On Fri, Aug 05, 2022 at 09:10:41AM -0300, Jason Gunthorpe wrote:
> On Fri, Aug 05, 2022 at 12:03:15PM +0200, Josef Johansson wrote:
> > On 2/14/22 11:07, Josef Johansson wrote:
> > > From: Josef Johansson <josef@xxxxxxxxxxx>
> > >
> > > PCI/MSI: Correct use of can_mask in msi_add_msi_desc()
> > > Commit 71020a3c0dff4 ("PCI/MSI: Use msi_add_msi_desc()") modifies
> > > the logic of checking msi_attrib.can_mask, without any reason.
> > > This commits restores that logic.
> > >
> > > Fixes: 71020a3c0dff4 ("PCI/MSI: Use msi_add_msi_desc()")
> > > Signed-off-by: Josef Johansson <josef@xxxxxxxxxxx>
> > >
> > > ---
> > > v2: Changing subject line to fit earlier commits.
> > >
> > > Trying to fix a NULL BUG in the NVMe MSIX implementation I stumbled upon this code,
> > > which ironically was what my last MSI patch resulted into.
> > >
> > > I don't see any reason why this logic was change, and it did not break anything
> > > correcting the logic.
> > >
> > > CC xen-devel since it very much relates to Xen kernel (via pci_msi_ignore_mask).
> > > ---
> > >
> > > diff --git a/drivers/pci/msi/msi.c b/drivers/pci/msi/msi.c
> > > index c19c7ca58186..146e7b9a01cc 100644
> > > --- a/drivers/pci/msi/msi.c
> > > +++ b/drivers/pci/msi/msi.c
> > > @@ -526,7 +526,7 @@ static int msix_setup_msi_descs(struct pci_dev *dev, void __iomem *base,
> > > desc.pci.msi_attrib.can_mask = !pci_msi_ignore_mask &&
> > > !desc.pci.msi_attrib.is_virtual;
> > > - if (!desc.pci.msi_attrib.can_mask) {
> > > + if (desc.pci.msi_attrib.can_mask) {
> > > addr = pci_msix_desc_addr(&desc);
> > > desc.pci.msix_ctrl = readl(addr + PCI_MSIX_ENTRY_VECTOR_CTRL);
> > > }
> > >
>
> Reviewed-by: Jason Gunthorpe <jgg@xxxxxxxxxx>
>
> Bjorn, please take it?
Thanks for the ping. Since 71020a3c0dff4 is by Thomas, and he merged
that along with a whole series of MSI work, I think I probably
expected him to take care of this.
This looks like a simple typo, so I think the commit log should be
reworded along that line, e.g., something like:
71020a3c0dff4 ("PCI/MSI: Use msi_add_msi_desc()") inadvertently
reversed the sense of "msi_attrib.can_mask" in one use:
- if (entry->pci.msi_attrib.can_mask) {
- addr = pci_msix_desc_addr(entry);
- entry->pci.msix_ctrl = readl(addr + PCI_MSIX_ENTRY_VECTOR_CTRL);
+ if (!desc.pci.msi_attrib.can_mask) {
+ addr = pci_msix_desc_addr(&desc);
+ desc.pci.msix_ctrl = readl(addr + PCI_MSIX_ENTRY_VECTOR_CTRL);
Restore the original test.
Thomas, do you want to take this? I'm happy to merge it, but would
like your reviewed-by or ack first.
Bjorn
