Re: [Qemu-devel] [PATCH v2 3/4] arm: Add PCIe host bridge in virt machine

[Alexander Graf]

On 27.01.15 17:52, Peter Maydell wrote:
> On 21 January 2015 at 16:18, Alexander Graf <address@hidden> wrote:
>> Now that we have a working "generic" PCIe host bridge driver, we can plug
>> it into ARMs virt machine to always have PCIe available to normal ARM VMs.
> 
> "ARM's".
> 
>>
>> I've successfully managed to expose a Bochs VGA device, XHCI and an e1000
>> into an AArch64 VM with this and they all lived happily ever after.
>>
>> Signed-off-by: Alexander Graf <address@hidden>
>> Tested-by: Claudio Fontana <address@hidden>
>>
>> ---
>>
>> Linux 3.19 only supports the generic PCIe host bridge driver for 32bit ARM
>> systems. If you want to use it with AArch64 guests, please apply the 
>> following
>> patch or wait until upstream cleaned up the code properly:
>>
>>   http://csgraf.de/agraf/pci/pci-3.19.patch
>>
>> v1 -> v2:
>>
>>   - Add define for pci range types
>>   - Remove mmio_window_size
>>   - Use 4 PCI INTX IRQ lines
>> ---
>>  default-configs/arm-softmmu.mak |   2 +
>>  hw/arm/virt.c                   | 112 
>> ++++++++++++++++++++++++++++++++++++++--
>>  include/sysemu/device_tree.h    |   9 ++++
>>  3 files changed, 118 insertions(+), 5 deletions(-)
>>
>> diff --git a/default-configs/arm-softmmu.mak 
>> b/default-configs/arm-softmmu.mak
>> index f3513fa..7671ee2 100644
>> --- a/default-configs/arm-softmmu.mak
>> +++ b/default-configs/arm-softmmu.mak
>> @@ -82,6 +82,8 @@ CONFIG_ZYNQ=y
>>  CONFIG_VERSATILE_PCI=y
>>  CONFIG_VERSATILE_I2C=y
>>
>> +CONFIG_PCI_GENERIC=y
>> +
>>  CONFIG_SDHCI=y
>>  CONFIG_INTEGRATOR_DEBUG=y
>>
>> diff --git a/hw/arm/virt.c b/hw/arm/virt.c
>> index 2353440..a727b4e 100644
>> --- a/hw/arm/virt.c
>> +++ b/hw/arm/virt.c
>> @@ -42,6 +42,7 @@
>>  #include "exec/address-spaces.h"
>>  #include "qemu/bitops.h"
>>  #include "qemu/error-report.h"
>> +#include "hw/pci-host/gpex.h"
>>
>>  #define NUM_VIRTIO_TRANSPORTS 32
>>
>> @@ -69,6 +70,7 @@ enum {
>>      VIRT_MMIO,
>>      VIRT_RTC,
>>      VIRT_FW_CFG,
>> +    VIRT_PCIE,
>>  };
>>
>>  typedef struct MemMapEntry {
>> @@ -129,13 +131,14 @@ static const MemMapEntry a15memmap[] = {
>>      [VIRT_FW_CFG] =     { 0x09020000, 0x0000000a },
>>      [VIRT_MMIO] =       { 0x0a000000, 0x00000200 },
>>      /* ...repeating for a total of NUM_VIRTIO_TRANSPORTS, each of that size 
>> */
>> -    /* 0x10000000 .. 0x40000000 reserved for PCI */
>> +    [VIRT_PCIE] =       { 0x10000000, 0x30000000 },
> 
> Might be nice to have a comment about how this is split up between
> MMIO, IO and config space.

I'd prefer to keep the details of that in the allocation code so that we
don't potentially have 2 places that diverge eventually. But I'll
document the allocation code a bit nicer and add a small overview split
here.

> 
>>      [VIRT_MEM] =        { 0x40000000, 30ULL * 1024 * 1024 * 1024 },
>>  };
>>
>>  static const int a15irqmap[] = {
>>      [VIRT_UART] = 1,
>>      [VIRT_RTC] = 2,
>> +    [VIRT_PCIE] = 3, /* ... to 6 */
>>      [VIRT_MMIO] = 16, /* ...to 16 + NUM_VIRTIO_TRANSPORTS - 1 */
>>  };
>>
>> @@ -312,7 +315,7 @@ static void fdt_add_cpu_nodes(const VirtBoardInfo *vbi)
>>      }
>>  }
>>
>> -static void fdt_add_gic_node(const VirtBoardInfo *vbi)
>> +static uint32_t fdt_add_gic_node(const VirtBoardInfo *vbi)
>>  {
>>      uint32_t gic_phandle;
>>
>> @@ -331,9 +334,11 @@ static void fdt_add_gic_node(const VirtBoardInfo *vbi)
>>                                       2, vbi->memmap[VIRT_GIC_CPU].base,
>>                                       2, vbi->memmap[VIRT_GIC_CPU].size);
>>      qemu_fdt_setprop_cell(vbi->fdt, "/intc", "phandle", gic_phandle);
>> +
>> +    return gic_phandle;
>>  }
>>
>> -static void create_gic(const VirtBoardInfo *vbi, qemu_irq *pic)
>> +static uint32_t create_gic(const VirtBoardInfo *vbi, qemu_irq *pic)
>>  {
>>      /* We create a standalone GIC v2 */
>>      DeviceState *gicdev;
>> @@ -380,7 +385,7 @@ static void create_gic(const VirtBoardInfo *vbi, 
>> qemu_irq *pic)
>>          pic[i] = qdev_get_gpio_in(gicdev, i);
>>      }
>>
>> -    fdt_add_gic_node(vbi);
>> +    return fdt_add_gic_node(vbi);
>>  }
>>
>>  static void create_uart(const VirtBoardInfo *vbi, qemu_irq *pic)
>> @@ -556,6 +561,100 @@ static void create_fw_cfg(const VirtBoardInfo *vbi)
>>      g_free(nodename);
>>  }
>>
>> +static void create_pcie_irq_map(const VirtBoardInfo *vbi, uint32_t 
>> gic_phandle,
>> +                                int first_irq, const char *nodename)
>> +{
>> +    int devfn, pin;
>> +    uint32_t full_irq_map[4 * 4 * 8] = { 0 };
>> +    uint32_t *irq_map = full_irq_map;
>> +
>> +    for (devfn = 0; devfn <= 0x18; devfn+=0x8) {
>> +        for (pin = 0; pin < 4; pin++) {
>> +            int irq_type = GIC_FDT_IRQ_TYPE_SPI;
>> +            int irq_nr = first_irq + ((pin + PCI_SLOT(devfn)) % 
>> PCI_NUM_PINS);
>> +            int irq_level = GIC_FDT_IRQ_FLAGS_LEVEL_HI;
>> +            int i;
>> +
>> +            uint32_t map[] = {
>> +                devfn << 8, 0, 0,                           /* devfn */
>> +                pin + 1,                                    /* PCI pin */
>> +                gic_phandle, irq_type, irq_nr, irq_level }; /* GIC irq */
>> +
>> +            /* Convert map to big endian */
>> +            for (i = 0; i < 8; i++) {
>> +                irq_map[i] = cpu_to_be32(map[i]);
>> +            }
>> +            irq_map += 8;
>> +        }
>> +    }
>> +
>> +    qemu_fdt_setprop(vbi->fdt, nodename, "interrupt-map",
>> +                     full_irq_map, sizeof(full_irq_map));
>> +}
>> +
>> +static void create_pcie(const VirtBoardInfo *vbi, qemu_irq *pic,
>> +                        uint32_t gic_phandle)
>> +{
>> +    hwaddr base = vbi->memmap[VIRT_PCIE].base;
>> +    hwaddr size = vbi->memmap[VIRT_PCIE].size;
>> +    hwaddr size_ioport = 64 * 1024;
>> +    hwaddr size_ecam = PCIE_MMCFG_SIZE_MIN;
> 
> See remarks in previous patch about wanting more than the minimum
> ECAM space.
> 
>> +    hwaddr size_mmio = size - size_ecam - size_ioport;
>> +    hwaddr base_mmio = base;
>> +    hwaddr base_ioport = base_mmio + size_mmio;
>> +    hwaddr base_ecam = base_ioport + size_ioport;
> 
> PCIe spec says the ECAM window has to be aligned to its size
> (ie if it's 1MB in size it has to be 1MB aligned, if it's
> 256MB aligned it has to be 256MB aligned). I think this is
> dumb, but it's what the spec says...

No problem, fortunately we have a good number of nice macros for this ;)

> 
>> +    int irq = vbi->irqmap[VIRT_PCIE];
>> +    MemoryRegion *mmio_alias;
>> +    MemoryRegion *mmio_reg;
>> +    DeviceState *dev;
>> +    char *nodename;
>> +    int i;
>> +
>> +    dev = qdev_create(NULL, TYPE_GPEX_HOST);
>> +    qdev_init_nofail(dev);
>> +
>> +    sysbus_mmio_map(SYS_BUS_DEVICE(dev), 0, base_ecam);
>> +    sysbus_mmio_map(SYS_BUS_DEVICE(dev), 2, base_ioport);
>> +
>> +    /* Map the MMIO window at the same spot in bus and cpu layouts */
>> +    mmio_alias = g_new0(MemoryRegion, 1);
>> +    mmio_reg = sysbus_mmio_get_region(SYS_BUS_DEVICE(dev), 1);
>> +    memory_region_init_alias(mmio_alias, OBJECT(dev), "pcie-mmio",
>> +                             mmio_reg, base_mmio, size_mmio);
>> +    memory_region_add_subregion(get_system_memory(), base_mmio, mmio_alias);
> 
> The comment claims to be mapping the MMIO window twice (in the
> system memory space and in the PCI mmio address space) but the
> code only seems to be mapping something into system memory space?

The comment claims to map it at the same spot. It means the offset in
system memory is the same offset as the one in the mmio window that gets
exported by the PHB.

The PHB exports a UINT64_MAX MMIO region as id 1. This is basically a
full window into the PCI address space. What we do here is to map a 1:1
window between CPU address space and PCI address space.

> 
>> +
>> +    for (i = 0; i < 4; i++) {
>> +        sysbus_connect_irq(SYS_BUS_DEVICE(dev), i, pic[irq + i]);
>> +    }
>> +
>> +    nodename = g_strdup_printf("/address@hidden" PRIx64, base);
>> +    qemu_fdt_add_subnode(vbi->fdt, nodename);
>> +    qemu_fdt_setprop_string(vbi->fdt, nodename,
>> +                            "compatible", "pci-host-ecam-generic");
>> +    qemu_fdt_setprop_string(vbi->fdt, nodename, "device_type", "pci");
>> +    qemu_fdt_setprop_cell(vbi->fdt, nodename, "#address-cells", 3);
>> +    qemu_fdt_setprop_cell(vbi->fdt, nodename, "#size-cells", 2);
>> +    qemu_fdt_setprop_cells(vbi->fdt, nodename, "bus-range", 0, 1);
>> +
>> +    qemu_fdt_setprop_sized_cells(vbi->fdt, nodename, "reg",
>> +                                 2, base_ecam, 2, size_ecam);
>> +    qemu_fdt_setprop_sized_cells(vbi->fdt, nodename, "ranges",
>> +                                 1, FDT_PCI_RANGE_IOPORT, 2, 0,
>> +                                 2, base_ioport, 2, size_ioport,
>> +
>> +                                 1, FDT_PCI_RANGE_MMIO, 2, base_mmio,
>> +                                 2, base_mmio, 2, size_mmio);
>> +
>> +    qemu_fdt_setprop_cell(vbi->fdt, nodename, "#interrupt-cells", 1);
>> +    create_pcie_irq_map(vbi, gic_phandle, irq, nodename);
>> +
>> +    qemu_fdt_setprop_cells(vbi->fdt, nodename, "interrupt-map-mask",
>> +                           0x1800, 0, 0, /* devfn (PCI_SLOT(3)) */
>> +                           0x7           /* PCI irq */);
> 
> I think at least the interrupt-map-mask and maybe also the
> #interrupt-cells setprop calls should go inside the
> create_pcie_irq_map() function. In particular the interrupt-map
> and interrupt-map-mask are meaningless unless interpreted together
> so they should go in the same function.

Sure, works for me.

> 
>> +
>> +    g_free(nodename);
>> +}
>> +
>>  static void *machvirt_dtb(const struct arm_boot_info *binfo, int *fdt_size)
>>  {
>>      const VirtBoardInfo *board = (const VirtBoardInfo *)binfo;
>> @@ -573,6 +672,7 @@ static void machvirt_init(MachineState *machine)
>>      MemoryRegion *ram = g_new(MemoryRegion, 1);
>>      const char *cpu_model = machine->cpu_model;
>>      VirtBoardInfo *vbi;
>> +    uint32_t gic_phandle;
>>
>>      if (!cpu_model) {
>>          cpu_model = "cortex-a15";
>> @@ -634,12 +734,14 @@ static void machvirt_init(MachineState *machine)
>>
>>      create_flash(vbi);
>>
>> -    create_gic(vbi, pic);
>> +    gic_phandle = create_gic(vbi, pic);
>>
>>      create_uart(vbi, pic);
>>
>>      create_rtc(vbi, pic);
>>
>> +    create_pcie(vbi, pic, gic_phandle);
>> +
>>      /* Create mmio transports, so the user can create virtio backends
>>       * (which will be automatically plugged in to the transports). If
>>       * no backend is created the transport will just sit harmlessly idle.
>> diff --git a/include/sysemu/device_tree.h b/include/sysemu/device_tree.h
>> index 899f05c..979169b 100644
>> --- a/include/sysemu/device_tree.h
>> +++ b/include/sysemu/device_tree.h
>> @@ -110,4 +110,13 @@ int qemu_fdt_setprop_sized_cells_from_array(void *fdt,
>>                                                  qdt_tmp);                 \
>>      })
>>
>> +#define FDT_PCI_RANGE_RELOCATABLE          0x80000000
>> +#define FDT_PCI_RANGE_PREFETCHABLE         0x40000000
>> +#define FDT_PCI_RANGE_ALIASED              0x20000000
>> +#define FDT_PCI_RANGE_TYPE                 0x03000000
>> +#define FDT_PCI_RANGE_MMIO_64BIT           0x03000000
> 
> I had to go dig up the spec to figure out that it wasn't an error that
> these two had the same values. Calling the first _TYPE_MASK might help?

Probably :)


Alex