Adaptec AAR-2410SA @ AACRAID crashes on PowerMac-G4 MDD

[legacy]

Code: Select all

insmod  /lib/modules/4.16.0-Fearless-Coyote-mdd-2xG4-dirty/kernel/drivers/scsi/aacraid/aacraid.ko

Code: Select all

[  493.719845] Adaptec aacraid driver 1.2.1[50877]-custom
[  493.719913] aacraid 0001:10:15.0: enabling device (0004 -> 0006)
[  613.726766] AAC0: adapter kernel panic'd fffffffd.
[  673.722761] AAC0: adapter kernel failed to start, init status = 48.

I am on

Code: Select all

# uname -r
4.16.0-Fearless-Coyote-mdd-2xG4-git

(it's the last kernel I got from the git repository)

this kind of bug is also present in these kernels
- kernel-4.9.72-2xG4
- kernel-4.9.90-2xG4

pci=nocrs doesn't help, I have already added it to cmdline, as it was suggested here (in a post written in 2015)

Code: Select all

cat /proc/cmdline
root=/dev/hda4 ro root=/dev/hda4 real_root=/dev/hda4 CONSOLE=/dev/tty1 pci=nocrs

I am using this PCI-X SATA card

Code: Select all

0001:10:15.0 RAID bus controller: Adaptec AAC-RAID (rev 01)
        Subsystem: Adaptec AAR-2410SA PCI SATA 4ch (Jaguar II)
        Flags: 66MHz, slow devsel, IRQ 58
        Memory at 84000000 (32-bit, prefetchable) [size=64M]
        Expansion ROM at 80088000 [disabled] [size=32K]
        Capabilities: [80] Power Management version 2

let me know.
Thanks.

[krinn]

why your command line is showing root=/dev/hda4, aacraid has always use scsi layer, while it might be valid on MAC (no idea) that disks do use old hd* layer, make sure your kernel do have CONFIG_BLK_DEV_SD=y
If people do use nocrs and it has help them (because crs were enable and disabling it has help them), then if we assume accraid is sensible to crs settings, my suggest is try "use_crs" to force them (assuming in your case using crs would help you).

[legacy]

why your command line is showing root=/dev/hda4

the powermac bootstraps for the first pATA harddrive, partition 4 is mounted as rootfs, and the kernel is then loaded and it loads the kernel-module for the AACRAID card

kernel do have CONFIG_BLK_DEV_SD=y

of course, it is

Code: Select all

#
# SCSI support type (disk, tape, CD-ROM)
#
CONFIG_BLK_DEV_SD=y

If people do use nocrs and it has help them (because crs were enable and disabling it has help them)

emm ...

pci=nocrs means the PCI ignores the host bridge windows from ACPI. If you need to use this, please report a bug

anyway, with/out crs doesn't work in both cases.

[krinn]

anyway, with/out crs doesn't work in both cases.

You didn't get it, you don't know the status of crs, people were having crs enable and force it disable with nocrs, so they know the status of crs for them, it's force disable.
But look at use_crs you'll see the help report it "prior 2008"... and a mac G4 was made prior 2008 no?
So it could be for your case that not using crs is the panic (because < 2008) and it then "could" also be that forcing crs may help instead.

I then expect a better answer with "use_crs" doesn't help instead of the blurry "with/out crs".

My next suggest is try a livecd (i think aacraid is nearly always on any) and see if the aacraid panic (to see if nothing could be done except hacking accraid, something for kernel.org bugzilla) or if it is just a kernel option that is missing (which we could help you with then).

[legacy]

mac G4 was made prior 2008 no?

2003.

thanks for the tip, I am going to force it, and see if it solves.

[legacy]

Code: Select all

[  491.162332] Adaptec aacraid driver 1.2.1[50877]-custom
[  491.162402] aacraid 0001:10:15.0: enabling device (0004 -> 0006)
[  611.170736] AAC0: adapter kernel panic'd fffffffd.
[  671.170734] AAC0: adapter kernel failed to start, init status = 48.

Code: Select all

 # cat /proc/cmdline
root=/dev/hda4 ro root=/dev/hda4 real_root=/dev/hda4 CONSOLE=/dev/tty1 pci=use_crs

[legacy]

Code: Select all

CONFIG_PCI_QUIRKS=y

PCI_QUIRKS is also set

insmod takes a lot of time, 2-3 minutes, then it crashes
on dmesg i see

immediately:

Code: Select all

[  491.162332] Adaptec aacraid driver 1.2.1[50877]-custom 
[  491.162402] aacraid 0001:10:15.0: enabling device (0004 -> 0006) 

2-3 minutes later ...

Code: Select all

[  611.170736] AAC0: adapter kernel panic'd fffffffd. 
[  671.170734] AAC0: adapter kernel failed to start, init status = 48. 

[krinn]

Try some livecd to see if the driver still panic (if really you need, i could provide my kernel config as i'm using aacraid too)

[legacy]

i'm using aacraid

which card, on which machine?

[legacy]

as far as I have understood, it's an (intel) i960 based AAC card.

[krinn]

I don't think it will help you, but here are the infos:

Code: Select all

05:00.0 RAID bus controller: Adaptec AAC-RAID (rev 09)
Subsystem: Adaptec 5405
[    0.198420] Adaptec aacraid driver 1.2.1[50834]-custom
[    0.198779] aacraid: Comm Interface enabled
[    0.407645] aacraid 0000:05:00.0: 64 Bit DAC enabled
[    0.417965] scsi host0: aacraid
Portage 2.3.24 (python 2.7.14-final-0, default/linux/amd64/17.0, gcc-7.3.0, glibc-2.26-r6, 4.14.11 x86_64)
=================================================================
System uname: Linux-4.14.11-x86_64-Intel-R-_Core-TM-_i7-4790K_CPU_@_4.00GHz-with-gentoo-2.4.1

[legacy]

You are running it on a x86 machine, where there is a PC-BIOS
it's a completely different matter.

[legacy]

it's an intel i960 based AAC card

giving a name to called functions inside the kernel module:

Code: Select all

Adaptec aacraid driver 1.2.1[50877]-custom
aac_rx_init initialize
aac_rx_init Check to see if the board panic'd while booting
aac_rx_init says the board has no panic while booting
aac_rx_init AAC0: adapter self-test OK
aac_rx_init AAC0: adapter monitor p
aac_rx_init Wait for the adapter to be up and running
aac_rx_init  Wait up to 3 minutes
aac_rx_restart_adapter AAC0: adapter kernel panic'd fffffffd.
aac_rx_init AAC0: adapter kernel failed to start, init status = 48.
aac_rx_init -> error_iounmap

[krinn]

You mean you expect someone using gentoo, with an AAR-2410SA card and a g4 coming to help you?
Well, better take lottery ticket no?

[legacy]

better fill a bug report to bugzilla, and I have already done it.

[krinn]

totally agree, looks like the best to do, maybe post the link for other?

[NeddySeagoon]

legacy,

root=/dev/hda4 with a 4.16.0 or even a 3.x kernel is bound to fail.

The drivers under

Code: Select all

< > ATA/ATAPI/MFM/RLL support (DEPRECATED)  ----

are no longer supported by udev and friends. At one time, /dev/hda* nodes were not created.

IDE disk drivers are under the

Code: Select all

<*> Serial ATA and Parallel ATA drivers (libata)  --->

Its still the IDE you are used to but you will get SCSI names ... /dev/sd*

Your

Code: Select all

Adaptec AAC-RAID

is a real SCSI device.

I suspect that the old IDE drivers on the < > ATA/ATAPI/MFM/RLL support (DEPRECATED) do not play nicely with the rest of the drivers.

[legacy]

IDE disk drivers are under the

Code: Select all

<*> Serial ATA and Parallel ATA drivers (libata)  --->

Its still the IDE you are used to but you will get SCSI names ... /dev/sd*

Where do you see support for the built-in pATA controller of the PowerMac-G4/MDD?

Your

Code: Select all

Adaptec AAC-RAID

is a real SCSI device.

indeed the AAC-RAID driver is located in the SCSI session of the kernel tree and has nothing to deal with libSATA; but the problem is different: AAC-RAID is based on i960 and it requires an x86 system in order to be properly initialized since it tries to emulate DMA-bus muster.

in short: it's an x86-only device.

[legacy]

Someone has ever used it?

[NeddySeagoon]

legacy,

I'm a bit reluctant to run

Code: Select all

ARCH=powerpc make menuconfig

on my amd64 as I just know I'll trash my .config.
Here's the next best thing.
I go into the

Code: Select all

Serial ATA and Parallel ATA drivers (libata)

menu and press 'z'
That shows me all the hidden menu items, including

Code: Select all

  ┌─────────────── Apple PowerMac/PowerBook internal 'MacIO' IDE ────────────────┐
  │ CONFIG_PATA_MACIO:                                                           │  
  │                                                                              │  
  │ Most IDE capable PowerMacs have IDE busses driven by a variant               │  
  │ of this controller which is part of the Apple chipset used on                │  
  │ most PowerMac models. Some models have multiple busses using                 │  
  │ different chipsets, though generally, MacIO is one of them. 
...
  │ Prompt: Apple PowerMac/PowerBook internal 'MacIO' IDE                        │  
  │   Location:                                                                  │  
  │     -> Device Drivers                                                        │  
  │       -> Serial ATA and Parallel ATA drivers (libata) (ATA [=y])             │  
  │         -> ATA SFF support (for legacy IDE and PATA) (ATA_SFF [=y])          │  
  │           -> ATA BMDMA support (ATA_BMDMA [=y])

Is that what you have?

You may well be able to use the Adaptec AAC-RAID card but not boot from it.

[Hu]

I seem to recall that you can override the name of the config to use. Looking in the source, it looks like setting $KCONFIG_CONFIG chooses the filename, with .config being the default if none is set. I have not tested this, so you should still archive a good .config before trying it. If it works, it would let you experiment for other people without overwriting your own.

[NeddySeagoon]

Hu,

All sound advice. I keep several kernel tree around too.

[legacy]

drivers/ata/pata_macio doesn't work for my MDD!

I have already tried it on kernel-4.16.0
lib_sata: PATA_MACIO is not set -> not in use, doesn't work
old_ata: BLK_DEV_IDE_PMAC -> currently in use, it works

You may well be able to use the Adaptec AAC-RAID card

AAC-RAID doesn't work and it doesn't depend on that: if you look at its driver, you see it requires an interaction with the BIOS, and it's all x86 firmware that can't be executed in a non-x86 machine. The kernel driver doesn't reinitialize the chip, it expects it already initialized by the PC-BIOS, and that can't happen on a PowerMAC since there is no interaction with any PC-BIOS-extension (this code is stored in the flash of the AAC-RAID card and it's probed and executed during the earlier step of the PC-bootrap, in fact, a menu appears on the PC's screen, and this is screen says " booting i960 chip on the AAC-RAID card" .. but this can't happen on a PowerMac), thus the kernel driver simply invokes a wait_until_time_out, it waits for the bootstrap of the i960 chip on the card, that doesn't happen, and it crashes.

It takes 3 minutes. That is exactly the timeout set in the while(1) {} loop.

Nothing of this process has anything to deal with the old_ata or lib_sata library in use, also you can have the pata_library for the built-in pata-controller without any issue if you also use the lib_sata for a SIL24 s/ATA controller: it works here pretty good!

but not boot from it.

I bootstrap from the lan.

[NeddySeagoon]

legacy,

I understand most of that. What got me going down the IDE rabbit hole was your

Code: Select all

root=/dev/hda4

You can have root on an IDE drive but not that way. Userspace has been gutted. The kernel will find and mount root but userspace will not be aware of /dev nodes with the old IDE names.

While that may be interesting, its not as you say, related to getting your AAR-2410SA card going on PowerPC.

There is a aacraid_linux_drivers_ppc_v1.2.1-52011.tgz advertised. However, it only mentions Adaptec RAID 81605Z Power as a compatible product. That's not to say its not backwards compatible with earlier cards too.

[legacy]

I understand most of that.

I don't think you understand HOW linux works

You can have root on an IDE drive but not that way. Userspace has been gutted.

Userspace-what? I am not using udev, in first place, and /dev/ is populated by mknod. Static devnames.
And what does this has to deal with a kernel module that explicitly reports a failure during the probe-step?

The kernel will find and mount root but userspace will not be aware of /dev nodes with the old IDE names.

For which reason on Earth? Black magic or what?

There is a aacraid_linux_drivers_ppc_v1.2.1-52011.tgz advertised. However, it only mentions Adaptec RAID 81605Z Power as a compatible product. That's not to say its not backwards compatible with earlier cards too.

Judging by what I am observing: well, there are also sources from Adaptec, but they don't add anything, thus probably there are/were (at least on the paper) versions of this card with a different firmware on the card that pre-initializes the i960-processor without the need of passing through the PC-BIOS. Those drivers are pure junk from my point of view since they simply wait for the i960-built-in controller booting. And this can't happen since the firmware can do nothing without the help of a PC-BIOS: all the booting and setting parts inside the flash card are all made with x86 code, and you will end with the silly need of an x86 emulator to initialize a sATA card: simply insane!

Anyway, I have already sent a few emails to Adaptec and they are not collaborating (as usual), but yesterday I a reply by Alan Cox who says ... "no dice".

In short: he is right, you'd best forget this card, it's pure junk. Time and money wasted on a non-x86 machine.

That said, I have replaced the card with a SIL24-SATA (syba re-branded) and it's fine without any need to modify anything in my configuration!