[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Xen-devel] [PATCH v3] xen/tools: Add 64 bits big bar support



Hi, Ian and Keir

Do you have other comments for this patch?

Thanks,
-Xudong

> -----Original Message-----
> From: Hao, Xudong
> Sent: Wednesday, September 26, 2012 4:36 PM
> To: ian.campbell@xxxxxxxxxx; keir@xxxxxxx
> Cc: xen-devel@xxxxxxxxxxxxx; Stefano.Stabellini@xxxxxxxxxxxxx;
> jbeulich@xxxxxxxx; Zhang, Xiantao; Hao, Xudong
> Subject: [PATCH v3] xen/tools: Add 64 bits big bar support
> 
> Currently it is assumed PCI device BAR access < 4G memory. If there is such a
> device whose BAR size is larger than 4G, it must access > 4G memory address.
> This patch enable the 64bits big BAR support on hvmloader.
> 
> v3 changes from v2:
> - Remain original print information
> 
> v2 changes from v1 as comments by Jan.
> 1) Set Dynamic MMIO high memory address instead of a fixed number 640G
> 2) Mask bar_sz earlier to avoid older code changes
> 3) Add bar size barrier to judge high memory resource
> 4) Clean up bar64_relocate code
> 
> Signed-off-by: Xiantao Zhang <xiantao.zhang@xxxxxxxxx>
> Signed-off-by: Xudong Hao <xudong.hao@xxxxxxxxx>
> ---
>  tools/firmware/hvmloader/cacheattr.c |   25 +++++++----
>  tools/firmware/hvmloader/config.h    |    2 +
>  tools/firmware/hvmloader/pci.c       |   77
> ++++++++++++++++++++++++++-------
>  tools/firmware/hvmloader/util.h      |    1 +
>  4 files changed, 80 insertions(+), 25 deletions(-)
> 
> diff --git a/tools/firmware/hvmloader/cacheattr.c
> b/tools/firmware/hvmloader/cacheattr.c
> index 646f07f..592b455 100644
> --- a/tools/firmware/hvmloader/cacheattr.c
> +++ b/tools/firmware/hvmloader/cacheattr.c
> @@ -40,24 +40,33 @@
>  #define MSR_PAT              0x0277
>  #define MSR_MTRRdefType      0x02ff
> 
> +unsigned int cpu_phys_addr(void)
> +{
> +    uint32_t eax, ebx, ecx, edx;
> +    unsigned int phys_bits = 36;
> +    /* Find the physical address size for this CPU. */
> +    cpuid(0x80000000, &eax, &ebx, &ecx, &edx);
> +    if ( eax >= 0x80000008 )
> +    {
> +        cpuid(0x80000008, &eax, &ebx, &ecx, &edx);
> +        phys_bits = (uint8_t)eax;
> +    }
> +
> +    return phys_bits;
> +}
> +
>  void cacheattr_init(void)
>  {
>      uint32_t eax, ebx, ecx, edx;
>      uint64_t mtrr_cap, mtrr_def, content, addr_mask;
> -    unsigned int i, nr_var_ranges, phys_bits = 36;
> +    unsigned int i, nr_var_ranges, phys_bits;
> 
>      /* Does the CPU support architectural MTRRs? */
>      cpuid(0x00000001, &eax, &ebx, &ecx, &edx);
>      if ( !(edx & (1u << 12)) )
>           return;
> 
> -    /* Find the physical address size for this CPU. */
> -    cpuid(0x80000000, &eax, &ebx, &ecx, &edx);
> -    if ( eax >= 0x80000008 )
> -    {
> -        cpuid(0x80000008, &eax, &ebx, &ecx, &edx);
> -        phys_bits = (uint8_t)eax;
> -    }
> +    phys_bits = cpu_phys_addr();
> 
>      printf("%u-bit phys ... ", phys_bits);
> 
> diff --git a/tools/firmware/hvmloader/config.h
> b/tools/firmware/hvmloader/config.h
> index 7c0180d..bbf2993 100644
> --- a/tools/firmware/hvmloader/config.h
> +++ b/tools/firmware/hvmloader/config.h
> @@ -53,6 +53,8 @@ extern struct bios_config ovmf_config;
>  /* MMIO hole: Hardcoded defaults, which can be dynamically expanded. */
>  #define PCI_MEM_START       0xf0000000
>  #define PCI_MEM_END         0xfc000000
> +#define PCI_MIN_BIG_BAR_SIZE          0x20000000
> +
>  extern unsigned long pci_mem_start, pci_mem_end;
> 
> 
> diff --git a/tools/firmware/hvmloader/pci.c b/tools/firmware/hvmloader/pci.c
> index fd56e50..0500db5 100644
> --- a/tools/firmware/hvmloader/pci.c
> +++ b/tools/firmware/hvmloader/pci.c
> @@ -36,19 +36,25 @@ unsigned long igd_opregion_pgbase = 0;
> 
>  void pci_setup(void)
>  {
> -    uint32_t base, devfn, bar_reg, bar_data, bar_sz, cmd, mmio_total = 0;
> +    uint8_t is_64bar, using_64bar, bar64_relocate = 0;
> +    uint32_t devfn, bar_reg, cmd, bar_data, bar_data_upper;
> +    uint64_t base, bar_sz, bar_sz_upper, mmio_total = 0;
>      uint32_t vga_devfn = 256;
>      uint16_t class, vendor_id, device_id;
>      unsigned int bar, pin, link, isa_irq;
> +    int64_t mmio_left;
> 
>      /* Resources assignable to PCI devices via BARs. */
>      struct resource {
> -        uint32_t base, max;
> -    } *resource, mem_resource, io_resource;
> +        uint64_t base, max;
> +    } *resource, mem_resource, high_mem_resource, io_resource;
> 
>      /* Create a list of device BARs in descending order of size. */
>      struct bars {
> -        uint32_t devfn, bar_reg, bar_sz;
> +        uint32_t is_64bar;
> +        uint32_t devfn;
> +        uint32_t bar_reg;
> +        uint64_t bar_sz;
>      } *bars = (struct bars *)scratch_start;
>      unsigned int i, nr_bars = 0;
> 
> @@ -133,22 +139,34 @@ void pci_setup(void)
>          /* Map the I/O memory and port resources. */
>          for ( bar = 0; bar < 7; bar++ )
>          {
> +            bar_sz_upper = 0;
>              bar_reg = PCI_BASE_ADDRESS_0 + 4*bar;
>              if ( bar == 6 )
>                  bar_reg = PCI_ROM_ADDRESS;
> 
>              bar_data = pci_readl(devfn, bar_reg);
> +            is_64bar = !!((bar_data & (PCI_BASE_ADDRESS_SPACE |
> +                         PCI_BASE_ADDRESS_MEM_TYPE_MASK)) ==
> +                         (PCI_BASE_ADDRESS_SPACE_MEMORY |
> +                         PCI_BASE_ADDRESS_MEM_TYPE_64));
>              pci_writel(devfn, bar_reg, ~0);
>              bar_sz = pci_readl(devfn, bar_reg);
>              pci_writel(devfn, bar_reg, bar_data);
> -            if ( bar_sz == 0 )
> -                continue;
> 
>              bar_sz &= (((bar_data & PCI_BASE_ADDRESS_SPACE) ==
>                          PCI_BASE_ADDRESS_SPACE_MEMORY) ?
>                         PCI_BASE_ADDRESS_MEM_MASK :
>                         (PCI_BASE_ADDRESS_IO_MASK & 0xffff));
> +            if (is_64bar) {
> +                bar_data_upper = pci_readl(devfn, bar_reg + 4);
> +                pci_writel(devfn, bar_reg + 4, ~0);
> +                bar_sz_upper = pci_readl(devfn, bar_reg + 4);
> +                pci_writel(devfn, bar_reg + 4, bar_data_upper);
> +                bar_sz = (bar_sz_upper << 32) | bar_sz;
> +            }
>              bar_sz &= ~(bar_sz - 1);
> +            if ( bar_sz == 0 )
> +                continue;
> 
>              for ( i = 0; i < nr_bars; i++ )
>                  if ( bars[i].bar_sz < bar_sz )
> @@ -157,6 +175,7 @@ void pci_setup(void)
>              if ( i != nr_bars )
>                  memmove(&bars[i+1], &bars[i], (nr_bars-i) *
> sizeof(*bars));
> 
> +            bars[i].is_64bar = is_64bar;
>              bars[i].devfn   = devfn;
>              bars[i].bar_reg = bar_reg;
>              bars[i].bar_sz  = bar_sz;
> @@ -167,11 +186,8 @@ void pci_setup(void)
> 
>              nr_bars++;
> 
> -            /* Skip the upper-half of the address for a 64-bit BAR. */
> -            if ( (bar_data & (PCI_BASE_ADDRESS_SPACE |
> -                              PCI_BASE_ADDRESS_MEM_TYPE_MASK))
> ==
> -                 (PCI_BASE_ADDRESS_SPACE_MEMORY |
> -                  PCI_BASE_ADDRESS_MEM_TYPE_64) )
> +            /*The upper half is already calculated, skip it! */
> +            if (is_64bar)
>                  bar++;
>          }
> 
> @@ -197,6 +213,9 @@ void pci_setup(void)
>              ((pci_mem_start << 1) != 0) )
>          pci_mem_start <<= 1;
> 
> +    if ( (pci_mem_start << 1) != 0 )
> +        bar64_relocate = 1;
> +
>      /* Relocate RAM that overlaps PCI space (in 64k-page chunks). */
>      while ( (pci_mem_start >> PAGE_SHIFT) < hvm_info->low_mem_pgend )
>      {
> @@ -218,11 +237,15 @@ void pci_setup(void)
>          hvm_info->high_mem_pgend += nr_pages;
>      }
> 
> +    high_mem_resource.base = ((uint64_t)hvm_info->high_mem_pgend) <<
> PAGE_SHIFT;
> +    high_mem_resource.max = 1ull << cpu_phys_addr();
>      mem_resource.base = pci_mem_start;
>      mem_resource.max = pci_mem_end;
>      io_resource.base = 0xc000;
>      io_resource.max = 0x10000;
> 
> +    mmio_left = pci_mem_end - pci_mem_start;
> +
>      /* Assign iomem and ioport resources in descending order of size. */
>      for ( i = 0; i < nr_bars; i++ )
>      {
> @@ -230,13 +253,29 @@ void pci_setup(void)
>          bar_reg = bars[i].bar_reg;
>          bar_sz  = bars[i].bar_sz;
> 
> +        using_64bar = bars[i].is_64bar && bar64_relocate && (mmio_left <
> bar_sz);
>          bar_data = pci_readl(devfn, bar_reg);
> 
>          if ( (bar_data & PCI_BASE_ADDRESS_SPACE) ==
>               PCI_BASE_ADDRESS_SPACE_MEMORY )
>          {
> -            resource = &mem_resource;
> -            bar_data &= ~PCI_BASE_ADDRESS_MEM_MASK;
> +            /* Mapping high memory if PCI deivce is 64 bits bar and the bar
> size
> +               is larger than 512M */
> +            if (using_64bar && (bar_sz > PCI_MIN_BIG_BAR_SIZE)) {
> +                if ( high_mem_resource.base & (bar_sz - 1) )
> +                    high_mem_resource.base =
> high_mem_resource.base -
> +                        (high_mem_resource.base & (bar_sz - 1)) +
> bar_sz;
> +                else
> +                    high_mem_resource.base =
> high_mem_resource.base -
> +                        (high_mem_resource.base & (bar_sz - 1));
> +                resource = &high_mem_resource;
> +                bar_data &= ~PCI_BASE_ADDRESS_MEM_MASK;
> +            }
> +            else {
> +                resource = &mem_resource;
> +                bar_data &= ~PCI_BASE_ADDRESS_MEM_MASK;
> +            }
> +            mmio_left -= bar_sz;
>          }
>          else
>          {
> @@ -244,13 +283,14 @@ void pci_setup(void)
>              bar_data &= ~PCI_BASE_ADDRESS_IO_MASK;
>          }
> 
> -        base = (resource->base + bar_sz - 1) & ~(bar_sz - 1);
> -        bar_data |= base;
> +        base = (resource->base  + bar_sz - 1) & ~(uint64_t)(bar_sz - 1);
> +        bar_data |= (uint32_t)base;
> +        bar_data_upper = (uint32_t)(base >> 32);
>          base += bar_sz;
> 
>          if ( (base < resource->base) || (base > resource->max) )
>          {
> -            printf("pci dev %02x:%x bar %02x size %08x: no space for "
> +            printf("pci dev %02x:%x bar %02x size %llx: no space for "
>                     "resource!\n", devfn>>3, devfn&7, bar_reg, bar_sz);
>              continue;
>          }
> @@ -258,8 +298,11 @@ void pci_setup(void)
>          resource->base = base;
> 
>          pci_writel(devfn, bar_reg, bar_data);
> -        printf("pci dev %02x:%x bar %02x size %08x: %08x\n",
> +        if (using_64bar)
> +            pci_writel(devfn, bar_reg + 4, bar_data_upper);
> +        printf("pci dev %02x:%x bar %02x size %llx: %08x\n",
>                 devfn>>3, devfn&7, bar_reg, bar_sz, bar_data);
> +
> 
>          /* Now enable the memory or I/O mapping. */
>          cmd = pci_readw(devfn, PCI_COMMAND);
> diff --git a/tools/firmware/hvmloader/util.h b/tools/firmware/hvmloader/util.h
> index 07a9d42..ff06071 100644
> --- a/tools/firmware/hvmloader/util.h
> +++ b/tools/firmware/hvmloader/util.h
> @@ -215,6 +215,7 @@ void pci_setup(void);
>  uint32_t rombios_highbios_setup(void);
> 
>  /* Miscellaneous. */
> +unsigned int cpu_phys_addr(void);
>  void cacheattr_init(void);
>  unsigned long create_mp_tables(void *table);
>  void hvm_write_smbios_tables(
> --
> 1.5.5


_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxx
http://lists.xen.org/xen-devel


 


Rackspace

Lists.xenproject.org is hosted with RackSpace, monitoring our
servers 24x7x365 and backed by RackSpace's Fanatical Support®.