[PATCH v5 0/2] PCI: AtomicOps: Fix pci_enable_atomic_ops_to_root()

[PATCH v5 0/2] PCI: AtomicOps: Fix pci_enable_atomic_ops_to_root()

[Gerd Bayer]

Hi Bjorn et al.

On s390, AtomicOp Requests are enabled on a PCI function that supports
them, despite the helper being ignorant about the root port's capability
to supporting their completion.

Patch 1: Fix the logic in pci_enable_atomic_ops_to_root()
Patch 2: Update references to PCIe spec in that function.

I did test that the issue is fixed with these patches. Also, I verified
that on a Mellanox/Nvidia ConnectX-6 adapter plugged straight into the
root port of a x86 system still gets AtomicOp Requests enabled.

Due to lacking the required hardware, I did not test this with any PCIe
switches between root port and endpoint. So test exposure to other
environments is highly appreciated. One particularly rare setup is a
RCiEP that might act as a AtomicOps Requestor - but was subject to a
regression in v4 as reported by Sashiko.

v5 of this series tries to make no functional changes for that class of
devices either.

Signed-off-by: Gerd Bayer <gbayer@linux.ibm.com>
---
Changes in v5:
- Introduce new pcibios_connects_to_atomicops_capable_rc() so arch's can
  declare AtomicOps support outside of PCIe config space. Defaults to
  "true" - except s390.
- rebase to 7.0-rc5
- Link to v4: https://lore.kernel.org/r/20260313-fix_pciatops-v4-0-93bc70a63935@linux.ibm.com

Changes in v4:
- drop patch 1 - it will become the base of a new series
- previous patch 2, now 1: reword commit message
- add a new patch to update references to PCI spec within
  pci_enable_atomic_ops_to_root()
- rebase to latest master
- Link to v3: https://lore.kernel.org/r/20260306-fix_pciatops-v3-0-99d12bcafb19@linux.ibm.com

Changes in v3:
- rebase to 7.0-rc2
- gentle ping
- add netdev and rdma lists for awareness
- Link to v2: https://lore.kernel.org/r/20251216-fix_pciatops-v2-0-d013e9b7e2ee@linux.ibm.com

Changes in v2:
- rebase to 6.19-rc1
- otherwise unchanged to v1
- Link to v1: https://lore.kernel.org/r/20251110-fix_pciatops-v1-0-edc58a57b62e@linux.ibm.com

---
Gerd Bayer (2):
      PCI: AtomicOps: Do not enable without support in root complex
      PCI: AtomicOps: Update references to PCIe spec

 arch/s390/pci/pci.c |  5 +++++
 drivers/pci/pci.c   | 52 ++++++++++++++++++++++++++++++++--------------------
 include/linux/pci.h |  1 +
 3 files changed, 38 insertions(+), 20 deletions(-)
---
base-commit: c369299895a591d96745d6492d4888259b004a9e
change-id: 20251106-fix_pciatops-7e8608eccb03

Best regards,
-- 
Gerd Bayer <gbayer@linux.ibm.com>

[PATCH v5 1/2] PCI: AtomicOps: Do not enable without support in root complex

[Gerd Bayer]

When inspecting the config space of a Connect-X physical function in an
s390 system after it was initialized by the mlx5_core device driver, we
found the function to be enabled to request AtomicOps despite the
system's root-complex lacking support for completing them:

1ed0:00:00.1 Ethernet controller: Mellanox Technologies MT2894 Family [ConnectX-6 Lx]
	Subsystem: Mellanox Technologies Device 0002
  [...]
	DevCtl2: Completion Timeout: 50us to 50ms, TimeoutDis-
		 AtomicOpsCtl: ReqEn+
		 IDOReq- IDOCompl- LTR- EmergencyPowerReductionReq-
		 10BitTagReq- OBFF Disabled, EETLPPrefixBlk-

Turns out the device driver calls pci_enable_atomic_ops_to_root() which
defaulted to enable AtomicOps requests even if it had no information
about the root-port that the PCIe device is attached to. Similarly,
AtomicOps requests are enabled for root-complex integrated endpoints
(RCiEPs) unconditionally.

Change the logic of pci_enable_atomic_ops_to_root() to fully traverse the
PCIe tree upwards, check that the bridge devices support delivering
AtomicOps transactions, and finally check that there is a root-port at
the end that does support completing AtomicOps - or that the support for
completing AtomicOps at the root complex is announced through some other
arch-specific way.

This announcement is implemented through the new
pcibios_connects_to_atomicops_capable_rc() function - with a default
implementation to always return "true" to leave the semantics for RCiEPs
intact. For s390, override pcibios_connects_to_atomicops_capable_rc() to
always return "false".

Do not change the enablement of AtomicOps requests if there is no
positive confirmation that the root complex can complete PCIe AtomicOps.

Reported-by: Alexander Schmidt <alexs@linux.ibm.com>
Cc: stable@vger.kernel.org
Fixes: 430a23689dea ("PCI: Add pci_enable_atomic_ops_to_root()")
Signed-off-by: Gerd Bayer <gbayer@linux.ibm.com>
---
 arch/s390/pci/pci.c |  5 +++++
 drivers/pci/pci.c   | 46 +++++++++++++++++++++++++++++-----------------
 include/linux/pci.h |  1 +
 3 files changed, 35 insertions(+), 17 deletions(-)

diff --git a/arch/s390/pci/pci.c b/arch/s390/pci/pci.c
index 2a430722cbe415dd56c92fed2e513e524f46481a..a13235d3218e8ca451e25fe8d9094500fa21aa26 100644
--- a/arch/s390/pci/pci.c
+++ b/arch/s390/pci/pci.c
@@ -265,6 +265,11 @@ static int zpci_cfg_store(struct zpci_dev *zdev, int offset, u32 val, u8 len)
 	return rc;
 }
 
+bool pcibios_connects_to_atomicops_capable_rc(struct pci_dev *pdev, u32 cap_mask)
+{
+	return false;
+}
+
 resource_size_t pcibios_align_resource(void *data, const struct resource *res,
 				       resource_size_t size,
 				       resource_size_t align)
diff --git a/drivers/pci/pci.c b/drivers/pci/pci.c
index 8479c2e1f74f1044416281aba11bf071ea89488a..c1143f8e6b2a0f029feb3c4390ac6f33837f6de1 100644
--- a/drivers/pci/pci.c
+++ b/drivers/pci/pci.c
@@ -3660,6 +3660,21 @@ void pci_acs_init(struct pci_dev *dev)
 	pci_disable_broken_acs_cap(dev);
 }
 
+
+static bool pci_is_atomicops_capable_rp(struct pci_dev *dev, u32 cap, u32 cap_mask)
+{
+	if ((!dev) ||
+	     !(pci_pcie_type(dev) != PCI_EXP_TYPE_ROOT_PORT))
+		return false;
+
+	return ((cap & cap_mask) == cap_mask);
+}
+
+bool __weak pcibios_connects_to_atomicops_capable_rc(struct pci_dev *pdev, u32 cap_mask)
+{
+	return true;
+}
+
 /**
  * pci_enable_atomic_ops_to_root - enable AtomicOp requests to root port
  * @dev: the PCI device
@@ -3676,7 +3691,7 @@ void pci_acs_init(struct pci_dev *dev)
 int pci_enable_atomic_ops_to_root(struct pci_dev *dev, u32 cap_mask)
 {
 	struct pci_bus *bus = dev->bus;
-	struct pci_dev *bridge;
+	struct pci_dev *bridge = NULL;
 	u32 cap, ctl2;
 
 	/*
@@ -3714,29 +3729,26 @@ int pci_enable_atomic_ops_to_root(struct pci_dev *dev, u32 cap_mask)
 		switch (pci_pcie_type(bridge)) {
 		/* Ensure switch ports support AtomicOp routing */
 		case PCI_EXP_TYPE_UPSTREAM:
-		case PCI_EXP_TYPE_DOWNSTREAM:
-			if (!(cap & PCI_EXP_DEVCAP2_ATOMIC_ROUTE))
-				return -EINVAL;
-			break;
-
-		/* Ensure root port supports all the sizes we care about */
-		case PCI_EXP_TYPE_ROOT_PORT:
-			if ((cap & cap_mask) != cap_mask)
-				return -EINVAL;
-			break;
-		}
-
-		/* Ensure upstream ports don't block AtomicOps on egress */
-		if (pci_pcie_type(bridge) == PCI_EXP_TYPE_UPSTREAM) {
+			/* Upstream ports must not block AtomicOps on egress */
 			pcie_capability_read_dword(bridge, PCI_EXP_DEVCTL2,
 						   &ctl2);
 			if (ctl2 & PCI_EXP_DEVCTL2_ATOMIC_EGRESS_BLOCK)
 				return -EINVAL;
+			fallthrough;
+		/* All switch ports need to route AtomicOps */
+		case PCI_EXP_TYPE_DOWNSTREAM:
+			if (!(cap & PCI_EXP_DEVCAP2_ATOMIC_ROUTE))
+				return -EINVAL;
+			break;
 		}
-
 		bus = bus->parent;
 	}
 
+	/* Finally, last bridge must be root port and support requested sizes */
+	if (!(pci_is_atomicops_capable_rp(bridge, cap, cap_mask) ||
+	     pcibios_connects_to_atomicops_capable_rc(dev, cap_mask)))
+		return -EINVAL;
+
 	pcie_capability_set_word(dev, PCI_EXP_DEVCTL2,
 				 PCI_EXP_DEVCTL2_ATOMIC_REQ);
 	return 0;
@@ -3813,7 +3825,7 @@ static int __pci_request_region(struct pci_dev *pdev, int bar,
 
 err_out:
 	pci_warn(pdev, "BAR %d: can't reserve %pR\n", bar,
-		 &pdev->resource[bar]);
+			&pdev->resource[bar]);
 	return -EBUSY;
 }
 
diff --git a/include/linux/pci.h b/include/linux/pci.h
index 1c270f1d512301de4d462fe7e5097c32af5c6f8d..498f266c9838c55e9b03d03fef49a82358047f4f 100644
--- a/include/linux/pci.h
+++ b/include/linux/pci.h
@@ -692,6 +692,7 @@ void pci_set_host_bridge_release(struct pci_host_bridge *bridge,
 				 void *release_data);
 
 int pcibios_root_bridge_prepare(struct pci_host_bridge *bridge);
+bool pcibios_connects_to_atomicops_capable_rc(struct pci_dev *pdev, u32 cap_mask);
 
 #define PCI_REGION_FLAG_MASK	0x0fU	/* These bits of resource flags tell us the PCI region flags */
 

-- 
2.51.0

[PATCH v5 2/2] PCI: AtomicOps: Update references to PCIe spec

[Gerd Bayer]

Point to the relevant sections in the most recent release 7.0 of the
PCIe spec. Text has mostly just moved around without any semantic
change.

Signed-off-by: Gerd Bayer <gbayer@linux.ibm.com>
---
 drivers/pci/pci.c | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/drivers/pci/pci.c b/drivers/pci/pci.c
index c1143f8e6b2a0f029feb3c4390ac6f33837f6de1..d0d52ccf1ac18ce5a1917a713af9ab87b61cb7f9 100644
--- a/drivers/pci/pci.c
+++ b/drivers/pci/pci.c
@@ -3695,7 +3695,7 @@ int pci_enable_atomic_ops_to_root(struct pci_dev *dev, u32 cap_mask)
 	u32 cap, ctl2;
 
 	/*
-	 * Per PCIe r5.0, sec 9.3.5.10, the AtomicOp Requester Enable bit
+	 * Per PCIe r7.0, sec 7.5.3.16, the AtomicOp Requester Enable bit
 	 * in Device Control 2 is reserved in VFs and the PF value applies
 	 * to all associated VFs.
 	 */
@@ -3706,9 +3706,9 @@ int pci_enable_atomic_ops_to_root(struct pci_dev *dev, u32 cap_mask)
 		return -EINVAL;
 
 	/*
-	 * Per PCIe r4.0, sec 6.15, endpoints and root ports may be
+	 * Per PCIe r7.0, sec 6.15, endpoints and root ports may be
 	 * AtomicOp requesters.  For now, we only support endpoints as
-	 * requesters and root ports as completers.  No endpoints as
+	 * requesters and root ports as completers. No endpoints as
 	 * completers, and no peer-to-peer.
 	 */
 

-- 
2.51.0

Re: [PATCH v5 1/2] PCI: AtomicOps: Do not enable without support in root complex

[Ilpo Järvinen]

On Mon, 23 Mar 2026, Gerd Bayer wrote:

> When inspecting the config space of a Connect-X physical function in an
> s390 system after it was initialized by the mlx5_core device driver, we
> found the function to be enabled to request AtomicOps despite the
> system's root-complex lacking support for completing them:
> 
> 1ed0:00:00.1 Ethernet controller: Mellanox Technologies MT2894 Family [ConnectX-6 Lx]
> 	Subsystem: Mellanox Technologies Device 0002
>   [...]
> 	DevCtl2: Completion Timeout: 50us to 50ms, TimeoutDis-
> 		 AtomicOpsCtl: ReqEn+
> 		 IDOReq- IDOCompl- LTR- EmergencyPowerReductionReq-
> 		 10BitTagReq- OBFF Disabled, EETLPPrefixBlk-
> 
> Turns out the device driver calls pci_enable_atomic_ops_to_root() which
> defaulted to enable AtomicOps requests even if it had no information
> about the root-port that the PCIe device is attached to. Similarly,
> AtomicOps requests are enabled for root-complex integrated endpoints
> (RCiEPs) unconditionally.
> 
> Change the logic of pci_enable_atomic_ops_to_root() to fully traverse the
> PCIe tree upwards, check that the bridge devices support delivering
> AtomicOps transactions, and finally check that there is a root-port at
> the end that does support completing AtomicOps - or that the support for
> completing AtomicOps at the root complex is announced through some other
> arch-specific way.
> 
> This announcement is implemented through the new
> pcibios_connects_to_atomicops_capable_rc() function - with a default
> implementation to always return "true" to leave the semantics for RCiEPs
> intact. For s390, override pcibios_connects_to_atomicops_capable_rc() to
> always return "false".
> 
> Do not change the enablement of AtomicOps requests if there is no
> positive confirmation that the root complex can complete PCIe AtomicOps.
> 
> Reported-by: Alexander Schmidt <alexs@linux.ibm.com>
> Cc: stable@vger.kernel.org
> Fixes: 430a23689dea ("PCI: Add pci_enable_atomic_ops_to_root()")
> Signed-off-by: Gerd Bayer <gbayer@linux.ibm.com>
> ---
>  arch/s390/pci/pci.c |  5 +++++
>  drivers/pci/pci.c   | 46 +++++++++++++++++++++++++++++-----------------
>  include/linux/pci.h |  1 +
>  3 files changed, 35 insertions(+), 17 deletions(-)
> 
> diff --git a/arch/s390/pci/pci.c b/arch/s390/pci/pci.c
> index 2a430722cbe415dd56c92fed2e513e524f46481a..a13235d3218e8ca451e25fe8d9094500fa21aa26 100644
> --- a/arch/s390/pci/pci.c
> +++ b/arch/s390/pci/pci.c
> @@ -265,6 +265,11 @@ static int zpci_cfg_store(struct zpci_dev *zdev, int offset, u32 val, u8 len)
>  	return rc;
>  }
>  
> +bool pcibios_connects_to_atomicops_capable_rc(struct pci_dev *pdev, u32 cap_mask)
> +{
> +	return false;
> +}
> +
>  resource_size_t pcibios_align_resource(void *data, const struct resource *res,
>  				       resource_size_t size,
>  				       resource_size_t align)
> diff --git a/drivers/pci/pci.c b/drivers/pci/pci.c
> index 8479c2e1f74f1044416281aba11bf071ea89488a..c1143f8e6b2a0f029feb3c4390ac6f33837f6de1 100644
> --- a/drivers/pci/pci.c
> +++ b/drivers/pci/pci.c
> @@ -3660,6 +3660,21 @@ void pci_acs_init(struct pci_dev *dev)
>  	pci_disable_broken_acs_cap(dev);
>  }
>  
> +

Extra newline.

> +static bool pci_is_atomicops_capable_rp(struct pci_dev *dev, u32 cap, u32 cap_mask)
> +{
> +	if ((!dev) ||

Unnecessary parenthesis.

> +	     !(pci_pcie_type(dev) != PCI_EXP_TYPE_ROOT_PORT))

This fits to one line.

> +		return false;
> +
> +	return ((cap & cap_mask) == cap_mask);

Extra parenthesis.

> +}
> +
> +bool __weak pcibios_connects_to_atomicops_capable_rc(struct pci_dev *pdev, u32 cap_mask)
> +{
> +	return true;
> +}
> +
>  /**
>   * pci_enable_atomic_ops_to_root - enable AtomicOp requests to root port
>   * @dev: the PCI device
> @@ -3676,7 +3691,7 @@ void pci_acs_init(struct pci_dev *dev)
>  int pci_enable_atomic_ops_to_root(struct pci_dev *dev, u32 cap_mask)
>  {
>  	struct pci_bus *bus = dev->bus;
> -	struct pci_dev *bridge;
> +	struct pci_dev *bridge = NULL;
>  	u32 cap, ctl2;
>  
>  	/*
> @@ -3714,29 +3729,26 @@ int pci_enable_atomic_ops_to_root(struct pci_dev *dev, u32 cap_mask)
>  		switch (pci_pcie_type(bridge)) {
>  		/* Ensure switch ports support AtomicOp routing */
>  		case PCI_EXP_TYPE_UPSTREAM:
> -		case PCI_EXP_TYPE_DOWNSTREAM:
> -			if (!(cap & PCI_EXP_DEVCAP2_ATOMIC_ROUTE))
> -				return -EINVAL;
> -			break;
> -
> -		/* Ensure root port supports all the sizes we care about */
> -		case PCI_EXP_TYPE_ROOT_PORT:
> -			if ((cap & cap_mask) != cap_mask)
> -				return -EINVAL;
> -			break;
> -		}
> -
> -		/* Ensure upstream ports don't block AtomicOps on egress */
> -		if (pci_pcie_type(bridge) == PCI_EXP_TYPE_UPSTREAM) {
> +			/* Upstream ports must not block AtomicOps on egress */
>  			pcie_capability_read_dword(bridge, PCI_EXP_DEVCTL2,
>  						   &ctl2);
>  			if (ctl2 & PCI_EXP_DEVCTL2_ATOMIC_EGRESS_BLOCK)
>  				return -EINVAL;
> +			fallthrough;
> +		/* All switch ports need to route AtomicOps */
> +		case PCI_EXP_TYPE_DOWNSTREAM:
> +			if (!(cap & PCI_EXP_DEVCAP2_ATOMIC_ROUTE))
> +				return -EINVAL;
> +			break;
>  		}
> -
>  		bus = bus->parent;
>  	}
>  
> +	/* Finally, last bridge must be root port and support requested sizes */
> +	if (!(pci_is_atomicops_capable_rp(bridge, cap, cap_mask) ||
> +	     pcibios_connects_to_atomicops_capable_rc(dev, cap_mask)))
> +		return -EINVAL;
> +
>  	pcie_capability_set_word(dev, PCI_EXP_DEVCTL2,
>  				 PCI_EXP_DEVCTL2_ATOMIC_REQ);
>  	return 0;
> @@ -3813,7 +3825,7 @@ static int __pci_request_region(struct pci_dev *pdev, int bar,
>  
>  err_out:
>  	pci_warn(pdev, "BAR %d: can't reserve %pR\n", bar,
> -		 &pdev->resource[bar]);
> +			&pdev->resource[bar]);

Unrelated change.

>  	return -EBUSY;
>  }
>  
> diff --git a/include/linux/pci.h b/include/linux/pci.h
> index 1c270f1d512301de4d462fe7e5097c32af5c6f8d..498f266c9838c55e9b03d03fef49a82358047f4f 100644
> --- a/include/linux/pci.h
> +++ b/include/linux/pci.h
> @@ -692,6 +692,7 @@ void pci_set_host_bridge_release(struct pci_host_bridge *bridge,
>  				 void *release_data);
>  
>  int pcibios_root_bridge_prepare(struct pci_host_bridge *bridge);
> +bool pcibios_connects_to_atomicops_capable_rc(struct pci_dev *pdev, u32 cap_mask);
>  
>  #define PCI_REGION_FLAG_MASK	0x0fU	/* These bits of resource flags tell us the PCI region flags */
>  
> 
> 

-- 
 i.