Re: MSI: Use correct data offset for 32-bit MSI in read_msi_msg()

Re: MSI: Use correct data offset for 32-bit MSI in read_msi_msg()

[Eric W. Biederman]

Roland Dreier <roland@digitalvampire.org> writes:

> While reading the MSI code trying to find a reason why MSI wouldn't
> work for devices that have a 32-bit MSI address capability, I noticed
> that read_msi_msg() seems to read the message data from the wrong
> offset in this case.

Doh!  Sorry about that.

Acked-by: "Eric W. Biederman" <ebiederm@xmission.com>
> Signed-off-by: Roland Dreier <roland@digitalvampire.org>
> ---
> Unfortunately MSI still doesn't seem to work with Intel 945GM
> integrated graphics on my laptop, even with this fix....
>
>  drivers/pci/msi.c |    2 +-
>  1 files changed, 1 insertions(+), 1 deletions(-)
>
> diff --git a/drivers/pci/msi.c b/drivers/pci/msi.c
> index be1df85..87e0161 100644
> --- a/drivers/pci/msi.c
> +++ b/drivers/pci/msi.c
> @@ -132,7 +132,7 @@ void read_msi_msg(unsigned int irq, struct msi_msg *msg)
>  			pci_read_config_word(dev, msi_data_reg(pos, 1), &data);
>  		} else {
>  			msg->address_hi = 0;
> -			pci_read_config_word(dev, msi_data_reg(pos, 1), &data);
> +			pci_read_config_word(dev, msi_data_reg(pos, 0), &data);
>  		}
>  		msg->data = data;
>  		break;

Re: MSI: Use correct data offset for 32-bit MSI in read_msi_msg()

[Roland Dreier]

 > > While reading the MSI code trying to find a reason why MSI wouldn't
 > > work for devices that have a 32-bit MSI address capability, I noticed
 > > that read_msi_msg() seems to read the message data from the wrong
 > > offset in this case.
 > 
 > Doh!  Sorry about that.

Doesn't matter, it wasn't the bug hitting me anyway :) the IRQ never
got affinity changed so read_msi_msg() never even got called.

I'm starting to think that Intel 945GM graphics just has a busted MSI
implementation, although there may be a bug hiding in the Linux code
that I'm not seeing.

 - R.

Re: MSI: Use correct data offset for 32-bit MSI in read_msi_msg()

[Eric W. Biederman]

Roland Dreier <roland@digitalvampire.org> writes:

>  > > While reading the MSI code trying to find a reason why MSI wouldn't
>  > > work for devices that have a 32-bit MSI address capability, I noticed
>  > > that read_msi_msg() seems to read the message data from the wrong
>  > > offset in this case.
>  > 
>  > Doh!  Sorry about that.
>
> Doesn't matter, it wasn't the bug hitting me anyway :) the IRQ never
> got affinity changed so read_msi_msg() never even got called.
>
> I'm starting to think that Intel 945GM graphics just has a busted MSI
> implementation, although there may be a bug hiding in the Linux code
> that I'm not seeing.

Well it is a bug worth fixing.  Who knows it may have something
to do with the disable_irq problems the forcedeth driver was seeing.

Right now MSI is still sufficiently new that everyone is still getting
the bugs out of their implementations.

Eric

Re: MSI: Use correct data offset for 32-bit MSI in read_msi_msg()

[Roland Dreier]

 > Well it is a bug worth fixing.  Who knows it may have something
 > to do with the disable_irq problems the forcedeth driver was seeing.

I agree completely, the fix looks obvious and there's no point in
leaving the bug there.  I just don't have a test case that I can point
to as being fixed by this, since I don't have a working 32-bit MSI
device...

Do you know whether the forcedeth devices have a 32-bit MSI address?

 - R.

Re: MSI: Use correct data offset for 32-bit MSI in read_msi_msg()

[Eric W. Biederman]

Roland Dreier <rdreier@cisco.com> writes:

>  > Well it is a bug worth fixing.  Who knows it may have something
>  > to do with the disable_irq problems the forcedeth driver was seeing.
>
> I agree completely, the fix looks obvious and there's no point in
> leaving the bug there.  I just don't have a test case that I can point
> to as being fixed by this, since I don't have a working 32-bit MSI
> device...
>
> Do you know whether the forcedeth devices have a 32-bit MSI address?

Sorry I don't.  The ones I have don't have an MSI capability.


Eric