[PATCH] PCI/MSI: msix_setup_msi_descs: Restore logic for msi_attrib.can_mask

[PATCH] PCI/MSI: msix_setup_msi_descs: Restore logic for msi_attrib.can_mask

[Josef Johansson]

From: Josef Johansson <josef@oderland.se>

PCI/MSI: msix_setup_msi_descs: Restore logic for msi_attrib.can_mask
    
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@oderland.se>

---
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, nor do I have the possibility
to see if anything works with my patch or without, since the kernel crashes
in other places.

As such this is still untested, but as far as I can tell it should restore
functionality.

Re-sending since it was rejected by linux-pci@vger.kernel.org due to HTML contents.
Sorry about that.

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);
 		}

--
2.31.1

Re: [PATCH] PCI/MSI: msix_setup_msi_descs: Restore logic for msi_attrib.can_mask

[Bjorn Helgaas]

[+cc Jason, since you reviewed the original commit]

On Sat, Jan 22, 2022 at 02:10:01AM +0100, Josef Johansson wrote:
> From: Josef Johansson <josef@oderland.se>
> 
> PCI/MSI: msix_setup_msi_descs: Restore logic for msi_attrib.can_mask

Please match the form and style of previous subject lines (in
particular, omit "msix_setup_msi_descs:").

> 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.

I agree, this looks like a typo in 71020a3c0dff4, but I might be
missing something, so Thomas should take a look, and I added Jason
since he reviewed it.

Since it was merged by Thomas, I'll let him take care of this, too.
If it *is* a typo, the fix looks like v5.17 material.

Before: https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/drivers/pci/msi/msi.c?id=71020a3c0dff4%5E#n522
After:  https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/drivers/pci/msi/msi.c?id=71020a3c0dff4#n520

> Fixes: 71020a3c0dff4 ("PCI/MSI: Use msi_add_msi_desc()")
> Signed-off-by: Josef Johansson <josef@oderland.se>
> 
> ---
> 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, nor do I have the possibility
> to see if anything works with my patch or without, since the kernel crashes
> in other places.
> 
> As such this is still untested, but as far as I can tell it should restore
> functionality.
> 
> Re-sending since it was rejected by linux-pci@vger.kernel.org due to HTML contents.
> Sorry about that.
> 
> 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);
>  		}
> 
> --
> 2.31.1
> 

Re: [PATCH] PCI/MSI: msix_setup_msi_descs: Restore logic for msi_attrib.can_mask

[Josef Johansson]

On 2/11/22 00:55, Bjorn Helgaas wrote:
> [+cc Jason, since you reviewed the original commit]
>
> On Sat, Jan 22, 2022 at 02:10:01AM +0100, Josef Johansson wrote:
>> From: Josef Johansson <josef@oderland.se>
>>
>> PCI/MSI: msix_setup_msi_descs: Restore logic for msi_attrib.can_mask
> Please match the form and style of previous subject lines (in
> particular, omit "msix_setup_msi_descs:").
Would this subject suffice?
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.
> I agree, this looks like a typo in 71020a3c0dff4, but I might be
> missing something, so Thomas should take a look, and I added Jason
> since he reviewed it.
>
> Since it was merged by Thomas, I'll let him take care of this, too.
> If it *is* a typo, the fix looks like v5.17 material.
>
> Before: https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/drivers/pci/msi/msi.c?id=71020a3c0dff4%5E#n522
> After:  https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/drivers/pci/msi/msi.c?id=71020a3c0dff4#n520
It should be noted that I successfully ran the kernel with this patch.
>> Fixes: 71020a3c0dff4 ("PCI/MSI: Use msi_add_msi_desc()")
>> Signed-off-by: Josef Johansson <josef@oderland.se>
>>
>> ---
>> 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, nor do I have the possibility
>> to see if anything works with my patch or without, since the kernel crashes
>> in other places.
>>
>> As such this is still untested, but as far as I can tell it should restore
>> functionality.
>>
>> Re-sending since it was rejected by linux-pci@vger.kernel.org due to HTML contents.
>> Sorry about that.
>>
>> 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);
>>  		}
>>
>> --
>> 2.31.1
>>

Re: [PATCH] PCI/MSI: msix_setup_msi_descs: Restore logic for msi_attrib.can_mask

[Bjorn Helgaas]

On Fri, Feb 11, 2022 at 01:10:22AM +0100, Josef Johansson wrote:
> On 2/11/22 00:55, Bjorn Helgaas wrote:
> > On Sat, Jan 22, 2022 at 02:10:01AM +0100, Josef Johansson wrote:
> >> From: Josef Johansson <josef@oderland.se>
> >>
> >> PCI/MSI: msix_setup_msi_descs: Restore logic for msi_attrib.can_mask
> > Please match the form and style of previous subject lines (in
> > particular, omit "msix_setup_msi_descs:").
> Would this subject suffice?
> PCI/MSI: Correct use of can_mask in msi_add_msi_desc()

Looks good to me!

Bjorn

Re: [PATCH] PCI/MSI: msix_setup_msi_descs: Restore logic for msi_attrib.can_mask

[Jason Gunthorpe]

On Thu, Feb 10, 2022 at 05:55:32PM -0600, Bjorn Helgaas wrote:
> > 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.
> 
> I agree, this looks like a typo in 71020a3c0dff4, but I might be
> missing something, so Thomas should take a look, and I added Jason
> since he reviewed it.

I concur

Jason