[PATCH v4 0/2] PCI: Fix isolated function probing and enable ARI for s390

[PATCH v4 0/2] PCI: Fix isolated function probing and enable ARI for s390

[Niklas Schnelle]

Hi Bjorn,

This series originally aimed to correctly detect ARI as being used on
s390. I had missed however that this, in a pretty obious manner, breaks
the isolated function probing resulting devices not getting probed. For
example if a partition/system only has a PF with devfn 1 passed-through.
Additionally the fix may (TBD) help with an issue that LoongArch has
encountered when using isolated function probing and tried to fix by
limiting it to bus 0 ([0]). If it does fix this it may make sense to
apply this separately from the second patch.

Besides the effect on s390 the second patch should also ensure that VFs
on a virtual bus created via virtfn_add_bus() correctly present ARI as
enabled. Sadly I don't have access to any device to test this.

Speaking of testing. For the first patch I can reproduce the AER
scenario described by only applying the second patch. The SR-IOV
scenario I encountered in the past before commit 25f39d3dcb48 ("s390/pci:
Ignore RID for isolated VFs") and tested the fix now with a partial
revert hack of that commit.

I also tested the series on x86_64 both on a Ryzen AI 340 based laptop
and a Xeon Sapphire Rapids based system (including with SR-IOV on
a ConnectX-6 DX).

Thanks,
Niklas

[0] https://lore.kernel.org/linux-pci/20251014074100.2149737-1-chenhuacai@loongson.cn/

Signed-off-by: Niklas Schnelle <schnelle@linux.ibm.com>
---
Changes in v4:
- Add fix of isolated function probing with ARI enabled and in certain
  cases with SR-IOV devices.
- Link to v3: https://lore.kernel.org/r/20250417-ari_no_bus_dev-v3-1-ba48f349aa47@linux.ibm.com

Changes in v3:
- Move setting of ari_enabled on s390 to bus creation and clear it in
  pcibios_add_device() if the capability is not available.
- Rebase on v6.15-rc2
- Link to v2: https://lore.kernel.org/r/20240918-ari_no_bus_dev-v2-1-83cfa991082f@linux.ibm.com

Changes in v2:
- Rebased on v6.11
- Link to v1: https://lore.kernel.org/r/20240730-ari_no_bus_dev-v1-1-7de17676f9fe@linux.ibm.com

---
Niklas Schnelle (2):
      PCI: Fix isolated PCI function probing with ARI and SR-IOV
      PCI: s390: Handle ARI on bus without associated struct pci_dev

 arch/s390/pci/pci.c     |  7 +++++++
 arch/s390/pci/pci_bus.c | 10 ++++++++++
 drivers/pci/pci.c       |  4 ++--
 drivers/pci/probe.c     | 21 +++++++++++++--------
 include/linux/pci.h     |  4 ++--
 5 files changed, 34 insertions(+), 12 deletions(-)
---
base-commit: 211ddde0823f1442e4ad052a2f30f050145ccada
change-id: 20240724-ari_no_bus_dev-52b2a27f3466

Best regards,
-- 
Niklas Schnelle

[PATCH v4 1/2] PCI: Fix isolated PCI function probing with ARI and SR-IOV

[Niklas Schnelle]

When the isolated PCI function probing mechanism is used in conjunction
with ARI or SR-IOV it may not find all available PCI functions. In the
case of ARI the problem is that next_ari_fn() always returns -ENODEV if
dev is NULL and thus if fn 0 is missing the scan stops.

For SR-IOV things are more complex. Here the problem is that the check
for multifunction may fail. One example where this can occur is if the
first passed-through function is a VF with devfn 8. Now in
pci_scan_slot() this means it is fn 0 and thus multifunction doesn't get
set. Since VFs don't get multifunction set via PCI_HEADER_TYPE_MFD it
remains unset and probing stops even if there is a devfn 9.

Now at the moment both of these issues are hidden on s390. The first one
because ARI is detected as disabled as struct pci_bus's self is NULL
even though firmware does enable and use ARI. The second issue is hidden
as a side effect of commit 25f39d3dcb48 ("s390/pci: Ignore RID for
isolated VFs"). This is because VFs are either put on their own virtual
bus if the parent PF is not passed-through to the same instance or VFs
are hotplugged once SR-IOV is enabled on the parent PF and then
pci_scan_single_device() is used.

Still especially the first issue prevents correct detection of ARI and
the second might be a problem for other users of isolated function
probing. Fix both issues by keeping things as simple as possible. If
isolated function probing is enabled simply scan every possible devfn.

Fixes: 189c6c33ff42 ("PCI: Extend isolated function probing to s390")
Link: https://lore.kernel.org/linux-pci/d3f11e8562f589ddb2c1c83e74161bd8948084c3.camel@linux.ibm.com/
Signed-off-by: Niklas Schnelle <schnelle@linux.ibm.com>
---
 drivers/pci/probe.c | 20 ++++++++++++--------
 1 file changed, 12 insertions(+), 8 deletions(-)

diff --git a/drivers/pci/probe.c b/drivers/pci/probe.c
index 0ce98e18b5a876afe72af35a9f4a44d598e8d500..13495b12fbcfae4b890bbd4b2f913742adf6dfed 100644
--- a/drivers/pci/probe.c
+++ b/drivers/pci/probe.c
@@ -2808,16 +2808,18 @@ static int next_ari_fn(struct pci_bus *bus, struct pci_dev *dev, int fn)
 	return next_fn;
 }
 
-static int next_fn(struct pci_bus *bus, struct pci_dev *dev, int fn)
+static int next_fn(struct pci_bus *bus, struct pci_dev *dev, int fn, bool isolated)
 {
-	if (pci_ari_enabled(bus))
-		return next_ari_fn(bus, dev, fn);
+	if (!isolated) {
+		if (pci_ari_enabled(bus))
+			return next_ari_fn(bus, dev, fn);
 
+		/* only multifunction devices may have more functions */
+		if (dev && !dev->multifunction)
+			return -ENODEV;
+	}
 	if (fn >= 7)
 		return -ENODEV;
-	/* only multifunction devices may have more functions */
-	if (dev && !dev->multifunction)
-		return -ENODEV;
 
 	return fn + 1;
 }
@@ -2857,12 +2859,14 @@ static int only_one_child(struct pci_bus *bus)
  */
 int pci_scan_slot(struct pci_bus *bus, int devfn)
 {
+	bool isolated_functions;
 	struct pci_dev *dev;
 	int fn = 0, nr = 0;
 
 	if (only_one_child(bus) && (devfn > 0))
 		return 0; /* Already scanned the entire slot */
 
+	isolated_functions = hypervisor_isolated_pci_functions();
 	do {
 		dev = pci_scan_single_device(bus, devfn + fn);
 		if (dev) {
@@ -2876,10 +2880,10 @@ int pci_scan_slot(struct pci_bus *bus, int devfn)
 			 * a hypervisor that passes through individual PCI
 			 * functions.
 			 */
-			if (!hypervisor_isolated_pci_functions())
+			if (!isolated_functions)
 				break;
 		}
-		fn = next_fn(bus, dev, fn);
+		fn = next_fn(bus, dev, fn, isolated_functions);
 	} while (fn >= 0);
 
 	/* Only one slot has PCIe device */

-- 
2.48.1

[PATCH v4 2/2] PCI: s390: Handle ARI on bus without associated struct pci_dev

[Niklas Schnelle]

On s390 PCI busses are virtualized and the downstream ports are
invisible to the OS and self in struct pci_bus is NULL. This associated
struct pci_dev is however relied upon in pci_ari_enabled() to check
whether ARI is enabled for the bus. ARI is therefor always detected as
disabled. At the same time, firmware on s390 always enables and relies
upon ARI thus causing a mismatch.

Despite simply being a mismatch this causes problems as some PCI devices
present a different SR-IOV topology depending on PCI_SRIOV_CTRL_ARI.

A similar mismatch may occur with SR-IOV when virtfn_add_bus() creates new
busses with no associated struct pci_dev. Here too pci_ari_enabled()
on these busses returns false even if ARI is actually used.

Prevent both mismatches by moving the ari_enabled flag from struct
pci_dev to struct pci_bus making it independent from self in struct
pci_bus. Let the bus inherit the ari_enabled state from its parent bus
when there is no bridge device such that busses added by virtfn_add_bus()
match their parent. For s390 set ari_enabled when the device supports
ARI in the awareness that all PCIe ports on s390 systems are ARI
capable.

Signed-off-by: Niklas Schnelle <schnelle@linux.ibm.com>
---
 arch/s390/pci/pci.c     |  7 +++++++
 arch/s390/pci/pci_bus.c | 10 ++++++++++
 drivers/pci/pci.c       |  4 ++--
 drivers/pci/probe.c     |  1 +
 include/linux/pci.h     |  4 ++--
 5 files changed, 22 insertions(+), 4 deletions(-)

diff --git a/arch/s390/pci/pci.c b/arch/s390/pci/pci.c
index c82c577db2bcd2143476cb8189fd89b9a4dc9836..773c0cbfc313ea1a6419a44d6158397dd13f6e76 100644
--- a/arch/s390/pci/pci.c
+++ b/arch/s390/pci/pci.c
@@ -609,6 +609,13 @@ int pcibios_device_add(struct pci_dev *pdev)
 			continue;
 		pci_claim_resource(pdev, i);
 	}
+	/*
+	 * The below is the s390 equivalent of pci_configure_ari()
+	 * which we can't use directly because the bridge devices
+	 * are hidden in firmware.
+	 */
+	if (!pci_find_ext_capability(pdev, PCI_EXT_CAP_ID_ARI))
+		zdev->zbus->bus->ari_enabled = 0;
 
 	return 0;
 }
diff --git a/arch/s390/pci/pci_bus.c b/arch/s390/pci/pci_bus.c
index 45a1c36c5a54e3a841e61cc365d3f36e9a94ba50..c887e61eb384ca98ff27d4f8af69e58c715b5002 100644
--- a/arch/s390/pci/pci_bus.c
+++ b/arch/s390/pci/pci_bus.c
@@ -207,6 +207,16 @@ static int zpci_bus_create_pci_bus(struct zpci_bus *zbus, struct zpci_dev *fr, s
 		return -EFAULT;
 	}
 
+	/*
+	 * On s390 PCI busses are virtualized and the bridge
+	 * devices are invisible to the OS. Furthermore busses
+	 * may exist without a devfn 0 function. Thus the normal
+	 * ARI detection does not work. At the same time fw/hw
+	 * has always enabled ARI when possible. Reflect the actual
+	 * state by setting ari_enabled whenever a device on the bus
+	 * supports it.
+	 */
+	bus->ari_enabled = 1;
 	zbus->bus = bus;
 
 	return 0;
diff --git a/drivers/pci/pci.c b/drivers/pci/pci.c
index b14dd064006cca80ec5275e45a35d6dc2b4d0bbc..8ef3c68280a629449e0a2176d938bf987c68dddf 100644
--- a/drivers/pci/pci.c
+++ b/drivers/pci/pci.c
@@ -3532,11 +3532,11 @@ void pci_configure_ari(struct pci_dev *dev)
 	if (pci_find_ext_capability(dev, PCI_EXT_CAP_ID_ARI)) {
 		pcie_capability_set_word(bridge, PCI_EXP_DEVCTL2,
 					 PCI_EXP_DEVCTL2_ARI);
-		bridge->ari_enabled = 1;
+		dev->bus->ari_enabled = 1;
 	} else {
 		pcie_capability_clear_word(bridge, PCI_EXP_DEVCTL2,
 					   PCI_EXP_DEVCTL2_ARI);
-		bridge->ari_enabled = 0;
+		dev->bus->ari_enabled = 0;
 	}
 }
 
diff --git a/drivers/pci/probe.c b/drivers/pci/probe.c
index 13495b12fbcfae4b890bbd4b2f913742adf6dfed..338bb7e6738d27865e3d50aa3094ca5ab29a6a47 100644
--- a/drivers/pci/probe.c
+++ b/drivers/pci/probe.c
@@ -1216,6 +1216,7 @@ static struct pci_bus *pci_alloc_child_bus(struct pci_bus *parent,
 
 	if (!bridge) {
 		child->dev.parent = parent->bridge;
+		child->ari_enabled = parent->ari_enabled;
 		goto add_dev;
 	}
 
diff --git a/include/linux/pci.h b/include/linux/pci.h
index d1fdf81fbe1e427aecbc951fa3fdf65c20450b05..a9c3dbf17339e523362bd179ad3c7c8c91293cf0 100644
--- a/include/linux/pci.h
+++ b/include/linux/pci.h
@@ -445,7 +445,6 @@ struct pci_dev {
 	unsigned int	irq_reroute_variant:2;	/* Needs IRQ rerouting variant */
 	unsigned int	msi_enabled:1;
 	unsigned int	msix_enabled:1;
-	unsigned int	ari_enabled:1;		/* ARI forwarding */
 	unsigned int	ats_enabled:1;		/* Address Translation Svc */
 	unsigned int	pasid_enabled:1;	/* Process Address Space ID */
 	unsigned int	pri_enabled:1;		/* Page Request Interface */
@@ -691,6 +690,7 @@ struct pci_bus {
 	unsigned int		is_added:1;
 	unsigned int		unsafe_warn:1;	/* warned about RW1C config write */
 	unsigned int		flit_mode:1;	/* Link in Flit mode */
+	unsigned int		ari_enabled:1;	/* ARI forwarding enabled */
 };
 
 #define to_pci_bus(n)	container_of(n, struct pci_bus, dev)
@@ -2740,7 +2740,7 @@ static inline bool pci_is_dev_assigned(struct pci_dev *pdev)
  */
 static inline bool pci_ari_enabled(struct pci_bus *bus)
 {
-	return bus->self && bus->self->ari_enabled;
+	return bus->ari_enabled;
 }
 
 /**

-- 
2.48.1

Re: [PATCH v4 1/2] PCI: Fix isolated PCI function probing with ARI and SR-IOV

[Niklas Schnelle]

On Thu, 2025-10-23 at 17:20 +0200, Niklas Schnelle wrote:
> When the isolated PCI function probing mechanism is used in conjunction
> with ARI or SR-IOV it may not find all available PCI functions. In the
> case of ARI the problem is that next_ari_fn() always returns -ENODEV if
> dev is NULL and thus if fn 0 is missing the scan stops.
> 
> For SR-IOV things are more complex. Here the problem is that the check
> for multifunction may fail. One example where this can occur is if the
> first passed-through function is a VF with devfn 8. Now in
> pci_scan_slot() this means it is fn 0 and thus multifunction doesn't get
> set. Since VFs don't get multifunction set via PCI_HEADER_TYPE_MFD it
> remains unset and probing stops even if there is a devfn 9.
> 
> Now at the moment both of these issues are hidden on s390. The first one
> because ARI is detected as disabled as struct pci_bus's self is NULL
> even though firmware does enable and use ARI. The second issue is hidden
> as a side effect of commit 25f39d3dcb48 ("s390/pci: Ignore RID for
> isolated VFs"). This is because VFs are either put on their own virtual
> bus if the parent PF is not passed-through to the same instance or VFs
> are hotplugged once SR-IOV is enabled on the parent PF and then
> pci_scan_single_device() is used.
> 
> Still especially the first issue prevents correct detection of ARI and
> the second might be a problem for other users of isolated function
> probing. Fix both issues by keeping things as simple as possible. If
> isolated function probing is enabled simply scan every possible devfn.
> 
> Fixes: 189c6c33ff42 ("PCI: Extend isolated function probing to s390")

Since I'm adding a fixes tag, I should also add Cc stable. Added for
next revision.

> Link: https://lore.kernel.org/linux-pci/d3f11e8562f589ddb2c1c83e74161bd8948084c3.camel@linux.ibm.com/
> Signed-off-by: Niklas Schnelle <schnelle@linux.ibm.com>
> ---
>  drivers/pci/probe.c | 20 ++++++++++++--------
>  1 file changed, 12 insertions(+), 8 deletions(-)
> 
--- snip ---