Zotac CI323 Installation - Controller Failures

[ac0hen]

Greetings!

I've been trying to install pfsense on a new ZBOX CI323 for a few days now.  I've got 8 Gig of Ram and a 120GB SSD.

Installing from the memstick image.

I get the same controller failures others have noted in earlier threads, and I've followed suggestions to pre-format the SSD using windows diskmanager before proceeding. 
I've also used Rufus to create partitions prior to install. 
I've tried the disk connected to SATA, and via USB 3.0.
I've re-imaged the USB installer, and tried a different USB stick.
I've imaged the install volume right onto the SSD (stupid idea, but thought I'd give it a shot)

The memstick boots fine, until I get a controller timeout and registry dump.

sdhci_pci0-slot0: Controller timeout

That happens a few times, and the boot process moves on and the installer runs.

I choose the simple install.

When the installer attempts a "bsdlabel -B -r -w ada0s1 auto" command, it stops at 77% (every time).  Then a "ahcich1" Timeout on slot 21 port 0" and "(ada0:ahcich1:0:0:0):  CAM status: Command timeout"

It seems like many people have been successful with this box, so perhaps I just got a bad one.

Let me know what other data I can provide.  The good folks at Zotac aren't interested in helping with "open source" software issues.

Thanks in advance, and my apologies for my BSD deficiencies.

Art

[dreamslacker]

Disable the SD card reader in BIOS.

SDHCI is the integrated card reader.

[IggyB]

@dreamslacker:

Disable the SD card reader in BIOS.

SDHCI is the integrated card reader.

No such thing in the BIOS. Is there a way to unload sdhci_pci0-slot0 controller module without recompiling?

[IggyB]

Only way to fix this is build a custom kernel.

Adding to loader.conf

mmc_load="NO"
  mmcsd_load="NO"
  sdhci_load="NO"

does not fix the problem.

neither does adding

hint.sdhci.0.disabled="1"
hint.sdhci.disabled="1"
hint.sdhci_pci0.disabled="1"
hint.sdhci_pci0.0.disabled="1"
hint.sdhci.0.disabled="YES"
hint.sdhci.disabled="YES"
hint.sdhci_pci0.disabled="YES"

to device.hints

[messerchmidt]

update to the latest bios and see

otherwise, run pfsense in a vm on the ci323:http://mobiletiger.jorba.de/vmware-esxi-6-0-n3150-itx-intel-celeron-braswell-platform-problem-solved/

[duren]

Just tried the latest beta of pfSense 2.4 / FreeBSD 11 and the timeouts have gotten worse.

Instead of there being 4, it's more like 10 now before it continues booting

I was hoping with 2.4, they would have disappeared :/

[grokdesigns]

@duren:

Just tried the latest beta of pfSense 2.4 / FreeBSD 11 and the timeouts have gotten worse.

Instead of there being 4, it's more like 10 now before it continues booting

I was hoping with 2.4, they would have disappeared :/

Not to resurrect an older thread, but I finally came across a solution that fixed this on 2.4 and wanted to document it here since this thread comes up when Googling this issue. Adding hw.sdhci.enable_msi=0 to /boot/loader.conf and hint.sdhci_pci.0.disabled="1" to /boot/device.hints got rid of the controller timeouts on boot in 2.4. I didn't try these directives one at a time, so you may not need both (another user said the line in device.hints didn't work, so maybe you only need the loader.conf one), but having both doesn't seem to cause any problems.

Edit: After applying an update, my device.hints file got reset and the timeouts came back, so that's definitely the one that needs to be added.

[Making_sense_of_pfSense]

@grokdesigns:

[…] Adding […] hint.sdhci_pci.0.disabled="1" to /boot/device.hints got rid of the controller timeouts on boot in 2.4.

Thank you, you just helped me fix the same problem on my Zotac CI323 box!

[IggyB]

Thanks
I can confirm this fix works on 2.3.4

[stephenw10]

Either of those files could be overwritten. Custom loader settings like that should go in /boot/loader.conf.local, create it if it doesn't exist.

I can see no reason why both those setting wouldn't work in that file but it would be good to hear from someone with access to this hardware to confirm it.

Steve

[dlm274]

just to document, because i found this thread on google searching for the same problem:

create the file /boot/loader.conf.local with following content:

hw.sdhci.enable_msi=0
hint.sdhci_pci.0.disabled=1
hint.sdhci_pci.1.disabled=1

That's update safe (tested on my own machine from today) and the second hint.sdhci is needed for the ZBOX CI327.

Now everything works like a charm on the pfsense with this little beast. Tested with 500MBit uplink and OpenVPN (AES CPU-Hardwareaccellerator enabled) - no limitations.

73 dlm274

[mukpfsense]

@dlm274 said in Zotac CI323 Installation - Controller Failures:

hw.sdhci.enable_msi=0
hint.sdhci_pci.0.disabled=1
hint.sdhci_pci.1.disabled=1

Well, I tried setting these at option 3 at pfsense boot (ESC ) to loader prompt. No change. It still hangs at the same spot.

[mukpfsense]

@dlm274 said in Zotac CI323 Installation - Controller Failures:

just to document, because i found this thread on google searching for the same problem:
create the file /boot/loader.conf.local with following content:

hw.sdhci.enable_msi=0
hint.sdhci_pci.0.disabled=1
hint.sdhci_pci.1.disabled=1

Also probably important, this is happening on a MINIX NEO Z83-4

[stephenw10]

@ac0hen said in Zotac CI323 Installation - Controller Failures:

bsdlabel -B -r -w ada0s1 auto" command, it stops at 77% (every time).  Then a "ahcich1" Timeout on slot 21 port 0" and "(ada0:ahcich1:0:0:0):  CAM status: Command timeout

With that same error?

[drzow]

I am having the same issue on initial boot on a Minis / Besstar Z83-F, which is another one of these tiny Intel Cherry Trail boxes. Trying to do an initial install, I am using a memstick, option 3 to launch the boot loader, then I type in:

set hw.sdhci.enable_msi=0
set hint.sdhci_pci.0.disabled=1
set hint.sdhci_pci.1.disabled=1
boot

and the kernel goes on its merry way, but I still end up hanging after the ppc0: cannot reserve I/O port range -- never actually get to the installer. Am I doing something wrong with setting the parameters?
Thanks!
Terry

[stephenw10]

The ppc0 (parallel port) error is common on many systems and should not prevent it booting.

What error are you seeing that makes you think you need to disable sdhci?

Steve

[drzow]

Thanks Steve! Agreed that the ppc0 error isn't the actual problem, but rather the hang after it, which you thought was sdhci related in this other thread. It seems to be endemic to these Cherry Trail boards. Some other forums also indicated that it's the eMMC, which if memory serves is the same thing as sdhci.
Thanks!
Terry

[stephenw10]

Hmm, I thought there was a thread for the UP board here. It's probably worth looking through that if you have not already.

Edit: This maybe? https://forum.netgate.com/post/743676

Steve

[drzow]

Ah-ha! Gave that a shot, but the Z83-F BIOS is just different enough from the UP boards that I couldn't get it to work. I was, however, able to uncover another lead that led me to this FreeBSD forum post, where it discusses the UARTs being the problem, so I:

set hint.uart.0.disabled="1"
set hint.uart.1.disabled="1"

and the kernel now boots and the installer is running!!!

I really appreciate your persistence on this Steve!

Terry

[Orbixx]

@dlm274 said in Zotac CI323 Installation - Controller Failures:

just to document, because i found this thread on google searching for the same problem:

create the file /boot/loader.conf.local with following content:

hw.sdhci.enable_msi=0
hint.sdhci_pci.0.disabled=1
hint.sdhci_pci.1.disabled=1

That's update safe (tested on my own machine from today) and the second hint.sdhci is needed for the ZBOX CI327.

Now everything works like a charm on the pfsense with this little beast. Tested with 500MBit uplink and OpenVPN (AES CPU-Hardwareaccellerator enabled) - no limitations.

73 dlm274

Reviving the topic again one year later!

I was installing on a new Fitlet 2 recently as my old home pfSense device started powering off when it hit around the 45 degree Celsius temperature mark. I encountered the same problem with the SD card causing command timeouts and the above boot params solved the problem.

Netgate, do the contents of /boot/loader.conf.local get processed in the pfSense config backup/restore function? Or would I have to reconfigure this again manually if say my disk died and I had to replace it?

I assume the root cause of this issue would have to be handled upstream in FreeBSD? Doesn't seem specific to pfSense. Did anybody submit a bug report?