[PATCH v4 0/5] PCI endpoint additional pci_epc_set_bar() checks

[PATCH v4 0/5] PCI endpoint additional pci_epc_set_bar() checks

[Niklas Cassel]

Hello all,

This series adds some extra checks to ensure that it is not possible to
program the iATU with an address which we did not intend to use.

If these checks were in place when testing some of the earlier revisions
of Frank's doorbell patches (which did not handle fixed BARs properly),
we would gotten an error, rather than silently using an address which we
did not intend to use.

Having these checks in place will hopefully avoid similar debugging in the
future.


Kind regards,
Niklas


Changes since v3:
-Added patch 1/5, which fixes iATU programming on DWC PCIe EP controllers.
-Reordered patches to have a more logical ordering.


Niklas Cassel (5):
  PCI: dwc: ep: iATU registers must be written after the BAR_MASK
  PCI: dwc: ep: Add 'address' alignment to 'size' check in
    dw_pcie_prog_ep_inbound_atu()
  PCI: artpec6: Implement dw_pcie_ep operation get_features
  PCI: endpoint: Add size check for fixed size BARs in pci_epc_set_bar()
  PCI: endpoint: Verify that requested BAR size is a power of two

 drivers/pci/controller/dwc/pcie-artpec6.c     | 13 +++++
 .../pci/controller/dwc/pcie-designware-ep.c   | 52 ++++++++++++++-----
 drivers/pci/controller/dwc/pcie-designware.c  |  5 +-
 drivers/pci/controller/dwc/pcie-designware.h  |  2 +-
 drivers/pci/endpoint/pci-epc-core.c           | 14 ++++-
 5 files changed, 67 insertions(+), 19 deletions(-)

-- 
2.47.0

[PATCH v4 1/5] PCI: dwc: ep: iATU registers must be written after the BAR_MASK

[Niklas Cassel]

The DWC Databook description for the LWR_TARGET_RW and LWR_TARGET_HW fields
in the IATU_LWR_TARGET_ADDR_OFF_INBOUND_i registers state that:
"Field size depends on log2(BAR_MASK+1) in BAR match mode."

I.e. only the upper bits are writable, and the number of writable bits is
dependent on the configured BAR_MASK.

If we do not write the BAR_MASK before writing the iATU registers, we are
relying the reset value of the BAR_MASK being larger than the requested
size of the first set_bar() call. The reset value of the BAR_MASK is SoC
dependent.

Thus, if the first set_bar() call requests a size that is larger than the
reset value of the BAR_MASK, the iATU will try to write to read-only bits,
which will cause the iATU to end up redirecting to a physical address that
is different from the address that was intended.

Thus, we should always write the iATU registers after writing the BAR_MASK.

Since set_bar() supports dynamically changing the physical address of a
BAR, this change is slightly bigger than a single line change.

While at it, add comments which clarifies the support for dynamically
changing the physical address of a BAR. (Which was previously missing.)

Fixes: f8aed6ec624f ("PCI: dwc: designware: Add EP mode support")
Signed-off-by: Niklas Cassel <cassel@kernel.org>
---
 .../pci/controller/dwc/pcie-designware-ep.c   | 46 ++++++++++++++-----
 1 file changed, 34 insertions(+), 12 deletions(-)

diff --git a/drivers/pci/controller/dwc/pcie-designware-ep.c b/drivers/pci/controller/dwc/pcie-designware-ep.c
index 507e40bd18c8..34d60142ffb5 100644
--- a/drivers/pci/controller/dwc/pcie-designware-ep.c
+++ b/drivers/pci/controller/dwc/pcie-designware-ep.c
@@ -222,19 +222,30 @@ static int dw_pcie_ep_set_bar(struct pci_epc *epc, u8 func_no, u8 vfunc_no,
 	if ((flags & PCI_BASE_ADDRESS_MEM_TYPE_64) && (bar & 1))
 		return -EINVAL;
 
-	reg = PCI_BASE_ADDRESS_0 + (4 * bar);
-
-	if (!(flags & PCI_BASE_ADDRESS_SPACE))
-		type = PCIE_ATU_TYPE_MEM;
-	else
-		type = PCIE_ATU_TYPE_IO;
+	/*
+	 * Certain EPF drivers dynamically change the physical address of a BAR
+	 * (i.e. they call set_bar() twice, without ever calling clear_bar(), as
+	 * calling clear_bar() would clear the BAR's PCI address assigned by the
+	 * host).
+	 */
+	if (ep->epf_bar[bar]) {
+		/*
+		 * We can only dynamically change a BAR if the new configuration
+		 * doesn't fundamentally differ from the existing configuration.
+		 */
+		if (ep->epf_bar[bar]->barno != bar ||
+		    ep->epf_bar[bar]->size != size ||
+		    ep->epf_bar[bar]->flags != flags)
+			return -EINVAL;
 
-	ret = dw_pcie_ep_inbound_atu(ep, func_no, type, epf_bar->phys_addr, bar);
-	if (ret)
-		return ret;
+		/*
+		 * When dynamically changing a BAR, skip writing the BAR reg, as
+		 * that would clear the BAR's PCI address assigned by the host.
+		 */
+		goto config_atu;
+	}
 
-	if (ep->epf_bar[bar])
-		return 0;
+	reg = PCI_BASE_ADDRESS_0 + (4 * bar);
 
 	dw_pcie_dbi_ro_wr_en(pci);
 
@@ -246,9 +257,20 @@ static int dw_pcie_ep_set_bar(struct pci_epc *epc, u8 func_no, u8 vfunc_no,
 		dw_pcie_ep_writel_dbi(ep, func_no, reg + 4, 0);
 	}
 
-	ep->epf_bar[bar] = epf_bar;
 	dw_pcie_dbi_ro_wr_dis(pci);
 
+config_atu:
+	if (!(flags & PCI_BASE_ADDRESS_SPACE))
+		type = PCIE_ATU_TYPE_MEM;
+	else
+		type = PCIE_ATU_TYPE_IO;
+
+	ret = dw_pcie_ep_inbound_atu(ep, func_no, type, epf_bar->phys_addr, bar);
+	if (ret)
+		return ret;
+
+	ep->epf_bar[bar] = epf_bar;
+
 	return 0;
 }
 
-- 
2.47.0

Re: [PATCH v4 1/5] PCI: dwc: ep: iATU registers must be written after the BAR_MASK

[Manivannan Sadhasivam]

On Fri, Nov 22, 2024 at 12:57:10PM +0100, Niklas Cassel wrote:
> The DWC Databook description for the LWR_TARGET_RW and LWR_TARGET_HW fields
> in the IATU_LWR_TARGET_ADDR_OFF_INBOUND_i registers state that:
> "Field size depends on log2(BAR_MASK+1) in BAR match mode."
> 
> I.e. only the upper bits are writable, and the number of writable bits is
> dependent on the configured BAR_MASK.
> 
> If we do not write the BAR_MASK before writing the iATU registers, we are
> relying the reset value of the BAR_MASK being larger than the requested
> size of the first set_bar() call. The reset value of the BAR_MASK is SoC
> dependent.
> 
> Thus, if the first set_bar() call requests a size that is larger than the
> reset value of the BAR_MASK, the iATU will try to write to read-only bits,
> which will cause the iATU to end up redirecting to a physical address that
> is different from the address that was intended.
> 
> Thus, we should always write the iATU registers after writing the BAR_MASK.
> 
> Since set_bar() supports dynamically changing the physical address of a
> BAR, this change is slightly bigger than a single line change.
> 
> While at it, add comments which clarifies the support for dynamically
> changing the physical address of a BAR. (Which was previously missing.)
> 

This is not strictly relevant to this patch, so should be separate.

> Fixes: f8aed6ec624f ("PCI: dwc: designware: Add EP mode support")
> Signed-off-by: Niklas Cassel <cassel@kernel.org>

LGTM! This patch should only have the iATU config change and should be
backported also. 

- Mani

> ---
>  .../pci/controller/dwc/pcie-designware-ep.c   | 46 ++++++++++++++-----
>  1 file changed, 34 insertions(+), 12 deletions(-)
> 
> diff --git a/drivers/pci/controller/dwc/pcie-designware-ep.c b/drivers/pci/controller/dwc/pcie-designware-ep.c
> index 507e40bd18c8..34d60142ffb5 100644
> --- a/drivers/pci/controller/dwc/pcie-designware-ep.c
> +++ b/drivers/pci/controller/dwc/pcie-designware-ep.c
> @@ -222,19 +222,30 @@ static int dw_pcie_ep_set_bar(struct pci_epc *epc, u8 func_no, u8 vfunc_no,
>  	if ((flags & PCI_BASE_ADDRESS_MEM_TYPE_64) && (bar & 1))
>  		return -EINVAL;
>  
> -	reg = PCI_BASE_ADDRESS_0 + (4 * bar);
> -
> -	if (!(flags & PCI_BASE_ADDRESS_SPACE))
> -		type = PCIE_ATU_TYPE_MEM;
> -	else
> -		type = PCIE_ATU_TYPE_IO;
> +	/*
> +	 * Certain EPF drivers dynamically change the physical address of a BAR
> +	 * (i.e. they call set_bar() twice, without ever calling clear_bar(), as
> +	 * calling clear_bar() would clear the BAR's PCI address assigned by the
> +	 * host).
> +	 */
> +	if (ep->epf_bar[bar]) {
> +		/*
> +		 * We can only dynamically change a BAR if the new configuration
> +		 * doesn't fundamentally differ from the existing configuration.
> +		 */
> +		if (ep->epf_bar[bar]->barno != bar ||
> +		    ep->epf_bar[bar]->size != size ||
> +		    ep->epf_bar[bar]->flags != flags)
> +			return -EINVAL;
>  
> -	ret = dw_pcie_ep_inbound_atu(ep, func_no, type, epf_bar->phys_addr, bar);
> -	if (ret)
> -		return ret;
> +		/*
> +		 * When dynamically changing a BAR, skip writing the BAR reg, as
> +		 * that would clear the BAR's PCI address assigned by the host.
> +		 */
> +		goto config_atu;
> +	}
>  
> -	if (ep->epf_bar[bar])
> -		return 0;
> +	reg = PCI_BASE_ADDRESS_0 + (4 * bar);
>  
>  	dw_pcie_dbi_ro_wr_en(pci);
>  
> @@ -246,9 +257,20 @@ static int dw_pcie_ep_set_bar(struct pci_epc *epc, u8 func_no, u8 vfunc_no,
>  		dw_pcie_ep_writel_dbi(ep, func_no, reg + 4, 0);
>  	}
>  
> -	ep->epf_bar[bar] = epf_bar;
>  	dw_pcie_dbi_ro_wr_dis(pci);
>  
> +config_atu:
> +	if (!(flags & PCI_BASE_ADDRESS_SPACE))
> +		type = PCIE_ATU_TYPE_MEM;
> +	else
> +		type = PCIE_ATU_TYPE_IO;
> +
> +	ret = dw_pcie_ep_inbound_atu(ep, func_no, type, epf_bar->phys_addr, bar);
> +	if (ret)
> +		return ret;
> +
> +	ep->epf_bar[bar] = epf_bar;
> +
>  	return 0;
>  }
>  
> -- 
> 2.47.0
> 

-- 
மணிவண்ணன் சதாசிவம்

[PATCH v4 2/5] PCI: dwc: ep: Add 'address' alignment to 'size' check in dw_pcie_prog_ep_inbound_atu()

[Niklas Cassel]

dw_pcie_prog_ep_inbound_atu() is used to program an inbound iATU in
"BAR Match Mode".

A memory address returned by e.g. kmalloc() is guaranteed to have natural
alignment (aligned to the size of the allocation). It is however not
guaranteed that pci_epc_set_bar() (and thus dw_pcie_prog_ep_inbound_atu())
is supplied an address that has natural alignment. (An EPF driver can send
in an arbitrary physical address to pci_epc_set_bar().)

The DWC Databook description for the LWR_TARGET_RW and LWR_TARGET_HW fields
in the IATU_LWR_TARGET_ADDR_OFF_INBOUND_i registers state that:
"Field size depends on log2(BAR_MASK+1) in BAR match mode."

I.e. only the upper bits are writable, and the number of writable bits is
dependent on the configured BAR_MASK.

Add a check to ensure that the physical address programmed in the iATU is
aligned to the size of the BAR (BAR_MASK+1), as without this, we can get
hard to debug errors, as we could write to bits that are read-only (without
getting a write error), which could cause the iATU to end up redirecting to
a physical address that is different from the address that we intended.

Signed-off-by: Niklas Cassel <cassel@kernel.org>
---
 drivers/pci/controller/dwc/pcie-designware-ep.c | 8 +++++---
 drivers/pci/controller/dwc/pcie-designware.c    | 5 +++--
 drivers/pci/controller/dwc/pcie-designware.h    | 2 +-
 3 files changed, 9 insertions(+), 6 deletions(-)

diff --git a/drivers/pci/controller/dwc/pcie-designware-ep.c b/drivers/pci/controller/dwc/pcie-designware-ep.c
index 34d60142ffb5..ff4350e7adb6 100644
--- a/drivers/pci/controller/dwc/pcie-designware-ep.c
+++ b/drivers/pci/controller/dwc/pcie-designware-ep.c
@@ -128,7 +128,8 @@ static int dw_pcie_ep_write_header(struct pci_epc *epc, u8 func_no, u8 vfunc_no,
 }
 
 static int dw_pcie_ep_inbound_atu(struct dw_pcie_ep *ep, u8 func_no, int type,
-				  dma_addr_t cpu_addr, enum pci_barno bar)
+				  dma_addr_t cpu_addr, enum pci_barno bar,
+				  size_t size)
 {
 	int ret;
 	u32 free_win;
@@ -145,7 +146,7 @@ static int dw_pcie_ep_inbound_atu(struct dw_pcie_ep *ep, u8 func_no, int type,
 	}
 
 	ret = dw_pcie_prog_ep_inbound_atu(pci, func_no, free_win, type,
-					  cpu_addr, bar);
+					  cpu_addr, bar, size);
 	if (ret < 0) {
 		dev_err(pci->dev, "Failed to program IB window\n");
 		return ret;
@@ -265,7 +266,8 @@ static int dw_pcie_ep_set_bar(struct pci_epc *epc, u8 func_no, u8 vfunc_no,
 	else
 		type = PCIE_ATU_TYPE_IO;
 
-	ret = dw_pcie_ep_inbound_atu(ep, func_no, type, epf_bar->phys_addr, bar);
+	ret = dw_pcie_ep_inbound_atu(ep, func_no, type, epf_bar->phys_addr, bar,
+				     size);
 	if (ret)
 		return ret;
 
diff --git a/drivers/pci/controller/dwc/pcie-designware.c b/drivers/pci/controller/dwc/pcie-designware.c
index 6d6cbc8b5b2c..3c683b6119c3 100644
--- a/drivers/pci/controller/dwc/pcie-designware.c
+++ b/drivers/pci/controller/dwc/pcie-designware.c
@@ -597,11 +597,12 @@ int dw_pcie_prog_inbound_atu(struct dw_pcie *pci, int index, int type,
 }
 
 int dw_pcie_prog_ep_inbound_atu(struct dw_pcie *pci, u8 func_no, int index,
-				int type, u64 cpu_addr, u8 bar)
+				int type, u64 cpu_addr, u8 bar, size_t size)
 {
 	u32 retries, val;
 
-	if (!IS_ALIGNED(cpu_addr, pci->region_align))
+	if (!IS_ALIGNED(cpu_addr, pci->region_align) ||
+	    !IS_ALIGNED(cpu_addr, size))
 		return -EINVAL;
 
 	dw_pcie_writel_atu_ib(pci, index, PCIE_ATU_LOWER_TARGET,
diff --git a/drivers/pci/controller/dwc/pcie-designware.h b/drivers/pci/controller/dwc/pcie-designware.h
index 347ab74ac35a..fc0872711672 100644
--- a/drivers/pci/controller/dwc/pcie-designware.h
+++ b/drivers/pci/controller/dwc/pcie-designware.h
@@ -491,7 +491,7 @@ int dw_pcie_prog_outbound_atu(struct dw_pcie *pci,
 int dw_pcie_prog_inbound_atu(struct dw_pcie *pci, int index, int type,
 			     u64 cpu_addr, u64 pci_addr, u64 size);
 int dw_pcie_prog_ep_inbound_atu(struct dw_pcie *pci, u8 func_no, int index,
-				int type, u64 cpu_addr, u8 bar);
+				int type, u64 cpu_addr, u8 bar, size_t size);
 void dw_pcie_disable_atu(struct dw_pcie *pci, u32 dir, int index);
 void dw_pcie_setup(struct dw_pcie *pci);
 void dw_pcie_iatu_detect(struct dw_pcie *pci);
-- 
2.47.0

Re: [PATCH v4 2/5] PCI: dwc: ep: Add 'address' alignment to 'size' check in dw_pcie_prog_ep_inbound_atu()

[Manivannan Sadhasivam]

On Fri, Nov 22, 2024 at 12:57:11PM +0100, Niklas Cassel wrote:
> dw_pcie_prog_ep_inbound_atu() is used to program an inbound iATU in
> "BAR Match Mode".
> 
> A memory address returned by e.g. kmalloc() is guaranteed to have natural
> alignment (aligned to the size of the allocation). It is however not
> guaranteed that pci_epc_set_bar() (and thus dw_pcie_prog_ep_inbound_atu())
> is supplied an address that has natural alignment. (An EPF driver can send
> in an arbitrary physical address to pci_epc_set_bar().)
> 
> The DWC Databook description for the LWR_TARGET_RW and LWR_TARGET_HW fields
> in the IATU_LWR_TARGET_ADDR_OFF_INBOUND_i registers state that:
> "Field size depends on log2(BAR_MASK+1) in BAR match mode."
> 
> I.e. only the upper bits are writable, and the number of writable bits is
> dependent on the configured BAR_MASK.
> 
> Add a check to ensure that the physical address programmed in the iATU is
> aligned to the size of the BAR (BAR_MASK+1), as without this, we can get
> hard to debug errors, as we could write to bits that are read-only (without
> getting a write error), which could cause the iATU to end up redirecting to
> a physical address that is different from the address that we intended.
> 
> Signed-off-by: Niklas Cassel <cassel@kernel.org>

Reviewed-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>

- Mani

> ---
>  drivers/pci/controller/dwc/pcie-designware-ep.c | 8 +++++---
>  drivers/pci/controller/dwc/pcie-designware.c    | 5 +++--
>  drivers/pci/controller/dwc/pcie-designware.h    | 2 +-
>  3 files changed, 9 insertions(+), 6 deletions(-)
> 
> diff --git a/drivers/pci/controller/dwc/pcie-designware-ep.c b/drivers/pci/controller/dwc/pcie-designware-ep.c
> index 34d60142ffb5..ff4350e7adb6 100644
> --- a/drivers/pci/controller/dwc/pcie-designware-ep.c
> +++ b/drivers/pci/controller/dwc/pcie-designware-ep.c
> @@ -128,7 +128,8 @@ static int dw_pcie_ep_write_header(struct pci_epc *epc, u8 func_no, u8 vfunc_no,
>  }
>  
>  static int dw_pcie_ep_inbound_atu(struct dw_pcie_ep *ep, u8 func_no, int type,
> -				  dma_addr_t cpu_addr, enum pci_barno bar)
> +				  dma_addr_t cpu_addr, enum pci_barno bar,
> +				  size_t size)
>  {
>  	int ret;
>  	u32 free_win;
> @@ -145,7 +146,7 @@ static int dw_pcie_ep_inbound_atu(struct dw_pcie_ep *ep, u8 func_no, int type,
>  	}
>  
>  	ret = dw_pcie_prog_ep_inbound_atu(pci, func_no, free_win, type,
> -					  cpu_addr, bar);
> +					  cpu_addr, bar, size);
>  	if (ret < 0) {
>  		dev_err(pci->dev, "Failed to program IB window\n");
>  		return ret;
> @@ -265,7 +266,8 @@ static int dw_pcie_ep_set_bar(struct pci_epc *epc, u8 func_no, u8 vfunc_no,
>  	else
>  		type = PCIE_ATU_TYPE_IO;
>  
> -	ret = dw_pcie_ep_inbound_atu(ep, func_no, type, epf_bar->phys_addr, bar);
> +	ret = dw_pcie_ep_inbound_atu(ep, func_no, type, epf_bar->phys_addr, bar,
> +				     size);
>  	if (ret)
>  		return ret;
>  
> diff --git a/drivers/pci/controller/dwc/pcie-designware.c b/drivers/pci/controller/dwc/pcie-designware.c
> index 6d6cbc8b5b2c..3c683b6119c3 100644
> --- a/drivers/pci/controller/dwc/pcie-designware.c
> +++ b/drivers/pci/controller/dwc/pcie-designware.c
> @@ -597,11 +597,12 @@ int dw_pcie_prog_inbound_atu(struct dw_pcie *pci, int index, int type,
>  }
>  
>  int dw_pcie_prog_ep_inbound_atu(struct dw_pcie *pci, u8 func_no, int index,
> -				int type, u64 cpu_addr, u8 bar)
> +				int type, u64 cpu_addr, u8 bar, size_t size)
>  {
>  	u32 retries, val;
>  
> -	if (!IS_ALIGNED(cpu_addr, pci->region_align))
> +	if (!IS_ALIGNED(cpu_addr, pci->region_align) ||
> +	    !IS_ALIGNED(cpu_addr, size))
>  		return -EINVAL;
>  
>  	dw_pcie_writel_atu_ib(pci, index, PCIE_ATU_LOWER_TARGET,
> diff --git a/drivers/pci/controller/dwc/pcie-designware.h b/drivers/pci/controller/dwc/pcie-designware.h
> index 347ab74ac35a..fc0872711672 100644
> --- a/drivers/pci/controller/dwc/pcie-designware.h
> +++ b/drivers/pci/controller/dwc/pcie-designware.h
> @@ -491,7 +491,7 @@ int dw_pcie_prog_outbound_atu(struct dw_pcie *pci,
>  int dw_pcie_prog_inbound_atu(struct dw_pcie *pci, int index, int type,
>  			     u64 cpu_addr, u64 pci_addr, u64 size);
>  int dw_pcie_prog_ep_inbound_atu(struct dw_pcie *pci, u8 func_no, int index,
> -				int type, u64 cpu_addr, u8 bar);
> +				int type, u64 cpu_addr, u8 bar, size_t size);
>  void dw_pcie_disable_atu(struct dw_pcie *pci, u32 dir, int index);
>  void dw_pcie_setup(struct dw_pcie *pci);
>  void dw_pcie_iatu_detect(struct dw_pcie *pci);
> -- 
> 2.47.0
> 

-- 
மணிவண்ணன் சதாசிவம்

[PATCH v4 3/5] PCI: artpec6: Implement dw_pcie_ep operation get_features

[Niklas Cassel]

All non-DWC EPC drivers implement (struct pci_epc *)->ops->get_features().
All DWC EPC drivers implement (struct dw_pcie_ep *)->ops->get_features(),
except for pcie-artpec6.c.

epc_features has been required in pci-epf-test.c since commit 6613bc2301ba
("PCI: endpoint: Fix NULL pointer dereference for ->get_features()").

A follow-up commit will make further use of epc_features in EPC core code.

Implement epc_features in the only EPC driver where it is currently not
implemented.

Signed-off-by: Niklas Cassel <cassel@kernel.org>
Reviewed-by: Frank Li <Frank.Li@nxp.com>
Acked-by: Jesper Nilsson <jesper.nilsson@axis.com>
---
 drivers/pci/controller/dwc/pcie-artpec6.c | 13 +++++++++++++
 1 file changed, 13 insertions(+)

diff --git a/drivers/pci/controller/dwc/pcie-artpec6.c b/drivers/pci/controller/dwc/pcie-artpec6.c
index f8e7283dacd4..234c8cbcae3a 100644
--- a/drivers/pci/controller/dwc/pcie-artpec6.c
+++ b/drivers/pci/controller/dwc/pcie-artpec6.c
@@ -369,9 +369,22 @@ static int artpec6_pcie_raise_irq(struct dw_pcie_ep *ep, u8 func_no,
 	return 0;
 }
 
+static const struct pci_epc_features artpec6_pcie_epc_features = {
+	.linkup_notifier = false,
+	.msi_capable = true,
+	.msix_capable = false,
+};
+
+static const struct pci_epc_features *
+artpec6_pcie_get_features(struct dw_pcie_ep *ep)
+{
+	return &artpec6_pcie_epc_features;
+}
+
 static const struct dw_pcie_ep_ops pcie_ep_ops = {
 	.init = artpec6_pcie_ep_init,
 	.raise_irq = artpec6_pcie_raise_irq,
+	.get_features = artpec6_pcie_get_features,
 };
 
 static int artpec6_pcie_probe(struct platform_device *pdev)
-- 
2.47.0

Re: [PATCH v4 3/5] PCI: artpec6: Implement dw_pcie_ep operation get_features

[Manivannan Sadhasivam]

On Fri, Nov 22, 2024 at 12:57:12PM +0100, Niklas Cassel wrote:
> All non-DWC EPC drivers implement (struct pci_epc *)->ops->get_features().
> All DWC EPC drivers implement (struct dw_pcie_ep *)->ops->get_features(),
> except for pcie-artpec6.c.
> 
> epc_features has been required in pci-epf-test.c since commit 6613bc2301ba
> ("PCI: endpoint: Fix NULL pointer dereference for ->get_features()").
> 
> A follow-up commit will make further use of epc_features in EPC core code.
> 
> Implement epc_features in the only EPC driver where it is currently not
> implemented.
> 
> Signed-off-by: Niklas Cassel <cassel@kernel.org>

Reviewed-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>

- Mani

> Reviewed-by: Frank Li <Frank.Li@nxp.com>
> Acked-by: Jesper Nilsson <jesper.nilsson@axis.com>
> ---
>  drivers/pci/controller/dwc/pcie-artpec6.c | 13 +++++++++++++
>  1 file changed, 13 insertions(+)
> 
> diff --git a/drivers/pci/controller/dwc/pcie-artpec6.c b/drivers/pci/controller/dwc/pcie-artpec6.c
> index f8e7283dacd4..234c8cbcae3a 100644
> --- a/drivers/pci/controller/dwc/pcie-artpec6.c
> +++ b/drivers/pci/controller/dwc/pcie-artpec6.c
> @@ -369,9 +369,22 @@ static int artpec6_pcie_raise_irq(struct dw_pcie_ep *ep, u8 func_no,
>  	return 0;
>  }
>  
> +static const struct pci_epc_features artpec6_pcie_epc_features = {
> +	.linkup_notifier = false,
> +	.msi_capable = true,
> +	.msix_capable = false,
> +};
> +
> +static const struct pci_epc_features *
> +artpec6_pcie_get_features(struct dw_pcie_ep *ep)
> +{
> +	return &artpec6_pcie_epc_features;
> +}
> +
>  static const struct dw_pcie_ep_ops pcie_ep_ops = {
>  	.init = artpec6_pcie_ep_init,
>  	.raise_irq = artpec6_pcie_raise_irq,
> +	.get_features = artpec6_pcie_get_features,
>  };
>  
>  static int artpec6_pcie_probe(struct platform_device *pdev)
> -- 
> 2.47.0
> 

-- 
மணிவண்ணன் சதாசிவம்

[PATCH v4 4/5] PCI: endpoint: Add size check for fixed size BARs in pci_epc_set_bar()

[Niklas Cassel]

A BAR of type BAR_FIXED has a fixed BAR size (the size cannot be changed).

When using pci_epf_alloc_space() to allocate backing memory for a BAR,
pci_epf_alloc_space() will always set the size to the fixed BAR size if
the BAR type is BAR_FIXED (and will give an error if you the requested size
is larger than the fixed BAR size).

However, some drivers might not call pci_epf_alloc_space() before calling
pci_epc_set_bar(), so add a check in pci_epc_set_bar() to ensure that an
EPF driver cannot set a size different from the fixed BAR size, if the BAR
type is BAR_FIXED.

The pci_epc_function_is_valid() check is removed because this check is now
done by pci_epc_get_features().

Signed-off-by: Niklas Cassel <cassel@kernel.org>
Reviewed-by: Frank Li <Frank.Li@nxp.com>
---
 drivers/pci/endpoint/pci-epc-core.c | 11 +++++++++--
 1 file changed, 9 insertions(+), 2 deletions(-)

diff --git a/drivers/pci/endpoint/pci-epc-core.c b/drivers/pci/endpoint/pci-epc-core.c
index bed7c7d1fe3c..c69c133701c9 100644
--- a/drivers/pci/endpoint/pci-epc-core.c
+++ b/drivers/pci/endpoint/pci-epc-core.c
@@ -609,10 +609,17 @@ EXPORT_SYMBOL_GPL(pci_epc_clear_bar);
 int pci_epc_set_bar(struct pci_epc *epc, u8 func_no, u8 vfunc_no,
 		    struct pci_epf_bar *epf_bar)
 {
-	int ret;
+	const struct pci_epc_features *epc_features;
+	enum pci_barno bar = epf_bar->barno;
 	int flags = epf_bar->flags;
+	int ret;
 
-	if (!pci_epc_function_is_valid(epc, func_no, vfunc_no))
+	epc_features = pci_epc_get_features(epc, func_no, vfunc_no);
+	if (!epc_features)
+		return -EINVAL;
+
+	if (epc_features->bar[bar].type == BAR_FIXED &&
+	    (epc_features->bar[bar].fixed_size != epf_bar->size))
 		return -EINVAL;
 
 	if ((epf_bar->barno == BAR_5 && flags & PCI_BASE_ADDRESS_MEM_TYPE_64) ||
-- 
2.47.0

Re: [PATCH v4 4/5] PCI: endpoint: Add size check for fixed size BARs in pci_epc_set_bar()

[Manivannan Sadhasivam]

On Fri, Nov 22, 2024 at 12:57:13PM +0100, Niklas Cassel wrote:
> A BAR of type BAR_FIXED has a fixed BAR size (the size cannot be changed).
> 
> When using pci_epf_alloc_space() to allocate backing memory for a BAR,
> pci_epf_alloc_space() will always set the size to the fixed BAR size if
> the BAR type is BAR_FIXED (and will give an error if you the requested size
> is larger than the fixed BAR size).
> 
> However, some drivers might not call pci_epf_alloc_space() before calling
> pci_epc_set_bar(), so add a check in pci_epc_set_bar() to ensure that an
> EPF driver cannot set a size different from the fixed BAR size, if the BAR
> type is BAR_FIXED.
> 
> The pci_epc_function_is_valid() check is removed because this check is now
> done by pci_epc_get_features().
> 
> Signed-off-by: Niklas Cassel <cassel@kernel.org>

Reviewed-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>

- Mani

> Reviewed-by: Frank Li <Frank.Li@nxp.com>
> ---
>  drivers/pci/endpoint/pci-epc-core.c | 11 +++++++++--
>  1 file changed, 9 insertions(+), 2 deletions(-)
> 
> diff --git a/drivers/pci/endpoint/pci-epc-core.c b/drivers/pci/endpoint/pci-epc-core.c
> index bed7c7d1fe3c..c69c133701c9 100644
> --- a/drivers/pci/endpoint/pci-epc-core.c
> +++ b/drivers/pci/endpoint/pci-epc-core.c
> @@ -609,10 +609,17 @@ EXPORT_SYMBOL_GPL(pci_epc_clear_bar);
>  int pci_epc_set_bar(struct pci_epc *epc, u8 func_no, u8 vfunc_no,
>  		    struct pci_epf_bar *epf_bar)
>  {
> -	int ret;
> +	const struct pci_epc_features *epc_features;
> +	enum pci_barno bar = epf_bar->barno;
>  	int flags = epf_bar->flags;
> +	int ret;
>  
> -	if (!pci_epc_function_is_valid(epc, func_no, vfunc_no))
> +	epc_features = pci_epc_get_features(epc, func_no, vfunc_no);
> +	if (!epc_features)
> +		return -EINVAL;
> +
> +	if (epc_features->bar[bar].type == BAR_FIXED &&
> +	    (epc_features->bar[bar].fixed_size != epf_bar->size))
>  		return -EINVAL;
>  
>  	if ((epf_bar->barno == BAR_5 && flags & PCI_BASE_ADDRESS_MEM_TYPE_64) ||
> -- 
> 2.47.0
> 

-- 
மணிவண்ணன் சதாசிவம்

[PATCH v4 5/5] PCI: endpoint: Verify that requested BAR size is a power of two

[Niklas Cassel]

When allocating a BAR using pci_epf_alloc_space(), there are checks that
round up the size to a power of two.

However, there is no check in pci_epc_set_bar() which verifies that the
requested BAR size is a power of two.

Add a power of two check in pci_epc_set_bar(), so that we don't need to
add such a check in each and every PCI endpoint controller driver.

Signed-off-by: Niklas Cassel <cassel@kernel.org>
---
 drivers/pci/endpoint/pci-epc-core.c | 3 +++
 1 file changed, 3 insertions(+)

diff --git a/drivers/pci/endpoint/pci-epc-core.c b/drivers/pci/endpoint/pci-epc-core.c
index c69c133701c9..6062677e9ffe 100644
--- a/drivers/pci/endpoint/pci-epc-core.c
+++ b/drivers/pci/endpoint/pci-epc-core.c
@@ -622,6 +622,9 @@ int pci_epc_set_bar(struct pci_epc *epc, u8 func_no, u8 vfunc_no,
 	    (epc_features->bar[bar].fixed_size != epf_bar->size))
 		return -EINVAL;
 
+	if (!is_power_of_2(epf_bar->size))
+		return -EINVAL;
+
 	if ((epf_bar->barno == BAR_5 && flags & PCI_BASE_ADDRESS_MEM_TYPE_64) ||
 	    (flags & PCI_BASE_ADDRESS_SPACE_IO &&
 	     flags & PCI_BASE_ADDRESS_IO_MASK) ||
-- 
2.47.0

Re: [PATCH v4 5/5] PCI: endpoint: Verify that requested BAR size is a power of two

[Manivannan Sadhasivam]

On Fri, Nov 22, 2024 at 12:57:14PM +0100, Niklas Cassel wrote:
> When allocating a BAR using pci_epf_alloc_space(), there are checks that
> round up the size to a power of two.
> 
> However, there is no check in pci_epc_set_bar() which verifies that the
> requested BAR size is a power of two.
> 
> Add a power of two check in pci_epc_set_bar(), so that we don't need to
> add such a check in each and every PCI endpoint controller driver.
> 
> Signed-off-by: Niklas Cassel <cassel@kernel.org>

Reviewed-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>

- Mani

> ---
>  drivers/pci/endpoint/pci-epc-core.c | 3 +++
>  1 file changed, 3 insertions(+)
> 
> diff --git a/drivers/pci/endpoint/pci-epc-core.c b/drivers/pci/endpoint/pci-epc-core.c
> index c69c133701c9..6062677e9ffe 100644
> --- a/drivers/pci/endpoint/pci-epc-core.c
> +++ b/drivers/pci/endpoint/pci-epc-core.c
> @@ -622,6 +622,9 @@ int pci_epc_set_bar(struct pci_epc *epc, u8 func_no, u8 vfunc_no,
>  	    (epc_features->bar[bar].fixed_size != epf_bar->size))
>  		return -EINVAL;
>  
> +	if (!is_power_of_2(epf_bar->size))
> +		return -EINVAL;
> +
>  	if ((epf_bar->barno == BAR_5 && flags & PCI_BASE_ADDRESS_MEM_TYPE_64) ||
>  	    (flags & PCI_BASE_ADDRESS_SPACE_IO &&
>  	     flags & PCI_BASE_ADDRESS_IO_MASK) ||
> -- 
> 2.47.0
> 

-- 
மணிவண்ணன் சதாசிவம்