Skip to contentSkip to conversation

PCI/VMD: Fix config addressing with bus offsets

PATCH In review Jon Derrick 2 participants 8 messages Jun 11, 2019 – Jul 24, 2019

Subsystems (4)
lore
Share & tools
lore
#0 lore raw

PATCH 1/1 PCI/VMD: Fix config addressing with bus offsets

VMD config space addressing relies on mapping the BDF of the target into the VMD config bar. When using bus number offsets to number the VMD domain, the offset needs to be ignored in order to correctly map devices to their config space.

Fixes 2a5a9c9a20f9 ("PCI: vmd: Add offset to bus numbers if necessary")
Cc <[email protected]> # v4.19
Cc <[email protected]> # v4.18
---
 drivers/pci/controller/vmd.c | 16 +++++++++-------
 1 file changed, 9 insertions(+), 7 deletions(-)
drivers/pci/controller/vmd.c +9 -7
index fd2dbd7..a59afec 100644
@@ -94,6 +94,7 @@ struct vmd_dev {
 	struct resource		resources[3];
 	struct irq_domain	*irq_domain;
 	struct pci_bus		*bus;
+	u8			busn_start;
 
 #ifdef CONFIG_X86_DEV_DMA_OPS
 	struct dma_map_ops	dma_ops;
@@ -465,7 +466,8 @@ static char __iomem *vmd_cfg_addr(struct vmd_dev *vmd, struct pci_bus *bus,
 				  unsigned int devfn, int reg, int len)
 {
 	char __iomem *addr = vmd->cfgbar +
-			     (bus->number << 20) + (devfn << 12) + reg;
+			     ((bus->number - vmd->busn_start) << 20) +
+			     (devfn << 12) + reg;
 
 	if ((addr - vmd->cfgbar) + len >=
 	    resource_size(&vmd->dev->resource[VMD_CFGBAR]))
@@ -588,7 +590,7 @@ static int vmd_enable_domain(struct vmd_dev *vmd, unsigned long features)
 	unsigned long flags;
 	LIST_HEAD(resources);
 	resource_size_t offset[2] = {0};
-	resource_size_t membar2_offset = 0x2000, busn_start = 0;
+	resource_size_t membar2_offset = 0x2000;
 
 	/*
 	 * Shadow registers may exist in certain VMD device ids which allow
@@ -630,14 +632,14 @@ static int vmd_enable_domain(struct vmd_dev *vmd, unsigned long features)
 		pci_read_config_dword(vmd->dev, PCI_REG_VMCONFIG, &vmconfig);
 		if (BUS_RESTRICT_CAP(vmcap) &&
 		    (BUS_RESTRICT_CFG(vmconfig) == 0x1))
-			busn_start = 128;
+			vmd->busn_start = 128;
 	}
 
 	res = &vmd->dev->resource[VMD_CFGBAR];
 	vmd->resources[0] = (struct resource) {
 		.name  = "VMD CFGBAR",
-		.start = busn_start,
-		.end   = busn_start + (resource_size(res) >> 20) - 1,
+		.start = vmd->busn_start,
+		.end   = vmd->busn_start + (resource_size(res) >> 20) - 1,
 		.flags = IORESOURCE_BUS | IORESOURCE_PCI_FIXED,
 	};
 
@@ -705,8 +707,8 @@ static int vmd_enable_domain(struct vmd_dev *vmd, unsigned long features)
 	pci_add_resource_offset(&resources, &vmd->resources[1], offset[0]);
 	pci_add_resource_offset(&resources, &vmd->resources[2], offset[1]);
 
-	vmd->bus = pci_create_root_bus(&vmd->dev->dev, busn_start, &vmd_ops,
-				       sd, &resources);
+	vmd->bus = pci_create_root_bus(&vmd->dev->dev, vmd->busn_start,
+				       &vmd_ops, sd, &resources);
 	if (!vmd->bus) {
 		pci_free_resource_list(&resources);
 		irq_domain_remove(vmd->irq_domain);
Signature · 1 line
1.8.3.1
#1 lore raw

[dropped CC stable]

Quoted 8 lines · Jon Derrick

VMD config space addressing relies on mapping the BDF of the target into the VMD config bar. When using bus number offsets to number the VMD domain, the offset needs to be ignored in order to correctly map devices to their config space.

Fixes 2a5a9c9a20f9 ("PCI: vmd: Add offset to bus numbers if necessary")
Cc <[email protected]> # v4.19
Cc <[email protected]> # v4.18

Hi Jon,

that's not how stable should be handled. You should always start by fixing mainline and if there are backports to be fixed too you should add patch dependencies in the CC area, see:

Documentation/process/stable-kernel-rules.rst

Never add stable to the CC list in the email header, only in the commit log.

When your patch hits mainline it will trickle back into stable, if you specified dependencies as described above there is nothing to do.

Thanks,
Lorenzo

Quoted 68 lines
---
 drivers/pci/controller/vmd.c | 16 +++++++++-------
 1 file changed, 9 insertions(+), 7 deletions(-)
drivers/pci/controller/vmd.c +9 -7
index fd2dbd7..a59afec 100644
@@ -94,6 +94,7 @@ struct vmd_dev {
 	struct resource		resources[3];
 	struct irq_domain	*irq_domain;
 	struct pci_bus		*bus;
+	u8			busn_start;
 
 #ifdef CONFIG_X86_DEV_DMA_OPS
 	struct dma_map_ops	dma_ops;
@@ -465,7 +466,8 @@ static char __iomem *vmd_cfg_addr(struct vmd_dev *vmd, struct pci_bus *bus,
 				  unsigned int devfn, int reg, int len)
 {
 	char __iomem *addr = vmd->cfgbar +
-			     (bus->number << 20) + (devfn << 12) + reg;
+			     ((bus->number - vmd->busn_start) << 20) +
+			     (devfn << 12) + reg;
 
 	if ((addr - vmd->cfgbar) + len >=
 	    resource_size(&vmd->dev->resource[VMD_CFGBAR]))
@@ -588,7 +590,7 @@ static int vmd_enable_domain(struct vmd_dev *vmd, unsigned long features)
 	unsigned long flags;
 	LIST_HEAD(resources);
 	resource_size_t offset[2] = {0};
-	resource_size_t membar2_offset = 0x2000, busn_start = 0;
+	resource_size_t membar2_offset = 0x2000;
 
 	/*
 	 * Shadow registers may exist in certain VMD device ids which allow
@@ -630,14 +632,14 @@ static int vmd_enable_domain(struct vmd_dev *vmd, unsigned long features)
 		pci_read_config_dword(vmd->dev, PCI_REG_VMCONFIG, &vmconfig);
 		if (BUS_RESTRICT_CAP(vmcap) &&
 		    (BUS_RESTRICT_CFG(vmconfig) == 0x1))
-			busn_start = 128;
+			vmd->busn_start = 128;
 	}
 
 	res = &vmd->dev->resource[VMD_CFGBAR];
 	vmd->resources[0] = (struct resource) {
 		.name  = "VMD CFGBAR",
-		.start = busn_start,
-		.end   = busn_start + (resource_size(res) >> 20) - 1,
+		.start = vmd->busn_start,
+		.end   = vmd->busn_start + (resource_size(res) >> 20) - 1,
 		.flags = IORESOURCE_BUS | IORESOURCE_PCI_FIXED,
 	};
 
@@ -705,8 +707,8 @@ static int vmd_enable_domain(struct vmd_dev *vmd, unsigned long features)
 	pci_add_resource_offset(&resources, &vmd->resources[1], offset[0]);
 	pci_add_resource_offset(&resources, &vmd->resources[2], offset[1]);
 
-	vmd->bus = pci_create_root_bus(&vmd->dev->dev, busn_start, &vmd_ops,
-				       sd, &resources);
+	vmd->bus = pci_create_root_bus(&vmd->dev->dev, vmd->busn_start,
+				       &vmd_ops, sd, &resources);
 	if (!vmd->bus) {
 		pci_free_resource_list(&resources);
 		irq_domain_remove(vmd->irq_domain);
Signature · 1 line
1.8.3.1
#2 lore raw
Quoted 23 lines · Lorenzo Pieralisi

[dropped CC stable]

Quoted 8 lines · Jon Derrick

VMD config space addressing relies on mapping the BDF of the target into the VMD config bar. When using bus number offsets to number the VMD domain, the offset needs to be ignored in order to correctly map devices to their config space.

Fixes 2a5a9c9a20f9 ("PCI: vmd: Add offset to bus numbers if necessary")
Cc <[email protected]> # v4.19
Cc <[email protected]> # v4.18

Hi Jon,

that's not how stable should be handled. You should always start by fixing mainline and if there are backports to be fixed too you should add patch dependencies in the CC area, see:

Documentation/process/stable-kernel-rules.rst

Never add stable to the CC list in the email header, only in the commit log.

Quoted 7 lines

When your patch hits mainline it will trickle back into stable, if you specified dependencies as described above there is nothing to do.

Thanks,
Lorenzo

Hi Lorenzo,

Thanks for correcting me here. It's not my normal flow so I apologize for getting this so egregiously incorrect.

#3 lore raw
In reply to Lorenzo Pieralisi
Quoted 23 lines · Lorenzo Pieralisi

[dropped CC stable]

Quoted 8 lines · Jon Derrick

VMD config space addressing relies on mapping the BDF of the target into the VMD config bar. When using bus number offsets to number the VMD domain, the offset needs to be ignored in order to correctly map devices to their config space.

Fixes 2a5a9c9a20f9 ("PCI: vmd: Add offset to bus numbers if necessary")
Cc <[email protected]> # v4.19
Cc <[email protected]> # v4.18

Hi Jon,

that's not how stable should be handled. You should always start by fixing mainline and if there are backports to be fixed too you should add patch dependencies in the CC area, see:

Documentation/process/stable-kernel-rules.rst

Never add stable to the CC list in the email header, only in the commit log.

Quoted 7 lines

When your patch hits mainline it will trickle back into stable, if you specified dependencies as described above there is nothing to do.

Thanks,
Lorenzo

Besides the stable issue, can we get this into 5.3?

#4 lore raw
Quoted 28 lines · Jonathan
Quoted 27 lines · Lorenzo Pieralisi

[dropped CC stable]

Quoted 8 lines · Jon Derrick

VMD config space addressing relies on mapping the BDF of the target into the VMD config bar. When using bus number offsets to number the VMD domain, the offset needs to be ignored in order to correctly map devices to their config space.

Fixes 2a5a9c9a20f9 ("PCI: vmd: Add offset to bus numbers if necessary")
Cc <[email protected]> # v4.19
Cc <[email protected]> # v4.18

Hi Jon,

that's not how stable should be handled. You should always start by fixing mainline and if there are backports to be fixed too you should add patch dependencies in the CC area, see:

Documentation/process/stable-kernel-rules.rst

Never add stable to the CC list in the email header, only in the commit log.

When your patch hits mainline it will trickle back into stable, if you specified dependencies as described above there is nothing to do.

Quoted 5 lines
Quoted 3 lines

Thanks,
Lorenzo

Besides the stable issue, can we get this into 5.3?

Usually we send fixes at -rc for patches that were merged in the previous merge window; this fix is not one of those so I think we will send it for v5.4 unless it is very urgent.

We should still update stable info in the log appropriately before queuing it.

Thanks,
Lorenzo

#5 lore raw
Quoted 39 lines · Lorenzo Pieralisi
Quoted 28 lines · Jonathan
Quoted 27 lines · Lorenzo Pieralisi

[dropped CC stable]

Quoted 8 lines · Jon Derrick

VMD config space addressing relies on mapping the BDF of the target into the VMD config bar. When using bus number offsets to number the VMD domain, the offset needs to be ignored in order to correctly map devices to their config space.

Fixes 2a5a9c9a20f9 ("PCI: vmd: Add offset to bus numbers if necessary")
Cc <[email protected]> # v4.19
Cc <[email protected]> # v4.18

Hi Jon,

that's not how stable should be handled. You should always start by fixing mainline and if there are backports to be fixed too you should add patch dependencies in the CC area, see:

Documentation/process/stable-kernel-rules.rst

Never add stable to the CC list in the email header, only in the commit log.

When your patch hits mainline it will trickle back into stable, if you specified dependencies as described above there is nothing to do.

Quoted 5 lines
Quoted 3 lines

Thanks,
Lorenzo

Besides the stable issue, can we get this into 5.3?

Usually we send fixes at -rc for patches that were merged in the previous merge window; this fix is not one of those so I think we will send it for v5.4 unless it is very urgent.

Quoted 5 lines

We should still update stable info in the log appropriately before queuing it.

Thanks,
Lorenzo

Sure. Had assumed it would be queued for 5.3 as it was submitted in 5.2 I will resubmit during 5.4 merge window and deal with stables later.

#6 lore raw

On Tue, Jul 23, 2019 at 03:12:51PM +0000, Derrick, Jonathan wrote:

[...]

Quoted 9 lines
Quoted 9 lines
Quoted 1 line

Besides the stable issue, can we get this into 5.3?

Usually we send fixes at -rc for patches that were merged in the previous merge window; this fix is not one of those so I think we will send it for v5.4 unless it is very urgent.

We should still update stable info in the log appropriately before queuing it.

Quoted 6 lines
Quoted 2 lines

Thanks,
Lorenzo

Sure. Had assumed it would be queued for 5.3 as it was submitted in 5.2 I will resubmit during 5.4 merge window and deal with stables later.

It should not be that complicated to add stable tags (possibly with commit dependencies on stable kernels where the patch does not apply and require additional patches to be pulled), there is plenty of time to sort this out.

Apologies for not picking it up for v5.3 - I should have asked you to just update the stable tag so that we could merge it.

Lorenzo

#7 lore raw
Quoted 24 lines · Lorenzo Pieralisi

On Tue, Jul 23, 2019 at 03:12:51PM +0000, Derrick, Jonathan wrote:

[...]

Quoted 6 lines
Quoted 6 lines
Quoted 1 line

Besides the stable issue, can we get this into 5.3?

Usually we send fixes at -rc for patches that were merged in the previous merge window; this fix is not one of those so I think we will send it for v5.4 unless it is very urgent.

Quoted 8 lines
Quoted 5 lines

We should still update stable info in the log appropriately before queuing it.

Thanks,
Lorenzo

Sure. Had assumed it would be queued for 5.3 as it was submitted in 5.2 I will resubmit during 5.4 merge window and deal with stables later.

It should not be that complicated to add stable tags (possibly with commit dependencies on stable kernels where the patch does not apply and require additional patches to be pulled), there is plenty of time to sort this out.

Quoted 4 lines

Apologies for not picking it up for v5.3 - I should have asked you to just update the stable tag so that we could merge it.

Lorenzo

No problem. I have another fix I need to include so I will do that during 5.4 merge window