Re: [PATCH v4 1/3] PCI: altera: Add Stratix 10 PCIe support

From: Ley Foon Tan
Date: Tue Feb 26 2019 - 02:07:44 EST


On Mon, Feb 25, 2019 at 5:35 PM Ley Foon Tan <ley.foon.tan@xxxxxxxxx> wrote:
>
> On Tue, 2019-02-19 at 16:23 +0000, Lorenzo Pieralisi wrote:
> > On Thu, Feb 14, 2019 at 11:20:36PM +0800, Ley Foon Tan wrote:
> > >
> > > Add PCIe Root Port support for Stratix 10 device.
> > >
> > > Main differences:
> > Main differences with what ? We need to rewrite this commit log.
> Differences compare with Cyclone V and Arria 10 devices.
> I will rewrite this.
> >
> > >
> > > - HIP interface to access Root Port configuration register.
> > > - TLP programming flow:
> > > - One REG0 register
> > > - Don't need to check alignment
> > >
> > > Signed-off-by: Ley Foon Tan <ley.foon.tan@xxxxxxxxx>
> > > ---
> > > drivers/pci/controller/pcie-altera.c | 246
> > > ++++++++++++++++++++++++++++++----
> > > 1 files changed, 222 insertions(+), 24 deletions(-)
> > >
> > > diff --git a/drivers/pci/controller/pcie-altera.c
> > > b/drivers/pci/controller/pcie-altera.c
> > > index 7d05e51..76bb6a6 100644
> > > --- a/drivers/pci/controller/pcie-altera.c
> > > +++ b/drivers/pci/controller/pcie-altera.c
> > > @@ -11,6 +11,7 @@
> > > #include <linux/irqchip/chained_irq.h>
> > > #include <linux/init.h>
> > > #include <linux/of_address.h>
> > > +#include <linux/of_device.h>
> > > #include <linux/of_irq.h>
> > > #include <linux/of_pci.h>
> > > #include <linux/pci.h>
> > > @@ -37,7 +38,12 @@
> > > #define RP_LTSSM_MASK 0x1f
> > > #define LTSSM_L0 0xf
> > >
> > > -#define PCIE_CAP_OFFSET 0x80
> > > +#define S10_RP_TX_CNTRL 0x2004
> > > +#define S10_RP_RXCPL_REG 0x2008
> > > +#define S10_RP_RXCPL_STATUS 0x200C
> > > +#define S10_RP_CFG_ADDR(pcie, reg) \
> > > + (((pcie)->hip_base) + (reg) + (1 << 20))
> > > +
> > > /* TLP configuration type 0 and 1 */
> > > #define TLP_FMTTYPE_CFGRD0 0x04 /*
> > > Configuration Read Type 0 */
> > > #define TLP_FMTTYPE_CFGWR0 0x44 /*
> > > Configuration Write Type 0 */
> > > @@ -49,18 +55,19 @@
> > > #define RP_DEVFN 0
> > > #define TLP_REQ_ID(bus, devfn) (((bus) << 8) |
> > > (devfn))
> > > #define TLP_CFGRD_DW0(pcie, bus)
> > > \
> > > - ((((bus == pcie->root_bus_nr) ? TLP_FMTTYPE_CFGRD0
> > > \
> > > - : TLP_FMTTYPE_CFGRD1) << 24) |
> > > \
> > > - TLP_PAYLOAD_SIZE)
> > > + ((((bus == pcie->root_bus_nr) ? pcie->pcie_data->cfgrd0
> > > \
> > > + : pcie->pcie_data->cfgrd1) << 24)
> > > | \
> > > + TLP_PAYLOAD_SIZE)
> > > #define TLP_CFGWR_DW0(pcie, bus)
> > > \
> > > - ((((bus == pcie->root_bus_nr) ? TLP_FMTTYPE_CFGWR0
> > > \
> > > - : TLP_FMTTYPE_CFGWR1) << 24) |
> > > \
> > > - TLP_PAYLOAD_SIZE)
> > > + ((((bus == pcie->root_bus_nr) ? pcie->pcie_data->cfgwr0
> > > \
> > > + : pcie->pcie_data->cfgwr1) << 24)
> > > | \
> > > + TLP_PAYLOAD_SIZE)
> > > #define TLP_CFG_DW1(pcie, tag, be) \
> > > - (((TLP_REQ_ID(pcie->root_bus_nr, RP_DEVFN)) << 16) | (tag <<
> > > 8) | (be))
> > > + (((TLP_REQ_ID(pcie->root_bus_nr, RP_DEVFN)) << 16) | (tag
> > > << 8) | (be))
> > > #define TLP_CFG_DW2(bus, devfn, offset) \
> > > (((bus) << 24) | ((devfn) << 16) |
> > > (offset))
> > > #define TLP_COMP_STATUS(s) (((s) >> 13) & 7)
> > > +#define TLP_BYTE_COUNT(s) (((s) >> 0) & 0xfff)
> > > #define TLP_HDR_SIZE 3
> > > #define TLP_LOOP 500
> > >
> > > @@ -69,14 +76,43 @@
> > >
> > > #define DWORD_MASK 3
> > >
> > > +#define S10_TLP_FMTTYPE_CFGRD0 0x05
> > > +#define S10_TLP_FMTTYPE_CFGRD1 0x04
> > > +#define S10_TLP_FMTTYPE_CFGWR0 0x45
> > > +#define S10_TLP_FMTTYPE_CFGWR1 0x44
> > > +
> > > +enum altera_pcie_version {
> > > + ALTERA_PCIE_V1 = 0,
> > > + ALTERA_PCIE_V2,
> > > +};
> > > +
> > > struct altera_pcie {
> > > struct platform_device *pdev;
> > > - void __iomem *cra_base; /* DT Cra */
> > > + void __iomem *cra_base;
> > > + void __iomem *hip_base;
> > > int irq;
> > > u8 root_bus_nr;
> > > struct irq_domain *irq_domain;
> > > struct resource bus_range;
> > > struct list_head resources;
> > > + const struct altera_pcie_data *pcie_data;
> > > +};
> > > +
> > > +struct altera_pcie_data {
> > > + int (*tlp_read_pkt)(struct altera_pcie *pcie, u32 *value);
> > > + void (*tlp_write_pkt)(struct altera_pcie *pcie, u32
> > > *headers,
> > > + u32 data, bool align);
> > > + bool (*get_link_status)(struct altera_pcie *pcie);
> > > + int (*rp_read_cfg)(struct altera_pcie *pcie, int where,
> > > + int size, u32 *value);
> > > + int (*rp_write_cfg)(struct altera_pcie *pcie, u8 bus, int
> > > where,
> > > + int size, u32 value);
> > > + enum altera_pcie_version version;
> > > + u32 cap_offset; /* PCIe capability
> > > structure register offset */
> > This is a duplication of struct pci_dev.pcie_cap right ?
>
> This is register offset in PCIe Rootport IP to access PCIe Cap
> structure register of directly RP. This allow us to access PCIe cap
> struct register without pci_dev.
>
> >
> > >
> > > + u32 cfgrd0;
> > > + u32 cfgrd1;
> > > + u32 cfgwr0;
> > > + u32 cfgwr1;
> > > };
> > >
> > > struct tlp_rp_regpair_t {
> > > @@ -101,6 +137,15 @@ static bool altera_pcie_link_up(struct
> > > altera_pcie *pcie)
> > > return !!((cra_readl(pcie, RP_LTSSM) & RP_LTSSM_MASK) ==
> > > LTSSM_L0);
> > > }
> > >
> > > +static bool s10_altera_pcie_link_up(struct altera_pcie *pcie)
> > > +{
> > > + void __iomem *addr = S10_RP_CFG_ADDR(pcie,
> > > + pcie->pcie_data->cap_offset +
> > > + PCI_EXP_LNKSTA);
> > > +
> > > + return !!(readw(addr) & PCI_EXP_LNKSTA_DLLLA);
> > > +}
> > > +
> > > /*
> > > * Altera PCIe port uses BAR0 of RC's configuration space as the
> > > translation
> > > * from PCI bus to native BUS. Entire DDR region is mapped into
> > > PCIe space
> > > @@ -128,12 +173,18 @@ static void tlp_write_tx(struct altera_pcie
> > > *pcie,
> > > cra_writel(pcie, tlp_rp_regdata->ctrl, RP_TX_CNTRL);
> > > }
> > >
> > > +static void s10_tlp_write_tx(struct altera_pcie *pcie, u32 reg0,
> > > u32 ctrl)
> > > +{
> > > + cra_writel(pcie, reg0, RP_TX_REG0);
> > > + cra_writel(pcie, ctrl, S10_RP_TX_CNTRL);
> > > +}
> > > +
> > > static bool altera_pcie_valid_device(struct altera_pcie *pcie,
> > > struct pci_bus *bus, int dev)
> > > {
> > > /* If there is no link, then there is no device */
> > > if (bus->number != pcie->root_bus_nr) {
> > > - if (!altera_pcie_link_up(pcie))
> > > + if (!pcie->pcie_data->get_link_status(pcie))
> > > return false;
> > > }
> > >
> > > @@ -183,6 +234,46 @@ static int tlp_read_packet(struct altera_pcie
> > > *pcie, u32 *value)
> > > return PCIBIOS_DEVICE_NOT_FOUND;
> > > }
> > >
> > > +static int s10_tlp_read_packet(struct altera_pcie *pcie, u32
> > > *value)
> > > +{
> > > + int i;
> > > + u32 ctrl;
> > > + u32 comp_status;
> > > + u32 dw[4];
> > > + u32 count = 0;
> > > +
> > > + for (i = 0; i < TLP_LOOP; i++) {
> > > + ctrl = cra_readl(pcie, S10_RP_RXCPL_STATUS);
> > > + if (!(ctrl & RP_RXCPL_SOP))
> > > + continue;
> > > +
> > > + /* Read first DW */
> > > + dw[count++] = cra_readl(pcie, S10_RP_RXCPL_REG);
> > > +
> > > + /* Poll for EOP */
> > > + for (i = 0; i < TLP_LOOP; i++) {
> > > + ctrl = cra_readl(pcie,
> > > S10_RP_RXCPL_STATUS);
> > > + dw[count++] = cra_readl(pcie,
> > > S10_RP_RXCPL_REG);
> > I think you'd better add a check on count lest it can overflow dw[].
> Okay, will check that.
> >
> > >
> > > + if (ctrl & RP_RXCPL_EOP) {
> > > + comp_status =
> > > TLP_COMP_STATUS(dw[1]);
> > > + if (comp_status)
> > > + return
> > > PCIBIOS_DEVICE_NOT_FOUND;
> > > +
> > > + if (value &&
> > > + TLP_BYTE_COUNT(dw[1]) ==
> > > sizeof(u32) &&
> > > + count >= 3)
> > > + *value = dw[3];
> > > +
> > > + return PCIBIOS_SUCCESSFUL;
> > > + }
> > > + }
> > > +
> > > + udelay(5);
> > > + }
> > > +
> > > + return PCIBIOS_DEVICE_NOT_FOUND;
> > > +}
> > > +
> > > static void tlp_write_packet(struct altera_pcie *pcie, u32
> > > *headers,
> > > u32 data, bool align)
> > > {
> > > @@ -210,6 +301,15 @@ static void tlp_write_packet(struct
> > > altera_pcie *pcie, u32 *headers,
> > > tlp_write_tx(pcie, &tlp_rp_regdata);
> > > }
> > >
> > > +static void s10_tlp_write_packet(struct altera_pcie *pcie, u32
> > > *headers,
> > > + u32 data, bool dummy)
> > > +{
> > > + s10_tlp_write_tx(pcie, headers[0], RP_TX_SOP);
> > > + s10_tlp_write_tx(pcie, headers[1], 0);
> > > + s10_tlp_write_tx(pcie, headers[2], 0);
> > > + s10_tlp_write_tx(pcie, data, RP_TX_EOP);
> > > +}
> > > +
> > > static int tlp_cfg_dword_read(struct altera_pcie *pcie, u8 bus,
> > > u32 devfn,
> > > int where, u8 byte_en, u32 *value)
> > > {
> > > @@ -219,9 +319,9 @@ static int tlp_cfg_dword_read(struct
> > > altera_pcie *pcie, u8 bus, u32 devfn,
> > > headers[1] = TLP_CFG_DW1(pcie, TLP_READ_TAG, byte_en);
> > > headers[2] = TLP_CFG_DW2(bus, devfn, where);
> > >
> > > - tlp_write_packet(pcie, headers, 0, false);
> > > + pcie->pcie_data->tlp_write_pkt(pcie, headers, 0, false);
> > >
> > > - return tlp_read_packet(pcie, value);
> > > + return pcie->pcie_data->tlp_read_pkt(pcie, value);
> > > }
> > >
> > > static int tlp_cfg_dword_write(struct altera_pcie *pcie, u8 bus,
> > > u32 devfn,
> > > @@ -236,11 +336,11 @@ static int tlp_cfg_dword_write(struct
> > > altera_pcie *pcie, u8 bus, u32 devfn,
> > >
> > > /* check alignment to Qword */
> > > if ((where & 0x7) == 0)
> > > - tlp_write_packet(pcie, headers, value, true);
> > > + pcie->pcie_data->tlp_write_pkt(pcie, headers,
> > > value, true);
> > > else
> > > - tlp_write_packet(pcie, headers, value, false);
> > > + pcie->pcie_data->tlp_write_pkt(pcie, headers,
> > > value, false);
> > >
> > > - ret = tlp_read_packet(pcie, NULL);
> > > + ret = pcie->pcie_data->tlp_read_pkt(pcie, NULL);
> > > if (ret != PCIBIOS_SUCCESSFUL)
> > > return ret;
> > >
> > > @@ -254,6 +354,53 @@ static int tlp_cfg_dword_write(struct
> > > altera_pcie *pcie, u8 bus, u32 devfn,
> > > return PCIBIOS_SUCCESSFUL;
> > > }
> > >
> > > +static int s10_rp_read_cfg(struct altera_pcie *pcie, int where,
> > > + int size, u32 *value)
> > > +{
> > > + void *addr = S10_RP_CFG_ADDR(pcie, where);
> > > +
> > > + switch (size) {
> > > + case 1:
> > > + *value = readb(addr);
> > > + break;
> > > + case 2:
> > > + *value = readw(addr);
> > > + break;
> > > + default:
> > > + *value = readl(addr);
> > > + break;
> > > + }
> > This boilerplace could be avoided if you used generic config
> > accessors (ie pci_generic_config_read()/write()).
> Okay. I will change this.
I just remembered that, we can't use pci_generic_config_read/write()
here because we need access to
Root port PCIe capabilities struct register to retrain PCIe link
before struct pci_bus *bus is available.
pci_generic_config_read/write() function needs struct pci_bus *bus.


Regards
Ley Foon