Bug 13036 (sony_laptop)
| Summary: | ACPI Error (hwvalid-0186): Denied AML access to port 0x00000080/4 (DMA1 0x0081-0x0083) [20090320] ACPI Exception (evregion-0422): AE_AML_ILLEGAL_ADDRESS, Returned by Handler for [SystemIO] [20090320] | ||
|---|---|---|---|
| Product: | ACPI | Reporter: | Giovanni Pellerano (giovanni.pellerano) |
| Component: | ACPICA-Core | Assignee: | acpi_acpica-core (acpi_acpica-core) |
| Status: | CLOSED CODE_FIX | ||
| Severity: | high | CC: | giovanni.pellerano, lenb, malattia, ming.m.lin, rodrigo |
| Priority: | P1 | ||
| Hardware: | i386 | ||
| OS: | Linux | ||
| Kernel Version: | 2.6.29-git14 | Subsystem: | |
| Regression: | Yes | Bisected commit-id: | |
| Attachments: |
acpi dump sony vaio sr19xn
patch vs 2.6.29-git14 acpidump vgn-fw235j patch vs 2.6.30-rc1 patch from comment #10, as applied to acpi tree |
||
|
Description
Giovanni Pellerano
2009-04-07 10:12:18 UTC
i've forgot to say I've forgot to say, i've a Sony Vaio SR 19XN; i'm going to attach the acpi dump. Created attachment 20866 [details]
acpi dump sony vaio sr19xn
This Sony has lots of writes to debug port 0x80. However, the AML is doing this with 32-bit writes rather than a 1-byte write. Linux is detecting the illegal access to ports 81-83, blocking them, and complaining. My guess is that Windows would allow a 1-byte write to port 80, but silently block a 4-byte write to port 80. The DSDT in this sony is full of debug writes to method P8XH, so you'll get these whenever the BIOS developer wanted to write to the 7-segment debug port on their prototype -- over 60 calls... is there something i can do to help you? Created attachment 20867 [details]
patch vs 2.6.29-git14
Please verify that this patch makes the error messages go away.
Created attachment 20872 [details]
acpidump vgn-fw235j
This patch works. There is no more errors. i think the patch works also on my sony vaio sr 19xn. at the time i did not have that message any more. instead i got: ACPI: EC: missing confirmations, switch off interrupt mode. ACPI Exception (evregion-0422): AE_TIME, Returned by Handler for [EmbeddedControl] [20090320] ACPI Error (psparse-0537): Method parse/execution failed [\_SB_.ADP1._PSR] (Node f701aea0), AE_TIME ACPI Exception (ac-0135): AE_TIME, Error reading AC Adapter state [20090320] is this related? Created attachment 20913 [details] patch vs 2.6.30-rc1 On Thu, Apr 9, 2009 at 10:56 PM, Moore, Robert <robert.moore@intel.com> wrote: > A 4-byte write to an I/O port means four one-byte writes to > four consecutive I/O ports (although, this can be done in one machine > instruction.) > > From our testing, Windows indeed is only allowing the single-byte write to > the (legal) port 0x80. The rest of the request is simply ignored. > In general, any part of an I/O request that overlaps the protected ports > is ignored. No error is returned. > > We are updating ACPICA to match this behavior. > > Bob As Bob said above, the attachment is the linux ACPICA patch. Giovanni, would you please help to test this patch? (only this patch, and do not apply the patch at comment #6) Thanks. Ok it does work Lin! =) Well done! thanks a lot. Does the error message you got at comment #9 also disappear? Thanks for testing. No, i got them randomly. Should them heve been disappeared? Did you got them before 2.6.30-rc1? yes, i think that is not a regression but an old bug never solved do you need someting ? Would you please file a new bug report for that error message since it's not related to this report? yes. i've just done it: http://bugzilla.kernel.org/show_bug.cgi?id=13055 Created attachment 21100 [details] patch from comment #10, as applied to acpi tree patch in comment #6 reverted from acpi tree. patch from comment #18 and revert from comment #19 shipping in linux-2.6.30-rc3-git4 closed |