mirror of https://lore.kernel.org/lkml/
 help / color / mirror / Atom feed
From: Neil Armstrong <neil.armstrong@linaro.org>
To: Manivannan Sadhasivam <manivannan.sadhasivam@oss.qualcomm.com>,
	mani@kernel.org, lpieralisi@kernel.org, kwilczynski@kernel.org,
	robh@kernel.org, bhelgaas@google.com
Cc: linux-pci@vger.kernel.org, linux-arm-msm@vger.kernel.org,
	linux-kernel@vger.kernel.org,
	Loic Poulain <loic.poulain@oss.qualcomm.com>
Subject: Re: [PATCH] PCI: qcom: Honor IOMMU provider's #iommu-cells in qcom_pcie_config_sid_1_9_0()
Date: Mon, 7 Sep 2026 17:46:49 +0200	[thread overview]
Message-ID: <e04b2fc0-710c-4cd3-8f38-e528d57b4487@linaro.org> (raw)
In-Reply-To: <20260907143349.317495-1-mani@kernel.org>

On 9/7/26 16:33, Manivannan Sadhasivam wrote:
> From: Manivannan Sadhasivam <manivannan.sadhasivam@oss.qualcomm.com>
> 
> qcom_pcie_config_sid_1_9_0() reads the "iommu-map" property as an array
> of fixed four-word {rid-base, phandle, sid, rid-len} entries to program
> the BDF to SID translation table. But that layout only holds for an
> IOMMU with '#iommu-cells = <1>'. The PCIe SMMUs on these SoCs use
> '#iommu-cells = <2>' (SID and mask), so per the pci-iommu binding each
> entry is really five cells long.
> 
> This used to work only because the DTs were themselves broken. They
> described iommu-map with four-cell entries that omitted the SID mask,
> which of_map_id() tolerated via its of_check_bad_map() fallback, and the
> four-word parsing coincidentally matched that malformed shape.
> 
> Since commit ccb2fd725d41 ("of: Respect #{iommu,msi}-cells in maps") the
> OF core parses such maps correctly, so the device trees were converted
> to proper five-cell entries, e.g. commit c41749e9554d ("arm64: dts:
> qcom: sm8250: Fix the PCIe iommu-map entries"). With five-cell entries
> the fixed four-word stride slips by one cell for each entry after the
> first, so qcom_pcie_config_sid_1_9_0() reads the endpoint mapping's RID
> from the preceding entry's length cell and its SID from the phandle cell.
> As the RID is the hash key, the endpoint's real RID is never programmed
> into the BDF to SID table. Its DMA then hashes to an unprogrammed slot,
> gets tagged with SID 0 and the SMMU faults like below on QCS8300:
> 
>    arm-smmu 15200000.iommu: Unhandled context fault: fsr=0x402, iova=0xffa00000, cbfrsynra=0x0, cb=1
> 
> To fix this, walk the map with a stride of 3 + '#iommu-cells' of the
> referenced IOMMU and take the SID from the first specifier cell, which is
> all the BDF to SID table needs. Validate the layout instead of trusting
> the array size. Also, preserve the legacy behavior of the old DTs by
> detecting the same pattern that of_check_bad_map() recognizes and
> falling back to a stride of four.
> 
> Fixes: 4c9398822106 ("PCI: qcom: Add support for configuring BDF to SID mapping for SM8250")
> Reported-by: Loic Poulain <loic.poulain@oss.qualcomm.com>
> Signed-off-by: Manivannan Sadhasivam <manivannan.sadhasivam@oss.qualcomm.com>
> ---
>   drivers/pci/controller/dwc/pcie-qcom.c | 90 ++++++++++++++++++--------
>   1 file changed, 64 insertions(+), 26 deletions(-)
> 
> diff --git a/drivers/pci/controller/dwc/pcie-qcom.c b/drivers/pci/controller/dwc/pcie-qcom.c
> index b58a607b713f..42cbeef083ca 100644
> --- a/drivers/pci/controller/dwc/pcie-qcom.c
> +++ b/drivers/pci/controller/dwc/pcie-qcom.c
> @@ -1143,50 +1143,90 @@ static void qcom_pcie_deinit_2_7_0(struct qcom_pcie *pcie)
>   
>   static int qcom_pcie_config_sid_1_9_0(struct qcom_pcie *pcie)
>   {
> -	/* iommu map structure */
> -	struct {
> -		u32 bdf;
> -		u32 phandle;
> -		u32 smmu_sid;
> -		u32 smmu_sid_len;
> -	} *map;
>   	void __iomem *bdf_to_sid_base = pcie->parf + PARF_BDF_TO_SID_TABLE_N;
>   	struct device *dev = pcie->pci->dev;
> +	struct device_node *iommu_np;
>   	u8 qcom_pcie_crc8_table[CRC8_TABLE_SIZE];
> -	int i, nr_map, size = 0;
> -	u32 smmu_sid_base;
> +	const __be32 *map;
> +	u32 iommu_cells, entry_cells, phandle, smmu_sid_base;
> +	int i, nr_cells, nr_map, size = 0;
>   	u32 val;
>   
> -	of_get_property(dev->of_node, "iommu-map", &size);
> -	if (!size)
> +	map = of_get_property(dev->of_node, "iommu-map", &size);
> +	if (!map || !size)
>   		return 0;
>   
> +	if (size % sizeof(*map)) {
> +		dev_err(dev, "Malformed iommu-map property\n");
> +		return -EINVAL;
> +	}
> +	nr_cells = size / sizeof(*map);
> +
> +	/*
> +	 * Each iommu-map entry is: rid-base (1 cell), phandle (1 cell),
> +	 * IOMMU specifier (#iommu-cells cells), length (1 cell). Read
> +	 * #iommu-cells from the IOMMU provider referenced by the first
> +	 * entry to compute the per-entry stride.
> +	 */
> +	phandle = be32_to_cpu(map[1]);
> +	iommu_np = of_find_node_by_phandle(phandle);
> +	if (!iommu_np) {
> +		dev_err(dev, "Failed to find IOMMU node in iommu-map\n");
> +		return -ENODEV;
> +	}
> +
> +	if (of_property_read_u32(iommu_np, "#iommu-cells", &iommu_cells))
> +		iommu_cells = 1;
> +	of_node_put(iommu_np);
> +
> +	entry_cells = 3 + iommu_cells;
> +
> +	/*
> +	 * Retain backward compatibility with DTs that describe iommu-map
> +	 * with 4-cell entries against an IOMMU declaring #iommu-cells = 2,
> +	 * matching the fallback in drivers/of/base.c::of_check_bad_map().
> +	 */
> +	if (iommu_cells == 2 && !(nr_cells % 4)) {
> +		bool legacy = true;
> +
> +		for (i = 0; i < nr_cells; i += 4) {
> +			if (be32_to_cpu(map[i + 1]) != phandle ||
> +			    be32_to_cpu(map[i + 3]) != 1) {
> +				legacy = false;
> +				break;
> +			}
> +		}
> +
> +		if (legacy) {
> +			dev_warn_once(dev, "iommu-map has 1-cell entries targeting 2-cell #iommu-cells, treating as 1-cell output\n");
> +			entry_cells = 4;
> +		}
> +	}
> +
> +	if (nr_cells % entry_cells) {
> +		dev_err(dev, "Malformed iommu-map property\n");
> +		return -EINVAL;
> +	}
> +	nr_map = nr_cells / entry_cells;
> +
>   	/* Enable BDF to SID translation by disabling bypass mode (default) */
>   	val = readl(pcie->parf + PARF_BDF_TO_SID_CFG);
>   	val &= ~BDF_TO_SID_BYPASS;
>   	writel(val, pcie->parf + PARF_BDF_TO_SID_CFG);
>   
> -	map = kzalloc(size, GFP_KERNEL);
> -	if (!map)
> -		return -ENOMEM;
> -
> -	of_property_read_u32_array(dev->of_node, "iommu-map", (u32 *)map,
> -				   size / sizeof(u32));
> -
> -	nr_map = size / (sizeof(*map));
> -
>   	crc8_populate_msb(qcom_pcie_crc8_table, QCOM_PCIE_CRC8_POLYNOMIAL);
>   
>   	/* Registers need to be zero out first */
>   	memset_io(bdf_to_sid_base, 0, CRC8_TABLE_SIZE * sizeof(u32));
>   
>   	/* Extract the SMMU SID base from the first entry of iommu-map */
> -	smmu_sid_base = map[0].smmu_sid;
> +	smmu_sid_base = be32_to_cpu(map[2]);
>   
>   	/* Look for an available entry to hold the mapping */
>   	for (i = 0; i < nr_map; i++) {
> -		__be16 bdf_be = cpu_to_be16(map[i].bdf);
> -		u32 val;
> +		u32 bdf = be32_to_cpu(map[i * entry_cells]);
> +		u32 sid = be32_to_cpu(map[i * entry_cells + 2]);
> +		__be16 bdf_be = cpu_to_be16(bdf);
>   		u8 hash;
>   
>   		hash = crc8(qcom_pcie_crc8_table, (u8 *)&bdf_be, sizeof(bdf_be), 0);
> @@ -1208,12 +1248,10 @@ static int qcom_pcie_config_sid_1_9_0(struct qcom_pcie *pcie)
>   		}
>   
>   		/* BDF [31:16] | SID [15:8] | NEXT [7:0] */
> -		val = map[i].bdf << 16 | (map[i].smmu_sid - smmu_sid_base) << 8 | 0;
> +		val = bdf << 16 | (sid - smmu_sid_base) << 8 | 0;
>   		writel(val, bdf_to_sid_base + hash * sizeof(u32));
>   	}
>   
> -	kfree(map);
> -
>   	return 0;
>   }
>   

Tested-by: Neil Armstrong <neil.armstrong@linaro.org> # on SM8650-HDK

Thanks,
Neil

  parent reply	other threads:[~2026-09-07 15:46 UTC|newest]

Thread overview: 4+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2026-09-07 14:33 Manivannan Sadhasivam
2026-09-07 15:43 ` Loic Poulain
2026-09-07 15:46 ` Neil Armstrong [this message]
2026-09-08  7:24 ` Konrad Dybcio

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=e04b2fc0-710c-4cd3-8f38-e528d57b4487@linaro.org \
    --to=neil.armstrong@linaro.org \
    --cc=bhelgaas@google.com \
    --cc=kwilczynski@kernel.org \
    --cc=linux-arm-msm@vger.kernel.org \
    --cc=linux-kernel@vger.kernel.org \
    --cc=linux-pci@vger.kernel.org \
    --cc=loic.poulain@oss.qualcomm.com \
    --cc=lpieralisi@kernel.org \
    --cc=mani@kernel.org \
    --cc=manivannan.sadhasivam@oss.qualcomm.com \
    --cc=robh@kernel.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox

all inboxes | Powered by JetHome®