Re: [RFC PATCH] PCI: export MSI mode using attributes, not kobjects

From: Greg Kroah-Hartman
Date: Sat Nov 02 2013 - 11:49:01 EST

On Fri, Nov 01, 2013 at 05:40:02PM -0600, Bjorn Helgaas wrote:
> On Tue, Oct 29, 2013 at 3:46 PM, Greg Kroah-Hartman
> <gregkh@xxxxxxxxxxxxxxxxxxx> wrote:
> > From: Greg Kroah-Hartman <gregkh@xxxxxxxxxxxxxxxxxxx>
> >
> > The PCI MSI sysfs code is a mess with kobjects for things that don't
> > really need to be kobjects. This patch creates attributes dynamically
> > for the MSI interrupts instead of using kobjects.
> >
> > Note, this does not delete the existing sysfs MSI code, but puts the
> > attributes under a "msi_irqs_2" directory for testing / example.
> >
> > Also note, this removes a directory from the current MSI interrupt sysfs
> > code:
> >
> > old MSI kobjects:
> > pci_device
> > âââ msi_irqs
> > âââ 40
> > âââ mode
> >
> > new MSI attributes:
> > pci_device
> > âââ msi_irqs_2
> > âââ 40
> >
> > As there was only one file "mode" with the kobject model, the interrupt
> > number is now a file that returns the "mode" of the interrupt (msi vs.
> > msix).
> >
> > Signed-off-by: Greg Kroah-Hartman <gregkh@xxxxxxxxxxxxxxxxxxx>
> > ---
> >
> > Bjorn, I can make up a patch that rips out the existing kobject code
> > here, but I figured this patch would make things easier to follow
> > instead of having to dig through the removed logic at the same time.
> >
> > I'll clean up the error handling path for the create attribute logic as
> > well, this was just a proof of concept that this could be done.
> >
> > Do you think that anyone cares about the current mode files in sysfs to
> > move things in this manner?
> I like this a lot better than trying to fix all the holes in the
> current kobject code.


> I have no idea who, if anybody, cares about the "mode" files. I
> assume there's a way to create the "mode" files with attributes, too?
> If so, we could replicate the existing structure with one patch, and
> simplify it with a second patch, so it would be easier to revert the
> directory change while keeping the fix.

No, we can't create a 2-level deep attribute at the moment, only one
level, like the patch does.

Based on Neil's comments, I think we should be fine with this as-is as
no one is messing with these files directly (which implies that we could
possibly just remove them entirely to save us the overall pain...)

Want me to redo this in a way that is acceptable (i.e. remove the
existing code at the same time?)


greg k-h
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at
Please read the FAQ at