Acer Aspire 4720, ACPI crashes

[Doug Bissett]

From: Doug Bissett

eComStation 2.0 RC4 installer

Message 1: I have a new Acer Aspire 4720 laptop. The general description is: Intel core duo T5450 2 GiB memory 250 GiB HDD other things, that seem to be irrelevant, at the moment.

If I leave ACPI enabled, I get a crash in ACPI. See the photo AcerTrap.jpg for the details. If I disable ACPI, and edit CONFIG.SYS to add a fifth USB controller (see other report), I can get booted to the GUI.

Message 2:

Okay, I got this to (partly) work. It seems that this laptop has an invalid ACPI table, but I got the system booted, in 2 processor mode, by using the line: PSD=ACPI.PSD /SMP /PIC /EIS /FS /SAP:0x20 I am not sure if /FS is actually needed. The ACPI wizard told me to use PSD=ACPI.PSD /SMP /APIC which didn't work. /APIC also doesn't work.

I still get a hang, if either of the NICS is enabled, but that could be because of the invalid table. More to be done, when I find a few minutes.

source: ​http://bugs.ecomstation.nl/view.php?id=1965

[Chuck McKinnis]

Okay, here is an update:

I have updated to ACPI 3.07.

/FS is not required.

The NICs are not assigned an IRQ in windows Vista, and it appears that that holds true in eCS too. However, after getting ACPI to load properly, I am now seeing both NICs get an IRQ attached (PCI.EXE output). I still have a problem, because the wired NIC (Broadcom BCM5787M) freezes when the B57.OS2 driver is loaded (it is the one that has been updated for ACPI). The driver for the wirelss NIC, (Intel 3945 a/b/g, using GENMAC 2.2 for 8086:4222) will now go past the driver loading, but it still won't work. Hardware manager, PCI.EXE, and RMView, all show it as attached to IRQ11, but when I do COPY WRND32$ CON, it says that the InterruptLevel is zero, indicating that the driver picked up the IRQ, before the device had one assigned. The system freezes, solid, as soon as anything attempts to communicate with the driver, which seems to be the same thing that happens with the other NIC (which shows as having IRQ10).

This machine has almost nothing useful in the BIOS settings, so it would seem that it is up to ACPI to get it all set up properly. I will try to assemble the logs, and attach them.

[eco]

Please collect ACPI logs, so we can see what happens inside the black box ​http://ecomstation.ru/projects/acpitools/?action=logs

first of all, call: copy acpica$ acpi.log

[Chuck McKinnis]

Log files, as collected by the acpo-logs-collector-20080219.zip package.

[Chuck McKinnis]

e-co:

ACPI for eComStation doesn't support Acer notebooks well. Acer is collaborating with Microsoft, they have different ACPI machine. ACPI for eCS is based on Intel machine.

So, the trap occurs inside ACPI Intel machine (?)

[Chuck McKinnis]

From: Dmitry Froloff

Hello Doug

I made a quick (very quick look :) ) into your DSDT ACPI table. It seems that your problem with trap inside ACPI PSD, is totally the same, as I had with NEC Versa laptop.

At that time, after some tests, I finally found that the trap in ACPI took place during executing CondRefOf() ACPI instruction. This fact was reported to ACPI developer in September 2007, but seems that it is not fixed yet. So ATM the only solution is to omit ACPI initialization (/SAP:0x20 ACPI.PSD switch) which, for sure, will cause other following disadvantages.

[Chuck McKinnis]

Thanks for looking. This is basically an Intel machine, so I don't know why they would use an alternate ACPI setup (but what do I know?). Anyway, more updates:

I have now updated to ACPI 3.08. No change.

Doing more research on the Vista support, for the two NICs, it appears that they are using something called MSI (Message Signaled Interrupt), rather than the older IRQ method (I would suspect that this really means "Microsoft winInterrupts"). The logs show this as being turned off, when booted to eCS, and that is supposed to trigger the use of the older IRQ method, but it doesn't work. I am beginning to suspect that the wires to connect the IRQ lines, just don't exist. I can get both NICs to work with Ubuntu Linux, but I haven't found any place where that tells me what interrupt method is being used. I guess I need to dig deeper into that, to see what is done.

[Chuck McKinnis]

Another update...

I got the wired NIC (a Broadcom BCM5787M) to work. The method to do that is:

Remove ACPI from CONFIG.SYS

Install GENMAC 2.2, using the method to install the driver for uninstalled hardware, fo the 14E4:167D chip set.

Copy the WRAPPER_14E4_167D directory to WRAPPER_14E4_1693

Copy the \IBMCOM\MACS\W14E4x167D.NIF file to W14E4x1693.NIF Do the appropriate modifications to the NIF file (including changing OPTIONS to "UNSUPPORTED"), to match the 14E4:1693 chip set.

Update the b57xp32.sys file, to the latest available from the Broadcom web site

Use MPTS to install the NIC, and now it works.

2 processor mode is enabled, by using OS2APIC.PSD, in place of ACAPI.PSD

More work is required, to make sure that everything else still works, but this is looking better.

If I can supply any information, to help fix ACPI, please ask.

[Chuck McKinnis]

I believe, that incident #233 is related to this.

[Chuck McKinnis]

ACPI13042008, with no parms on ACPI.PSD line

[Chuck McKinnis]

ACPI13042008, with only /SMP parm

[Chuck McKinnis]

ACPI13042008, with only /SMP parm

[Chuck McKinnis]

ACPI13042008, with /SMP /APIC parms

[Chuck McKinnis]

Update: I tried the ACPI13042008 version, as requested by Eugene Gorbunoff, Roderick Klein, and Dimitry Frolof. The results were interesting:

New install of eCS 2.0 RC4, with the contents of the new ACPI version inserted.

With no parameters on the ACPI.PSD line, and with both APM.ADD, and ACPIDAMN.EXE, REMed, I was able to boot the system, using the SMP kernel. The logs in the file acpi-log-Model-of-your-PC.ziq, contains the results of collecting logs.

If I add the "/SMP" parameter, it boots, and the logs are collected in the file s_acpi-log-Model-of-your-PC.ziq.

If I also add the "/APIC" parameter, it hangs with the screen showing the contents of smpapic.jpg.

I hope this helps...

[Chuck McKinnis]

Update: I took the REM off of APM.ADD, and ACPIDAMN.EXE, and I am able to boot. The battery widget also shows battery status, apparently correctly, and the Power object shows APM 1.2 as active.

I tried to install GENMAC, for the Broadcom NIC, and I get a repeating Trap 0008, somewhere near the time when the driver loads.

I tried the (modified for ACPI) B57.OS2 driver, and it hangs after it displays two lines from the driver.

I tried installing the GENMAC driver for the Intel wireless (8086:4222) NIC. The system loads, but doing COPY WRND32$ CON, shows the InterruptLevel to be 0 (zero), which is what I was seeing before. Any attempt to enable the wireless, results in a hard hang.

I tried adding "/CD", but that made no difference.

[Chuck McKinnis]

More progress... I have installed ACPI version 3.09. Now, I can boot the system in APIC mode, with the parameters: /SMP /APIC /CD "/CD" never helped before, and it seems that I no longer need "/EIS". I am still unable to use either NIC, with ACPI, but I haven't tried every possible combination yet. Sound, using Paul Smedley's UniAud 1.92, seems to be working better too.

[eco]

What is the problem with NIC if loaded ACPI?

show us pci.exe output

lantran.log

did you tried patched drivers? ​http://ecomstation.ru/projects/acpitools/?action=acpi-smp-drv

[Chuck McKinnis]

When I use ACPI, both NICs show as having an IRQ, in RMVIEW, Hardware Manager, and PCI.EXE. I am attempting to use the GENMAC 2.2 driver with them both, but doing COPY WRND32$ shows InterruptLevel = 0 (zero). Any attempt to use either NIC results in a hard hang, with a 5 second power off switch, to break out of it. When I don't use ACPI, I can get the Broadcom (wired) NIC to work, but the Intel (wireless) NIC goes into some sort of a loop when NETBIND.EXE runs (I am sure that this is a GENMAC problem).

LANTRAN.LOG is not yet created, when I have the problem, because the system hangs at the first attempt to access the NICs.

The PCI output will be uploaded shortly.

Pathched drivers are irrelevant, since I am attempting to use GENMAC 2.2, although I have tried the patched B57.OS2 driver, with the Broadcom NIC, which freezes the system, immediately after it puts the copyright notice on the screen.

I suspect that the devices are getting the IRQ assignment too late, and the drivers are seeing IRQ 0 (zero) as assigned, and they are trying to use IRQ 0 (zero), which, of course, causes an IRQ hang.

[Chuck McKinnis]

PCI output, with ACPI and APIC mode

[Chuck McKinnis]

Where is 3.10? I have the latest posted ACPI07052008.ZIP, which says it is 3.09.

[Steven Levine]

3.10 has not been released yet. Your ticket was moved to 3.10 because that's when the code changes that might resolve your issue will be available. This assumes that Pasha does not make a testcase available before then. If he does, he will add a comment to this ticket.

[dgbisse]

I updated to v3.10. No obvious changes to my problem.

[dgbisse]

Updated to 3.11. No change to the operation. New logs collected, and uploaded.

[pasha]

remove pls from \os2\boot\snoop.lst string with acpisnp.snp and give me new log. Questions, have you drivers for your cardbus/firewire support? Given log show some problem with this devices (1180:xxxx).

[dgbisse]

New logs uploaded, as requested.

Answer: No, I have no support for the 1180:xxxx devices. From what I can tell, that is a Ricoh multifunction device, of some sort. There is nothing much of interest in what it does, although it would be nice if it would work in eCS. I don't know if the firewire part is separate from the card reader (xD, MMC, Memory stick).

[dgbisse]

Updated to ACPI 3.13, using acpifu-20081127.wpi. No changes to the operation. New log file to be uploaded...

[dgbisse]

Installed eCS 2.0 RC6a (includes ACPI 3.14). No changes. I will upload a new log file, if you like, but I don't think there will be anything new...

[pasha]

Acer notebook has very bad acpi table. Wait and try new version of acpi

[pasha]

Please download experimental ACPI build from Mensys site:

• Experimental build for you:

ACPI-APIC-BAT.ZIP

• Read instructions: ​http://ecomstation.ru/projects/acpitools/?action=testcase

[dgbisse]

Replying to pasha:

Please download experimental ACPI build from Mensys site:

• Experimental build for you:

ACPI-APIC-BAT.ZIP

• Read instructions: ​http://ecomstation.ru/projects/acpitools/?action=testcase

No change. The GENMAC support for the NICS still doesn't see the proper IRQ. They see IRQ 0. They both seem to get the correct IRQ, when I use OS2APIC.PSD. The Hardware manager does show the correct IRQs being used, in the Hardware Manager, when I use APIC.PSD.

[pasha]

Please download experimental ACPI build from Mensys site:

• Experimental build for you:

ACPI315PRE.ZIP

• Read instructions: ​http://ecomstation.ru/projects/acpitools/?action=testcase

Need acpi log

[dgbisse]

Replying to pasha:

Please download experimental ACPI build from Mensys site:

• Experimental build for you:

ACPI315PRE.ZIP

• Read instructions: ​http://ecomstation.ru/projects/acpitools/?action=testcase

Need acpi log

The system appears to load properly, but will not respond to mouse clicks, even though the mouse cursor moves. The only keyboard thing that works is CAD, which opens CADH. The only thing that seems to work in CADH is R for reboot. ACPI.log is not produced.

[pasha]

Add to %ETC%\AcpiD.cfg:

AcpiLog = c:\AcpiCA.log

This will create log. Attach this log here

[dgbisse]

0311AcpiCA.log attached.

It had occurred to me, that I have done so many things to try to get this to work, that I had lost track of it all. I decided to re-install eCS, so I did a clean install of eCS 2.0 RC6a. Now, the machine boots, and seems to work, with the new ACPI315PRE. I haven't had enough time to try everything, yet. I did not install UniAud, and the NIC driver is REMed, so I can boot the machine. It appears that the NICs still don't see the assigned IRQ, and they try to use IRQ0, even though the Hardware Manager shows them as having IRQ17.

[dgbisse]

I give up. I found another use for this machine, using windows. It will not be running eCS.

I must say, that I am somewhat disappointed, because this machine should work, but after more than a year, I have come to the conclusion, that The ACPI project is not capable of producing working code.

[eco]

e-co:

Yes, ACPI notebooks don't work well with Intel ACPI. The research was made in 2008.

So, what is the main problem? Troubles with network adapters? It's common problem for all Acer notebooks: http://svn.netlabs.org/acpi/ticket/233

What to do? PCIWRITE is aimed to fix this. We can ask Dmitry Frolov to fill the command. It's necessary collect pci.exe output a) with ACPI and b) without. I don't remember, which option is used to query pci config space dump.

[pasha]

Please download experimental ACPI build from Mensys site:

• Experimental build for you:

acpi319PollingFullOnly.zip

• Read instructions: ​http://ecomstation.ru/projects/acpitools/?action=testcase

[pasha]

feedback pending -> mensys stage

[David Azarewicz]

This ticket is really not relevant to the current software. Acer systems are known to have serious compatibility problems and you may not be able to get it to work completely. Have you tried the current release?

[dgbisse]

Replying to dazarewicz:

This ticket is really not relevant to the current software. Acer systems are known to have serious compatibility problems and you may not be able to get it to work completely. Have you tried the current release?

As noted above, I no longer have this system, so I cannot try anything with it.

[David Azarewicz]

Ok, thanks for letting me know.