Hi all,
running Linus' latest (2b12a4c) git kernel, acer-wmi triggers the
following ACPI Error on my Acer TravelMate 6492:
acer-wmi: Acer Laptop ACPI-WMI Extras
ACPI Error (nseval-0159): Insufficient arguments - method [WQAA] needs 1, found 0 [20080609]
acer-wmi: Unable to detect available WMID devices
Reverting commit f3454ae8104efb2dbf0d08ec42c6f5d0fe9225bc (ACPICA: Add
argument count checking to control method invocation via
acpi_evaluate_object) makes acer_wmi work as before.
Doing a little debugging the call trace starting at acer_wmi_init() is:
WMID_set_capabilities() (called from drivers/misc/acer-wmi.c:1192)
wmi_query_block(WMID_GUID2, 1, &out) (called from drivers/misc/acer-wmi.c:747)
acpi_evaluate_object(handle, method, NULL, out) (called from drivers/acpi/wmi.c:350)
acpi_ns_evaluate(info) (called from drivers/acpi/namespace/nsxfeval.c:258)
Is it a bug in acer-wmi or in Acer's ACPI implementation? I can provide
more information as necessary.
Sven
On Saturday 02 August 2008 16:50:54 Sven Wegener wrote:
> Is it a bug in acer-wmi or in Acer's ACPI implementation? I can provide
> more information as necessary.
Neither - it's a bug in wmi.c
Try the following patch:
---
ACPI: WMI: Set instance for query block calls
From: Carlos Corbacho <[email protected]>
Although the necessary data structure was set up, it was never actually
passed in, so data block calls have only been working by sheer chance.
(On Acer laptops. the data block methods we've been calling never look at
the instance value, hence acer-wmi never triggered this before).
f3454ae8104efb2dbf0d08ec42c6f5d0fe9225bc brought this to light.
Signed-off-by: Carlos Corbacho <[email protected]>
---
drivers/acpi/wmi.c | 2 +-
1 files changed, 1 insertions(+), 1 deletions(-)
diff --git a/drivers/acpi/wmi.c b/drivers/acpi/wmi.c
index c33b1c6..cfe2c83 100644
--- a/drivers/acpi/wmi.c
+++ b/drivers/acpi/wmi.c
@@ -347,7 +347,7 @@ struct acpi_buffer *out)
strcpy(method, "WQ");
strncat(method, block->object_id, 2);
- status = acpi_evaluate_object(handle, method, NULL, out);
+ status = acpi_evaluate_object(handle, method, &input, out);
/*
* If ACPI_WMI_EXPENSIVE, call the relevant WCxx method, even if
On Sat, 2 Aug 2008, Carlos Corbacho wrote:
> On Saturday 02 August 2008 16:50:54 Sven Wegener wrote:
> > Is it a bug in acer-wmi or in Acer's ACPI implementation? I can provide
> > more information as necessary.
>
> Neither - it's a bug in wmi.c
>
> Try the following patch:
> ---
> ACPI: WMI: Set instance for query block calls
>
> From: Carlos Corbacho <[email protected]>
>
> Although the necessary data structure was set up, it was never actually
> passed in, so data block calls have only been working by sheer chance.
>
> (On Acer laptops. the data block methods we've been calling never look at
> the instance value, hence acer-wmi never triggered this before).
>
> f3454ae8104efb2dbf0d08ec42c6f5d0fe9225bc brought this to light.
>
> Signed-off-by: Carlos Corbacho <[email protected]>
Yep, that fixes it, thanks.
Tested-by: Sven Wegener <[email protected]>
> diff --git a/drivers/acpi/wmi.c b/drivers/acpi/wmi.c
> index c33b1c6..cfe2c83 100644
> --- a/drivers/acpi/wmi.c
> +++ b/drivers/acpi/wmi.c
> @@ -347,7 +347,7 @@ struct acpi_buffer *out)
> strcpy(method, "WQ");
> strncat(method, block->object_id, 2);
>
> - status = acpi_evaluate_object(handle, method, NULL, out);
> + status = acpi_evaluate_object(handle, method, &input, out);
>
> /*
> * If ACPI_WMI_EXPENSIVE, call the relevant WCxx method, even if
On Sat, Aug 02, 2008 at 05:28:45PM +0100, Carlos Corbacho wrote:
> ACPI: WMI: Set instance for query block calls
>
> From: Carlos Corbacho <[email protected]>
>
> Although the necessary data structure was set up, it was never actually
> passed in, so data block calls have only been working by sheer chance.
Hm. Has this gone upstream? Looks pretty important for .27.
--
Matthew Garrett | [email protected]
On Wed, 27 Aug 2008, Matthew Garrett wrote:
> On Sat, Aug 02, 2008 at 05:28:45PM +0100, Carlos Corbacho wrote:
> > ACPI: WMI: Set instance for query block calls
> >
> > From: Carlos Corbacho <[email protected]>
> >
> > Although the necessary data structure was set up, it was never actually
> > passed in, so data block calls have only been working by sheer chance.
>
> Hm. Has this gone upstream? Looks pretty important for .27.
Yes, it's dab36ad8d50dc9424dfc4926f62aaf9bd52dcf13.
Sven