Bug 1444

Created attachment 1242 [details]
acpidmp from 2.6.0-test9 kernel

Created attachment 1243 [details]
dmesg from 2.6.0-test9 kernel

Created attachment 1244 [details]
dmidecode from 2.6.0-test9 kernel

Created attachment 1245 [details]
lspci -vv from 2.6.0-test9 kernel

Created attachment 1246 [details]
/proc/interrupts from 2.6.0-test9 kernel

I want to add something very important for me. I am not able to enter my
notebook in sleep mode with the following command : echo 1 > /proc/acpi/sleep .
My computer report good time remaining for using it on the battery for two
minutes after the boot but when I want to verify the time remaining after e.g.
ten minutes, it reports a wrong value like 30 minutes whereas I have 80% battery
capacity available.

Thank you in advance.

Olivier..

same warnings as in bug #1404 -- a different Compaq laptop. 
 
    ACPI-1120: *** Error: Method execution failed [\_SB_.C046.C052] (Node cffbdc40), 
AE_AML_BUFFER_LIMIT 
    ACPI-1120: *** Error: Method execution failed [\_SB_.C046.C056] (Node cffbdc20), 
AE_AML_BUFFER_LIMIT 
    ACPI-1120: *** Error: Method execution failed [\_SB_.C046._CRS] (Node cffbdc00), 
AE_AML_BUFFER_LIMIT 
    ACPI-0098: *** Error: Method execution failed [\_SB_.C046._CRS] (Node cffbdc00), 
AE_AML_BUFFER_LIMIT 
 

Would you please try at http://bugzilla.kernel.org/show_bug.cgi?id=1520

Created attachment 1407 [details]
dmesg from 2.6.0-test8 kernel with patch from bug 1520

Created attachment 1408 [details]
dmesg from 2.6.0-test9 with patch from bug 1520

As you can see in the last two messages, the patch doesn't correct the problem.
I have a bunch of error messages related to acpi methods. In fact, usb devices
don't work at all, the battery is detected but my acpid daemon says that the
charge is null whereas the capacity of the battery wasn't null.

Olivier.

Thanks for your testing. I just update patch at bug 1520. And I hope you try it
again. The updated patch has 1 fix for error of replacing length of buffer. This
patch is try to fix bug of accessing region field through buffer. If there is no 
AE_AML_BUFFER_LIMIT, it means the patch works. Thanks a lot.

Created attachment 1413 [details]
dmesg from 2.6.0-test8 with 2nd patch from bug n

Created attachment 1414 [details]
dmesg from 2.6.0-test9 with 2nd patch from bug n

Created attachment 1427 [details]
a updated version

Maybe there are other things wrong in your DSDT. 
I want to look the ACPI debug information. Would you please try this patch with
ACPI_DEBUG option enabled.
Without this patch, C046.C052 will report AE_AML_BUFFER_LIMIT when accessing
7th object in C049.

Created attachment 1428 [details]
dmesg for n620c linux-2.6-test9 with patch (id=1427)

I join this bug report: I have a compaq Evo N620C laptop which has a similar
problem (AE_AML_BUFFER_LIMIT)
This laptop does not really support APM so i'd really like to get ACPI working.

I tried several patches but nothing worked for the moment.
I precise that despite of these messages I can have the acpi state (acpi -V)
and all seems to work BUT I often get processes reading acpi state stuck in D
state after a moment (gkrellm, wmbattery or wmacpi). After that everything
trying to read /proc/acpi gets stuck too.
sleep states does not work either ...
As a programmer (not a kernel one, but i can learn ;) perhaps I can offer my
help too ?

Thanks
(attached: dmesg with latest proposed patch)

Created attachment 1429 [details]
dmesg from 2.6.0-test9 kernel with 3rd patch

I was obliged to use 2.6.0-test9 kernel with the latest acpi patch because of
the acpi debug bug 1341 that block the startup of my Linux system. I follow
your instructions and the results are in the following attachement.

Thanks, Olivier.

We need a trace of the following methods with debug enabled (acpi_dbg_level = 
0x00FFFFFF)

    ACPI-1120: *** Error: Method execution failed [\_SB_.C046.C052] (Node 
cffbdc40), 
AE_AML_BUFFER_LIMIT 
    ACPI-1120: *** Error: Method execution failed [\_SB_.C046.C056] (Node 
cffbdc20), 
AE_AML_BUFFER_LIMIT 
    ACPI-1120: *** Error: Method execution failed [\_SB_.C046._CRS] (Node 
cffbdc00), 
AE_AML_BUFFER_LIMIT 
    ACPI-0098: *** Error: Method execution failed [\_SB_.C046._CRS] (Node 
cffbdc00), 
AE_AML_BUFFER_LIMIT 

How do I trace these methods properly ?

Thanks
Olivier.

The below message appeared in your latest dmesg. This error is not what my 
patch try to address. It looks like a new one.

"dsopcode-0524 [19] ds_init_buffer_field  : Field size 1392 exceeds Buffer size 
1296 (bits)"

In your dsdt, there are something confusing me.
In method C056,
sizeof (C050) = 136
Buffer size of Local0 = 1296 (Your dmesg shows this value)
Obviously, 1296/8 - 136 = 26 = 1A
Obviously, C04D = 1
So below loop in C056 will only execute once.

                While (LLess (Local1, C04D))
                {
                    Store (C052 (Local0, Local2, Local1, Local3), Local2)
                    Increment (Local1)
                }

And Local0 is 1296-bits buffer,
Loca2 is sizeof(C050)-2 = 134
Local1 is sizeof(C050= 136
Local3 is 134 -26 = 108

In C052, before executing CreateDWord (Arg0, Local2 ..)
below two statements will be executed.
Store (Arg1, local2)
Add (Local2, 0xA, Loca2)

So, the local2 should be 134 + 10 =144 = 1152bits.
Obviously It should never fail to create a DWord in a 1296 bits from 1152bits
offset.

So, there must be something wrong in value of Arg1, which has been changed
somewhere.

Here is a small test case to exhabits this issue.
Method (C056, 0, NotSerialized)
{
Store (Buffer(0x6) {}, Local0)
C888(Local0, 2, 0, 0)
}


           Method (C888, 4, NotSerialized) {
                Store (Arg1, Local7)
                Store (Arg3, Local4)

                Add (Local4, 0x2, Local5)
                While (LLess (Local4, Local5))

      {Executing all Device _STA and_INI methods:.\_SB_._INI failed: AE_OK
.\_SB_.C00C._INI failed: AE_OK
.dsopcode-0526 [15] ds_init_buffer_field  : Field [C0EE] size 64 exceeds Buffer
[<NUL] size 48 (bits)
 psparse-1120: *** Error: Method execution failed [\_SB_.C888] (Node cfc53aa0),
AE_AML_BUFFER_LIMIT
 psparse-1120: *** Error: Method execution failed [\_SB_.C04B] (Node cfc5395c),
AE_AML_BUFFER_LIMIT
 psparse-1120: *** Error: Method execution failed [\_SB_.C046._INI] (Node
cfc53884), AE_AML_BUFFER_LIMIT
ment (Local4)
                        Increment (Local7)
                }
                Store (Arg1, Local7)
                CreateDWordField (Arg0, Local7, C0EE)
           }

In my testing machine, below message shows:

[I messed up above comments, repost it here]
Here is a small test case to exhabits this issue.

Method (C056, 0, NotSerialized)
{
Store (Buffer(0x6) {}, Local0)
C888(Local0, 2, 0, 0)
}


           Method (C888, 4, NotSerialized) {
                Store (Arg1, Local7)
                Store (Arg3, Local4)

                Add (Local4, 0x2, Local5)
                While (LLess (Local4, Local5))
               {
 
                        Increment (Local4)
                        Increment (Local7)
                }
                Store (Arg1, Local7)
                CreateDWordField (Arg0, Local7, C0EE)
           }

Dmesg error from my testing machine: (IBM T21)
.dsopcode-0526 [15] ds_init_buffer_field  : Field [C0EE] size 64 exceeds Buffer
[<NUL] size 48 (bits)
 psparse-1120: *** Error: Method execution failed [\_SB_.C888] (Node cfc53aa0),
AE_AML_BUFFER_LIMIT
 psparse-1120: *** Error: Method execution failed [\_SB_.C04B] (Node cfc5395c),
AE_AML_BUFFER_LIMIT
 psparse-1120: *** Error: Method execution failed [\_SB_.C046._INI] (Node
cfc53884), AE_AML_BUFFER_LIMIT

From above message, It is clear that , In current implementation of CA, 
store (Arg1, local7) will cause Arg1 and local7 reference same object.
So Add 2 to local7 will change Arg1 too. The second statement of " store (Arg1,
local7) " will not reset local7 to the initial value.

Created attachment 1453 [details]
a proposal patch for fixing this issue

Created attachment 1454 [details]
dmesg from 2.6.0-test9 kernel with lastest proposed patch

It seems that error messages have disappeared.

Good job. Thank you.

However, when I enter my laptop in sleep mode, it wake up immediatly. Why ?
(I use this command : echo 1 > /proc/acpi/sleep)

Thank you. Olivier.

I think this would be a better fix, it also handles the more general case:

    /*
     * If the reference count on the object is more than one, we must
     * take a copy of the object before we store.
     */
    if (ObjDesc->Common.ReferenceCount > 1)
    {
        Status = AcpiUtCopyIobjectToIobject (ObjDesc, &NewObjDesc, WalkState);
        ObjDesc = NewObjDesc;
        if (ACPI_FAILURE (Status))
        {
            return_ACPI_STATUS (Status);
        }
    }

Where should I place this fix, in which file ? Please.

Olivier.

*** Bug 1592 has been marked as a duplicate of this bug. ***

Thank you for response for comment #28.

Olivier.

For comment #27, where have I to place this fix ? In which file (exstore.c or
another) ? At which lines ?

Thank you

 Please add those code at drivers/acpi/executer/exstore.c @@ -167,6 +167,38 @@

Excuse me but I have a lot of compilling errors with the code submitted in
comment #27 (variables not defined, functions not recognized, etc...)

Thanks Olivier

Created attachment 1554 [details]
exstore.c modified (2.6.0-test10 kernel)

Is my exstore modified good? My kernel is compiling without errors. However, it
isn't working because my kernel crash at startup.
Could you help me please ?
Thanks, Olivier.

Created attachment 1628 [details]
patch bug #1444 for testing

Could you test this patch because I don't know if it's good or not.

Thank you.
Olivier.

Official patch has been merged into latest ACPI CA relase, and It will be found 
in the comming linux acpi patch release soon. 

I've tried the lastest official ACPI patch released today and my notebook works
great (no error messages related to ACPI now).

Thank you.
Olivier.

Created attachment 1659 [details]
dmesg from 2.6.0-test11-bk8 with lastest official ACPI patch

Good job. Really Thank you.

Olivier.

With the latest patches on kernel 2.4.23, I have these same messages.  I will 
attach the command outputs requested on acpi.sourceforge.net 
 
Distribution: N/A (LFS 5.0) 
Hardware Environment: Compaq Evo N800w (latest BIOS) 
Software Environment: kernel boot 

Created attachment 1673 [details]
acpidmp output

Created attachment 1674 [details]
dmesg output

Created attachment 1675 [details]
dmidecode output

Created attachment 1676 [details]
/proc/interrupts output

Created attachment 1677 [details]
lspci output

Brant, 
actually, your dmesg show different messages for a different issue. 
Please file a new bug.  Closing the AE_AML_BUFFER_LIMIT bug. 
 
thanks, 
-Len 
 

Brant, 
I have made a patch to address your problem. please visit
http://bugzilla.kernel.org/show_bug.cgi?id=1690.

Thanks,
Luming

As an FYI, I tried applying this patch to my kernel in addition to the one in bug 1690 and my 
kernel does not boot without acpi=off.