[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <4dd920dd-5e2c-2a00-17ae-61cf4d154ddd@opensource.wdc.com>
Date: Wed, 15 Feb 2023 10:20:42 +0900
From: Damien Le Moal <damien.lemoal@...nsource.wdc.com>
To: Rick Wertenbroek <rick.wertenbroek@...il.com>,
alberto.dassatti@...g-vd.ch
Cc: xxm@...k-chips.com, rick.wertenbroek@...g-vd.ch,
stable@...r.kernel.org, Rob Herring <robh+dt@...nel.org>,
Krzysztof Kozlowski <krzysztof.kozlowski+dt@...aro.org>,
Heiko Stuebner <heiko@...ech.de>,
Shawn Lin <shawn.lin@...k-chips.com>,
Lorenzo Pieralisi <lpieralisi@...nel.org>,
Krzysztof WilczyĆski <kw@...ux.com>,
Bjorn Helgaas <bhelgaas@...gle.com>,
Jani Nikula <jani.nikula@...el.com>,
Rodrigo Vivi <rodrigo.vivi@...el.com>,
Mikko Kovanen <mikko.kovanen@...amobile.com>,
Greg Kroah-Hartman <gregkh@...uxfoundation.org>,
devicetree@...r.kernel.org, linux-arm-kernel@...ts.infradead.org,
linux-rockchip@...ts.infradead.org, linux-kernel@...r.kernel.org,
linux-pci@...r.kernel.org
Subject: Re: [PATCH v2 6/9] PCI: rockchip: Fix window mapping and address
translation for endpoint
On 2/14/23 23:08, Rick Wertenbroek wrote:
> The RK3399 PCI endpoint core has 33 windows for PCIe space, now in the
> driver up to 32 fixed size (1M) windows are used and pages are allocated
> and mapped accordingly. The driver first used a single window and allocated
> space inside which caused translation issues (between CPU space and PCI
> space) because a window can only have a single translation at a given
> time, which if multiple pages are allocated inside will cause conflicts.
> Now each window is a single region of 1M which will always guarantee that
> the translation is not in conflict.
>
> Set the translation register addresses for physical function. As documented
> in the technical reference manual (TRM) section 17.5.5 "PCIe Address
> Translation" and section 17.6.8 "Address Translation Registers Description"
>
> Fixes: cf590b078391 ("PCI: rockchip: Add EP driver for Rockchip PCIe controller")
> Cc: stable@...r.kernel.org
> Signed-off-by: Rick Wertenbroek <rick.wertenbroek@...il.com>
> ---
> drivers/pci/controller/pcie-rockchip-ep.c | 67 ++++++++++++-----------
> drivers/pci/controller/pcie-rockchip.h | 25 +++++----
> 2 files changed, 49 insertions(+), 43 deletions(-)
>
> diff --git a/drivers/pci/controller/pcie-rockchip-ep.c b/drivers/pci/controller/pcie-rockchip-ep.c
> index 4c84e403e..cbc281a6a 100644
> --- a/drivers/pci/controller/pcie-rockchip-ep.c
> +++ b/drivers/pci/controller/pcie-rockchip-ep.c
> @@ -76,11 +76,17 @@ static void rockchip_pcie_prog_ep_ob_atu(struct rockchip_pcie *rockchip, u8 fn,
> if (num_pass_bits < 8)
> num_pass_bits = 8;
>
> - cpu_addr -= rockchip->mem_res->start;
> - addr0 = ((is_nor_msg ? 0x10 : (num_pass_bits - 1)) &
> - PCIE_CORE_OB_REGION_ADDR0_NUM_BITS) |
> - (lower_32_bits(cpu_addr) & PCIE_CORE_OB_REGION_ADDR0_LO_ADDR);
> - addr1 = upper_32_bits(is_nor_msg ? cpu_addr : pci_addr);
> + if (is_nor_msg) {
> + dev_warn(rockchip->dev, "NOR MSG\n");
I do not think this warning is needed.
In fact, if you move your patch 7 before this one, we could probably drop
the is_nor_msg == true case entirely since with your patch 7, only the
host driver uses AXI_WRAPPER_NOR_MSG. So warning and returning for that
case should be enough.
> + cpu_addr -= rockchip->mem_res->start;
This needs to be done for the !is_nor_msg case too.
> + addr0 = (0x10 & PCIE_CORE_OB_REGION_ADDR0_NUM_BITS) |
> + (lower_32_bits(cpu_addr) & PCIE_CORE_OB_REGION_ADDR0_LO_ADDR);
> + addr1 = upper_32_bits(cpu_addr);
> + } else {
> + addr0 = (num_pass_bits & PCIE_CORE_OB_REGION_ADDR0_NUM_BITS) |
> + (lower_32_bits(pci_addr) & PCIE_CORE_OB_REGION_ADDR0_LO_ADDR);
> + addr1 = upper_32_bits(pci_addr);
> + }
> desc0 = ROCKCHIP_PCIE_AT_OB_REGION_DESC0_DEVFN(fn) | type;
> desc1 = 0;
>
> @@ -103,12 +109,6 @@ static void rockchip_pcie_prog_ep_ob_atu(struct rockchip_pcie *rockchip, u8 fn,
> ROCKCHIP_PCIE_AT_OB_REGION_DESC0(r));
> rockchip_pcie_write(rockchip, desc1,
> ROCKCHIP_PCIE_AT_OB_REGION_DESC1(r));
> -
> - addr0 =
> - ((num_pass_bits - 1) & PCIE_CORE_OB_REGION_ADDR0_NUM_BITS) |
> - (lower_32_bits(cpu_addr) &
> - PCIE_CORE_OB_REGION_ADDR0_LO_ADDR);
> - addr1 = upper_32_bits(cpu_addr);
This hunk should have been removed in patch 1. But as commented, this is
needed, at least for me. Without setting the cpu addr to OB region cpu
addr register, mmio/dma does not work for me, despite the TRM saying that
these registers are unused.
> }
> }
>
> @@ -256,15 +256,7 @@ static int rockchip_pcie_ep_map_addr(struct pci_epc *epc, u8 fn, u8 vfn,
> struct rockchip_pcie *pcie = &ep->rockchip;
> u32 r;
>
> - r = find_first_zero_bit(&ep->ob_region_map, BITS_PER_LONG);
> - /*
> - * Region 0 is reserved for configuration space and shouldn't
> - * be used elsewhere per TRM, so leave it out.
> - */
> - if (r >= ep->max_regions - 1) {
> - dev_err(&epc->dev, "no free outbound region\n");
> - return -EINVAL;
> - }
> + r = (addr >> ilog2(SZ_1M)) & 0x1f;
Locally, I added a smal helper:
static inline int rockchip_ob_region(u64 addr)
{
return (addr >> ilog2(SZ_1M)) & 0x1f;
}
That makes the code nicer and avoids having this open coded repeatedly in
different places.
>
> rockchip_pcie_prog_ep_ob_atu(pcie, fn, r, AXI_WRAPPER_MEM_WRITE, addr,
> pci_addr, size);
> @@ -282,15 +274,11 @@ static void rockchip_pcie_ep_unmap_addr(struct pci_epc *epc, u8 fn, u8 vfn,
> struct rockchip_pcie *rockchip = &ep->rockchip;
> u32 r;
>
> - for (r = 0; r < ep->max_regions - 1; r++)
> + for (r = 0; r < ep->max_regions; r++)
> if (ep->ob_addr[r] == addr)
> break;
>
> - /*
> - * Region 0 is reserved for configuration space and shouldn't
> - * be used elsewhere per TRM, so leave it out.
> - */
> - if (r == ep->max_regions - 1)
> + if (r == ep->max_regions)
> return;
>
> rockchip_pcie_clear_ep_ob_atu(rockchip, r);
> @@ -411,6 +399,7 @@ static int rockchip_pcie_ep_send_msi_irq(struct rockchip_pcie_ep *ep, u8 fn,
> u16 flags, mme, data, data_mask;
> u8 msi_count;
> u64 pci_addr, pci_addr_mask = 0xff;
> + u32 r;
>
> /* Check MSI enable bit */
> flags = rockchip_pcie_read(&ep->rockchip,
> @@ -444,12 +433,12 @@ static int rockchip_pcie_ep_send_msi_irq(struct rockchip_pcie_ep *ep, u8 fn,
> ROCKCHIP_PCIE_EP_FUNC_BASE(fn) +
> ROCKCHIP_PCIE_EP_MSI_CTRL_REG +
> PCI_MSI_ADDRESS_LO);
> - pci_addr &= GENMASK_ULL(63, 2);
>
> /* Set the outbound region if needed. */
> if (unlikely(ep->irq_pci_addr != (pci_addr & ~pci_addr_mask) ||
> ep->irq_pci_fn != fn)) {
> - rockchip_pcie_prog_ep_ob_atu(rockchip, fn, ep->max_regions - 1,
> + r = (ep->irq_phys_addr >> ilog2(SZ_1M)) & 0x1f;
> + rockchip_pcie_prog_ep_ob_atu(rockchip, fn, r,
> AXI_WRAPPER_MEM_WRITE,
> ep->irq_phys_addr,
> pci_addr & ~pci_addr_mask,
> @@ -539,6 +528,8 @@ static int rockchip_pcie_parse_ep_dt(struct rockchip_pcie *rockchip,
> if (err < 0 || ep->max_regions > MAX_REGION_LIMIT)
> ep->max_regions = MAX_REGION_LIMIT;
>
> + ep->ob_region_map = 0;
> +
> err = of_property_read_u8(dev->of_node, "max-functions",
> &ep->epc->max_functions);
> if (err < 0)
> @@ -559,7 +550,10 @@ static int rockchip_pcie_ep_probe(struct platform_device *pdev)
> struct rockchip_pcie *rockchip;
> struct pci_epc *epc;
> size_t max_regions;
> + struct pci_epc_mem_window *windows = NULL;
> int err;
> + u32 cfg;
> + int i;
Nit: instead of declaring this with another line, you could declare it
together with "err" above.
>
> ep = devm_kzalloc(dev, sizeof(*ep), GFP_KERNEL);
> if (!ep)
> @@ -606,15 +600,26 @@ static int rockchip_pcie_ep_probe(struct platform_device *pdev)
> /* Only enable function 0 by default */
> rockchip_pcie_write(rockchip, BIT(0), PCIE_CORE_PHY_FUNC_CFG);
>
> - err = pci_epc_mem_init(epc, rockchip->mem_res->start,
> - resource_size(rockchip->mem_res), PAGE_SIZE);
> + windows = devm_kcalloc(dev, ep->max_regions, sizeof(struct pci_epc_mem_window), GFP_KERNEL);
> + if (!windows) {
> + err = -ENOMEM;
> + goto err_uninit_port;
> + }
> + for (i = 0; i < ep->max_regions; i++) {
> + windows[i].phys_base = rockchip->mem_res->start + (SZ_1M * i);
> + windows[i].size = SZ_1M;
> + windows[i].page_size = SZ_1M;
> + }
> + err = pci_epc_multi_mem_init(epc, windows, ep->max_regions);
> + devm_kfree(dev, windows);
> +
> if (err < 0) {
> dev_err(dev, "failed to initialize the memory space\n");
> goto err_uninit_port;
> }
>
> ep->irq_cpu_addr = pci_epc_mem_alloc_addr(epc, &ep->irq_phys_addr,
> - SZ_128K);
> + SZ_1M);
The region for this needs to be marked as allocated in the ob_region_map. So:
set_bit(rockchip_ob_region(ep->irq_phys_addr),
&ep->ob_region_map);
Of note though is that this ob_region_bitmap is used to set and clear bits
*only*, it is actually never checked at all to see if there is a bug and a
mapped region is being remapped without an unmap first. Not sure it is
very useful in the end.
> if (!ep->irq_cpu_addr) {
> dev_err(dev, "failed to reserve memory space for MSI\n");
> err = -ENOMEM;
> diff --git a/drivers/pci/controller/pcie-rockchip.h b/drivers/pci/controller/pcie-rockchip.h
> index f3a5ff1cf..72e427a0f 100644
> --- a/drivers/pci/controller/pcie-rockchip.h
> +++ b/drivers/pci/controller/pcie-rockchip.h
> @@ -134,6 +134,7 @@
>
> #define PCIE_RC_RP_ATS_BASE 0x400000
> #define PCIE_RC_CONFIG_NORMAL_BASE 0x800000
> +#define PCIE_EP_PF_CONFIG_REGS_BASE 0x800000
> #define PCIE_RC_CONFIG_BASE 0xa00000
> #define PCIE_EP_CONFIG_BASE 0xa00000
> #define PCIE_EP_CONFIG_DID_VID (PCIE_EP_CONFIG_BASE + 0x00)
> @@ -228,13 +229,14 @@
> #define ROCKCHIP_PCIE_EP_MSI_CTRL_MASK_MSI_CAP BIT(24)
> #define ROCKCHIP_PCIE_EP_DUMMY_IRQ_ADDR 0x1
> #define ROCKCHIP_PCIE_EP_PCI_LEGACY_IRQ_ADDR 0x3
> -#define ROCKCHIP_PCIE_EP_FUNC_BASE(fn) (((fn) << 12) & GENMASK(19, 12))
> +#define ROCKCHIP_PCIE_EP_FUNC_BASE(fn) \
> + (PCIE_EP_PF_CONFIG_REGS_BASE + (((fn) << 12) & GENMASK(19, 12)))
> +#define ROCKCHIP_PCIE_EP_VIRT_FUNC_BASE(fn) \
> + (PCIE_EP_PF_CONFIG_REGS_BASE + 0x10000 + (((fn) << 12) & GENMASK(19, 12)))
> #define ROCKCHIP_PCIE_AT_IB_EP_FUNC_BAR_ADDR0(fn, bar) \
> - (PCIE_RC_RP_ATS_BASE + 0x0840 + (fn) * 0x0040 + (bar) * 0x0008)
> + (PCIE_CORE_AXI_CONF_BASE + 0x0828 + (fn) * 0x0040 + (bar) * 0x0008)
> #define ROCKCHIP_PCIE_AT_IB_EP_FUNC_BAR_ADDR1(fn, bar) \
> - (PCIE_RC_RP_ATS_BASE + 0x0844 + (fn) * 0x0040 + (bar) * 0x0008)
> -#define ROCKCHIP_PCIE_AT_OB_REGION_PCI_ADDR0(r) \
> - (PCIE_RC_RP_ATS_BASE + 0x0000 + ((r) & 0x1f) * 0x0020)
> + (PCIE_CORE_AXI_CONF_BASE + 0x082c + (fn) * 0x0040 + (bar) * 0x0008)
> #define ROCKCHIP_PCIE_AT_OB_REGION_PCI_ADDR0_DEVFN_MASK GENMASK(19, 12)
> #define ROCKCHIP_PCIE_AT_OB_REGION_PCI_ADDR0_DEVFN(devfn) \
> (((devfn) << 12) & \
> @@ -242,20 +244,19 @@
> #define ROCKCHIP_PCIE_AT_OB_REGION_PCI_ADDR0_BUS_MASK GENMASK(27, 20)
> #define ROCKCHIP_PCIE_AT_OB_REGION_PCI_ADDR0_BUS(bus) \
> (((bus) << 20) & ROCKCHIP_PCIE_AT_OB_REGION_PCI_ADDR0_BUS_MASK)
> +#define PCIE_RC_EP_ATR_OB_REGIONS_1_32 (PCIE_CORE_AXI_CONF_BASE + 0x0020)
> +#define ROCKCHIP_PCIE_AT_OB_REGION_PCI_ADDR0(r) \
> + (PCIE_RC_EP_ATR_OB_REGIONS_1_32 + 0x0000 + ((r) & 0x1f) * 0x0020)
> #define ROCKCHIP_PCIE_AT_OB_REGION_PCI_ADDR1(r) \
> - (PCIE_RC_RP_ATS_BASE + 0x0004 + ((r) & 0x1f) * 0x0020)
> + (PCIE_RC_EP_ATR_OB_REGIONS_1_32 + 0x0004 + ((r) & 0x1f) * 0x0020)
> #define ROCKCHIP_PCIE_AT_OB_REGION_DESC0_HARDCODED_RID BIT(23)
> #define ROCKCHIP_PCIE_AT_OB_REGION_DESC0_DEVFN_MASK GENMASK(31, 24)
> #define ROCKCHIP_PCIE_AT_OB_REGION_DESC0_DEVFN(devfn) \
> (((devfn) << 24) & ROCKCHIP_PCIE_AT_OB_REGION_DESC0_DEVFN_MASK)
> #define ROCKCHIP_PCIE_AT_OB_REGION_DESC0(r) \
> - (PCIE_RC_RP_ATS_BASE + 0x0008 + ((r) & 0x1f) * 0x0020)
> + (PCIE_RC_EP_ATR_OB_REGIONS_1_32 + 0x0008 + ((r) & 0x1f) * 0x0020)
> #define ROCKCHIP_PCIE_AT_OB_REGION_DESC1(r) \
> - (PCIE_RC_RP_ATS_BASE + 0x000c + ((r) & 0x1f) * 0x0020)
> -#define ROCKCHIP_PCIE_AT_OB_REGION_CPU_ADDR0(r) \
> - (PCIE_RC_RP_ATS_BASE + 0x0018 + ((r) & 0x1f) * 0x0020)
> -#define ROCKCHIP_PCIE_AT_OB_REGION_CPU_ADDR1(r) \
> - (PCIE_RC_RP_ATS_BASE + 0x001c + ((r) & 0x1f) * 0x0020)
> + (PCIE_RC_EP_ATR_OB_REGIONS_1_32 + 0x000c + ((r) & 0x1f) * 0x0020)
>
> #define ROCKCHIP_PCIE_CORE_EP_FUNC_BAR_CFG0(fn) \
> (PCIE_CORE_CTRL_MGMT_BASE + 0x0240 + (fn) * 0x0008)
--
Damien Le Moal
Western Digital Research
Powered by blists - more mailing lists