Opened 11 years ago

Closed 5 years ago

#19 closed defect (wontfix)

3945 device isn't initialized, when ACPI.PSD is loaded

Reported by: yoda Owned by: anonymous
Priority: blocker Milestone: Release 4.00
Component: ACPI PSD Version: 3.21.02
Keywords: 3945 wlan Cc: yoda

Description

My wlan device Intel 3945 a/b/g does not get initialized, when ACPI.PSD is loaded. If ACPI.PSD is not loaded, OS/2 initializes device correctly, and it works with GenMac? 2.1 driver. When ACPI.PSD is used, PCI output does not show any memmory address for the device, and it says bus access disabled. Result is that GenMac? can't access the device. Laptop is Acer Travelmate 4672

Attachments (1)

acpi299test-Acer-info10.zip (115.0 KB) - added by yoda 11 years ago.

Download all attachments as: .zip

Change History (20)

Changed 11 years ago by yoda

comment:1 Changed 11 years ago by anonymous

  • Owner changed from pasha to anonymous
  • Status changed from new to assigned

This notebbok has bad acpi table for Intel ACPI CA. This is SMP. Fisrt need make reading file acpi,cfg at SMP kernel, then this can be fixed

comment:2 follow-up: Changed 11 years ago by anonymous

  • Resolution set to fixed
  • Status changed from assigned to closed
  • Version set to 3.01

Can't be fixed now. This notebook has ACPI table with error in Intel interpretator. After start ACPI, PCI config space was changed. As I understood, changes PCI Config space was't ending, because acpi Intel interpretator error is going.

comment:3 in reply to: ↑ 2 Changed 11 years ago by yoda

  • Cc yoda added
  • Milestone changed from eCS 2.0 to eCS 2.x
  • Resolution fixed deleted
  • Status changed from closed to reopened
  • Version changed from 3.01 to 3.02

Replying to anonymous:

Can't be fixed now.

Well, I don't find that closing ticket because of this is the right way.

I reopened it, and assigned it to Milestone eCS 2.x instead

This notebook has ACPI table with error in Intel interpretator. After start ACPI, PCI config space was changed. As I understood, changes PCI Config space was't ending, because acpi Intel interpretator error is going.

This might be right for eCS ACPI - but device works fine under both Win and Linux, without any special modifications, so I assume, the same can be possible for eCS.

comment:4 Changed 11 years ago by David McKenna

You might want to try adding:

LINK LNKA 10

LINK LNKD 5

LINK LNKF 11

to your acpi.cfg file... see if it helps.

comment:5 Changed 10 years ago by guest

  • Resolution set to fixed
  • Status changed from reopened to closed

e-co:

Acer is a Windows notebook. Pasha said that the tables are bad. PCI config space is empty because the processing of ACPI tables failed.

What to do? Search for modified tables on Linux sites.

Currently we add the Acer notebook to blacklist.

comment:6 Changed 9 years ago by stevenhl

  • Resolution fixed deleted
  • Status changed from closed to reopened

comment:7 Changed 9 years ago by stevenhl

  • Resolution set to unknown
  • Status changed from reopened to closed

comment:8 Changed 9 years ago by yoda

  • Priority changed from critical to blocker
  • Resolution unknown deleted
  • Status changed from closed to reopened
  • Version changed from 3.02 to 3.17

There are no errors with ACPI table on this notebook. Yes, there used to be - due to erros in this ACPI driver.

ACPI drive does still not initialize the device correctly, it does not initialize the PCI config space correctly, and because of this the device doesn't work - and system TRAPS.

Removing ACPI driver lets OS/2 initialize the device correctly, and it does work this way.

comment:9 Changed 8 years ago by pasha

you need tra ecs 2.0

comment:10 Changed 8 years ago by yoda

There is no difference using eCS 2.0 GA

The PCI table is still wrongly configured, and GenMac? will trap or hang the system, when the Wlan device gets activated when eCenter loads (XWlan starts).

If I modify the PCI table using PCIWrite, I can get it working.

comment:11 Changed 6 years ago by yoda

  • Milestone changed from eCS 2.x to Release Version 4.0.0
  • Version changed from 3.17 to 3.19.02

This problem also exists in the new line of ACPI drivers from 3.19.02 and up but with later drivers driver (3.19.14 and up I think) the system Traps while loading ACPI driver, so I can not test this again, before the other problem is fixed.

comment:12 Changed 6 years ago by yoda

  • Version changed from 3.19.02 to 3.20.04

Problem still exists in 3.20.05 (test version)

comment:13 Changed 6 years ago by yoda

  • Keywords wlan added
  • Version changed from 3.20.04 to 3.21.01

Problem still exists in 3.21.01

comment:14 Changed 6 years ago by dazarewicz

  • Milestone changed from Release Version 4.0.0 to Feedback pending

There is no need to keep testing new PSD versions regarding this issue since this is not a PSD problem. This is a problem with the hardware and/or the BIOS, and/or the ACPI tables. None of these problems can be addressed in the PSD.

The adapter in question (8086:4222 at 3:0:0 "PRO/Wireless 3945ABG [Golan] Network Connection") is already known to have problems, and the ACPI tables in this system are already known to be faulty. There is no appropriate action that can be taken in the PSD to address these issues.

You might be able to use one of the old tools that were written for setting up PCMCIA adapters to fixup your wireless adapter, too. I know of two but there are probably others: <http://kannegieser.net/veit/quelle/wpcicfg.arj> and <http://88watts.net/download/pcicfgwr-0.01.zip>. The latter has better documentation and is easier to use.

comment:15 Changed 6 years ago by yoda

I tried your pcicfgwr.sys driver.

It refuses to load.

Krnl just reports something like BASEDEV in line 1 was not loaded.

comment:16 follow-up: Changed 6 years ago by yoda

  • Milestone changed from Feedback pending to Release Version 4.0.0
  • Version changed from 3.21.01 to 3.21.02

Well, it seems to be a bug in your "better documentation".

Got it right now. Used the same values as I used to use with PCIWRITE, and now I can enable Wlan device again and use it.

You told me before, that these values were very wrong, and I was just lucky it was working; so you are of course welcome to help me find better values.

comment:17 in reply to: ↑ 16 ; follow-up: Changed 6 years ago by dazarewicz

Replying to yoda:

Well, it seems to be a bug in your "better documentation".

I guess the documentation can always get even better.

Got it right now. Used the same values as I used to use with PCIWRITE, and now I can enable Wlan device again and use it.

You told me before, that these values were very wrong, and I was just lucky it was working; so you are of course welcome to help me find better values.

As I recall, the problem was that you were writing dword values to registers that were byte and word sizes which caused you to write values to other registers that you should not have been writing to. The "luck" is because in that specific case, writing to these other registers either did not matter or did not cause a problem. In general, if you are going to do something like this you should only write what you need to change and leave everything else alone.

comment:18 in reply to: ↑ 17 Changed 6 years ago by yoda

Replying to dazarewicz:

Replying to yoda:

Well, it seems to be a bug in your "better documentation".

I guess the documentation can always get even better.

Last line in doc:

BASEDEV=pcicfgwr.sys 0:28:2 B0x1a=5 W6:0x1001 D16:0x8000C000

I unfortunately used that as prototype; but it has wrong syntax, and doesn't work.

Got it right now. Used the same values as I used to use with PCIWRITE, and now I can enable Wlan device again and use it.

You told me before, that these values were very wrong, and I was just lucky it was working; so you are of course welcome to help me find better values.

As I recall, the problem was that you were writing dword values to registers that were byte and word sizes which caused you to write values to other registers that you should not have been writing to. The "luck" is because in that specific case, writing to these other registers either did not matter or did not cause a problem. In general, if you are going to do something like this you should only write what you need to change and leave everything else alone.

Well, I didn't invent what I am writing to those registers, someone else did. I just tested the lines he suggested, until it worked.

So better values are very welcome :-)

comment:19 Changed 5 years ago by dazarewicz

  • Resolution set to wontfix
  • Status changed from reopened to closed

This issue will not be addressed by the ACPI software. See comment #14.

Note: See TracTickets for help on using tickets.