TI's J721E SoC uses Cadence PCIe core to implement both RC mode and EP mode. The high level features are: *) Supports Legacy, MSI and MSI-X interrupt *) Supports upto GEN4 speed mode *) Supports SR-IOV *) Supports multiple physical function *) Ability to route all transactions via SMMU This patch series *) Add support in Cadence PCIe core to be used for TI's J721E SoC *) Add a driver for J721E PCIe wrapper v1 of the series can be found @ [1] v2 of the series can be found @ [2] v3 of the series can be found @ [5] v4 of the series can be found @ [6] Changes from v4: 1) Added Reviewed-by: & Acked-by: tags from RobH 2) Removed un-used accessors for pcie-cadence.h and removed having ops for read/write accessors 3) Updated cdns,cdns-pcie-host.yaml to remove "mem" from reg Changes from v3: 1) Changed the order of files in MAINTAINTERS file to fix Joe's comments 2) Fixed indentation and added Reviewed-by: Rob Herring <robh@kernel.org> 3) Cleaned up computing msix_tbl 4) Fixed RobH's comment on J721E driver Changes from v2: 1) Converting Cadence binding to YAML schema was done as a separate series [3] & [4]. [3] is merged and [4] is pending. 2) Included MSI-X support in this series 3) Added link down interrupt handling (only error message) 4) Rebased to latest 5.7-rc1 5) Adapted TI J721E binding to [3] & [4] Changes from v1: 1) Added DT schemas cdns-pcie-host.yaml, cdns-pcie-ep.yaml and cdns-pcie.yaml for Cadence PCIe core and included it in TI's PCIe DT schema. 2) Added cpu_addr_fixup() for Cadence Platform driver. 3) Fixed subject/description/renamed functions as commented by Andrew Murray. [1] -> http://lore.kernel.org/r/20191209092147.22901-1-kishon@ti.com [2] -> http://lore.kernel.org/r/20200106102058.19183-1-kishon@ti.com [3] -> http://lore.kernel.org/r/20200305103017.16706-1-kishon@ti.com [4] -> http://lore.kernel.org/r/20200417114322.31111-1-kishon@ti.com [5] -> http://lore.kernel.org/r/20200417125753.13021-1-kishon@ti.com [6] -> http://lore.kernel.org/r/20200506151429.12255-1-kishon@ti.com Alan Douglas (1): PCI: cadence: Add MSI-X support to Endpoint driver Kishon Vijay Abraham I (13): PCI: cadence: Fix cdns_pcie_{host|ep}_setup() error path linux/kernel.h: Add PTR_ALIGN_DOWN macro PCI: cadence: Convert all r/w accessors to perform only 32-bit accesses PCI: cadence: Add support to start link and verify link status PCI: cadence: Allow pci_host_bridge to have custom pci_ops dt-bindings: PCI: cadence: Remove "mem" from reg binding PCI: cadence: Add new *ops* for CPU addr fixup PCI: cadence: Fix updating Vendor ID and Subsystem Vendor ID register dt-bindings: PCI: Add host mode dt-bindings for TI's J721E SoC dt-bindings: PCI: Add EP mode dt-bindings for TI's J721E SoC PCI: j721e: Add TI J721E PCIe driver misc: pci_endpoint_test: Add J721E in pci_device_id table MAINTAINERS: Add Kishon Vijay Abraham I for TI J721E SoC PCIe .../bindings/pci/cdns,cdns-pcie-host.yaml | 8 +- .../bindings/pci/ti,j721e-pci-ep.yaml | 89 ++++ .../bindings/pci/ti,j721e-pci-host.yaml | 113 ++++ MAINTAINERS | 4 +- drivers/misc/pci_endpoint_test.c | 9 + drivers/pci/controller/cadence/Kconfig | 23 + drivers/pci/controller/cadence/Makefile | 1 + drivers/pci/controller/cadence/pci-j721e.c | 490 ++++++++++++++++++ .../pci/controller/cadence/pcie-cadence-ep.c | 125 ++++- .../controller/cadence/pcie-cadence-host.c | 59 ++- .../controller/cadence/pcie-cadence-plat.c | 13 + drivers/pci/controller/cadence/pcie-cadence.c | 8 +- drivers/pci/controller/cadence/pcie-cadence.h | 127 ++++- include/linux/kernel.h | 1 + 14 files changed, 1017 insertions(+), 53 deletions(-) create mode 100644 Documentation/devicetree/bindings/pci/ti,j721e-pci-ep.yaml create mode 100644 Documentation/devicetree/bindings/pci/ti,j721e-pci-host.yaml create mode 100644 drivers/pci/controller/cadence/pci-j721e.c -- 2.17.1
commit bd22885aa188 ("PCI: cadence: Refactor driver to use as a core library") while refactoring the Cadence PCIe driver to be used as library, removed pm_runtime_get_sync() from cdns_pcie_ep_setup() and cdns_pcie_host_setup() but missed to remove the corresponding pm_runtime_put_sync() in the error path. Fix it here. Fixes: bd22885aa188 ("PCI: cadence: Refactor driver to use as a core library") Reviewed-by: Rob Herring <robh@kernel.org> Signed-off-by: Kishon Vijay Abraham I <kishon@ti.com> --- drivers/pci/controller/cadence/pcie-cadence-ep.c | 9 ++------- drivers/pci/controller/cadence/pcie-cadence-host.c | 6 +----- 2 files changed, 3 insertions(+), 12 deletions(-) diff --git a/drivers/pci/controller/cadence/pcie-cadence-ep.c b/drivers/pci/controller/cadence/pcie-cadence-ep.c index 1c173dad67d1..1fdae37843ef 100644 --- a/drivers/pci/controller/cadence/pcie-cadence-ep.c +++ b/drivers/pci/controller/cadence/pcie-cadence-ep.c @@ -8,7 +8,6 @@ #include <linux/of.h> #include <linux/pci-epc.h> #include <linux/platform_device.h> -#include <linux/pm_runtime.h> #include <linux/sizes.h> #include "pcie-cadence.h" @@ -440,8 +439,7 @@ int cdns_pcie_ep_setup(struct cdns_pcie_ep *ep) epc = devm_pci_epc_create(dev, &cdns_pcie_epc_ops); if (IS_ERR(epc)) { dev_err(dev, "failed to create epc device\n"); - ret = PTR_ERR(epc); - goto err_init; + return PTR_ERR(epc); } epc_set_drvdata(epc, ep); @@ -453,7 +451,7 @@ int cdns_pcie_ep_setup(struct cdns_pcie_ep *ep) resource_size(pcie->mem_res)); if (ret < 0) { dev_err(dev, "failed to initialize the memory space\n"); - goto err_init; + return ret; } ep->irq_cpu_addr = pci_epc_mem_alloc_addr(epc, &ep->irq_phys_addr, @@ -472,8 +470,5 @@ int cdns_pcie_ep_setup(struct cdns_pcie_ep *ep) free_epc_mem: pci_epc_mem_exit(epc); - err_init: - pm_runtime_put_sync(dev); - return ret; } diff --git a/drivers/pci/controller/cadence/pcie-cadence-host.c b/drivers/pci/controller/cadence/pcie-cadence-host.c index 70e0eaa15bf9..8e73a680b567 100644 --- a/drivers/pci/controller/cadence/pcie-cadence-host.c +++ b/drivers/pci/controller/cadence/pcie-cadence-host.c @@ -7,7 +7,6 @@ #include <linux/of_address.h> #include <linux/of_pci.h> #include <linux/platform_device.h> -#include <linux/pm_runtime.h> #include "pcie-cadence.h" @@ -476,7 +475,7 @@ int cdns_pcie_host_setup(struct cdns_pcie_rc *rc) ret = cdns_pcie_host_init(dev, &resources, rc); if (ret) - goto err_init; + return ret; list_splice_init(&resources, &bridge->windows); bridge->dev.parent = dev; @@ -494,8 +493,5 @@ int cdns_pcie_host_setup(struct cdns_pcie_rc *rc) err_host_probe: pci_free_resource_list(&resources); - err_init: - pm_runtime_put_sync(dev); - return ret; } -- 2.17.1
Add a macro for aligning down a pointer. This is useful to get an aligned register address when a device allows only word access and doesn't allow half word or byte access. Acked-by: Rob Herring <robh@kernel.org> Signed-off-by: Kishon Vijay Abraham I <kishon@ti.com> --- include/linux/kernel.h | 1 + 1 file changed, 1 insertion(+) diff --git a/include/linux/kernel.h b/include/linux/kernel.h index 9b7a8d74a9d6..c3b361b5be54 100644 --- a/include/linux/kernel.h +++ b/include/linux/kernel.h @@ -34,6 +34,7 @@ #define ALIGN_DOWN(x, a) __ALIGN_KERNEL((x) - ((a) - 1), (a)) #define __ALIGN_MASK(x, mask) __ALIGN_KERNEL_MASK((x), (mask)) #define PTR_ALIGN(p, a) ((typeof(p))ALIGN((unsigned long)(p), (a))) +#define PTR_ALIGN_DOWN(p, a) ((typeof(p))ALIGN_DOWN((unsigned long)(p), (a))) #define IS_ALIGNED(x, a) (((x) & ((typeof(x))(a) - 1)) == 0) /* generic data direction definitions */ -- 2.17.1
Certain platforms like TI's J721E using Cadence PCIe IP can perform only 32-bit accesses for reading or writing to Cadence registers. Convert all read and write accesses to 32-bit in Cadence PCIe driver in preparation for adding PCIe support in TI's J721E SoC. Signed-off-by: Kishon Vijay Abraham I <kishon@ti.com> --- drivers/pci/controller/cadence/pcie-cadence.h | 71 ++++++++++++++----- 1 file changed, 53 insertions(+), 18 deletions(-) diff --git a/drivers/pci/controller/cadence/pcie-cadence.h b/drivers/pci/controller/cadence/pcie-cadence.h index bc49c22e48a9..737e9561092b 100644 --- a/drivers/pci/controller/cadence/pcie-cadence.h +++ b/drivers/pci/controller/cadence/pcie-cadence.h @@ -319,50 +319,88 @@ struct cdns_pcie_ep { /* Register access */ -static inline void cdns_pcie_writeb(struct cdns_pcie *pcie, u32 reg, u8 value) +static inline void cdns_pcie_writel(struct cdns_pcie *pcie, u32 reg, u32 value) { - writeb(value, pcie->reg_base + reg); + writel(value, pcie->reg_base + reg); } -static inline void cdns_pcie_writew(struct cdns_pcie *pcie, u32 reg, u16 value) +static inline u32 cdns_pcie_readl(struct cdns_pcie *pcie, u32 reg) { - writew(value, pcie->reg_base + reg); + return readl(pcie->reg_base + reg); } -static inline void cdns_pcie_writel(struct cdns_pcie *pcie, u32 reg, u32 value) +static inline u32 cdns_pcie_read_sz(void __iomem *addr, int size) { - writel(value, pcie->reg_base + reg); + void __iomem *aligned_addr = PTR_ALIGN_DOWN(addr, 0x4); + unsigned int offset = (unsigned long)addr & 0x3; + u32 val = readl(aligned_addr); + + if (!IS_ALIGNED((uintptr_t)addr, size)) { + WARN(1, "Address %p and size %d are not aligned\n", addr, size); + return 0; + } + + if (size > 2) + return val; + + return (val >> (8 * offset)) & ((1 << (size * 8)) - 1); } -static inline u32 cdns_pcie_readl(struct cdns_pcie *pcie, u32 reg) +static inline void cdns_pcie_write_sz(void __iomem *addr, int size, u32 value) { - return readl(pcie->reg_base + reg); + void __iomem *aligned_addr = PTR_ALIGN_DOWN(addr, 0x4); + unsigned int offset = (unsigned long)addr & 0x3; + u32 mask; + u32 val; + + if (!IS_ALIGNED((uintptr_t)addr, size)) { + WARN(1, "Address %p and size %d are not aligned\n", addr, size); + return; + } + + if (size > 2) { + writel(value, addr); + return; + } + + mask = ~(((1 << (size * 8)) - 1) << (offset * 8)); + val = readl(aligned_addr) & mask; + val |= value << (offset * 8); + writel(val, aligned_addr); } /* Root Port register access */ static inline void cdns_pcie_rp_writeb(struct cdns_pcie *pcie, u32 reg, u8 value) { - writeb(value, pcie->reg_base + CDNS_PCIE_RP_BASE + reg); + void __iomem *addr = pcie->reg_base + CDNS_PCIE_RP_BASE + reg; + + cdns_pcie_write_sz(addr, 0x1, value); } static inline void cdns_pcie_rp_writew(struct cdns_pcie *pcie, u32 reg, u16 value) { - writew(value, pcie->reg_base + CDNS_PCIE_RP_BASE + reg); + void __iomem *addr = pcie->reg_base + CDNS_PCIE_RP_BASE + reg; + + cdns_pcie_write_sz(addr, 0x2, value); } /* Endpoint Function register access */ static inline void cdns_pcie_ep_fn_writeb(struct cdns_pcie *pcie, u8 fn, u32 reg, u8 value) { - writeb(value, pcie->reg_base + CDNS_PCIE_EP_FUNC_BASE(fn) + reg); + void __iomem *addr = pcie->reg_base + CDNS_PCIE_EP_FUNC_BASE(fn) + reg; + + cdns_pcie_write_sz(addr, 0x1, value); } static inline void cdns_pcie_ep_fn_writew(struct cdns_pcie *pcie, u8 fn, u32 reg, u16 value) { - writew(value, pcie->reg_base + CDNS_PCIE_EP_FUNC_BASE(fn) + reg); + void __iomem *addr = pcie->reg_base + CDNS_PCIE_EP_FUNC_BASE(fn) + reg; + + cdns_pcie_write_sz(addr, 0x2, value); } static inline void cdns_pcie_ep_fn_writel(struct cdns_pcie *pcie, u8 fn, @@ -371,14 +409,11 @@ static inline void cdns_pcie_ep_fn_writel(struct cdns_pcie *pcie, u8 fn, writel(value, pcie->reg_base + CDNS_PCIE_EP_FUNC_BASE(fn) + reg); } -static inline u8 cdns_pcie_ep_fn_readb(struct cdns_pcie *pcie, u8 fn, u32 reg) -{ - return readb(pcie->reg_base + CDNS_PCIE_EP_FUNC_BASE(fn) + reg); -} - static inline u16 cdns_pcie_ep_fn_readw(struct cdns_pcie *pcie, u8 fn, u32 reg) { - return readw(pcie->reg_base + CDNS_PCIE_EP_FUNC_BASE(fn) + reg); + void __iomem *addr = pcie->reg_base + CDNS_PCIE_EP_FUNC_BASE(fn) + reg; + + return cdns_pcie_read_sz(addr, 0x2); } static inline u32 cdns_pcie_ep_fn_readl(struct cdns_pcie *pcie, u8 fn, u32 reg) -- 2.17.1
Add cdns_pcie_ops to start link and verify link status. The registers to start link and to check link status is in Platform specific PCIe wrapper. Add support for platform specific drivers to add callback functions for the PCIe Cadence core to start link and verify link status. Signed-off-by: Kishon Vijay Abraham I <kishon@ti.com> Reviewed-by: Rob Herring <robh@kernel.org> --- .../pci/controller/cadence/pcie-cadence-ep.c | 8 ++++ .../controller/cadence/pcie-cadence-host.c | 28 ++++++++++++++ drivers/pci/controller/cadence/pcie-cadence.h | 37 ++++++++++++++++++- 3 files changed, 72 insertions(+), 1 deletion(-) diff --git a/drivers/pci/controller/cadence/pcie-cadence-ep.c b/drivers/pci/controller/cadence/pcie-cadence-ep.c index 1fdae37843ef..14021d760482 100644 --- a/drivers/pci/controller/cadence/pcie-cadence-ep.c +++ b/drivers/pci/controller/cadence/pcie-cadence-ep.c @@ -354,8 +354,10 @@ static int cdns_pcie_ep_start(struct pci_epc *epc) { struct cdns_pcie_ep *ep = epc_get_drvdata(epc); struct cdns_pcie *pcie = &ep->pcie; + struct device *dev = pcie->dev; struct pci_epf *epf; u32 cfg; + int ret; /* * BIT(0) is hardwired to 1, hence function 0 is always enabled @@ -366,6 +368,12 @@ static int cdns_pcie_ep_start(struct pci_epc *epc) cfg |= BIT(epf->func_no); cdns_pcie_writel(pcie, CDNS_PCIE_LM_EP_FUNC_CFG, cfg); + ret = cdns_pcie_start_link(pcie); + if (ret) { + dev_err(dev, "Failed to start link\n"); + return ret; + } + return 0; } diff --git a/drivers/pci/controller/cadence/pcie-cadence-host.c b/drivers/pci/controller/cadence/pcie-cadence-host.c index 8e73a680b567..93a9414932a9 100644 --- a/drivers/pci/controller/cadence/pcie-cadence-host.c +++ b/drivers/pci/controller/cadence/pcie-cadence-host.c @@ -3,6 +3,7 @@ // Cadence PCIe host controller driver. // Author: Cyrille Pitchen <cyrille.pitchen@free-electrons.com> +#include <linux/delay.h> #include <linux/kernel.h> #include <linux/of_address.h> #include <linux/of_pci.h> @@ -422,6 +423,23 @@ static int cdns_pcie_host_init(struct device *dev, return err; } +static int cdns_pcie_host_wait_for_link(struct cdns_pcie *pcie) +{ + struct device *dev = pcie->dev; + int retries; + + /* Check if the link is up or not */ + for (retries = 0; retries < LINK_WAIT_MAX_RETRIES; retries++) { + if (cdns_pcie_link_up(pcie)) { + dev_info(dev, "Link up\n"); + return 0; + } + usleep_range(LINK_WAIT_USLEEP_MIN, LINK_WAIT_USLEEP_MAX); + } + + return -ETIMEDOUT; +} + int cdns_pcie_host_setup(struct cdns_pcie_rc *rc) { struct device *dev = rc->pcie.dev; @@ -470,6 +488,16 @@ int cdns_pcie_host_setup(struct cdns_pcie_rc *rc) pcie->mem_res = res; + ret = cdns_pcie_start_link(pcie); + if (ret) { + dev_err(dev, "Failed to start link\n"); + return ret; + } + + ret = cdns_pcie_host_wait_for_link(pcie); + if (ret) + dev_dbg(dev, "PCIe link never came up\n"); + for (bar = RP_BAR0; bar <= RP_NO_BAR; bar++) rc->avail_ib_bar[bar] = true; diff --git a/drivers/pci/controller/cadence/pcie-cadence.h b/drivers/pci/controller/cadence/pcie-cadence.h index 737e9561092b..c013e629e9fa 100644 --- a/drivers/pci/controller/cadence/pcie-cadence.h +++ b/drivers/pci/controller/cadence/pcie-cadence.h @@ -10,6 +10,11 @@ #include <linux/pci.h> #include <linux/phy/phy.h> +/* Parameters for the waiting for link up routine */ +#define LINK_WAIT_MAX_RETRIES 10 +#define LINK_WAIT_USLEEP_MIN 90000 +#define LINK_WAIT_USLEEP_MAX 100000 + /* * Local Management Registers */ @@ -245,12 +250,20 @@ enum cdns_pcie_msg_routing { MSG_ROUTING_GATHER, }; +struct cdns_pcie_ops { + int (*start_link)(struct cdns_pcie *pcie); + void (*stop_link)(struct cdns_pcie *pcie); + bool (*link_up)(struct cdns_pcie *pcie); +}; + /** * struct cdns_pcie - private data for Cadence PCIe controller drivers * @reg_base: IO mapped register base * @mem_res: start/end offsets in the physical system memory to map PCI accesses * @is_rc: tell whether the PCIe controller mode is Root Complex or Endpoint. * @bus: In Root Complex mode, the bus number + * @ops: Platform specific ops to control various inputs from Cadence PCIe + * wrapper */ struct cdns_pcie { void __iomem *reg_base; @@ -261,7 +274,7 @@ struct cdns_pcie { int phy_count; struct phy **phy; struct device_link **link; - const struct cdns_pcie_common_ops *ops; + const struct cdns_pcie_ops *ops; }; /** @@ -421,6 +434,28 @@ static inline u32 cdns_pcie_ep_fn_readl(struct cdns_pcie *pcie, u8 fn, u32 reg) return readl(pcie->reg_base + CDNS_PCIE_EP_FUNC_BASE(fn) + reg); } +static inline int cdns_pcie_start_link(struct cdns_pcie *pcie) +{ + if (pcie->ops->start_link) + return pcie->ops->start_link(pcie); + + return 0; +} + +static inline void cdns_pcie_stop_link(struct cdns_pcie *pcie) +{ + if (pcie->ops->stop_link) + pcie->ops->stop_link(pcie); +} + +static inline bool cdns_pcie_link_up(struct cdns_pcie *pcie) +{ + if (pcie->ops->link_up) + return pcie->ops->link_up(pcie); + + return true; +} + #ifdef CONFIG_PCIE_CADENCE_HOST int cdns_pcie_host_setup(struct cdns_pcie_rc *rc); #else -- 2.17.1
Certain platforms like TI's J721E allows only 32-bit configuration space access. In such cases pci_generic_config_read and pci_generic_config_write cannot be used. Add support in Cadence core to let pci_host_bridge have custom pci_ops. Signed-off-by: Kishon Vijay Abraham I <kishon@ti.com> --- drivers/pci/controller/cadence/pcie-cadence-host.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) diff --git a/drivers/pci/controller/cadence/pcie-cadence-host.c b/drivers/pci/controller/cadence/pcie-cadence-host.c index 93a9414932a9..62796791f02c 100644 --- a/drivers/pci/controller/cadence/pcie-cadence-host.c +++ b/drivers/pci/controller/cadence/pcie-cadence-host.c @@ -508,7 +508,8 @@ int cdns_pcie_host_setup(struct cdns_pcie_rc *rc) list_splice_init(&resources, &bridge->windows); bridge->dev.parent = dev; bridge->busnr = pcie->bus; - bridge->ops = &cdns_pcie_host_ops; + if (!bridge->ops) + bridge->ops = &cdns_pcie_host_ops; bridge->map_irq = of_irq_parse_and_map_pci; bridge->swizzle_irq = pci_common_swizzle; -- 2.17.1
"mem" is not a memory resource and it overlaps with PCIe config space and memory region. Removve "mem" from reg binding. Signed-off-by: Kishon Vijay Abraham I <kishon@ti.com> --- .../devicetree/bindings/pci/cdns,cdns-pcie-host.yaml | 8 +++----- 1 file changed, 3 insertions(+), 5 deletions(-) diff --git a/Documentation/devicetree/bindings/pci/cdns,cdns-pcie-host.yaml b/Documentation/devicetree/bindings/pci/cdns,cdns-pcie-host.yaml index 84a8f095d031..6d67067843bf 100644 --- a/Documentation/devicetree/bindings/pci/cdns,cdns-pcie-host.yaml +++ b/Documentation/devicetree/bindings/pci/cdns,cdns-pcie-host.yaml @@ -18,13 +18,12 @@ properties: const: cdns,cdns-pcie-host reg: - maxItems: 3 + maxItems: 2 reg-names: items: - const: reg - const: cfg - - const: mem msi-parent: true @@ -49,9 +48,8 @@ examples: device-id = <0x0200>; reg = <0x0 0xfb000000 0x0 0x01000000>, - <0x0 0x41000000 0x0 0x00001000>, - <0x0 0x40000000 0x0 0x04000000>; - reg-names = "reg", "cfg", "mem"; + <0x0 0x41000000 0x0 0x00001000>; + reg-names = "reg", "cfg"; ranges = <0x02000000 0x0 0x42000000 0x0 0x42000000 0x0 0x1000000>, <0x01000000 0x0 0x43000000 0x0 0x43000000 0x0 0x0010000>; -- 2.17.1
Cadence driver uses "mem" memory resource to obtain the offset of configuration space address region, memory space address region and message space address region. The obtained offset is used to program the Address Translation Unit (ATU). However certain platforms like TI's J721E SoC require the absolute address to be programmed in the ATU and not just the offset. Signed-off-by: Kishon Vijay Abraham I <kishon@ti.com> --- .../pci/controller/cadence/pcie-cadence-host.c | 15 ++++----------- .../pci/controller/cadence/pcie-cadence-plat.c | 13 +++++++++++++ drivers/pci/controller/cadence/pcie-cadence.c | 8 ++++++-- drivers/pci/controller/cadence/pcie-cadence.h | 1 + 4 files changed, 24 insertions(+), 13 deletions(-) diff --git a/drivers/pci/controller/cadence/pcie-cadence-host.c b/drivers/pci/controller/cadence/pcie-cadence-host.c index 62796791f02c..3003fafa3bfa 100644 --- a/drivers/pci/controller/cadence/pcie-cadence-host.c +++ b/drivers/pci/controller/cadence/pcie-cadence-host.c @@ -330,15 +330,14 @@ static int cdns_pcie_host_map_dma_ranges(struct cdns_pcie_rc *rc) static int cdns_pcie_host_init_address_translation(struct cdns_pcie_rc *rc) { struct cdns_pcie *pcie = &rc->pcie; - struct resource *mem_res = pcie->mem_res; struct resource *bus_range = rc->bus_range; struct resource *cfg_res = rc->cfg_res; struct device *dev = pcie->dev; struct device_node *np = dev->of_node; struct of_pci_range_parser parser; + u64 cpu_addr = cfg_res->start; struct of_pci_range range; u32 addr0, addr1, desc1; - u64 cpu_addr; int r, err; /* @@ -351,7 +350,9 @@ static int cdns_pcie_host_init_address_translation(struct cdns_pcie_rc *rc) cdns_pcie_writel(pcie, CDNS_PCIE_AT_OB_REGION_PCI_ADDR1(0), addr1); cdns_pcie_writel(pcie, CDNS_PCIE_AT_OB_REGION_DESC1(0), desc1); - cpu_addr = cfg_res->start - mem_res->start; + if (pcie->ops->cpu_addr_fixup) + cpu_addr = pcie->ops->cpu_addr_fixup(pcie, cpu_addr); + addr0 = CDNS_PCIE_AT_OB_REGION_CPU_ADDR0_NBITS(12) | (lower_32_bits(cpu_addr) & GENMASK(31, 8)); addr1 = upper_32_bits(cpu_addr); @@ -480,14 +481,6 @@ int cdns_pcie_host_setup(struct cdns_pcie_rc *rc) } rc->cfg_res = res; - res = platform_get_resource_byname(pdev, IORESOURCE_MEM, "mem"); - if (!res) { - dev_err(dev, "missing \"mem\"\n"); - return -EINVAL; - } - - pcie->mem_res = res; - ret = cdns_pcie_start_link(pcie); if (ret) { dev_err(dev, "Failed to start link\n"); diff --git a/drivers/pci/controller/cadence/pcie-cadence-plat.c b/drivers/pci/controller/cadence/pcie-cadence-plat.c index f5c6bf6dfcb8..6f5f07b3eed1 100644 --- a/drivers/pci/controller/cadence/pcie-cadence-plat.c +++ b/drivers/pci/controller/cadence/pcie-cadence-plat.c @@ -13,6 +13,8 @@ #include <linux/of_device.h> #include "pcie-cadence.h" +#define CDNS_PLAT_CPU_TO_BUS_ADDR 0x0FFFFFFF + /** * struct cdns_plat_pcie - private data for this PCIe platform driver * @pcie: Cadence PCIe controller @@ -30,6 +32,15 @@ struct cdns_plat_pcie_of_data { static const struct of_device_id cdns_plat_pcie_of_match[]; +static u64 cdns_plat_cpu_addr_fixup(struct cdns_pcie *pcie, u64 cpu_addr) +{ + return cpu_addr & CDNS_PLAT_CPU_TO_BUS_ADDR; +} + +static const struct cdns_pcie_ops cdns_plat_ops = { + .cpu_addr_fixup = cdns_plat_cpu_addr_fixup, +}; + static int cdns_plat_pcie_probe(struct platform_device *pdev) { const struct cdns_plat_pcie_of_data *data; @@ -66,6 +77,7 @@ static int cdns_plat_pcie_probe(struct platform_device *pdev) rc = pci_host_bridge_priv(bridge); rc->pcie.dev = dev; + rc->pcie.ops = &cdns_plat_ops; cdns_plat_pcie->pcie = &rc->pcie; cdns_plat_pcie->is_rc = is_rc; @@ -93,6 +105,7 @@ static int cdns_plat_pcie_probe(struct platform_device *pdev) return -ENOMEM; ep->pcie.dev = dev; + ep->pcie.ops = &cdns_plat_ops; cdns_plat_pcie->pcie = &ep->pcie; cdns_plat_pcie->is_rc = is_rc; diff --git a/drivers/pci/controller/cadence/pcie-cadence.c b/drivers/pci/controller/cadence/pcie-cadence.c index cd795f6fc1e2..8a02981fd456 100644 --- a/drivers/pci/controller/cadence/pcie-cadence.c +++ b/drivers/pci/controller/cadence/pcie-cadence.c @@ -73,7 +73,9 @@ void cdns_pcie_set_outbound_region(struct cdns_pcie *pcie, u8 fn, cdns_pcie_writel(pcie, CDNS_PCIE_AT_OB_REGION_DESC1(r), desc1); /* Set the CPU address */ - cpu_addr -= pcie->mem_res->start; + if (pcie->ops->cpu_addr_fixup) + cpu_addr = pcie->ops->cpu_addr_fixup(pcie, cpu_addr); + addr0 = CDNS_PCIE_AT_OB_REGION_CPU_ADDR0_NBITS(nbits) | (lower_32_bits(cpu_addr) & GENMASK(31, 8)); addr1 = upper_32_bits(cpu_addr); @@ -100,7 +102,9 @@ void cdns_pcie_set_outbound_region_for_normal_msg(struct cdns_pcie *pcie, u8 fn, } /* Set the CPU address */ - cpu_addr -= pcie->mem_res->start; + if (pcie->ops->cpu_addr_fixup) + cpu_addr = pcie->ops->cpu_addr_fixup(pcie, cpu_addr); + addr0 = CDNS_PCIE_AT_OB_REGION_CPU_ADDR0_NBITS(17) | (lower_32_bits(cpu_addr) & GENMASK(31, 8)); addr1 = upper_32_bits(cpu_addr); diff --git a/drivers/pci/controller/cadence/pcie-cadence.h b/drivers/pci/controller/cadence/pcie-cadence.h index c013e629e9fa..3490723169c6 100644 --- a/drivers/pci/controller/cadence/pcie-cadence.h +++ b/drivers/pci/controller/cadence/pcie-cadence.h @@ -254,6 +254,7 @@ struct cdns_pcie_ops { int (*start_link)(struct cdns_pcie *pcie); void (*stop_link)(struct cdns_pcie *pcie); bool (*link_up)(struct cdns_pcie *pcie); + u64 (*cpu_addr_fixup)(struct cdns_pcie *pcie, u64 cpu_addr); }; /** -- 2.17.1
Commit 1b79c5284439 ("PCI: cadence: Add host driver for Cadence PCIe controller") in order to update Vendor ID, directly wrote to PCI_VENDOR_ID register. However PCI_VENDOR_ID in root port configuration space is read-only register and writing to it will have no effect. Use local management register to configure Vendor ID and Subsystem Vendor ID. Fixes: 1b79c5284439 ("PCI: cadence: Add host driver for Cadence PCIe controller") Reviewed-by: Rob Herring <robh@kernel.org> Signed-off-by: Kishon Vijay Abraham I <kishon@ti.com> --- drivers/pci/controller/cadence/pcie-cadence-host.c | 9 +++++++-- 1 file changed, 7 insertions(+), 2 deletions(-) diff --git a/drivers/pci/controller/cadence/pcie-cadence-host.c b/drivers/pci/controller/cadence/pcie-cadence-host.c index 3003fafa3bfa..7ee9e06f1285 100644 --- a/drivers/pci/controller/cadence/pcie-cadence-host.c +++ b/drivers/pci/controller/cadence/pcie-cadence-host.c @@ -76,6 +76,7 @@ static int cdns_pcie_host_init_root_port(struct cdns_pcie_rc *rc) { struct cdns_pcie *pcie = &rc->pcie; u32 value, ctrl; + u32 id; /* * Set the root complex BAR configuration register: @@ -95,8 +96,12 @@ static int cdns_pcie_host_init_root_port(struct cdns_pcie_rc *rc) cdns_pcie_writel(pcie, CDNS_PCIE_LM_RC_BAR_CFG, value); /* Set root port configuration space */ - if (rc->vendor_id != 0xffff) - cdns_pcie_rp_writew(pcie, PCI_VENDOR_ID, rc->vendor_id); + if (rc->vendor_id != 0xffff) { + id = CDNS_PCIE_LM_ID_VENDOR(rc->vendor_id) | + CDNS_PCIE_LM_ID_SUBSYS(rc->vendor_id); + cdns_pcie_writel(pcie, CDNS_PCIE_LM_ID, id); + } + if (rc->device_id != 0xffff) cdns_pcie_rp_writew(pcie, PCI_DEVICE_ID, rc->device_id); -- 2.17.1
From: Alan Douglas <adouglas@cadence.com> Implement ->set_msix() and ->get_msix() callback functions in order to configure MSIX capability in the PCIe endpoint controller. Add cdns_pcie_ep_send_msix_irq() to send MSIX interrupts to Host. cdns_pcie_ep_send_msix_irq() gets the MSIX table address (virtual address) from "struct cdns_pcie_epf" that gets initialized in ->set_bar() call back function. Signed-off-by: Alan Douglas <adouglas@cadence.com> [kishon@ti.com: Re-implement MSIX support in accordance with the re-designed core MSI-X interfaces] Signed-off-by: Kishon Vijay Abraham I <kishon@ti.com> --- .../pci/controller/cadence/pcie-cadence-ep.c | 108 +++++++++++++++++- drivers/pci/controller/cadence/pcie-cadence.h | 10 ++ 2 files changed, 117 insertions(+), 1 deletion(-) diff --git a/drivers/pci/controller/cadence/pcie-cadence-ep.c b/drivers/pci/controller/cadence/pcie-cadence-ep.c index 14021d760482..c5696274d81f 100644 --- a/drivers/pci/controller/cadence/pcie-cadence-ep.c +++ b/drivers/pci/controller/cadence/pcie-cadence-ep.c @@ -51,6 +51,7 @@ static int cdns_pcie_ep_set_bar(struct pci_epc *epc, u8 fn, struct pci_epf_bar *epf_bar) { struct cdns_pcie_ep *ep = epc_get_drvdata(epc); + struct cdns_pcie_epf *epf = &ep->epf[fn]; struct cdns_pcie *pcie = &ep->pcie; dma_addr_t bar_phys = epf_bar->phys_addr; enum pci_barno bar = epf_bar->barno; @@ -111,6 +112,8 @@ static int cdns_pcie_ep_set_bar(struct pci_epc *epc, u8 fn, CDNS_PCIE_LM_EP_FUNC_BAR_CFG_BAR_CTRL(b, ctrl)); cdns_pcie_writel(pcie, reg, cfg); + epf->epf_bar[bar] = epf_bar; + return 0; } @@ -118,6 +121,7 @@ static void cdns_pcie_ep_clear_bar(struct pci_epc *epc, u8 fn, struct pci_epf_bar *epf_bar) { struct cdns_pcie_ep *ep = epc_get_drvdata(epc); + struct cdns_pcie_epf *epf = &ep->epf[fn]; struct cdns_pcie *pcie = &ep->pcie; enum pci_barno bar = epf_bar->barno; u32 reg, cfg, b, ctrl; @@ -139,6 +143,8 @@ static void cdns_pcie_ep_clear_bar(struct pci_epc *epc, u8 fn, cdns_pcie_writel(pcie, CDNS_PCIE_AT_IB_EP_FUNC_BAR_ADDR0(fn, bar), 0); cdns_pcie_writel(pcie, CDNS_PCIE_AT_IB_EP_FUNC_BAR_ADDR1(fn, bar), 0); + + epf->epf_bar[bar] = NULL; } static int cdns_pcie_ep_map_addr(struct pci_epc *epc, u8 fn, phys_addr_t addr, @@ -224,6 +230,50 @@ static int cdns_pcie_ep_get_msi(struct pci_epc *epc, u8 fn) return mme; } +static int cdns_pcie_ep_get_msix(struct pci_epc *epc, u8 func_no) +{ + struct cdns_pcie_ep *ep = epc_get_drvdata(epc); + struct cdns_pcie *pcie = &ep->pcie; + u32 cap = CDNS_PCIE_EP_FUNC_MSIX_CAP_OFFSET; + u32 val, reg; + + reg = cap + PCI_MSIX_FLAGS; + val = cdns_pcie_ep_fn_readw(pcie, func_no, reg); + if (!(val & PCI_MSIX_FLAGS_ENABLE)) + return -EINVAL; + + val &= PCI_MSIX_FLAGS_QSIZE; + + return val; +} + +static int cdns_pcie_ep_set_msix(struct pci_epc *epc, u8 fn, u16 interrupts, + enum pci_barno bir, u32 offset) +{ + struct cdns_pcie_ep *ep = epc_get_drvdata(epc); + struct cdns_pcie *pcie = &ep->pcie; + u32 cap = CDNS_PCIE_EP_FUNC_MSIX_CAP_OFFSET; + u32 val, reg; + + reg = cap + PCI_MSIX_FLAGS; + val = cdns_pcie_ep_fn_readw(pcie, fn, reg); + val &= ~PCI_MSIX_FLAGS_QSIZE; + val |= interrupts; + cdns_pcie_ep_fn_writew(pcie, fn, reg, val); + + /* Set MSIX BAR and offset */ + reg = cap + PCI_MSIX_TABLE; + val = offset | bir; + cdns_pcie_ep_fn_writel(pcie, fn, reg, val); + + /* Set PBA BAR and offset. BAR must match MSIX BAR */ + reg = cap + PCI_MSIX_PBA; + val = (offset + (interrupts * PCI_MSIX_ENTRY_SIZE)) | bir; + cdns_pcie_ep_fn_writel(pcie, fn, reg, val); + + return 0; +} + static void cdns_pcie_ep_assert_intx(struct cdns_pcie_ep *ep, u8 fn, u8 intx, bool is_asserted) { @@ -330,6 +380,52 @@ static int cdns_pcie_ep_send_msi_irq(struct cdns_pcie_ep *ep, u8 fn, return 0; } +static int cdns_pcie_ep_send_msix_irq(struct cdns_pcie_ep *ep, u8 fn, + u16 interrupt_num) +{ + u32 cap = CDNS_PCIE_EP_FUNC_MSIX_CAP_OFFSET; + u32 tbl_offset, msg_data, reg, vec_ctrl; + struct cdns_pcie *pcie = &ep->pcie; + struct pci_epf_msix_tbl *msix_tbl; + struct cdns_pcie_epf *epf; + u64 pci_addr_mask = 0xff; + u64 msg_addr; + u16 flags; + u8 bir; + + /* Check whether the MSI-X feature has been enabled by the PCI host. */ + flags = cdns_pcie_ep_fn_readw(pcie, fn, cap + PCI_MSIX_FLAGS); + if (!(flags & PCI_MSIX_FLAGS_ENABLE)) + return -EINVAL; + + reg = cap + PCI_MSIX_TABLE; + tbl_offset = cdns_pcie_ep_fn_readl(pcie, fn, reg); + bir = tbl_offset & PCI_MSIX_TABLE_BIR; + tbl_offset &= PCI_MSIX_TABLE_OFFSET; + + epf = &ep->epf[fn]; + msix_tbl = epf->epf_bar[bir]->addr + tbl_offset; + msg_addr = msix_tbl[(interrupt_num - 1)].msg_addr; + msg_data = msix_tbl[(interrupt_num - 1)].msg_data; + vec_ctrl = msix_tbl[(interrupt_num - 1)].vector_ctrl; + + /* Set the outbound region if needed. */ + if (ep->irq_pci_addr != (msg_addr & ~pci_addr_mask) || + ep->irq_pci_fn != fn) { + /* First region was reserved for IRQ writes. */ + cdns_pcie_set_outbound_region(pcie, fn, 0, + false, + ep->irq_phys_addr, + msg_addr & ~pci_addr_mask, + pci_addr_mask + 1); + ep->irq_pci_addr = (msg_addr & ~pci_addr_mask); + ep->irq_pci_fn = fn; + } + writel(msg_data, ep->irq_cpu_addr + (msg_addr & pci_addr_mask)); + + return 0; +} + static int cdns_pcie_ep_raise_irq(struct pci_epc *epc, u8 fn, enum pci_epc_irq_type type, u16 interrupt_num) @@ -343,6 +439,9 @@ static int cdns_pcie_ep_raise_irq(struct pci_epc *epc, u8 fn, case PCI_EPC_IRQ_MSI: return cdns_pcie_ep_send_msi_irq(ep, fn, interrupt_num); + case PCI_EPC_IRQ_MSIX: + return cdns_pcie_ep_send_msix_irq(ep, fn, interrupt_num); + default: break; } @@ -380,7 +479,7 @@ static int cdns_pcie_ep_start(struct pci_epc *epc) static const struct pci_epc_features cdns_pcie_epc_features = { .linkup_notifier = false, .msi_capable = true, - .msix_capable = false, + .msix_capable = true, }; static const struct pci_epc_features* @@ -397,6 +496,8 @@ static const struct pci_epc_ops cdns_pcie_epc_ops = { .unmap_addr = cdns_pcie_ep_unmap_addr, .set_msi = cdns_pcie_ep_set_msi, .get_msi = cdns_pcie_ep_get_msi, + .set_msix = cdns_pcie_ep_set_msix, + .get_msix = cdns_pcie_ep_get_msix, .raise_irq = cdns_pcie_ep_raise_irq, .start = cdns_pcie_ep_start, .get_features = cdns_pcie_ep_get_features, @@ -455,6 +556,11 @@ int cdns_pcie_ep_setup(struct cdns_pcie_ep *ep) if (of_property_read_u8(np, "max-functions", &epc->max_functions) < 0) epc->max_functions = 1; + ep->epf = devm_kcalloc(dev, epc->max_functions, sizeof(*ep->epf), + GFP_KERNEL); + if (!ep->epf) + return -ENOMEM; + ret = pci_epc_mem_init(epc, pcie->mem_res->start, resource_size(pcie->mem_res)); if (ret < 0) { diff --git a/drivers/pci/controller/cadence/pcie-cadence.h b/drivers/pci/controller/cadence/pcie-cadence.h index 3490723169c6..1bf7d4e09b8a 100644 --- a/drivers/pci/controller/cadence/pcie-cadence.h +++ b/drivers/pci/controller/cadence/pcie-cadence.h @@ -113,6 +113,7 @@ #define CDNS_PCIE_EP_FUNC_BASE(fn) (((fn) << 12) & GENMASK(19, 12)) #define CDNS_PCIE_EP_FUNC_MSI_CAP_OFFSET 0x90 +#define CDNS_PCIE_EP_FUNC_MSIX_CAP_OFFSET 0xb0 /* * Root Port Registers (PCI configuration space for the root port function) @@ -302,6 +303,14 @@ struct cdns_pcie_rc { bool avail_ib_bar[CDNS_PCIE_RP_MAX_IB]; }; +/** + * struct cdns_pcie_epf - Structure to hold info about endpoint function + * @epf_bar: reference to the pci_epf_bar for the six Base Address Registers + */ +struct cdns_pcie_epf { + struct pci_epf_bar *epf_bar[PCI_STD_NUM_BARS]; +}; + /** * struct cdns_pcie_ep - private data for this PCIe endpoint controller driver * @pcie: Cadence PCIe controller @@ -329,6 +338,7 @@ struct cdns_pcie_ep { u64 irq_pci_addr; u8 irq_pci_fn; u8 irq_pending; + struct cdns_pcie_epf *epf; }; -- 2.17.1
Add host mode dt-bindings for TI's J721E SoC. Signed-off-by: Kishon Vijay Abraham I <kishon@ti.com> Reviewed-by: Rob Herring <robh@kernel.org> --- .../bindings/pci/ti,j721e-pci-host.yaml | 113 ++++++++++++++++++ 1 file changed, 113 insertions(+) create mode 100644 Documentation/devicetree/bindings/pci/ti,j721e-pci-host.yaml diff --git a/Documentation/devicetree/bindings/pci/ti,j721e-pci-host.yaml b/Documentation/devicetree/bindings/pci/ti,j721e-pci-host.yaml new file mode 100644 index 000000000000..d7b60487c6c3 --- /dev/null +++ b/Documentation/devicetree/bindings/pci/ti,j721e-pci-host.yaml @@ -0,0 +1,113 @@ +# SPDX-License-Identifier: (GPL-2.0-only OR BSD-2-Clause) +# Copyright (C) 2019 Texas Instruments Incorporated - http://www.ti.com/ +%YAML 1.2 +--- +$id: "http://devicetree.org/schemas/pci/ti,j721e-pci-host.yaml#" +$schema: "http://devicetree.org/meta-schemas/core.yaml#" + +title: TI J721E PCI Host (PCIe Wrapper) + +maintainers: + - Kishon Vijay Abraham I <kishon@ti.com> + +allOf: + - $ref: "cdns-pcie-host.yaml#" + +properties: + compatible: + enum: + - ti,j721e-pcie-host + + reg: + maxItems: 4 + + reg-names: + items: + - const: intd_cfg + - const: user_cfg + - const: reg + - const: cfg + + ti,syscon-pcie-ctrl: + description: Phandle to the SYSCON entry required for configuring PCIe mode + and link speed. + allOf: + - $ref: /schemas/types.yaml#/definitions/phandle + + power-domains: + maxItems: 1 + + clocks: + maxItems: 1 + description: clock-specifier to represent input to the PCIe + + clock-names: + items: + - const: fck + + vendor-id: + const: 0x104c + + device-id: + const: 0xb00d + + msi-map: true + +required: + - compatible + - reg + - reg-names + - ti,syscon-pcie-ctrl + - max-link-speed + - num-lanes + - power-domains + - clocks + - clock-names + - vendor-id + - device-id + - msi-map + - dma-coherent + - dma-ranges + - ranges + - reset-gpios + - phys + - phy-names + +examples: + - | + #include <dt-bindings/soc/ti,sci_pm_domain.h> + #include <dt-bindings/gpio/gpio.h> + + bus { + #address-cells = <2>; + #size-cells = <2>; + + pcie0_rc: pcie@2900000 { + compatible = "ti,j721e-pcie-host"; + reg = <0x00 0x02900000 0x00 0x1000>, + <0x00 0x02907000 0x00 0x400>, + <0x00 0x0d000000 0x00 0x00800000>, + <0x00 0x10000000 0x00 0x00001000>; + reg-names = "intd_cfg", "user_cfg", "reg", "cfg"; + ti,syscon-pcie-ctrl = <&pcie0_ctrl>; + max-link-speed = <3>; + num-lanes = <2>; + power-domains = <&k3_pds 239 TI_SCI_PD_EXCLUSIVE>; + clocks = <&k3_clks 239 1>; + clock-names = "fck"; + device_type = "pci"; + #address-cells = <3>; + #size-cells = <2>; + bus-range = <0x0 0xf>; + vendor-id = <0x104c>; + device-id = <0xb00d>; + msi-map = <0x0 &gic_its 0x0 0x10000>; + dma-coherent; + reset-gpios = <&exp1 6 GPIO_ACTIVE_HIGH>; + phys = <&serdes0_pcie_link>; + phy-names = "pcie-phy"; + ranges = <0x01000000 0x0 0x10001000 0x00 0x10001000 0x0 0x0010000>, + <0x02000000 0x0 0x10011000 0x00 0x10011000 0x0 0x7fef000>; + dma-ranges = <0x02000000 0x0 0x0 0x0 0x0 0x10000 0x0>; + }; + }; -- 2.17.1
Add PCIe EP mode dt-bindings for TI's J721E SoC. Signed-off-by: Kishon Vijay Abraham I <kishon@ti.com> Reviewed-by: Rob Herring <robh@kernel.org> --- .../bindings/pci/ti,j721e-pci-ep.yaml | 89 +++++++++++++++++++ 1 file changed, 89 insertions(+) create mode 100644 Documentation/devicetree/bindings/pci/ti,j721e-pci-ep.yaml diff --git a/Documentation/devicetree/bindings/pci/ti,j721e-pci-ep.yaml b/Documentation/devicetree/bindings/pci/ti,j721e-pci-ep.yaml new file mode 100644 index 000000000000..c09d25b2c1b2 --- /dev/null +++ b/Documentation/devicetree/bindings/pci/ti,j721e-pci-ep.yaml @@ -0,0 +1,89 @@ +# SPDX-License-Identifier: (GPL-2.0-only OR BSD-2-Clause) +# Copyright (C) 2020 Texas Instruments Incorporated - http://www.ti.com/ +%YAML 1.2 +--- +$id: "http://devicetree.org/schemas/pci/ti,j721e-pci-ep.yaml#" +$schema: "http://devicetree.org/meta-schemas/core.yaml#" + +title: TI J721E PCI EP (PCIe Wrapper) + +maintainers: + - Kishon Vijay Abraham I <kishon@ti.com> + +allOf: + - $ref: "cdns-pcie-ep.yaml#" + +properties: + compatible: + enum: + - ti,j721e-pcie-ep + + reg: + maxItems: 4 + + reg-names: + items: + - const: intd_cfg + - const: user_cfg + - const: reg + - const: mem + + ti,syscon-pcie-ctrl: + description: Phandle to the SYSCON entry required for configuring PCIe mode + and link speed. + allOf: + - $ref: /schemas/types.yaml#/definitions/phandle + + power-domains: + maxItems: 1 + + clocks: + maxItems: 1 + description: clock-specifier to represent input to the PCIe + + clock-names: + items: + - const: fck + + dma-coherent: + description: Indicates that the PCIe IP block can ensure the coherency + +required: + - compatible + - reg + - reg-names + - ti,syscon-pcie-ctrl + - max-link-speed + - num-lanes + - power-domains + - clocks + - clock-names + - cdns,max-outbound-regions + - dma-coherent + - max-functions + - phys + - phy-names + +examples: + - | + #include <dt-bindings/soc/ti,sci_pm_domain.h> + + pcie0_ep: pcie-ep@d000000 { + compatible = "ti,j721e-pcie-ep"; + reg = <0x00 0x02900000 0x00 0x1000>, + <0x00 0x02907000 0x00 0x400>, + <0x00 0x0d000000 0x00 0x00800000>, + <0x00 0x10000000 0x00 0x08000000>; + reg-names = "intd_cfg", "user_cfg", "reg", "mem"; + ti,syscon-pcie-ctrl = <&pcie0_ctrl>; + max-link-speed = <3>; + num-lanes = <2>; + power-domains = <&k3_pds 239 TI_SCI_PD_EXCLUSIVE>; + clocks = <&k3_clks 239 1>; + clock-names = "fck"; + cdns,max-outbound-regions = <16>; + max-functions = /bits/ 8 <6>; + dma-coherent; + phys = <&serdes0_pcie_link>; + phy-names = "pcie-phy"; + }; -- 2.17.1
Add support for PCIe controller in J721E SoC. The controller uses the Cadence PCIe core programmed by pcie-cadence*.c. The PCIe controller will work in both host mode and device mode. Some of the features of the controller are: *) Supports both RC mode and EP mode *) Supports MSI and MSI-X support *) Supports upto GEN3 speed mode *) Supports SR-IOV capability *) Ability to route all transactions via SMMU (support will be added in a later patch). Signed-off-by: Kishon Vijay Abraham I <kishon@ti.com> --- drivers/pci/controller/cadence/Kconfig | 23 + drivers/pci/controller/cadence/Makefile | 1 + drivers/pci/controller/cadence/pci-j721e.c | 490 ++++++++++++++++++ .../controller/cadence/pcie-cadence-host.c | 4 +- drivers/pci/controller/cadence/pcie-cadence.h | 8 + 5 files changed, 524 insertions(+), 2 deletions(-) create mode 100644 drivers/pci/controller/cadence/pci-j721e.c diff --git a/drivers/pci/controller/cadence/Kconfig b/drivers/pci/controller/cadence/Kconfig index b76b3cf55ce5..5d30564190e1 100644 --- a/drivers/pci/controller/cadence/Kconfig +++ b/drivers/pci/controller/cadence/Kconfig @@ -42,4 +42,27 @@ config PCIE_CADENCE_PLAT_EP endpoint mode. This PCIe controller may be embedded into many different vendors SoCs. +config PCI_J721E + bool + +config PCI_J721E_HOST + bool "TI J721E PCIe platform host controller" + depends on OF + select PCIE_CADENCE_HOST + select PCI_J721E + help + Say Y here if you want to support the TI J721E PCIe platform + controller in host mode. TI J721E PCIe controller uses Cadence PCIe + core. + +config PCI_J721E_EP + bool "TI J721E PCIe platform endpoint controller" + depends on OF + depends on PCI_ENDPOINT + select PCIE_CADENCE_EP + select PCI_J721E + help + Say Y here if you want to support the TI J721E PCIe platform + controller in endpoint mode. TI J721E PCIe controller uses Cadence PCIe + core. endmenu diff --git a/drivers/pci/controller/cadence/Makefile b/drivers/pci/controller/cadence/Makefile index 232a3f20876a..9bac5fb2f13d 100644 --- a/drivers/pci/controller/cadence/Makefile +++ b/drivers/pci/controller/cadence/Makefile @@ -3,3 +3,4 @@ obj-$(CONFIG_PCIE_CADENCE) += pcie-cadence.o obj-$(CONFIG_PCIE_CADENCE_HOST) += pcie-cadence-host.o obj-$(CONFIG_PCIE_CADENCE_EP) += pcie-cadence-ep.o obj-$(CONFIG_PCIE_CADENCE_PLAT) += pcie-cadence-plat.o +obj-$(CONFIG_PCI_J721E) += pci-j721e.o diff --git a/drivers/pci/controller/cadence/pci-j721e.c b/drivers/pci/controller/cadence/pci-j721e.c new file mode 100644 index 000000000000..9b3ab880a3c5 --- /dev/null +++ b/drivers/pci/controller/cadence/pci-j721e.c @@ -0,0 +1,490 @@ +// SPDX-License-Identifier: GPL-2.0 +/** + * pci-j721e - PCIe controller driver for TI's J721E SoCs + * + * Copyright (C) 2020 Texas Instruments Incorporated - http://www.ti.com + * Author: Kishon Vijay Abraham I <kishon@ti.com> + */ + +#include <linux/delay.h> +#include <linux/gpio/consumer.h> +#include <linux/io.h> +#include <linux/irqchip/chained_irq.h> +#include <linux/irqdomain.h> +#include <linux/mfd/syscon.h> +#include <linux/of_device.h> +#include <linux/of_irq.h> +#include <linux/pci.h> +#include <linux/pm_runtime.h> +#include <linux/regmap.h> + +#include "../../pci.h" +#include "pcie-cadence.h" + +#define ENABLE_REG_SYS_2 0x108 +#define STATUS_REG_SYS_2 0x508 +#define STATUS_CLR_REG_SYS_2 0x708 +#define LINK_DOWN BIT(1) + +#define J721E_PCIE_USER_CMD_STATUS 0x4 +#define LINK_TRAINING_ENABLE BIT(0) + +#define J721E_PCIE_USER_LINKSTATUS 0x14 +#define LINK_STATUS GENMASK(1, 0) + +enum link_status { + NO_RECEIVERS_DETECTED, + LINK_TRAINING_IN_PROGRESS, + LINK_UP_DL_IN_PROGRESS, + LINK_UP_DL_COMPLETED, +}; + +#define J721E_MODE_RC BIT(7) +#define LANE_COUNT_MASK BIT(8) +#define LANE_COUNT(n) ((n) << 8) + +#define GENERATION_SEL_MASK GENMASK(1, 0) + +#define MAX_LANES 2 + +struct j721e_pcie { + struct device *dev; + u32 mode; + u32 num_lanes; + struct cdns_pcie *cdns_pcie; + void __iomem *user_cfg_base; + void __iomem *intd_cfg_base; +}; + +enum j721e_pcie_mode { + PCI_MODE_RC, + PCI_MODE_EP, +}; + +struct j721e_pcie_data { + enum j721e_pcie_mode mode; +}; + +static inline u32 j721e_pcie_user_readl(struct j721e_pcie *pcie, u32 offset) +{ + return readl(pcie->user_cfg_base + offset); +} + +static inline void j721e_pcie_user_writel(struct j721e_pcie *pcie, u32 offset, + u32 value) +{ + writel(value, pcie->user_cfg_base + offset); +} + +static inline u32 j721e_pcie_intd_readl(struct j721e_pcie *pcie, u32 offset) +{ + return readl(pcie->intd_cfg_base + offset); +} + +static inline void j721e_pcie_intd_writel(struct j721e_pcie *pcie, u32 offset, + u32 value) +{ + writel(value, pcie->intd_cfg_base + offset); +} + +static irqreturn_t j721e_pcie_link_irq_handler(int irq, void *priv) +{ + struct j721e_pcie *pcie = priv; + struct device *dev = pcie->dev; + u32 reg; + + reg = j721e_pcie_intd_readl(pcie, STATUS_REG_SYS_2); + if (!(reg & LINK_DOWN)) + return IRQ_NONE; + + dev_err(dev, "LINK DOWN!\n"); + + j721e_pcie_intd_writel(pcie, STATUS_CLR_REG_SYS_2, LINK_DOWN); + return IRQ_HANDLED; +} + +static void j721e_pcie_config_link_irq(struct j721e_pcie *pcie) +{ + u32 reg; + + reg = j721e_pcie_intd_readl(pcie, ENABLE_REG_SYS_2); + reg |= LINK_DOWN; + j721e_pcie_intd_writel(pcie, ENABLE_REG_SYS_2, reg); +} + +static int j721e_pcie_start_link(struct cdns_pcie *cdns_pcie) +{ + struct j721e_pcie *pcie = dev_get_drvdata(cdns_pcie->dev); + u32 reg; + + reg = j721e_pcie_user_readl(pcie, J721E_PCIE_USER_CMD_STATUS); + reg |= LINK_TRAINING_ENABLE; + j721e_pcie_user_writel(pcie, J721E_PCIE_USER_CMD_STATUS, reg); + + return 0; +} + +static void j721e_pcie_stop_link(struct cdns_pcie *cdns_pcie) +{ + struct j721e_pcie *pcie = dev_get_drvdata(cdns_pcie->dev); + u32 reg; + + reg = j721e_pcie_user_readl(pcie, J721E_PCIE_USER_CMD_STATUS); + reg &= ~LINK_TRAINING_ENABLE; + j721e_pcie_user_writel(pcie, J721E_PCIE_USER_CMD_STATUS, reg); +} + +static bool j721e_pcie_link_up(struct cdns_pcie *cdns_pcie) +{ + struct j721e_pcie *pcie = dev_get_drvdata(cdns_pcie->dev); + u32 reg; + + reg = j721e_pcie_user_readl(pcie, J721E_PCIE_USER_LINKSTATUS); + reg &= LINK_STATUS; + if (reg == LINK_UP_DL_COMPLETED) + return true; + + return false; +} + +static const struct cdns_pcie_ops j721e_ops_ops = { + .start_link = j721e_pcie_start_link, + .stop_link = j721e_pcie_stop_link, + .link_up = j721e_pcie_link_up, +}; + +static int j721e_pcie_set_mode(struct j721e_pcie *pcie, struct regmap *syscon) +{ + struct device *dev = pcie->dev; + u32 mask = J721E_MODE_RC; + u32 mode = pcie->mode; + u32 val = 0; + int ret = 0; + + if (mode == PCI_MODE_RC) + val = J721E_MODE_RC; + + ret = regmap_update_bits(syscon, 0, mask, val); + if (ret) + dev_err(dev, "failed to set pcie mode\n"); + + return ret; +} + +static int j721e_pcie_set_link_speed(struct j721e_pcie *pcie, + struct regmap *syscon) +{ + struct device *dev = pcie->dev; + struct device_node *np = dev->of_node; + int link_speed; + u32 val = 0; + int ret; + + link_speed = of_pci_get_max_link_speed(np); + if (link_speed < 2) + link_speed = 2; + + val = link_speed - 1; + ret = regmap_update_bits(syscon, 0, GENERATION_SEL_MASK, val); + if (ret) + dev_err(dev, "failed to set link speed\n"); + + return ret; +} + +static int j721e_pcie_set_lane_count(struct j721e_pcie *pcie, + struct regmap *syscon) +{ + struct device *dev = pcie->dev; + u32 lanes = pcie->num_lanes; + u32 val = 0; + int ret; + + val = LANE_COUNT(lanes - 1); + ret = regmap_update_bits(syscon, 0, LANE_COUNT_MASK, val); + if (ret) + dev_err(dev, "failed to set link count\n"); + + return ret; +} + +static int j721e_pcie_ctrl_init(struct j721e_pcie *pcie) +{ + struct device *dev = pcie->dev; + struct device_node *node = dev->of_node; + struct regmap *syscon; + int ret; + + syscon = syscon_regmap_lookup_by_phandle(node, "ti,syscon-pcie-ctrl"); + if (IS_ERR(syscon)) { + dev_err(dev, "Unable to get ti,syscon-pcie-ctrl regmap\n"); + return PTR_ERR(syscon); + } + + ret = j721e_pcie_set_mode(pcie, syscon); + if (ret < 0) { + dev_err(dev, "Failed to set pci mode\n"); + return ret; + } + + ret = j721e_pcie_set_link_speed(pcie, syscon); + if (ret < 0) { + dev_err(dev, "Failed to set link speed\n"); + return ret; + } + + ret = j721e_pcie_set_lane_count(pcie, syscon); + if (ret < 0) { + dev_err(dev, "Failed to set num-lanes\n"); + return ret; + } + + return 0; +} + +static int cdns_ti_pcie_config_read(struct pci_bus *bus, unsigned int devfn, + int where, int size, u32 *value) +{ + struct pci_host_bridge *bridge = pci_find_host_bridge(bus); + struct cdns_pcie_rc *rc = pci_host_bridge_priv(bridge); + unsigned int busn = bus->number; + + if (busn == rc->bus_range->start) + return pci_generic_config_read32(bus, devfn, where, size, + value); + + return pci_generic_config_read(bus, devfn, where, size, value); +} + +static int cdns_ti_pcie_config_write(struct pci_bus *bus, unsigned int devfn, + int where, int size, u32 value) +{ + struct pci_host_bridge *bridge = pci_find_host_bridge(bus); + struct cdns_pcie_rc *rc = pci_host_bridge_priv(bridge); + unsigned int busn = bus->number; + + if (busn == rc->bus_range->start) + return pci_generic_config_write32(bus, devfn, where, size, + value); + + return pci_generic_config_write(bus, devfn, where, size, value); +} + +static struct pci_ops cdns_ti_pcie_host_ops = { + .map_bus = cdns_pci_map_bus, + .read = cdns_ti_pcie_config_read, + .write = cdns_ti_pcie_config_write, +}; + +static const struct j721e_pcie_data j721e_pcie_rc_data = { + .mode = PCI_MODE_RC, +}; + +static const struct j721e_pcie_data j721e_pcie_ep_data = { + .mode = PCI_MODE_EP, +}; + +static const struct of_device_id of_j721e_pcie_match[] = { + { + .compatible = "ti,j721e-pcie-host", + .data = &j721e_pcie_rc_data, + }, + { + .compatible = "ti,j721e-pcie-ep", + .data = &j721e_pcie_ep_data, + }, + {}, +}; + +static int j721e_pcie_probe(struct platform_device *pdev) +{ + struct device *dev = &pdev->dev; + struct device_node *node = dev->of_node; + struct pci_host_bridge *bridge; + struct j721e_pcie_data *data; + struct cdns_pcie *cdns_pcie; + struct j721e_pcie *pcie; + struct cdns_pcie_rc *rc; + struct cdns_pcie_ep *ep; + struct gpio_desc *gpiod; + void __iomem *base; + u32 num_lanes; + u32 mode; + int ret; + int irq; + + data = (struct j721e_pcie_data *)of_device_get_match_data(dev); + if (!data) + return -EINVAL; + + mode = (u32)data->mode; + + pcie = devm_kzalloc(dev, sizeof(*pcie), GFP_KERNEL); + if (!pcie) + return -ENOMEM; + + pcie->dev = dev; + pcie->mode = mode; + + base = devm_platform_ioremap_resource_byname(pdev, "intd_cfg"); + if (IS_ERR(base)) + return PTR_ERR(base); + pcie->intd_cfg_base = base; + + base = devm_platform_ioremap_resource_byname(pdev, "user_cfg"); + if (IS_ERR(base)) + return PTR_ERR(base); + pcie->user_cfg_base = base; + + ret = of_property_read_u32(node, "num-lanes", &num_lanes); + if (ret || num_lanes > MAX_LANES) + num_lanes = 1; + pcie->num_lanes = num_lanes; + + irq = platform_get_irq_byname(pdev, "link_state"); + if (irq < 0) + return irq; + + dev_set_drvdata(dev, pcie); + pm_runtime_enable(dev); + ret = pm_runtime_get_sync(dev); + if (ret < 0) { + dev_err(dev, "pm_runtime_get_sync failed\n"); + goto err_get_sync; + } + + ret = j721e_pcie_ctrl_init(pcie); + if (ret < 0) { + dev_err(dev, "pm_runtime_get_sync failed\n"); + goto err_get_sync; + } + + ret = devm_request_irq(dev, irq, j721e_pcie_link_irq_handler, 0, + "j721e-pcie-link-down-irq", pcie); + if (ret < 0) { + dev_err(dev, "failed to request link state IRQ %d\n", irq); + goto err_get_sync; + } + + j721e_pcie_config_link_irq(pcie); + + switch (mode) { + case PCI_MODE_RC: + if (!IS_ENABLED(CONFIG_PCIE_CADENCE_HOST)) { + ret = -ENODEV; + goto err_get_sync; + } + + bridge = devm_pci_alloc_host_bridge(dev, sizeof(*rc)); + if (!bridge) { + ret = -ENOMEM; + goto err_get_sync; + } + + bridge->ops = &cdns_ti_pcie_host_ops; + rc = pci_host_bridge_priv(bridge); + + cdns_pcie = &rc->pcie; + cdns_pcie->dev = dev; + cdns_pcie->ops = &j721e_ops_ops; + pcie->cdns_pcie = cdns_pcie; + + gpiod = devm_gpiod_get_optional(dev, "reset", GPIOD_OUT_LOW); + if (IS_ERR(gpiod)) { + ret = PTR_ERR(gpiod); + if (ret != -EPROBE_DEFER) + dev_err(dev, "Failed to get reset GPIO\n"); + goto err_get_sync; + } + + ret = cdns_pcie_init_phy(dev, cdns_pcie); + if (ret) { + dev_err(dev, "Failed to init phy\n"); + goto err_get_sync; + } + + /* + * "Power Sequencing and Reset Signal Timings" table in + * PCI EXPRESS CARD ELECTROMECHANICAL SPECIFICATION, REV. 3.0 + * indicates PERST# should be deasserted after minimum of 100us + * once REFCLK is stable. The REFCLK to the connector in RC + * mode is selected while enabling the PHY. So deassert PERST# + * after 100 us. + */ + if (gpiod) { + usleep_range(100, 200); + gpiod_set_value_cansleep(gpiod, 1); + } + + ret = cdns_pcie_host_setup(rc); + if (ret < 0) + goto err_pcie_setup; + + break; + case PCI_MODE_EP: + if (!IS_ENABLED(CONFIG_PCIE_CADENCE_EP)) { + ret = -ENODEV; + goto err_get_sync; + } + + ep = devm_kzalloc(dev, sizeof(*ep), GFP_KERNEL); + if (!ep) { + ret = -ENOMEM; + goto err_get_sync; + } + + cdns_pcie = &ep->pcie; + cdns_pcie->dev = dev; + cdns_pcie->ops = &j721e_ops_ops; + pcie->cdns_pcie = cdns_pcie; + + ret = cdns_pcie_init_phy(dev, cdns_pcie); + if (ret) { + dev_err(dev, "Failed to init phy\n"); + goto err_get_sync; + } + + ret = cdns_pcie_ep_setup(ep); + if (ret < 0) + goto err_pcie_setup; + + break; + default: + dev_err(dev, "INVALID device type %d\n", mode); + } + + return 0; + +err_pcie_setup: + cdns_pcie_disable_phy(cdns_pcie); + +err_get_sync: + pm_runtime_put(dev); + pm_runtime_disable(dev); + + return ret; +} + +static int j721e_pcie_remove(struct platform_device *pdev) +{ + struct j721e_pcie *pcie = platform_get_drvdata(pdev); + struct cdns_pcie *cdns_pcie = pcie->cdns_pcie; + struct device *dev = &pdev->dev; + + cdns_pcie_disable_phy(cdns_pcie); + pm_runtime_put(dev); + pm_runtime_disable(dev); + + return 0; +} + +static struct platform_driver j721e_pcie_driver = { + .probe = j721e_pcie_probe, + .remove = j721e_pcie_remove, + .driver = { + .name = "j721e-pcie", + .of_match_table = of_j721e_pcie_match, + .suppress_bind_attrs = true, + }, +}; +builtin_platform_driver(j721e_pcie_driver); diff --git a/drivers/pci/controller/cadence/pcie-cadence-host.c b/drivers/pci/controller/cadence/pcie-cadence-host.c index 7ee9e06f1285..cf8b34b71b8f 100644 --- a/drivers/pci/controller/cadence/pcie-cadence-host.c +++ b/drivers/pci/controller/cadence/pcie-cadence-host.c @@ -17,8 +17,8 @@ static u64 bar_max_size[] = { [RP_NO_BAR] = _BITULL(63), }; -static void __iomem *cdns_pci_map_bus(struct pci_bus *bus, unsigned int devfn, - int where) +void __iomem *cdns_pci_map_bus(struct pci_bus *bus, unsigned int devfn, + int where) { struct pci_host_bridge *bridge = pci_find_host_bridge(bus); struct cdns_pcie_rc *rc = pci_host_bridge_priv(bridge); diff --git a/drivers/pci/controller/cadence/pcie-cadence.h b/drivers/pci/controller/cadence/pcie-cadence.h index 1bf7d4e09b8a..f8e8144e222d 100644 --- a/drivers/pci/controller/cadence/pcie-cadence.h +++ b/drivers/pci/controller/cadence/pcie-cadence.h @@ -469,11 +469,19 @@ static inline bool cdns_pcie_link_up(struct cdns_pcie *pcie) #ifdef CONFIG_PCIE_CADENCE_HOST int cdns_pcie_host_setup(struct cdns_pcie_rc *rc); +void __iomem *cdns_pci_map_bus(struct pci_bus *bus, unsigned int devfn, + int where); #else static inline int cdns_pcie_host_setup(struct cdns_pcie_rc *rc) { return 0; } + +static void __iomem *cdns_pci_map_bus(struct pci_bus *bus, unsigned int devfn, + int where) +{ + return NULL; +} #endif #ifdef CONFIG_PCIE_CADENCE_EP -- 2.17.1
Add J721E in pci_device_id table so that pci-epf-test can be used for testing PCIe EP in J721E. Reviewed-by: Rob Herring <robh@kernel.org> Signed-off-by: Kishon Vijay Abraham I <kishon@ti.com> --- drivers/misc/pci_endpoint_test.c | 9 +++++++++ 1 file changed, 9 insertions(+) diff --git a/drivers/misc/pci_endpoint_test.c b/drivers/misc/pci_endpoint_test.c index ef5a1af6bab7..a70b17e5dd9a 100644 --- a/drivers/misc/pci_endpoint_test.c +++ b/drivers/misc/pci_endpoint_test.c @@ -68,6 +68,7 @@ #define PCI_ENDPOINT_TEST_FLAGS 0x2c #define FLAG_USE_DMA BIT(0) +#define PCI_DEVICE_ID_TI_J721E 0xb00d #define PCI_DEVICE_ID_TI_AM654 0xb00c #define is_am654_pci_dev(pdev) \ @@ -930,6 +931,11 @@ static const struct pci_endpoint_test_data am654_data = { .irq_type = IRQ_TYPE_MSI, }; +static const struct pci_endpoint_test_data j721e_data = { + .alignment = 256, + .irq_type = IRQ_TYPE_MSI, +}; + static const struct pci_device_id pci_endpoint_test_tbl[] = { { PCI_DEVICE(PCI_VENDOR_ID_TI, PCI_DEVICE_ID_TI_DRA74x), .driver_data = (kernel_ulong_t)&default_data, @@ -942,6 +948,9 @@ static const struct pci_device_id pci_endpoint_test_tbl[] = { { PCI_DEVICE(PCI_VENDOR_ID_TI, PCI_DEVICE_ID_TI_AM654), .driver_data = (kernel_ulong_t)&am654_data }, + { PCI_DEVICE(PCI_VENDOR_ID_TI, PCI_DEVICE_ID_TI_J721E), + .driver_data = (kernel_ulong_t)&j721e_data, + }, { } }; MODULE_DEVICE_TABLE(pci, pci_endpoint_test_tbl); -- 2.17.1
Add Kishon Vijay Abraham I as MAINTAINER for TI J721E SoC PCIe. Acked-by: Rob Herring <robh@kernel.org> Signed-off-by: Kishon Vijay Abraham I <kishon@ti.com> --- MAINTAINERS | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-) diff --git a/MAINTAINERS b/MAINTAINERS index 2926327e4976..9d40e1318f7c 100644 --- a/MAINTAINERS +++ b/MAINTAINERS @@ -12978,12 +12978,14 @@ S: Maintained F: Documentation/devicetree/bindings/pci/designware-pcie.txt F: drivers/pci/controller/dwc/*designware* -PCI DRIVER FOR TI DRA7XX +PCI DRIVER FOR TI DRA7XX/J721E M: Kishon Vijay Abraham I <kishon@ti.com> L: linux-omap@vger.kernel.org L: linux-pci@vger.kernel.org +L: linux-arm-kernel@lists.infradead.org S: Supported F: Documentation/devicetree/bindings/pci/ti-pci.txt +F: drivers/pci/controller/cadence/pci-j721e.c F: drivers/pci/controller/dwc/pci-dra7xx.c PCI DRIVER FOR TI KEYSTONE -- 2.17.1
On Thu, May 21, 2020 at 9:37 PM Kishon Vijay Abraham I <kishon@ti.com> wrote:
>
> Certain platforms like TI's J721E using Cadence PCIe IP can perform only
> 32-bit accesses for reading or writing to Cadence registers. Convert all
> read and write accesses to 32-bit in Cadence PCIe driver in preparation
> for adding PCIe support in TI's J721E SoC.
Looking more closely I don't think cdns_pcie_ep_assert_intx is okay
with this and never can be given the PCI_COMMAND and PCI_STATUS
registers are in the same word (IIRC, that's the main reason 32-bit
config space accesses are broken). So this isn't going to work at
least for EP accesses. And maybe you need a custom .raise_irq() hook
to minimize any problems (such as making the RMW atomic at least from
the endpoint's perspective).
Rob
Hi Rob, On 5/22/2020 9:24 PM, Rob Herring wrote: > On Thu, May 21, 2020 at 9:37 PM Kishon Vijay Abraham I <kishon@ti.com> wrote: >> >> Certain platforms like TI's J721E using Cadence PCIe IP can perform only >> 32-bit accesses for reading or writing to Cadence registers. Convert all >> read and write accesses to 32-bit in Cadence PCIe driver in preparation >> for adding PCIe support in TI's J721E SoC. > > Looking more closely I don't think cdns_pcie_ep_assert_intx is okay > with this and never can be given the PCI_COMMAND and PCI_STATUS > registers are in the same word (IIRC, that's the main reason 32-bit > config space accesses are broken). So this isn't going to work at right, PCI_STATUS has write '1' to clear bits and there's a chance that it could be reset while raising legacy interrupt. While this cannot be avoided for TI's J721E, other platforms doesn't have to have this limitation. > least for EP accesses. And maybe you need a custom .raise_irq() hook > to minimize any problems (such as making the RMW atomic at least from > the endpoint's perspective). This is to make sure EP doesn't update in-consistent state when RC is updating the PCI_STATUS register? Since this involves two different systems, how do we make this atomic? Thanks Kishon
On Sun, May 24, 2020 at 9:30 PM Kishon Vijay Abraham I <kishon@ti.com> wrote:
>
> Hi Rob,
>
> On 5/22/2020 9:24 PM, Rob Herring wrote:
> > On Thu, May 21, 2020 at 9:37 PM Kishon Vijay Abraham I <kishon@ti.com> wrote:
> >>
> >> Certain platforms like TI's J721E using Cadence PCIe IP can perform only
> >> 32-bit accesses for reading or writing to Cadence registers. Convert all
> >> read and write accesses to 32-bit in Cadence PCIe driver in preparation
> >> for adding PCIe support in TI's J721E SoC.
> >
> > Looking more closely I don't think cdns_pcie_ep_assert_intx is okay
> > with this and never can be given the PCI_COMMAND and PCI_STATUS
> > registers are in the same word (IIRC, that's the main reason 32-bit
> > config space accesses are broken). So this isn't going to work at
>
> right, PCI_STATUS has write '1' to clear bits and there's a chance that it
> could be reset while raising legacy interrupt. While this cannot be avoided for
> TI's J721E, other platforms doesn't have to have this limitation.
> > least for EP accesses. And maybe you need a custom .raise_irq() hook
> > to minimize any problems (such as making the RMW atomic at least from
> > the endpoint's perspective).
>
> This is to make sure EP doesn't update in-consistent state when RC is updating
> the PCI_STATUS register? Since this involves two different systems, how do we
> make this atomic?
You can't make it atomic WRT both systems, but is there locking around
each RMW? Specifically, are preemption and interrupts disabled to
ensure time between a read and write are minimized? You wouldn't want
interrupts disabled during the delay too though (i.e. around
.raise_irq()).
BTW, I've asked this question before, but aren't PCI legacy interrupts
level triggered? If so, isn't generating a pulse wrong?
Rob
On Fri, 22 May 2020 09:06:23 +0530, Kishon Vijay Abraham I wrote:
> "mem" is not a memory resource and it overlaps with PCIe config space
> and memory region. Removve "mem" from reg binding.
>
> Signed-off-by: Kishon Vijay Abraham I <kishon@ti.com>
> ---
> .../devicetree/bindings/pci/cdns,cdns-pcie-host.yaml | 8 +++-----
> 1 file changed, 3 insertions(+), 5 deletions(-)
>
Reviewed-by: Rob Herring <robh@kernel.org>
Hi Rob, On 5/26/2020 8:42 PM, Rob Herring wrote: > On Sun, May 24, 2020 at 9:30 PM Kishon Vijay Abraham I <kishon@ti.com> wrote: >> >> Hi Rob, >> >> On 5/22/2020 9:24 PM, Rob Herring wrote: >>> On Thu, May 21, 2020 at 9:37 PM Kishon Vijay Abraham I <kishon@ti.com> wrote: >>>> >>>> Certain platforms like TI's J721E using Cadence PCIe IP can perform only >>>> 32-bit accesses for reading or writing to Cadence registers. Convert all >>>> read and write accesses to 32-bit in Cadence PCIe driver in preparation >>>> for adding PCIe support in TI's J721E SoC. >>> >>> Looking more closely I don't think cdns_pcie_ep_assert_intx is okay >>> with this and never can be given the PCI_COMMAND and PCI_STATUS >>> registers are in the same word (IIRC, that's the main reason 32-bit >>> config space accesses are broken). So this isn't going to work at >> >> right, PCI_STATUS has write '1' to clear bits and there's a chance that it >> could be reset while raising legacy interrupt. While this cannot be avoided for >> TI's J721E, other platforms doesn't have to have this limitation. >>> least for EP accesses. And maybe you need a custom .raise_irq() hook >>> to minimize any problems (such as making the RMW atomic at least from >>> the endpoint's perspective). >> >> This is to make sure EP doesn't update in-consistent state when RC is updating >> the PCI_STATUS register? Since this involves two different systems, how do we >> make this atomic? > > You can't make it atomic WRT both systems, but is there locking around > each RMW? Specifically, are preemption and interrupts disabled to > ensure time between a read and write are minimized? You wouldn't want > interrupts disabled during the delay too though (i.e. around > .raise_irq()). Okay, I'll add spin spin_lock_irqsave() in cdns_pcie_write_sz(). As you also pointed below that delay for legacy interrupt is wrong and it has to be fixed (with a later series). How do you want to handle cdns_pcie_ep_fn_writew() now? Because now we are changing the default implementation to perform only 32-bit access (used for legacy interrupt, msi-x interrupt and while writing standard headers) and it's not okay only for legacy interrupts for platforms other than TI. So just for legacy interrupt, you want me to add a different accessor which does not perform 32-bit writes (while we add a different .raise_irq for TI platform? > > BTW, I've asked this question before, but aren't PCI legacy interrupts > level triggered? If so, isn't generating a pulse wrong? You are right. This is wrong and it has to be fixed. I'll work on this later. Thanks Kishon
On Wed, May 27, 2020 at 4:49 AM Kishon Vijay Abraham I <kishon@ti.com> wrote: > > Hi Rob, > > On 5/26/2020 8:42 PM, Rob Herring wrote: > > On Sun, May 24, 2020 at 9:30 PM Kishon Vijay Abraham I <kishon@ti.com> wrote: > >> > >> Hi Rob, > >> > >> On 5/22/2020 9:24 PM, Rob Herring wrote: > >>> On Thu, May 21, 2020 at 9:37 PM Kishon Vijay Abraham I <kishon@ti.com> wrote: > >>>> > >>>> Certain platforms like TI's J721E using Cadence PCIe IP can perform only > >>>> 32-bit accesses for reading or writing to Cadence registers. Convert all > >>>> read and write accesses to 32-bit in Cadence PCIe driver in preparation > >>>> for adding PCIe support in TI's J721E SoC. > >>> > >>> Looking more closely I don't think cdns_pcie_ep_assert_intx is okay > >>> with this and never can be given the PCI_COMMAND and PCI_STATUS > >>> registers are in the same word (IIRC, that's the main reason 32-bit > >>> config space accesses are broken). So this isn't going to work at > >> > >> right, PCI_STATUS has write '1' to clear bits and there's a chance that it > >> could be reset while raising legacy interrupt. While this cannot be avoided for > >> TI's J721E, other platforms doesn't have to have this limitation. > >>> least for EP accesses. And maybe you need a custom .raise_irq() hook > >>> to minimize any problems (such as making the RMW atomic at least from > >>> the endpoint's perspective). > >> > >> This is to make sure EP doesn't update in-consistent state when RC is updating > >> the PCI_STATUS register? Since this involves two different systems, how do we > >> make this atomic? > > > > You can't make it atomic WRT both systems, but is there locking around > > each RMW? Specifically, are preemption and interrupts disabled to > > ensure time between a read and write are minimized? You wouldn't want > > interrupts disabled during the delay too though (i.e. around > > .raise_irq()). > > Okay, I'll add spin spin_lock_irqsave() in cdns_pcie_write_sz(). As you also > pointed below that delay for legacy interrupt is wrong and it has to be fixed > (with a later series). But you don't need a lock everywhere. You need locks in the callers (and only sometimes). > How do you want to handle cdns_pcie_ep_fn_writew() now? Because now we are > changing the default implementation to perform only 32-bit access (used for > legacy interrupt, msi-x interrupt and while writing standard headers) and it's > not okay only for legacy interrupts for platforms other than TI. Now I'm wondering how set_msi is not racy in the current code with the host setting/clearing PCI_MSI_FLAGS_ENABLE? Maybe that bit is RO from the EP side? Ultimately I think you're going to have to provide your own endpoint functions or you need accessors for specific registers like PCI_MSI_FLAGS. Then for example, you just rely on the 2 bytes before PCI_MSI_FLAGS being reserved and do a 32-bit access without a RMW. Trying to abstract this at the register read/write level is going to be fragile. Rob
Hi Rob, On 5/27/2020 10:07 PM, Rob Herring wrote: > On Wed, May 27, 2020 at 4:49 AM Kishon Vijay Abraham I <kishon@ti.com> wrote: >> >> Hi Rob, >> >> On 5/26/2020 8:42 PM, Rob Herring wrote: >>> On Sun, May 24, 2020 at 9:30 PM Kishon Vijay Abraham I <kishon@ti.com> wrote: >>>> >>>> Hi Rob, >>>> >>>> On 5/22/2020 9:24 PM, Rob Herring wrote: >>>>> On Thu, May 21, 2020 at 9:37 PM Kishon Vijay Abraham I <kishon@ti.com> wrote: >>>>>> >>>>>> Certain platforms like TI's J721E using Cadence PCIe IP can perform only >>>>>> 32-bit accesses for reading or writing to Cadence registers. Convert all >>>>>> read and write accesses to 32-bit in Cadence PCIe driver in preparation >>>>>> for adding PCIe support in TI's J721E SoC. >>>>> >>>>> Looking more closely I don't think cdns_pcie_ep_assert_intx is okay >>>>> with this and never can be given the PCI_COMMAND and PCI_STATUS >>>>> registers are in the same word (IIRC, that's the main reason 32-bit >>>>> config space accesses are broken). So this isn't going to work at >>>> >>>> right, PCI_STATUS has write '1' to clear bits and there's a chance that it >>>> could be reset while raising legacy interrupt. While this cannot be avoided for >>>> TI's J721E, other platforms doesn't have to have this limitation. >>>>> least for EP accesses. And maybe you need a custom .raise_irq() hook >>>>> to minimize any problems (such as making the RMW atomic at least from >>>>> the endpoint's perspective). >>>> >>>> This is to make sure EP doesn't update in-consistent state when RC is updating >>>> the PCI_STATUS register? Since this involves two different systems, how do we >>>> make this atomic? >>> >>> You can't make it atomic WRT both systems, but is there locking around >>> each RMW? Specifically, are preemption and interrupts disabled to >>> ensure time between a read and write are minimized? You wouldn't want >>> interrupts disabled during the delay too though (i.e. around >>> .raise_irq()). >> >> Okay, I'll add spin spin_lock_irqsave() in cdns_pcie_write_sz(). As you also >> pointed below that delay for legacy interrupt is wrong and it has to be fixed >> (with a later series). > > But you don't need a lock everywhere. You need locks in the callers > (and only sometimes). Okay, the locks should be added only for registers where HOST can also write to the same register? Maybe only raise_irq then.. > >> How do you want to handle cdns_pcie_ep_fn_writew() now? Because now we are >> changing the default implementation to perform only 32-bit access (used for >> legacy interrupt, msi-x interrupt and while writing standard headers) and it's >> not okay only for legacy interrupts for platforms other than TI. > > Now I'm wondering how set_msi is not racy in the current code with the > host setting/clearing PCI_MSI_FLAGS_ENABLE? Maybe that bit is RO from > the EP side? set_msi/set_msix is a one time configuration that is invoked before the host establishes the link with the endpoint. I don't think we have to consider this as racy. Thanks Kishon > > Ultimately I think you're going to have to provide your own endpoint > functions or you need accessors for specific registers like > PCI_MSI_FLAGS. Then for example, you just rely on the 2 bytes before > PCI_MSI_FLAGS being reserved and do a 32-bit access without a RMW. > Trying to abstract this at the register read/write level is going to > be fragile > > Rob >
Hi Rob, On 5/28/2020 3:36 AM, Kishon Vijay Abraham I wrote: > Hi Rob, > > On 5/27/2020 10:07 PM, Rob Herring wrote: >> On Wed, May 27, 2020 at 4:49 AM Kishon Vijay Abraham I <kishon@ti.com> wrote: >>> >>> Hi Rob, >>> >>> On 5/26/2020 8:42 PM, Rob Herring wrote: >>>> On Sun, May 24, 2020 at 9:30 PM Kishon Vijay Abraham I <kishon@ti.com> wrote: >>>>> >>>>> Hi Rob, >>>>> >>>>> On 5/22/2020 9:24 PM, Rob Herring wrote: >>>>>> On Thu, May 21, 2020 at 9:37 PM Kishon Vijay Abraham I <kishon@ti.com> wrote: >>>>>>> >>>>>>> Certain platforms like TI's J721E using Cadence PCIe IP can perform only >>>>>>> 32-bit accesses for reading or writing to Cadence registers. Convert all >>>>>>> read and write accesses to 32-bit in Cadence PCIe driver in preparation >>>>>>> for adding PCIe support in TI's J721E SoC. >>>>>> >>>>>> Looking more closely I don't think cdns_pcie_ep_assert_intx is okay >>>>>> with this and never can be given the PCI_COMMAND and PCI_STATUS >>>>>> registers are in the same word (IIRC, that's the main reason 32-bit >>>>>> config space accesses are broken). So this isn't going to work at >>>>> >>>>> right, PCI_STATUS has write '1' to clear bits and there's a chance that it >>>>> could be reset while raising legacy interrupt. While this cannot be avoided for >>>>> TI's J721E, other platforms doesn't have to have this limitation. >>>>>> least for EP accesses. And maybe you need a custom .raise_irq() hook >>>>>> to minimize any problems (such as making the RMW atomic at least from >>>>>> the endpoint's perspective). >>>>> >>>>> This is to make sure EP doesn't update in-consistent state when RC is updating >>>>> the PCI_STATUS register? Since this involves two different systems, how do we >>>>> make this atomic? >>>> >>>> You can't make it atomic WRT both systems, but is there locking around >>>> each RMW? Specifically, are preemption and interrupts disabled to >>>> ensure time between a read and write are minimized? You wouldn't want >>>> interrupts disabled during the delay too though (i.e. around >>>> .raise_irq()). >>> >>> Okay, I'll add spin spin_lock_irqsave() in cdns_pcie_write_sz(). As you also >>> pointed below that delay for legacy interrupt is wrong and it has to be fixed >>> (with a later series). >> >> But you don't need a lock everywhere. You need locks in the callers >> (and only sometimes). > > Okay, the locks should be added only for registers where HOST can also write to > the same register? Maybe only raise_irq then.. > >> >>> How do you want to handle cdns_pcie_ep_fn_writew() now? Because now we are >>> changing the default implementation to perform only 32-bit access (used for >>> legacy interrupt, msi-x interrupt and while writing standard headers) and it's >>> not okay only for legacy interrupts for platforms other than TI. >> >> Now I'm wondering how set_msi is not racy in the current code with the >> host setting/clearing PCI_MSI_FLAGS_ENABLE? Maybe that bit is RO from >> the EP side? > > set_msi/set_msix is a one time configuration that is invoked before the host > establishes the link with the endpoint. I don't think we have to consider this > as racy. Can we try to close on this discussion please? Thanks Kishon > > Thanks > Kishon > >> >> Ultimately I think you're going to have to provide your own endpoint >> functions or you need accessors for specific registers like >> PCI_MSI_FLAGS. Then for example, you just rely on the 2 bytes before >> PCI_MSI_FLAGS being reserved and do a 32-bit access without a RMW. >> Trying to abstract this at the register read/write level is going to >> be fragile >> >> Rob >>
Hi Rob, On 6/1/2020 6:46 AM, Kishon Vijay Abraham I wrote: > Hi Rob, > > On 5/28/2020 3:36 AM, Kishon Vijay Abraham I wrote: >> Hi Rob, >> >> On 5/27/2020 10:07 PM, Rob Herring wrote: >>> On Wed, May 27, 2020 at 4:49 AM Kishon Vijay Abraham I <kishon@ti.com> wrote: >>>> >>>> Hi Rob, >>>> >>>> On 5/26/2020 8:42 PM, Rob Herring wrote: >>>>> On Sun, May 24, 2020 at 9:30 PM Kishon Vijay Abraham I <kishon@ti.com> wrote: >>>>>> >>>>>> Hi Rob, >>>>>> >>>>>> On 5/22/2020 9:24 PM, Rob Herring wrote: >>>>>>> On Thu, May 21, 2020 at 9:37 PM Kishon Vijay Abraham I <kishon@ti.com> wrote: >>>>>>>> >>>>>>>> Certain platforms like TI's J721E using Cadence PCIe IP can perform only >>>>>>>> 32-bit accesses for reading or writing to Cadence registers. Convert all >>>>>>>> read and write accesses to 32-bit in Cadence PCIe driver in preparation >>>>>>>> for adding PCIe support in TI's J721E SoC. >>>>>>> >>>>>>> Looking more closely I don't think cdns_pcie_ep_assert_intx is okay >>>>>>> with this and never can be given the PCI_COMMAND and PCI_STATUS >>>>>>> registers are in the same word (IIRC, that's the main reason 32-bit >>>>>>> config space accesses are broken). So this isn't going to work at >>>>>> >>>>>> right, PCI_STATUS has write '1' to clear bits and there's a chance that it >>>>>> could be reset while raising legacy interrupt. While this cannot be avoided for >>>>>> TI's J721E, other platforms doesn't have to have this limitation. >>>>>>> least for EP accesses. And maybe you need a custom .raise_irq() hook >>>>>>> to minimize any problems (such as making the RMW atomic at least from >>>>>>> the endpoint's perspective). >>>>>> >>>>>> This is to make sure EP doesn't update in-consistent state when RC is updating >>>>>> the PCI_STATUS register? Since this involves two different systems, how do we >>>>>> make this atomic? >>>>> >>>>> You can't make it atomic WRT both systems, but is there locking around >>>>> each RMW? Specifically, are preemption and interrupts disabled to >>>>> ensure time between a read and write are minimized? You wouldn't want >>>>> interrupts disabled during the delay too though (i.e. around >>>>> .raise_irq()). >>>> >>>> Okay, I'll add spin spin_lock_irqsave() in cdns_pcie_write_sz(). As you also >>>> pointed below that delay for legacy interrupt is wrong and it has to be fixed >>>> (with a later series). >>> >>> But you don't need a lock everywhere. You need locks in the callers >>> (and only sometimes). >> >> Okay, the locks should be added only for registers where HOST can also write to >> the same register? Maybe only raise_irq then.. >> >>> >>>> How do you want to handle cdns_pcie_ep_fn_writew() now? Because now we are >>>> changing the default implementation to perform only 32-bit access (used for >>>> legacy interrupt, msi-x interrupt and while writing standard headers) and it's >>>> not okay only for legacy interrupts for platforms other than TI. >>> >>> Now I'm wondering how set_msi is not racy in the current code with the >>> host setting/clearing PCI_MSI_FLAGS_ENABLE? Maybe that bit is RO from >>> the EP side? >> >> set_msi/set_msix is a one time configuration that is invoked before the host >> establishes the link with the endpoint. I don't think we have to consider this >> as racy. > > Can we try to close on this discussion please? Should we just try to handle .raise_irq() separately for TI platform and all the other accesses remain as 32-bit access? Thanks Kishon > > Thanks > Kishon > >> >> Thanks >> Kishon >> >>> >>> Ultimately I think you're going to have to provide your own endpoint >>> functions or you need accessors for specific registers like >>> PCI_MSI_FLAGS. Then for example, you just rely on the 2 bytes before >>> PCI_MSI_FLAGS being reserved and do a 32-bit access without a RMW. >>> Trying to abstract this at the register read/write level is going to >>> be fragile >>> >>> Rob >>>