mirror of https://lore.kernel.org/lkml/
 help / color / mirror / Atom feed
From: Krishna Chaitanya Chundru <krishna.chundru@oss.qualcomm.com>
To: "Mayank Rana" <mayank.rana@oss.qualcomm.com>,
	cros-qcom-dts-watchers@chromium.org,
	"Bjorn Andersson" <andersson@kernel.org>,
	"Konrad Dybcio" <konradybcio@kernel.org>,
	"Rob Herring" <robh@kernel.org>,
	"Krzysztof Kozlowski" <krzk+dt@kernel.org>,
	"Conor Dooley" <conor+dt@kernel.org>,
	"Lorenzo Pieralisi" <lpieralisi@kernel.org>,
	"Manivannan Sadhasivam" <mani@kernel.org>,
	"Krzysztof Wilczyński" <kwilczynski@kernel.org>,
	"Bjorn Helgaas" <bhelgaas@google.com>,
	"Jingoo Han" <jingoohan1@gmail.com>
Cc: linux-arm-msm@vger.kernel.org, devicetree@vger.kernel.org,
	linux-kernel@vger.kernel.org, linux-pci@vger.kernel.org,
	quic_vbadigan@quicinc.com, quic_mrana@quicinc.com,
	quic_vpernami@quicinc.com, mmareddy@quicinc.com
Subject: Re: [PATCH v7 4/5] PCI: dwc: Add ECAM support with iATU configuration
Date: Tue, 26 Aug 2025 18:16:44 +0530	[thread overview]
Message-ID: <c89fc295-0dfa-4910-838d-3520272cb26b@oss.qualcomm.com> (raw)
In-Reply-To: <a158c4f5-e9c3-48c2-b440-fa9dc281b276@oss.qualcomm.com>



On 8/25/2025 11:35 PM, Mayank Rana wrote:
> Hi Krishna
> 
> On 8/22/2025 2:27 AM, Krishna Chaitanya Chundru wrote:
>> The current implementation requires iATU for every configuration
>> space access which increases latency & cpu utilization.
>>
>> Designware databook 5.20a, section 3.10.10.3 says about CFG Shift 
>> Feature,
>> which shifts/maps the BDF (bits [31:16] of the third header DWORD, which
>> would be matched against the Base and Limit addresses) of the incoming
>> CfgRd0/CfgWr0 down to bits[27:12]of the translated address.
>>
>> Configuring iATU in config shift feature enables ECAM feature to 
>> access the
>> config space, which avoids iATU configuration for every config access.
>>
>> Add "ctrl2" into struct dw_pcie_ob_atu_cfg  to enable config shift 
>> feature.
>>
>> As DBI comes under config space, this avoids remapping of DBI space
>> separately. Instead, it uses the mapped config space address returned 
>> from
>> ECAM initialization. Change the order of dw_pcie_get_resources() 
>> execution
>> to achieve this.
>>
>> Enable the ECAM feature if the config space size is equal to size 
>> required
>> to represent number of buses in the bus range property.
> 
> Also add 256 MB alignment requirement for using iATU config shift mode 
> here.
> 
>> Signed-off-by: Krishna Chaitanya Chundru 
>> <krishna.chundru@oss.qualcomm.com>
>> ---
>>   drivers/pci/controller/dwc/Kconfig                |   1 +
>>   drivers/pci/controller/dwc/pcie-designware-host.c | 131 
>> +++++++++++++++++++---
>>   drivers/pci/controller/dwc/pcie-designware.c      |   2 +-
>>   drivers/pci/controller/dwc/pcie-designware.h      |   5 +
>>   4 files changed, 124 insertions(+), 15 deletions(-)
>>
>> diff --git a/drivers/pci/controller/dwc/Kconfig 
>> b/drivers/pci/controller/dwc/Kconfig
>> index 
>> ff6b6d9e18ecfa44273e87931551f9e63fbe3cba..a0e7ad3fb5afec63b0f919732a50147229623186 100644
>> --- a/drivers/pci/controller/dwc/Kconfig
>> +++ b/drivers/pci/controller/dwc/Kconfig
>> @@ -20,6 +20,7 @@ config PCIE_DW_HOST
>>       bool
>>       select PCIE_DW
>>       select IRQ_MSI_LIB
>> +    select PCI_HOST_COMMON
>>   config PCIE_DW_EP
>>       bool
>> diff --git a/drivers/pci/controller/dwc/pcie-designware-host.c 
>> b/drivers/pci/controller/dwc/pcie-designware-host.c
>> index 
>> 952f8594b501254d2b2de5d5e056e16d2aa8d4b7..abb93265a19fd62d3fecc64f29f37baf67291b40 100644
>> --- a/drivers/pci/controller/dwc/pcie-designware-host.c
>> +++ b/drivers/pci/controller/dwc/pcie-designware-host.c
>> @@ -413,6 +413,81 @@ static void 
>> dw_pcie_host_request_msg_tlp_res(struct dw_pcie_rp *pp)
>>       }
>>   }
>> +static int dw_pcie_config_ecam_iatu(struct dw_pcie_rp *pp)
>> +{
>> +    struct dw_pcie *pci = to_dw_pcie_from_pp(pp);
>> +    struct dw_pcie_ob_atu_cfg atu = {0};
>> +    resource_size_t bus_range_max;
>> +    struct resource_entry *bus;
>> +    int ret;
>> +
>> +    bus = resource_list_first_type(&pp->bridge->windows, 
>> IORESOURCE_BUS);
>> +
>> +    /*
>> +     * Root bus under the host bridge doesn't require any iATU 
>> configuration
>> +     * as DBI region will be used to access root bus config space.
>> +     * Immediate bus under Root Bus, needs type 0 iATU configuration and
>> +     * remaining buses need type 1 iATU configuration.
>> +     */
>> +    atu.index = 0;
>> +    atu.type = PCIE_ATU_TYPE_CFG0;
>> +    atu.parent_bus_addr = pp->cfg0_base + SZ_1M;
>> +    /* 1MiB is to cover 1 (bus) * 32 (devices) * 8 (functions) */
>> +    atu.size = SZ_1M;
>> +    atu.ctrl2 = PCIE_ATU_CFG_SHIFT_MODE_ENABLE;
>> +    ret = dw_pcie_prog_outbound_atu(pci, &atu);
>> +    if (ret)
>> +        return ret;
>> +
>> +    bus_range_max = resource_size(bus->res);
>> +
>> +    if (bus_range_max < 2)
>> +        return 0;
>> +
>> +    /* Configure remaining buses in type 1 iATU configuration */
>> +    atu.index = 1;
>> +    atu.type = PCIE_ATU_TYPE_CFG1;
>> +    atu.parent_bus_addr = pp->cfg0_base + SZ_2M;
>> +    atu.size = (SZ_1M * bus_range_max) - SZ_2M;
>> +    atu.ctrl2 = PCIE_ATU_CFG_SHIFT_MODE_ENABLE;
>> +
>> +    return dw_pcie_prog_outbound_atu(pci, &atu);
>> +}
>> +
>> +static int dw_pcie_create_ecam_window(struct dw_pcie_rp *pp, struct 
>> resource *res)
>> +{
>> +    struct dw_pcie *pci = to_dw_pcie_from_pp(pp);
>> +    struct device *dev = pci->dev;
>> +    struct resource_entry *bus;
>> +
>> +    bus = resource_list_first_type(&pp->bridge->windows, 
>> IORESOURCE_BUS);
>> +    if (!bus)
>> +        return -ENODEV;
>> +    pp->cfg = pci_ecam_create(dev, res, bus->res, 
>> &pci_generic_ecam_ops);
>> +    if (IS_ERR(pp->cfg))
>> +        return PTR_ERR(pp->cfg);
>> +
>> +    pci->dbi_base = pp->cfg->win;
>> +    pci->dbi_phys_addr = res->start;
>> +
>> +    return 0;
>> +}
>> +
>> +static bool dw_pcie_ecam_enabled(struct dw_pcie_rp *pp, struct 
>> resource *config_res)
>> +{
>> +    struct resource *bus_range;
>> +    u64 nr_buses;
> 
> As change is using Synopsis IP based iATU config shift mode 
> functionality, it is must that ECAM/DBI base address has to be 256 MB 
> aligned. Hence add change to check against alignment.
> 
Just to clarify this is not Synopsis IP requirement it is PCie
requirement. PCIe spec 6, sec 7.2.2 says "base address of the range is
aligned to a 2(n+20)-byte memory address boundary". n is 8 here.

I will add this info as a comment.
> #define IS_256MB_ALIGNED(x) IS_ALIGNED(x, SZ_256M)
> 
> if (!IS_256MB_ALIGNED(config_res->start))
>            return false;
> 
Ack.

- Krishna Chaitanya.
>> +
>> +    bus_range = resource_list_first_type(&pp->bridge->windows, 
>> IORESOURCE_BUS)->res;
>> +    if (!bus_range)
>> +        return false;
>> +
>> +    nr_buses = resource_size(config_res) >> PCIE_ECAM_BUS_SHIFT;
>> +
>> +    return !!(nr_buses >= resource_size(bus_range));
>> +}
>> +
>>   static int dw_pcie_host_get_resources(struct dw_pcie_rp *pp)
>>   {
>>       struct dw_pcie *pci = to_dw_pcie_from_pp(pp);
>> @@ -422,10 +497,6 @@ static int dw_pcie_host_get_resources(struct 
>> dw_pcie_rp *pp)
>>       struct resource *res;
>>       int ret;
>> -    ret = dw_pcie_get_resources(pci);
>> -    if (ret)
>> -        return ret;
>> -
>>       res = platform_get_resource_byname(pdev, IORESOURCE_MEM, "config");
>>       if (!res) {
>>           dev_err(dev, "Missing \"config\" reg space\n");
>> @@ -435,9 +506,32 @@ static int dw_pcie_host_get_resources(struct 
>> dw_pcie_rp *pp)
>>       pp->cfg0_size = resource_size(res);
>>       pp->cfg0_base = res->start;
>> -    pp->va_cfg0_base = devm_pci_remap_cfg_resource(dev, res);
>> -    if (IS_ERR(pp->va_cfg0_base))
>> -        return PTR_ERR(pp->va_cfg0_base);
>> +    pp->ecam_enabled = dw_pcie_ecam_enabled(pp, res);
>> +    if (pp->ecam_enabled) {
>> +        ret = dw_pcie_create_ecam_window(pp, res);
>> +        if (ret)
>> +            return ret;
>> +
>> +        pp->bridge->ops = (struct pci_ops 
>> *)&pci_generic_ecam_ops.pci_ops;
>> +        pp->bridge->sysdata = pp->cfg;
>> +        pp->cfg->priv = pp;
>> +    } else {
>> +        pp->va_cfg0_base = devm_pci_remap_cfg_resource(dev, res);
>> +        if (IS_ERR(pp->va_cfg0_base))
>> +            return PTR_ERR(pp->va_cfg0_base);
>> +
>> +        /* Set default bus ops */
>> +        pp->bridge->ops = &dw_pcie_ops;
>> +        pp->bridge->child_ops = &dw_child_pcie_ops;
>> +        pp->bridge->sysdata = pp;
>> +    }
>> +
>> +    ret = dw_pcie_get_resources(pci);
>> +    if (ret) {
>> +        if (pp->cfg)
>> +            pci_ecam_free(pp->cfg);
>> +        return ret;
>> +    }
>>       /* Get the I/O range from DT */
>>       win = resource_list_first_type(&pp->bridge->windows, 
>> IORESOURCE_IO);
>> @@ -476,14 +570,10 @@ int dw_pcie_host_init(struct dw_pcie_rp *pp)
>>       if (ret)
>>           return ret;
>> -    /* Set default bus ops */
>> -    bridge->ops = &dw_pcie_ops;
>> -    bridge->child_ops = &dw_child_pcie_ops;
>> -
>>       if (pp->ops->init) {
>>           ret = pp->ops->init(pp);
>>           if (ret)
>> -            return ret;
>> +            goto err_free_ecam;
>>       }
>>       if (pci_msi_enabled()) {
>> @@ -525,6 +615,14 @@ int dw_pcie_host_init(struct dw_pcie_rp *pp)
>>       if (ret)
>>           goto err_free_msi;
>> +    if (pp->ecam_enabled) {
>> +        ret = dw_pcie_config_ecam_iatu(pp);
>> +        if (ret) {
>> +            dev_err(dev, "Failed to configure iATU in ECAM mode\n");
>> +            goto err_free_msi;
>> +        }
>> +    }
>> +
>>       /*
>>        * Allocate the resource for MSG TLP before programming the iATU
>>        * outbound window in dw_pcie_setup_rc(). Since the allocation 
>> depends
>> @@ -560,8 +658,6 @@ int dw_pcie_host_init(struct dw_pcie_rp *pp)
>>           /* Ignore errors, the link may come up later */
>>           dw_pcie_wait_for_link(pci);
>> -    bridge->sysdata = pp;
>> -
>>       ret = pci_host_probe(bridge);
>>       if (ret)
>>           goto err_stop_link;
>> @@ -587,6 +683,10 @@ int dw_pcie_host_init(struct dw_pcie_rp *pp)
>>       if (pp->ops->deinit)
>>           pp->ops->deinit(pp);
>> +err_free_ecam:
>> +    if (pp->cfg)
>> +        pci_ecam_free(pp->cfg);
>> +
>>       return ret;
>>   }
>>   EXPORT_SYMBOL_GPL(dw_pcie_host_init);
>> @@ -609,6 +709,9 @@ void dw_pcie_host_deinit(struct dw_pcie_rp *pp)
>>       if (pp->ops->deinit)
>>           pp->ops->deinit(pp);
>> +
>> +    if (pp->cfg)
>> +        pci_ecam_free(pp->cfg);
>>   }
>>   EXPORT_SYMBOL_GPL(dw_pcie_host_deinit);
>> diff --git a/drivers/pci/controller/dwc/pcie-designware.c 
>> b/drivers/pci/controller/dwc/pcie-designware.c
>> index 
>> 4684c671a81bee468f686a83cc992433b38af59d..6826ddb9478d41227fa011018cffa8d2242336a9 100644
>> --- a/drivers/pci/controller/dwc/pcie-designware.c
>> +++ b/drivers/pci/controller/dwc/pcie-designware.c
>> @@ -576,7 +576,7 @@ int dw_pcie_prog_outbound_atu(struct dw_pcie *pci,
>>           val = dw_pcie_enable_ecrc(val);
>>       dw_pcie_writel_atu_ob(pci, atu->index, PCIE_ATU_REGION_CTRL1, val);
>> -    val = PCIE_ATU_ENABLE;
>> +    val = PCIE_ATU_ENABLE | atu->ctrl2;
>>       if (atu->type == PCIE_ATU_TYPE_MSG) {
>>           /* The data-less messages only for now */
>>           val |= PCIE_ATU_INHIBIT_PAYLOAD | atu->code;
>> diff --git a/drivers/pci/controller/dwc/pcie-designware.h 
>> b/drivers/pci/controller/dwc/pcie-designware.h
>> index 
>> ceb022506c3191cd8fe580411526e20cc3758fed..f770e160ce7c538e0835e7cf80bae9ed099f906c 100644
>> --- a/drivers/pci/controller/dwc/pcie-designware.h
>> +++ b/drivers/pci/controller/dwc/pcie-designware.h
>> @@ -20,6 +20,7 @@
>>   #include <linux/irq.h>
>>   #include <linux/msi.h>
>>   #include <linux/pci.h>
>> +#include <linux/pci-ecam.h>
>>   #include <linux/reset.h>
>>   #include <linux/pci-epc.h>
>> @@ -169,6 +170,7 @@
>>   #define PCIE_ATU_REGION_CTRL2        0x004
>>   #define PCIE_ATU_ENABLE            BIT(31)
>>   #define PCIE_ATU_BAR_MODE_ENABLE    BIT(30)
>> +#define PCIE_ATU_CFG_SHIFT_MODE_ENABLE    BIT(28)
>>   #define PCIE_ATU_INHIBIT_PAYLOAD    BIT(22)
>>   #define PCIE_ATU_FUNC_NUM_MATCH_EN      BIT(19)
>>   #define PCIE_ATU_LOWER_BASE        0x008
>> @@ -387,6 +389,7 @@ struct dw_pcie_ob_atu_cfg {
>>       u8 func_no;
>>       u8 code;
>>       u8 routing;
>> +    u32 ctrl2;
>>       u64 parent_bus_addr;
>>       u64 pci_addr;
>>       u64 size;
>> @@ -425,6 +428,8 @@ struct dw_pcie_rp {
>>       struct resource        *msg_res;
>>       bool            use_linkup_irq;
>>       struct pci_eq_presets    presets;
>> +    bool            ecam_enabled;
>> +    struct pci_config_window *cfg;
>>   };
>>   struct dw_pcie_ep_ops {
>>
> Regards,
> Mayank

  reply	other threads:[~2025-08-26 12:46 UTC|newest]

Thread overview: 9+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2025-08-22  9:27 [PATCH v7 0/5] " Krishna Chaitanya Chundru
2025-08-22  9:27 ` [PATCH v7 1/5] arm64: dts: qcom: sc7280: Increase config size to 256MB for ECAM feature Krishna Chaitanya Chundru
2025-08-22  9:27 ` [PATCH v7 2/5] PCI: dwc: Add support for ELBI resource mapping Krishna Chaitanya Chundru
2025-08-22  9:27 ` [PATCH v7 3/5] PCI: dwc: qcom: Switch to dwc " Krishna Chaitanya Chundru
2025-08-22  9:27 ` [PATCH v7 4/5] PCI: dwc: Add ECAM support with iATU configuration Krishna Chaitanya Chundru
2025-08-25 18:05   ` Mayank Rana
2025-08-26 12:46     ` Krishna Chaitanya Chundru [this message]
2025-08-26 17:43       ` Mayank Rana
2025-08-22  9:27 ` [PATCH v7 5/5] PCI: qcom: Add support for ECAM feature Krishna Chaitanya Chundru

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=c89fc295-0dfa-4910-838d-3520272cb26b@oss.qualcomm.com \
    --to=krishna.chundru@oss.qualcomm.com \
    --cc=andersson@kernel.org \
    --cc=bhelgaas@google.com \
    --cc=conor+dt@kernel.org \
    --cc=cros-qcom-dts-watchers@chromium.org \
    --cc=devicetree@vger.kernel.org \
    --cc=jingoohan1@gmail.com \
    --cc=konradybcio@kernel.org \
    --cc=krzk+dt@kernel.org \
    --cc=kwilczynski@kernel.org \
    --cc=linux-arm-msm@vger.kernel.org \
    --cc=linux-kernel@vger.kernel.org \
    --cc=linux-pci@vger.kernel.org \
    --cc=lpieralisi@kernel.org \
    --cc=mani@kernel.org \
    --cc=mayank.rana@oss.qualcomm.com \
    --cc=mmareddy@quicinc.com \
    --cc=quic_mrana@quicinc.com \
    --cc=quic_vbadigan@quicinc.com \
    --cc=quic_vpernami@quicinc.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®