer isn't exactly on our "recommended list". Asus makes great desktop boards. But for a server, that's a fail waiting to happen.
3. That board was $36 from Newegg. That's not exactly high-end, not exactly the definition of reliable or well supported by any stretch of the imagination. We've told people that even with $70 boards you are taking risks because you're getting what you paid for.

Even the forum mod that uses AMD had some problems with 9.3. In fact there were lots of problems with AMD and I wasn't sure if 9.3 was going to be the release that kills AMD for FreeNAS forever or not. I'm kind of expecting that someday there will be some update that breaks AMD in a fatal way. AMD sells CPUs and motherboard manufacturers are on a race to the bottom to sell their boards. That race to the bottom comes at a price... and as the customer you'll pay that price if you don't plan ahead.

Do I feel bad for you? Yep.

Do I think your problems are also self-inflicted wounds? I do.

Does my opinion really matter when the facts of the matter are so obvious? Not really.
 

eire1274

Dabbler
Joined
Jan 1, 2015
Messages
13
Not to say anything critical, but FreeNAS has a huge following for the low-power, cheap systems. This machine is primarily used as a file backup for my photography, a lite web server, and a Plex Media server, and functionally it works PERFECTLY on 9.2.1.9. Have yet to see the system peak over 25% CPU utilization even with multiple transcoding sessions running. Investing in far more expensive components makes absolutely no sense to me, and the only reason I chose the Asus board is because it was verified to be ECC-U compliant (as are all of the Kabini/AM1 APUs).

Why you think FreeNAS is going to break AMD makes absolutely no sense to me, as they are fully compliant with FreeBSD. I respected FreeNAS for being on such a wide platform, as BSD is.

It makes no difference. I'll sit on 9.2.1.9 on my hardware until I find another solution (other than FreeNAS) that will better support me. I prefer to invest my money in memory and drives, not on Intel's over-inflated price structure.

Oh, BTW, with ISOLinux it boots fine. Just need to know the interrupt of the boot drive. The issue is that Asus' UEFI is looking for the boot flag on the partition, and is not seeing it. Doing work that it doesn't need to do. Get it out of the way and it's fine. I want to point out that this is at no level an AMD issue, it is an Asus UEFI software issue. Get around that and FreeNAS works like it was built for the hardware.
 

derekzchu

Dabbler
Joined
Dec 5, 2014
Messages
23
I think I see what Big is getting at.

If it is a timing issue, are the 2.0 drivers being loaded sooner than the 3.0 drivers? If so, would a delay of 90 sec...120 sec...180 sec make them work? If not, then would it rule out the "timing" hypothesis (for the 3.0 side of things at least).

Has anyone booted 9.3 via 3.0 USB port successfully?

after installing 9.3 stable via USB2.0, I set the tunable for USB3.0 and moved my bootable USB to 3.0 and it booted up fine.
 

9C1 Newbee

Patron
Joined
Oct 9, 2012
Messages
485
There are LOADS of misleading information out on the interwebs about FreeNAS. I just cringe when I see "turn that old 486 into a file slinging machine with FreeNAS" kinda stuff. If you are unwilling to get proper hardware, FreeNAS is not for you. If I were were you I would look at other options now while the switch is simple. But hell, what do I know.
 

danb35

Hall of Famer
Joined
Aug 16, 2011
Messages
15,455
Not to say anything critical, but FreeNAS has a huge following for the low-power, cheap systems.
...and this is the problem. The product now called FreeNAS was never designed or intended for this use case. There's another product that used to be called FreeNAS. It would run from a 256-MB USB stick on whatever old hardware you had laying around. Development on that product went dormant, iXSystems bought its name, and built a whole new product on that name. The only similarities between the old FreeNAS and the new FreeNAS are that they both run on FreeBSD, and they're both NAS OS distributions. The products are otherwise completely unrelated. Then the old project revived and changed its name to NAS4Free (www.nas4free.org, not www.nas4free.com--the latter redirects to freenas.org).

Unfortunately, iXSystems' decision to repurpose the name of an existing, popular project to a new product with very different goals has resulted in a lot of confusion. The new product is great; the confusion not so much.
 

jgreco

Resident Grinch
Joined
May 29, 2011
Messages
18,681
I don't know that they're "very different goals". I think FreeNAS .7 was always conflicted because it wanted to make ZFS viable on early-to-mid nought'ies era random hardware and recycled drives, but anyone who has worked with this stuff for awhile can recognize that ZFS is oriented more towards being a larger software SAN or NAS system. Providing a supportable system that can reliably do GUI-oriented tasks like disk replacement means that you need defined structure; the anything-goes cowboy sysadmin'ing often required on a FreeNAS .7 box is a result of the failure to provide structure.

I'd suggest that iXsystems looked long and hard at both the successes and - more importantly - the weaknesses and failures of FreeNAS .7. I got *much* more active in the FreeNAS community the instant it was clear that they actually wanted to make a real appliance, because before that, if I was going to have to do most of my crap on FreeNAS .7 from the CLI anyways, it was just as easy to make a FreeBSD filer that met local design criteria as well.

FreeNAS 8.0 was kind of a bridge in that it actually did support a lot of the same hardware platforms that FreeNAS .7 did, but it is no longer 2010 and the platform has matured into a true appliance. Ironically, NAS4Free has followed in a somewhat similar manner, in that it is now preferred to use its "embedded" install, which shows that the choices iX made to appliance-ify it, which I agree was unpopular with many in the community, are actually part of the natural evolutionary process of a NAS platform.
 

9C1 Newbee

Patron
Joined
Oct 9, 2012
Messages
485
What has impressed me about FreeNAS, if I have a problem, I can talk directly with a developer. That is really really cool. Support for any product nowadays is next to nonexistent.
 

cyberjock

Inactive Account
Joined
Mar 25, 2012
Messages
19,526
He says that with the same hardware but different USB stick it boots fine, but with the Sandisk it doesn't. Sounds like the Sandisk is broken to me. No?
 

derekzchu

Dabbler
Joined
Dec 5, 2014
Messages
23
works for me running Freenas 9.3 on a Sandisk Fit 8GB. Once i moved it to USB2.0 port, it was fine.
 

Apollo

Wizard
Joined
Jun 13, 2013
Messages
1,449
This timing thing makes me wonder whether the USB 3.0 controller wouldn't be in fact part of a USB hub. I think depending on how the IC is wired on the board, it could act as a stand alone unit or would wait for the CPU takes control of the communication, meaning setting the hub to behave a specific way. In a standalone configuration, there are timing requirement that needs to be met after a hub reset for which the I2C or SPI interface can succesfully establish communication with he hub. Outside of this time window, the device becomes autonomous and communication is performed across a different high speed link. I don't remeber on top of my head the part number for the hub in question.
For the folks working on the driver side, does it sound somewhat familiar?
As for the comment regarding delays in driver/hardware handshacking, well as an Electronic Engineer, I find this to me a big non-sense. I agree there are some minimum timing requirement that needs to be met, for instance PCI express requires PCI express devices to be ready something like 100ms after reset has been deasserted. If timing is not met the PCI express may not be available.
The better approch in system design is based on feedback theory, and in the digital domain the CPU during peripheral enumeration, or discovery will know what device it should expect. This is I think part of the BIOS, if not it will be the driver, or application to establish device stage.
The basic for handshaking would be somewhat similar to checking if a device such as USB key has been inserted, by the generation of an interrupt (possibly), or trough data polling. If a device as been detected, then it is a matter for the system to establish communication by monitoring specific registers in the controller. Going this route will guarantee proper sequencing irrelevant of the amount of time it has to wait. In the event no errors has been flagged by the controller (through polling or interrupt) and let's say the device died, then only a watchdog would allow the controller to be reset to exit the failed state.
I don't want to insult any unix like programmers, but to me using timing as a pass/fail condition to execute the next piece of code is simply in my opinion a bad programming practice. Inherently, this sounds like a solution to a problem or process that is not understood.
 

mark_

Cadet
Joined
Oct 4, 2014
Messages
6
This does not work for me on an Asrock C2750D4I.
Last working version: FreeNAS 9.3 STABLE 201412240734
Not working versions: several after that, e.g. 201501301837 which I tried right now
 

cyberjock

Inactive Account
Joined
Mar 25, 2012
Messages
19,526
This does not work for me on an Asrock C2750D4I.
Last working version: FreeNAS 9.3 STABLE 201412240734
Not working versions: several after that, e.g. 201501301837 which I tried right now

That's interesting. I'm running that board and having zero problems with 201501301837... I'm using a SATA DOM though.

Do you have the latest BIOS installed?

So what USB port are you using (front/back/top/bottom) and what USB device are you using?
 

mark_

Cadet
Joined
Oct 4, 2014
Messages
6
I have the current BIOS (2.80) and BMC (0.23) installed. I installed FreeNAS on two identical USB sticks, inserted on the backside. This worked with the FreeNAS 9.3 versions since I got this hardware in November until I now try to upgrade further.
Which SATA DOM do you have? I want to get rid of the USB sticks because they sometimes degrade. I think that they are too unreliable.
 

DrKK

FreeNAS Generalissimo
Joined
Oct 15, 2013
Messages
3,630
what makes it more "interesting" is we probably have over a thousand users with that board, none of whom have reported this.
 

jgreco

Resident Grinch
Joined
May 29, 2011
Messages
18,681
I don't want to insult any unix like programmers, but to me using timing as a pass/fail condition to execute the next piece of code is simply in my opinion a bad programming practice. Inherently, this sounds like a solution to a problem or process that is not understood.

Congrats, you've discovered a fundamental flaw of PC's ... hardware design by morons, idiots, and arseholes intent on putting out poorly designed CHEAP hardware which require botchy software hacks.
 
Status
Not open for further replies.
Top