[v5,04/12] PCI: brcmstb: add dma-range mapping for inbound traffic

Message ID 1537367527-20773-5-git-send-email-jim2101024@gmail.com
State New
Delegated to: Lorenzo Pieralisi
Headers show
Series
  • PCI: brcmstb: Add Broadcom Settopbox PCIe support (resend)
Related show

Commit Message

Jim Quinlan Sept. 19, 2018, 2:31 p.m.
The Broadcom STB PCIe host controller is intimately related to the
memory subsystem.  This close relationship adds complexity to how cpu
system memory is mapped to PCIe memory.  Ideally, this mapping is an
identity mapping, or an identity mapping off by a constant.  Not so in
this case.

Consider the Broadcom reference board BCM97445LCC_4X8 which has 6 GB
of system memory.  Here is how the PCIe controller maps the
system memory to PCIe memory:

  memc0-a@[        0....3fffffff] <=> pci@[        0....3fffffff]
  memc0-b@[100000000...13fffffff] <=> pci@[ 40000000....7fffffff]
  memc1-a@[ 40000000....7fffffff] <=> pci@[ 80000000....bfffffff]
  memc1-b@[300000000...33fffffff] <=> pci@[ c0000000....ffffffff]
  memc2-a@[ 80000000....bfffffff] <=> pci@[100000000...13fffffff]
  memc2-b@[c00000000...c3fffffff] <=> pci@[140000000...17fffffff]

Although there are some "gaps" that can be added between the
individual mappings by software, the permutation of memory regions for
the most part is fixed by HW.  The solution of having something close
to an identity mapping is not possible.

The idea behind this HW design is that the same PCIe module can
act as an RC or EP, and if it acts as an EP it concatenates all
of system memory into a BAR so anything can be accessed.  Unfortunately,
when the PCIe block is in the role of an RC it also presents this
"BAR" to downstream PCIe devices, rather than offering an identity map
between its system memory and PCIe space.

Suppose that an endpoint driver allocs some DMA memory.  Suppose this
memory is located at 0x6000_0000, which is in the middle of memc1-a.
The driver wants a dma_addr_t value that it can pass on to the EP to
use.  Without doing any custom mapping, the EP will use this value for
DMA: the driver will get a dma_addr_t equal to 0x6000_0000.  But this
won't work; the device needs a dma_addr_t that reflects the PCIe space
address, namely 0xa000_0000.

So, essentially the solution to this problem must modify the
dma_addr_t returned by the DMA routines routines.  The method to do
this is to redefine the __dma_to_phys() and __phys_to_dma() functions
of the ARM, ARM64, and MIPS architectures.  This commit sets up the
infrastructure in the Brcm PCIe controller to prepare for this, while
there is three other subsequent commits to implement/redefine these
two functions for the three target architectures.

Signed-off-by: Jim Quinlan <jim2101024@gmail.com>
---
 drivers/pci/controller/pcie-brcmstb.c | 130 ++++++++++++++++++++++++++++++----
 include/soc/brcmstb/common.h          |  16 +++++
 2 files changed, 133 insertions(+), 13 deletions(-)

Comments

Ard Biesheuvel Sept. 20, 2018, 2:19 a.m. | #1
On 19 September 2018 at 07:31, Jim Quinlan <jim2101024@gmail.com> wrote:
> The Broadcom STB PCIe host controller is intimately related to the
> memory subsystem.  This close relationship adds complexity to how cpu
> system memory is mapped to PCIe memory.  Ideally, this mapping is an
> identity mapping, or an identity mapping off by a constant.  Not so in
> this case.
>
> Consider the Broadcom reference board BCM97445LCC_4X8 which has 6 GB
> of system memory.  Here is how the PCIe controller maps the
> system memory to PCIe memory:
>
>   memc0-a@[        0....3fffffff] <=> pci@[        0....3fffffff]
>   memc0-b@[100000000...13fffffff] <=> pci@[ 40000000....7fffffff]
>   memc1-a@[ 40000000....7fffffff] <=> pci@[ 80000000....bfffffff]
>   memc1-b@[300000000...33fffffff] <=> pci@[ c0000000....ffffffff]
>   memc2-a@[ 80000000....bfffffff] <=> pci@[100000000...13fffffff]
>   memc2-b@[c00000000...c3fffffff] <=> pci@[140000000...17fffffff]
>

So is describing this as

dma-ranges = <0x0 0x0 0x0 0x0 0x0 0x40000000>,
             <0x0 0x40000000 0x1 0x0 0x0 0x40000000>,
             <0x0 0x80000000 0x0 0x40000000 0x0 0x40000000>,
             <0x0 0xc0000000 0x3 0x0 0x0 0x40000000>,
             <0x1 0x0 0x0 0x80000000 0x0 0x40000000>,
             <0x1 0x40000000 0x0 0xc0000000 0x0 0x40000000>;

not working for you? I haven't tried this myself, but since DT permits
describing the inbound mappings this way, we should fix the code if it
doesn't work at the moment.


> Although there are some "gaps" that can be added between the
> individual mappings by software, the permutation of memory regions for
> the most part is fixed by HW.  The solution of having something close
> to an identity mapping is not possible.
>
> The idea behind this HW design is that the same PCIe module can
> act as an RC or EP, and if it acts as an EP it concatenates all
> of system memory into a BAR so anything can be accessed.  Unfortunately,
> when the PCIe block is in the role of an RC it also presents this
> "BAR" to downstream PCIe devices, rather than offering an identity map
> between its system memory and PCIe space.
>
> Suppose that an endpoint driver allocs some DMA memory.  Suppose this
> memory is located at 0x6000_0000, which is in the middle of memc1-a.
> The driver wants a dma_addr_t value that it can pass on to the EP to
> use.  Without doing any custom mapping, the EP will use this value for
> DMA: the driver will get a dma_addr_t equal to 0x6000_0000.  But this
> won't work; the device needs a dma_addr_t that reflects the PCIe space
> address, namely 0xa000_0000.
>
> So, essentially the solution to this problem must modify the
> dma_addr_t returned by the DMA routines routines.  The method to do
> this is to redefine the __dma_to_phys() and __phys_to_dma() functions
> of the ARM, ARM64, and MIPS architectures.  This commit sets up the
> infrastructure in the Brcm PCIe controller to prepare for this, while
> there is three other subsequent commits to implement/redefine these
> two functions for the three target architectures.
>
> Signed-off-by: Jim Quinlan <jim2101024@gmail.com>
> ---
>  drivers/pci/controller/pcie-brcmstb.c | 130 ++++++++++++++++++++++++++++++----
>  include/soc/brcmstb/common.h          |  16 +++++
>  2 files changed, 133 insertions(+), 13 deletions(-)
>
> diff --git a/drivers/pci/controller/pcie-brcmstb.c b/drivers/pci/controller/pcie-brcmstb.c
> index 9c87d10..abfa429 100644
> --- a/drivers/pci/controller/pcie-brcmstb.c
> +++ b/drivers/pci/controller/pcie-brcmstb.c
> @@ -21,6 +21,7 @@
>  #include <linux/printk.h>
>  #include <linux/sizes.h>
>  #include <linux/slab.h>
> +#include <soc/brcmstb/common.h>
>  #include <soc/brcmstb/memory_api.h>
>  #include <linux/string.h>
>  #include <linux/types.h>
> @@ -321,6 +322,7 @@ static void __iomem *brcm_pcie_map_conf(struct pci_bus *bus, unsigned int devfn,
>         (((val) & ~reg##_##field##_MASK) | \
>          (reg##_##field##_MASK & (field_val << reg##_##field##_SHIFT)))
>
> +static struct of_pci_range *brcm_dma_ranges;
>  static phys_addr_t scb_size[BRCM_MAX_SCB];
>  static int num_memc;
>  static int num_pcie;
> @@ -599,6 +601,79 @@ static inline void brcm_pcie_perst_set(struct brcm_pcie *pcie,
>                 WR_FLD_RB(pcie->base, PCIE_MISC_PCIE_CTRL, PCIE_PERSTB, !val);
>  }
>
> +static int brcm_pcie_parse_map_dma_ranges(struct brcm_pcie *pcie)
> +{
> +       int i;
> +       struct of_pci_range_parser parser;
> +       struct device_node *dn = pcie->dn;
> +
> +       /*
> +        * Parse dma-ranges property if present.  If there are multiple
> +        * PCIe controllers, we only have to parse from one of them since
> +        * the others will have an identical mapping.
> +        */
> +       if (!of_pci_dma_range_parser_init(&parser, dn)) {
> +               struct of_pci_range *p;
> +               unsigned int max_ranges = (parser.end - parser.range)
> +                       / parser.np;
> +
> +               /* Add a null entry to indicate the end of the array */
> +               brcm_dma_ranges = kcalloc(max_ranges + 1,
> +                                         sizeof(struct of_pci_range),
> +                                         GFP_KERNEL);
> +               if (!brcm_dma_ranges)
> +                       return -ENOMEM;
> +
> +               p = brcm_dma_ranges;
> +               while (of_pci_range_parser_one(&parser, p))
> +                       p++;
> +       }
> +
> +       for (i = 0, num_memc = 0; i < BRCM_MAX_SCB; i++) {
> +               u64 size = brcmstb_memory_memc_size(i);
> +
> +               if (size == (u64)-1) {
> +                       dev_err(pcie->dev, "cannot get memc%d size", i);
> +                       return -EINVAL;
> +               } else if (size) {
> +                       scb_size[i] = roundup_pow_of_two_64(size);
> +                       num_memc++;
> +               } else {
> +                       break;
> +               }
> +       }
> +
> +       return 0;
> +}
> +
> +dma_addr_t brcm_phys_to_dma(struct device *dev, phys_addr_t paddr)
> +{
> +       struct of_pci_range *p;
> +
> +       if (!dev || !dev_is_pci(dev))
> +               return (dma_addr_t)paddr;
> +       for (p = brcm_dma_ranges; p && p->size; p++)
> +               if (paddr >= p->cpu_addr && paddr < (p->cpu_addr + p->size))
> +                       return (dma_addr_t)(paddr - p->cpu_addr + p->pci_addr);
> +
> +       return (dma_addr_t)paddr;
> +}
> +
> +phys_addr_t brcm_dma_to_phys(struct device *dev, dma_addr_t dev_addr)
> +{
> +       struct of_pci_range *p;
> +
> +       if (!dev || !dev_is_pci(dev))
> +               return (phys_addr_t)dev_addr;
> +       for (p = brcm_dma_ranges; p && p->size; p++)
> +               if (dev_addr >= p->pci_addr
> +                   && dev_addr < (p->pci_addr + p->size))
> +                       return (phys_addr_t)
> +                               (dev_addr - p->pci_addr + p->cpu_addr);
> +
> +       return (phys_addr_t)dev_addr;
> +}
> +
>  static int brcm_pcie_add_controller(struct brcm_pcie *pcie)
>  {
>         int i, ret = 0;
> @@ -610,6 +685,10 @@ static int brcm_pcie_add_controller(struct brcm_pcie *pcie)
>                 goto done;
>         }
>
> +       ret = brcm_pcie_parse_map_dma_ranges(pcie);
> +       if (ret)
> +               goto done;
> +
>         /* Determine num_memc and their sizes */
>         for (i = 0, num_memc = 0; i < BRCM_MAX_SCB; i++) {
>                 u64 size = brcmstb_memory_memc_size(i);
> @@ -639,8 +718,13 @@ static int brcm_pcie_add_controller(struct brcm_pcie *pcie)
>  static void brcm_pcie_remove_controller(struct brcm_pcie *pcie)
>  {
>         mutex_lock(&brcm_pcie_lock);
> -       if (--num_pcie == 0)
> -               num_memc = 0;
> +       if (--num_pcie > 0)
> +               goto out;
> +
> +       kfree(brcm_dma_ranges);
> +       brcm_dma_ranges = NULL;
> +       num_memc = 0;
> +out:
>         mutex_unlock(&brcm_pcie_lock);
>  }
>
> @@ -747,11 +831,37 @@ static int brcm_pcie_setup(struct brcm_pcie *pcie)
>          */
>         rc_bar2_size = roundup_pow_of_two_64(total_mem_size);
>
> -       /*
> -        * Set simple configuration based on memory sizes
> -        * only.  We always start the viewport at address 0.
> -        */
> -       rc_bar2_offset = 0;
> +       if (brcm_dma_ranges) {
> +               /*
> +                * The best-case scenario is to place the inbound
> +                * region in the first 4GB of pcie-space, as some
> +                * legacy devices can only address 32bits.
> +                * We would also like to put the MSI under 4GB
> +                * as well, since some devices require a 32bit
> +                * MSI target address.
> +                */
> +               if (total_mem_size <= 0xc0000000ULL &&
> +                   rc_bar2_size <= 0x100000000ULL) {
> +                       rc_bar2_offset = 0;
> +               } else {
> +                       /*
> +                        * The system memory is 4GB or larger so we
> +                        * cannot start the inbound region at location
> +                        * 0 (since we have to allow some space for
> +                        * outbound memory @ 3GB).  So instead we
> +                        * start it at the 1x multiple of its size
> +                        */
> +                       rc_bar2_offset = rc_bar2_size;
> +               }
> +
> +       } else {
> +               /*
> +                * Set simple configuration based on memory sizes
> +                * only.  We always start the viewport at address 0,
> +                * and set the MSI target address accordingly.
> +                */
> +               rc_bar2_offset = 0;
> +       }
>
>         tmp = lower_32_bits(rc_bar2_offset);
>         tmp = INSERT_FIELD(tmp, PCIE_MISC_RC_BAR2_CONFIG_LO, SIZE,
> @@ -969,7 +1079,6 @@ static int brcm_pcie_probe(struct platform_device *pdev)
>         struct brcm_pcie *pcie;
>         struct resource *res;
>         void __iomem *base;
> -       u32 tmp;
>         struct pci_host_bridge *bridge;
>         struct pci_bus *child;
>
> @@ -986,11 +1095,6 @@ static int brcm_pcie_probe(struct platform_device *pdev)
>                 return -EINVAL;
>         }
>
> -       if (of_property_read_u32(dn, "dma-ranges", &tmp) == 0) {
> -               dev_err(&pdev->dev, "cannot yet handle dma-ranges\n");
> -               return -EINVAL;
> -       }
> -
>         data = of_id->data;
>         pcie->reg_offsets = data->offsets;
>         pcie->reg_field_info = data->reg_field_info;
> diff --git a/include/soc/brcmstb/common.h b/include/soc/brcmstb/common.h
> index cfb5335..a7f19e0 100644
> --- a/include/soc/brcmstb/common.h
> +++ b/include/soc/brcmstb/common.h
> @@ -12,4 +12,20 @@
>
>  bool soc_is_brcmstb(void);
>
> +#if defined(CONFIG_PCIE_BRCMSTB)
> +dma_addr_t brcm_phys_to_dma(struct device *dev, phys_addr_t paddr);
> +phys_addr_t brcm_dma_to_phys(struct device *dev, dma_addr_t dev_addr);
> +#else
> +static inline dma_addr_t brcm_phys_to_dma(struct device *dev, phys_addr_t paddr)
> +{
> +       return (dma_addr_t)paddr;
> +}
> +
> +static inline phys_addr_t brcm_dma_to_phys(struct device *dev,
> +                                          dma_addr_t dev_addr)
> +{
> +       return (phys_addr_t)dev_addr;
> +}
> +#endif
> +
>  #endif /* __SOC_BRCMSTB_COMMON_H__ */
> --
> 1.9.0.138.g2de3478
>
>
> _______________________________________________
> linux-arm-kernel mailing list
> linux-arm-kernel@lists.infradead.org
> http://lists.infradead.org/mailman/listinfo/linux-arm-kernel
Florian Fainelli Sept. 20, 2018, 8:55 p.m. | #2
On 09/19/2018 07:19 PM, Ard Biesheuvel wrote:
> On 19 September 2018 at 07:31, Jim Quinlan <jim2101024@gmail.com> wrote:
>> The Broadcom STB PCIe host controller is intimately related to the
>> memory subsystem.  This close relationship adds complexity to how cpu
>> system memory is mapped to PCIe memory.  Ideally, this mapping is an
>> identity mapping, or an identity mapping off by a constant.  Not so in
>> this case.
>>
>> Consider the Broadcom reference board BCM97445LCC_4X8 which has 6 GB
>> of system memory.  Here is how the PCIe controller maps the
>> system memory to PCIe memory:
>>
>>   memc0-a@[        0....3fffffff] <=> pci@[        0....3fffffff]
>>   memc0-b@[100000000...13fffffff] <=> pci@[ 40000000....7fffffff]
>>   memc1-a@[ 40000000....7fffffff] <=> pci@[ 80000000....bfffffff]
>>   memc1-b@[300000000...33fffffff] <=> pci@[ c0000000....ffffffff]
>>   memc2-a@[ 80000000....bfffffff] <=> pci@[100000000...13fffffff]
>>   memc2-b@[c00000000...c3fffffff] <=> pci@[140000000...17fffffff]
>>
> 
> So is describing this as
> 
> dma-ranges = <0x0 0x0 0x0 0x0 0x0 0x40000000>,
>              <0x0 0x40000000 0x1 0x0 0x0 0x40000000>,
>              <0x0 0x80000000 0x0 0x40000000 0x0 0x40000000>,
>              <0x0 0xc0000000 0x3 0x0 0x0 0x40000000>,
>              <0x1 0x0 0x0 0x80000000 0x0 0x40000000>,
>              <0x1 0x40000000 0x0 0xc0000000 0x0 0x40000000>;
> 
> not working for you? I haven't tried this myself, but since DT permits
> describing the inbound mappings this way, we should fix the code if it
> doesn't work at the moment.

You mean encoding the memory controller index in the first cell? If that
works, that's indeed a much cleaner solution, though is it standard
compliant in any form?
Ard Biesheuvel Sept. 20, 2018, 9:04 p.m. | #3
On 20 September 2018 at 13:55, Florian Fainelli <f.fainelli@gmail.com> wrote:
> On 09/19/2018 07:19 PM, Ard Biesheuvel wrote:
>> On 19 September 2018 at 07:31, Jim Quinlan <jim2101024@gmail.com> wrote:
>>> The Broadcom STB PCIe host controller is intimately related to the
>>> memory subsystem.  This close relationship adds complexity to how cpu
>>> system memory is mapped to PCIe memory.  Ideally, this mapping is an
>>> identity mapping, or an identity mapping off by a constant.  Not so in
>>> this case.
>>>
>>> Consider the Broadcom reference board BCM97445LCC_4X8 which has 6 GB
>>> of system memory.  Here is how the PCIe controller maps the
>>> system memory to PCIe memory:
>>>
>>>   memc0-a@[        0....3fffffff] <=> pci@[        0....3fffffff]
>>>   memc0-b@[100000000...13fffffff] <=> pci@[ 40000000....7fffffff]
>>>   memc1-a@[ 40000000....7fffffff] <=> pci@[ 80000000....bfffffff]
>>>   memc1-b@[300000000...33fffffff] <=> pci@[ c0000000....ffffffff]
>>>   memc2-a@[ 80000000....bfffffff] <=> pci@[100000000...13fffffff]
>>>   memc2-b@[c00000000...c3fffffff] <=> pci@[140000000...17fffffff]
>>>
>>
>> So is describing this as
>>
>> dma-ranges = <0x0 0x0 0x0 0x0 0x0 0x40000000>,
>>              <0x0 0x40000000 0x1 0x0 0x0 0x40000000>,
>>              <0x0 0x80000000 0x0 0x40000000 0x0 0x40000000>,
>>              <0x0 0xc0000000 0x3 0x0 0x0 0x40000000>,
>>              <0x1 0x0 0x0 0x80000000 0x0 0x40000000>,
>>              <0x1 0x40000000 0x0 0xc0000000 0x0 0x40000000>;
>>
>> not working for you? I haven't tried this myself, but since DT permits
>> describing the inbound mappings this way, we should fix the code if it
>> doesn't work at the moment.
>
> You mean encoding the memory controller index in the first cell? If that
> works, that's indeed a much cleaner solution, though is it standard
> compliant in any form?

No those are just memory addresses (although I may have screwed up the
order). From Documentation/devicetree/booting-without-of.txt:

"""
Optional property:
- dma-ranges: <prop-encoded-array> encoded as arbitrary number of triplets of
        (child-bus-address, parent-bus-address, length). Each triplet specified
        describes a contiguous DMA address range.
"""
Florian Fainelli Sept. 20, 2018, 9:31 p.m. | #4
On 09/20/2018 02:04 PM, Ard Biesheuvel wrote:
> On 20 September 2018 at 13:55, Florian Fainelli <f.fainelli@gmail.com> wrote:
>> On 09/19/2018 07:19 PM, Ard Biesheuvel wrote:
>>> On 19 September 2018 at 07:31, Jim Quinlan <jim2101024@gmail.com> wrote:
>>>> The Broadcom STB PCIe host controller is intimately related to the
>>>> memory subsystem.  This close relationship adds complexity to how cpu
>>>> system memory is mapped to PCIe memory.  Ideally, this mapping is an
>>>> identity mapping, or an identity mapping off by a constant.  Not so in
>>>> this case.
>>>>
>>>> Consider the Broadcom reference board BCM97445LCC_4X8 which has 6 GB
>>>> of system memory.  Here is how the PCIe controller maps the
>>>> system memory to PCIe memory:
>>>>
>>>>   memc0-a@[        0....3fffffff] <=> pci@[        0....3fffffff]
>>>>   memc0-b@[100000000...13fffffff] <=> pci@[ 40000000....7fffffff]
>>>>   memc1-a@[ 40000000....7fffffff] <=> pci@[ 80000000....bfffffff]
>>>>   memc1-b@[300000000...33fffffff] <=> pci@[ c0000000....ffffffff]
>>>>   memc2-a@[ 80000000....bfffffff] <=> pci@[100000000...13fffffff]
>>>>   memc2-b@[c00000000...c3fffffff] <=> pci@[140000000...17fffffff]
>>>>
>>>
>>> So is describing this as
>>>
>>> dma-ranges = <0x0 0x0 0x0 0x0 0x0 0x40000000>,
>>>              <0x0 0x40000000 0x1 0x0 0x0 0x40000000>,
>>>              <0x0 0x80000000 0x0 0x40000000 0x0 0x40000000>,
>>>              <0x0 0xc0000000 0x3 0x0 0x0 0x40000000>,
>>>              <0x1 0x0 0x0 0x80000000 0x0 0x40000000>,
>>>              <0x1 0x40000000 0x0 0xc0000000 0x0 0x40000000>;
>>>
>>> not working for you? I haven't tried this myself, but since DT permits
>>> describing the inbound mappings this way, we should fix the code if it
>>> doesn't work at the moment.
>>
>> You mean encoding the memory controller index in the first cell? If that
>> works, that's indeed a much cleaner solution, though is it standard
>> compliant in any form?
> 
> No those are just memory addresses (although I may have screwed up the
> order). From Documentation/devicetree/booting-without-of.txt:
> 
> """
> Optional property:
> - dma-ranges: <prop-encoded-array> encoded as arbitrary number of triplets of
>         (child-bus-address, parent-bus-address, length). Each triplet specified
>         describes a contiguous DMA address range.
> """
> 

Then I am confused by your comment, that's what this patch does, it adds
support for reading "dma-ranges" from Device Tree and setting up inbound
windows using that. The only caveat is that because the PCIe root
complex has some ties with the memory bus architecture it is connected
to (SCB in our case) there is still a requirement to know the
translation between a given physical address and its backing memory
controller/aperture.
Ard Biesheuvel Sept. 20, 2018, 9:33 p.m. | #5
On 20 September 2018 at 14:31, Florian Fainelli <f.fainelli@gmail.com> wrote:
> On 09/20/2018 02:04 PM, Ard Biesheuvel wrote:
>> On 20 September 2018 at 13:55, Florian Fainelli <f.fainelli@gmail.com> wrote:
>>> On 09/19/2018 07:19 PM, Ard Biesheuvel wrote:
>>>> On 19 September 2018 at 07:31, Jim Quinlan <jim2101024@gmail.com> wrote:
>>>>> The Broadcom STB PCIe host controller is intimately related to the
>>>>> memory subsystem.  This close relationship adds complexity to how cpu
>>>>> system memory is mapped to PCIe memory.  Ideally, this mapping is an
>>>>> identity mapping, or an identity mapping off by a constant.  Not so in
>>>>> this case.
>>>>>
>>>>> Consider the Broadcom reference board BCM97445LCC_4X8 which has 6 GB
>>>>> of system memory.  Here is how the PCIe controller maps the
>>>>> system memory to PCIe memory:
>>>>>
>>>>>   memc0-a@[        0....3fffffff] <=> pci@[        0....3fffffff]
>>>>>   memc0-b@[100000000...13fffffff] <=> pci@[ 40000000....7fffffff]
>>>>>   memc1-a@[ 40000000....7fffffff] <=> pci@[ 80000000....bfffffff]
>>>>>   memc1-b@[300000000...33fffffff] <=> pci@[ c0000000....ffffffff]
>>>>>   memc2-a@[ 80000000....bfffffff] <=> pci@[100000000...13fffffff]
>>>>>   memc2-b@[c00000000...c3fffffff] <=> pci@[140000000...17fffffff]
>>>>>
>>>>
>>>> So is describing this as
>>>>
>>>> dma-ranges = <0x0 0x0 0x0 0x0 0x0 0x40000000>,
>>>>              <0x0 0x40000000 0x1 0x0 0x0 0x40000000>,
>>>>              <0x0 0x80000000 0x0 0x40000000 0x0 0x40000000>,
>>>>              <0x0 0xc0000000 0x3 0x0 0x0 0x40000000>,
>>>>              <0x1 0x0 0x0 0x80000000 0x0 0x40000000>,
>>>>              <0x1 0x40000000 0x0 0xc0000000 0x0 0x40000000>;
>>>>
>>>> not working for you? I haven't tried this myself, but since DT permits
>>>> describing the inbound mappings this way, we should fix the code if it
>>>> doesn't work at the moment.
>>>
>>> You mean encoding the memory controller index in the first cell? If that
>>> works, that's indeed a much cleaner solution, though is it standard
>>> compliant in any form?
>>
>> No those are just memory addresses (although I may have screwed up the
>> order). From Documentation/devicetree/booting-without-of.txt:
>>
>> """
>> Optional property:
>> - dma-ranges: <prop-encoded-array> encoded as arbitrary number of triplets of
>>         (child-bus-address, parent-bus-address, length). Each triplet specified
>>         describes a contiguous DMA address range.
>> """
>>
>
> Then I am confused by your comment, that's what this patch does, it adds
> support for reading "dma-ranges" from Device Tree and setting up inbound
> windows using that. The only caveat is that because the PCIe root
> complex has some ties with the memory bus architecture it is connected
> to (SCB in our case) there is still a requirement to know the
> translation between a given physical address and its backing memory
> controller/aperture.
>

Ah ok, apologies for the noise then.

I was hoping that having working support for dma-ranges would remove
the need for the special phys<->dma conversion routines.
Florian Fainelli Sept. 20, 2018, 9:39 p.m. | #6
On 09/20/2018 02:33 PM, Ard Biesheuvel wrote:
> On 20 September 2018 at 14:31, Florian Fainelli <f.fainelli@gmail.com> wrote:
>> On 09/20/2018 02:04 PM, Ard Biesheuvel wrote:
>>> On 20 September 2018 at 13:55, Florian Fainelli <f.fainelli@gmail.com> wrote:
>>>> On 09/19/2018 07:19 PM, Ard Biesheuvel wrote:
>>>>> On 19 September 2018 at 07:31, Jim Quinlan <jim2101024@gmail.com> wrote:
>>>>>> The Broadcom STB PCIe host controller is intimately related to the
>>>>>> memory subsystem.  This close relationship adds complexity to how cpu
>>>>>> system memory is mapped to PCIe memory.  Ideally, this mapping is an
>>>>>> identity mapping, or an identity mapping off by a constant.  Not so in
>>>>>> this case.
>>>>>>
>>>>>> Consider the Broadcom reference board BCM97445LCC_4X8 which has 6 GB
>>>>>> of system memory.  Here is how the PCIe controller maps the
>>>>>> system memory to PCIe memory:
>>>>>>
>>>>>>   memc0-a@[        0....3fffffff] <=> pci@[        0....3fffffff]
>>>>>>   memc0-b@[100000000...13fffffff] <=> pci@[ 40000000....7fffffff]
>>>>>>   memc1-a@[ 40000000....7fffffff] <=> pci@[ 80000000....bfffffff]
>>>>>>   memc1-b@[300000000...33fffffff] <=> pci@[ c0000000....ffffffff]
>>>>>>   memc2-a@[ 80000000....bfffffff] <=> pci@[100000000...13fffffff]
>>>>>>   memc2-b@[c00000000...c3fffffff] <=> pci@[140000000...17fffffff]
>>>>>>
>>>>>
>>>>> So is describing this as
>>>>>
>>>>> dma-ranges = <0x0 0x0 0x0 0x0 0x0 0x40000000>,
>>>>>              <0x0 0x40000000 0x1 0x0 0x0 0x40000000>,
>>>>>              <0x0 0x80000000 0x0 0x40000000 0x0 0x40000000>,
>>>>>              <0x0 0xc0000000 0x3 0x0 0x0 0x40000000>,
>>>>>              <0x1 0x0 0x0 0x80000000 0x0 0x40000000>,
>>>>>              <0x1 0x40000000 0x0 0xc0000000 0x0 0x40000000>;
>>>>>
>>>>> not working for you? I haven't tried this myself, but since DT permits
>>>>> describing the inbound mappings this way, we should fix the code if it
>>>>> doesn't work at the moment.
>>>>
>>>> You mean encoding the memory controller index in the first cell? If that
>>>> works, that's indeed a much cleaner solution, though is it standard
>>>> compliant in any form?
>>>
>>> No those are just memory addresses (although I may have screwed up the
>>> order). From Documentation/devicetree/booting-without-of.txt:
>>>
>>> """
>>> Optional property:
>>> - dma-ranges: <prop-encoded-array> encoded as arbitrary number of triplets of
>>>         (child-bus-address, parent-bus-address, length). Each triplet specified
>>>         describes a contiguous DMA address range.
>>> """
>>>
>>
>> Then I am confused by your comment, that's what this patch does, it adds
>> support for reading "dma-ranges" from Device Tree and setting up inbound
>> windows using that. The only caveat is that because the PCIe root
>> complex has some ties with the memory bus architecture it is connected
>> to (SCB in our case) there is still a requirement to know the
>> translation between a given physical address and its backing memory
>> controller/aperture.
>>
> 
> Ah ok, apologies for the noise then.
> 
> I was hoping that having working support for dma-ranges would remove
> the need for the special phys<->dma conversion routines.

What you describe definitively works with platform devices, but I am not
sure this is working for PCIe devices, although, conceptually it should,
yes.
Jim Quinlan Sept. 21, 2018, 5:40 p.m. | #7
On Thu, Sep 20, 2018 at 5:39 PM Florian Fainelli <f.fainelli@gmail.com> wrote:
>
> On 09/20/2018 02:33 PM, Ard Biesheuvel wrote:
> > On 20 September 2018 at 14:31, Florian Fainelli <f.fainelli@gmail.com> wrote:
> >> On 09/20/2018 02:04 PM, Ard Biesheuvel wrote:
> >>> On 20 September 2018 at 13:55, Florian Fainelli <f.fainelli@gmail.com> wrote:
> >>>> On 09/19/2018 07:19 PM, Ard Biesheuvel wrote:
> >>>>> On 19 September 2018 at 07:31, Jim Quinlan <jim2101024@gmail.com> wrote:
> >>>>>> The Broadcom STB PCIe host controller is intimately related to the
> >>>>>> memory subsystem.  This close relationship adds complexity to how cpu
> >>>>>> system memory is mapped to PCIe memory.  Ideally, this mapping is an
> >>>>>> identity mapping, or an identity mapping off by a constant.  Not so in
> >>>>>> this case.
> >>>>>>
> >>>>>> Consider the Broadcom reference board BCM97445LCC_4X8 which has 6 GB
> >>>>>> of system memory.  Here is how the PCIe controller maps the
> >>>>>> system memory to PCIe memory:
> >>>>>>
> >>>>>>   memc0-a@[        0....3fffffff] <=> pci@[        0....3fffffff]
> >>>>>>   memc0-b@[100000000...13fffffff] <=> pci@[ 40000000....7fffffff]
> >>>>>>   memc1-a@[ 40000000....7fffffff] <=> pci@[ 80000000....bfffffff]
> >>>>>>   memc1-b@[300000000...33fffffff] <=> pci@[ c0000000....ffffffff]
> >>>>>>   memc2-a@[ 80000000....bfffffff] <=> pci@[100000000...13fffffff]
> >>>>>>   memc2-b@[c00000000...c3fffffff] <=> pci@[140000000...17fffffff]
> >>>>>>
> >>>>>
> >>>>> So is describing this as
> >>>>>
> >>>>> dma-ranges = <0x0 0x0 0x0 0x0 0x0 0x40000000>,
> >>>>>              <0x0 0x40000000 0x1 0x0 0x0 0x40000000>,
> >>>>>              <0x0 0x80000000 0x0 0x40000000 0x0 0x40000000>,
> >>>>>              <0x0 0xc0000000 0x3 0x0 0x0 0x40000000>,
> >>>>>              <0x1 0x0 0x0 0x80000000 0x0 0x40000000>,
> >>>>>              <0x1 0x40000000 0x0 0xc0000000 0x0 0x40000000>;
> >>>>>
> >>>>> not working for you? I haven't tried this myself, but since DT permits
> >>>>> describing the inbound mappings this way, we should fix the code if it
> >>>>> doesn't work at the moment.
> >>>>
> >>>> You mean encoding the memory controller index in the first cell? If that
> >>>> works, that's indeed a much cleaner solution, though is it standard
> >>>> compliant in any form?
> >>>
> >>> No those are just memory addresses (although I may have screwed up the
> >>> order). From Documentation/devicetree/booting-without-of.txt:
> >>>
> >>> """
> >>> Optional property:
> >>> - dma-ranges: <prop-encoded-array> encoded as arbitrary number of triplets of
> >>>         (child-bus-address, parent-bus-address, length). Each triplet specified
> >>>         describes a contiguous DMA address range.
> >>> """
> >>>
> >>
> >> Then I am confused by your comment, that's what this patch does, it adds
> >> support for reading "dma-ranges" from Device Tree and setting up inbound
> >> windows using that. The only caveat is that because the PCIe root
> >> complex has some ties with the memory bus architecture it is connected
> >> to (SCB in our case) there is still a requirement to know the
> >> translation between a given physical address and its backing memory
> >> controller/aperture.
> >>
> >
> > Ah ok, apologies for the noise then.
> >
> > I was hoping that having working support for dma-ranges would remove
> > the need for the special phys<->dma conversion routines.
>
> What you describe definitively works with platform devices, but I am not
> sure this is working for PCIe devices, although, conceptually it should,
> yes.
Sorry for my delay in responding.  One problem is that
of_dma_configure() only looks at the first dma-range given and then
converts it to dev->dma_pfn_offset which is respected by the DMA API.
However, we often have multiple dma-ranges, not just one.  This is the
big issue.

There is another issue with of_dma_configure() being invoked by the EP
driver on "bridge->parent->of_node", which is our RC device,
Of_dma_configure() calls of_dma_range() on the of_get_next_parent() of
our RC's device node and this misses the dma-ranges property which is
contained within the RC.  I think I could workaround this but there is
no getting around the first problem.

Thanks,
Jim

> --
> Florian
Ard Biesheuvel Sept. 24, 2018, 8:25 a.m. | #8
On Fri, 21 Sep 2018 at 19:41, Jim Quinlan <jim2101024@gmail.com> wrote:
>
> On Thu, Sep 20, 2018 at 5:39 PM Florian Fainelli <f.fainelli@gmail.com> wrote:
> >
> > On 09/20/2018 02:33 PM, Ard Biesheuvel wrote:
> > > On 20 September 2018 at 14:31, Florian Fainelli <f.fainelli@gmail.com> wrote:
> > >> On 09/20/2018 02:04 PM, Ard Biesheuvel wrote:
> > >>> On 20 September 2018 at 13:55, Florian Fainelli <f.fainelli@gmail.com> wrote:
> > >>>> On 09/19/2018 07:19 PM, Ard Biesheuvel wrote:
> > >>>>> On 19 September 2018 at 07:31, Jim Quinlan <jim2101024@gmail.com> wrote:
> > >>>>>> The Broadcom STB PCIe host controller is intimately related to the
> > >>>>>> memory subsystem.  This close relationship adds complexity to how cpu
> > >>>>>> system memory is mapped to PCIe memory.  Ideally, this mapping is an
> > >>>>>> identity mapping, or an identity mapping off by a constant.  Not so in
> > >>>>>> this case.
> > >>>>>>
> > >>>>>> Consider the Broadcom reference board BCM97445LCC_4X8 which has 6 GB
> > >>>>>> of system memory.  Here is how the PCIe controller maps the
> > >>>>>> system memory to PCIe memory:
> > >>>>>>
> > >>>>>>   memc0-a@[        0....3fffffff] <=> pci@[        0....3fffffff]
> > >>>>>>   memc0-b@[100000000...13fffffff] <=> pci@[ 40000000....7fffffff]
> > >>>>>>   memc1-a@[ 40000000....7fffffff] <=> pci@[ 80000000....bfffffff]
> > >>>>>>   memc1-b@[300000000...33fffffff] <=> pci@[ c0000000....ffffffff]
> > >>>>>>   memc2-a@[ 80000000....bfffffff] <=> pci@[100000000...13fffffff]
> > >>>>>>   memc2-b@[c00000000...c3fffffff] <=> pci@[140000000...17fffffff]
> > >>>>>>
> > >>>>>
> > >>>>> So is describing this as
> > >>>>>
> > >>>>> dma-ranges = <0x0 0x0 0x0 0x0 0x0 0x40000000>,
> > >>>>>              <0x0 0x40000000 0x1 0x0 0x0 0x40000000>,
> > >>>>>              <0x0 0x80000000 0x0 0x40000000 0x0 0x40000000>,
> > >>>>>              <0x0 0xc0000000 0x3 0x0 0x0 0x40000000>,
> > >>>>>              <0x1 0x0 0x0 0x80000000 0x0 0x40000000>,
> > >>>>>              <0x1 0x40000000 0x0 0xc0000000 0x0 0x40000000>;
> > >>>>>
> > >>>>> not working for you? I haven't tried this myself, but since DT permits
> > >>>>> describing the inbound mappings this way, we should fix the code if it
> > >>>>> doesn't work at the moment.
> > >>>>
> > >>>> You mean encoding the memory controller index in the first cell? If that
> > >>>> works, that's indeed a much cleaner solution, though is it standard
> > >>>> compliant in any form?
> > >>>
> > >>> No those are just memory addresses (although I may have screwed up the
> > >>> order). From Documentation/devicetree/booting-without-of.txt:
> > >>>
> > >>> """
> > >>> Optional property:
> > >>> - dma-ranges: <prop-encoded-array> encoded as arbitrary number of triplets of
> > >>>         (child-bus-address, parent-bus-address, length). Each triplet specified
> > >>>         describes a contiguous DMA address range.
> > >>> """
> > >>>
> > >>
> > >> Then I am confused by your comment, that's what this patch does, it adds
> > >> support for reading "dma-ranges" from Device Tree and setting up inbound
> > >> windows using that. The only caveat is that because the PCIe root
> > >> complex has some ties with the memory bus architecture it is connected
> > >> to (SCB in our case) there is still a requirement to know the
> > >> translation between a given physical address and its backing memory
> > >> controller/aperture.
> > >>
> > >
> > > Ah ok, apologies for the noise then.
> > >
> > > I was hoping that having working support for dma-ranges would remove
> > > the need for the special phys<->dma conversion routines.
> >
> > What you describe definitively works with platform devices, but I am not
> > sure this is working for PCIe devices, although, conceptually it should,
> > yes.
> Sorry for my delay in responding.  One problem is that
> of_dma_configure() only looks at the first dma-range given and then
> converts it to dev->dma_pfn_offset which is respected by the DMA API.
> However, we often have multiple dma-ranges, not just one.  This is the
> big issue.
>

Given the recent attention to getting these APIs in shape, this may be
something Robin or Christoph may care to look into?

In any case, the description of dma-ranges should be in sync with the
way Linux interprets it, so this is either a documentation bug or a
DMA layer bug.

> There is another issue with of_dma_configure() being invoked by the EP
> driver on "bridge->parent->of_node", which is our RC device,
> Of_dma_configure() calls of_dma_range() on the of_get_next_parent() of
> our RC's device node and this misses the dma-ranges property which is
> contained within the RC.  I think I could workaround this but there is
> no getting around the first problem.
>

IIUC dma-ranges should be added to the parent bus of a device, which I
guess is slightly ambiguous for a root complex that incorporates a
host bridge.
Jim Quinlan Sept. 24, 2018, 3:01 p.m. | #9
On Mon, Sep 24, 2018 at 4:25 AM Ard Biesheuvel
<ard.biesheuvel@linaro.org> wrote:
>
> On Fri, 21 Sep 2018 at 19:41, Jim Quinlan <jim2101024@gmail.com> wrote:
> >
> > On Thu, Sep 20, 2018 at 5:39 PM Florian Fainelli <f.fainelli@gmail.com> wrote:
> > >
> > > On 09/20/2018 02:33 PM, Ard Biesheuvel wrote:
> > > > On 20 September 2018 at 14:31, Florian Fainelli <f.fainelli@gmail.com> wrote:
> > > >> On 09/20/2018 02:04 PM, Ard Biesheuvel wrote:
> > > >>> On 20 September 2018 at 13:55, Florian Fainelli <f.fainelli@gmail.com> wrote:
> > > >>>> On 09/19/2018 07:19 PM, Ard Biesheuvel wrote:
> > > >>>>> On 19 September 2018 at 07:31, Jim Quinlan <jim2101024@gmail.com> wrote:
> > > >>>>>> The Broadcom STB PCIe host controller is intimately related to the
> > > >>>>>> memory subsystem.  This close relationship adds complexity to how cpu
> > > >>>>>> system memory is mapped to PCIe memory.  Ideally, this mapping is an
> > > >>>>>> identity mapping, or an identity mapping off by a constant.  Not so in
> > > >>>>>> this case.
> > > >>>>>>
> > > >>>>>> Consider the Broadcom reference board BCM97445LCC_4X8 which has 6 GB
> > > >>>>>> of system memory.  Here is how the PCIe controller maps the
> > > >>>>>> system memory to PCIe memory:
> > > >>>>>>
> > > >>>>>>   memc0-a@[        0....3fffffff] <=> pci@[        0....3fffffff]
> > > >>>>>>   memc0-b@[100000000...13fffffff] <=> pci@[ 40000000....7fffffff]
> > > >>>>>>   memc1-a@[ 40000000....7fffffff] <=> pci@[ 80000000....bfffffff]
> > > >>>>>>   memc1-b@[300000000...33fffffff] <=> pci@[ c0000000....ffffffff]
> > > >>>>>>   memc2-a@[ 80000000....bfffffff] <=> pci@[100000000...13fffffff]
> > > >>>>>>   memc2-b@[c00000000...c3fffffff] <=> pci@[140000000...17fffffff]
> > > >>>>>>
> > > >>>>>
> > > >>>>> So is describing this as
> > > >>>>>
> > > >>>>> dma-ranges = <0x0 0x0 0x0 0x0 0x0 0x40000000>,
> > > >>>>>              <0x0 0x40000000 0x1 0x0 0x0 0x40000000>,
> > > >>>>>              <0x0 0x80000000 0x0 0x40000000 0x0 0x40000000>,
> > > >>>>>              <0x0 0xc0000000 0x3 0x0 0x0 0x40000000>,
> > > >>>>>              <0x1 0x0 0x0 0x80000000 0x0 0x40000000>,
> > > >>>>>              <0x1 0x40000000 0x0 0xc0000000 0x0 0x40000000>;
> > > >>>>>
> > > >>>>> not working for you? I haven't tried this myself, but since DT permits
> > > >>>>> describing the inbound mappings this way, we should fix the code if it
> > > >>>>> doesn't work at the moment.
> > > >>>>
> > > >>>> You mean encoding the memory controller index in the first cell? If that
> > > >>>> works, that's indeed a much cleaner solution, though is it standard
> > > >>>> compliant in any form?
> > > >>>
> > > >>> No those are just memory addresses (although I may have screwed up the
> > > >>> order). From Documentation/devicetree/booting-without-of.txt:
> > > >>>
> > > >>> """
> > > >>> Optional property:
> > > >>> - dma-ranges: <prop-encoded-array> encoded as arbitrary number of triplets of
> > > >>>         (child-bus-address, parent-bus-address, length). Each triplet specified
> > > >>>         describes a contiguous DMA address range.
> > > >>> """
> > > >>>
> > > >>
> > > >> Then I am confused by your comment, that's what this patch does, it adds
> > > >> support for reading "dma-ranges" from Device Tree and setting up inbound
> > > >> windows using that. The only caveat is that because the PCIe root
> > > >> complex has some ties with the memory bus architecture it is connected
> > > >> to (SCB in our case) there is still a requirement to know the
> > > >> translation between a given physical address and its backing memory
> > > >> controller/aperture.
> > > >>
> > > >
> > > > Ah ok, apologies for the noise then.
> > > >
> > > > I was hoping that having working support for dma-ranges would remove
> > > > the need for the special phys<->dma conversion routines.
> > >
> > > What you describe definitively works with platform devices, but I am not
> > > sure this is working for PCIe devices, although, conceptually it should,
> > > yes.
> > Sorry for my delay in responding.  One problem is that
> > of_dma_configure() only looks at the first dma-range given and then
> > converts it to dev->dma_pfn_offset which is respected by the DMA API.
> > However, we often have multiple dma-ranges, not just one.  This is the
> > big issue.
> >
>
> Given the recent attention to getting these APIs in shape, this may be
> something Robin or Christoph may care to look into?

It looks like this has been brought up before in the "[RFC PATCH] of:
Fix DMA configuration for non-DT masters" thread aka

https://lists.linuxfoundation.org/pipermail/iommu/2017-April/021325.html

In the thread "Oza Oza", a Broadcom coworker probably dealing with the
same exact problem as I,  enumerates three problems.   #1 and #2 are
the exact same ones I've just given: the "dma-ranges" prop of the RC
DT node is "skipped", and of_dma_get_range() only considers the first
entry in any "dma-ranges".

Thanks, Jim

>
> In any case, the description of dma-ranges should be in sync with the
> way Linux interprets it, so this is either a documentation bug or a
> DMA layer bug.
>
> > There is another issue with of_dma_configure() being invoked by the EP
> > driver on "bridge->parent->of_node", which is our RC device,
> > Of_dma_configure() calls of_dma_range() on the of_get_next_parent() of
> > our RC's device node and this misses the dma-ranges property which is
> > contained within the RC.  I think I could workaround this but there is
> > no getting around the first problem.
> >
>
> IIUC dma-ranges should be added to the parent bus of a device, which I
> guess is slightly ambiguous for a root complex that incorporates a
> host bridge.
Florian Fainelli Sept. 26, 2018, 2:59 a.m. | #10
On 9/24/2018 8:01 AM, Jim Quinlan wrote:
> On Mon, Sep 24, 2018 at 4:25 AM Ard Biesheuvel
> <ard.biesheuvel@linaro.org> wrote:
>>
>> On Fri, 21 Sep 2018 at 19:41, Jim Quinlan <jim2101024@gmail.com> wrote:
>>>
>>> On Thu, Sep 20, 2018 at 5:39 PM Florian Fainelli <f.fainelli@gmail.com> wrote:
>>>>
>>>> On 09/20/2018 02:33 PM, Ard Biesheuvel wrote:
>>>>> On 20 September 2018 at 14:31, Florian Fainelli <f.fainelli@gmail.com> wrote:
>>>>>> On 09/20/2018 02:04 PM, Ard Biesheuvel wrote:
>>>>>>> On 20 September 2018 at 13:55, Florian Fainelli <f.fainelli@gmail.com> wrote:
>>>>>>>> On 09/19/2018 07:19 PM, Ard Biesheuvel wrote:
>>>>>>>>> On 19 September 2018 at 07:31, Jim Quinlan <jim2101024@gmail.com> wrote:
>>>>>>>>>> The Broadcom STB PCIe host controller is intimately related to the
>>>>>>>>>> memory subsystem.  This close relationship adds complexity to how cpu
>>>>>>>>>> system memory is mapped to PCIe memory.  Ideally, this mapping is an
>>>>>>>>>> identity mapping, or an identity mapping off by a constant.  Not so in
>>>>>>>>>> this case.
>>>>>>>>>>
>>>>>>>>>> Consider the Broadcom reference board BCM97445LCC_4X8 which has 6 GB
>>>>>>>>>> of system memory.  Here is how the PCIe controller maps the
>>>>>>>>>> system memory to PCIe memory:
>>>>>>>>>>
>>>>>>>>>>    memc0-a@[        0....3fffffff] <=> pci@[        0....3fffffff]
>>>>>>>>>>    memc0-b@[100000000...13fffffff] <=> pci@[ 40000000....7fffffff]
>>>>>>>>>>    memc1-a@[ 40000000....7fffffff] <=> pci@[ 80000000....bfffffff]
>>>>>>>>>>    memc1-b@[300000000...33fffffff] <=> pci@[ c0000000....ffffffff]
>>>>>>>>>>    memc2-a@[ 80000000....bfffffff] <=> pci@[100000000...13fffffff]
>>>>>>>>>>    memc2-b@[c00000000...c3fffffff] <=> pci@[140000000...17fffffff]
>>>>>>>>>>
>>>>>>>>>
>>>>>>>>> So is describing this as
>>>>>>>>>
>>>>>>>>> dma-ranges = <0x0 0x0 0x0 0x0 0x0 0x40000000>,
>>>>>>>>>               <0x0 0x40000000 0x1 0x0 0x0 0x40000000>,
>>>>>>>>>               <0x0 0x80000000 0x0 0x40000000 0x0 0x40000000>,
>>>>>>>>>               <0x0 0xc0000000 0x3 0x0 0x0 0x40000000>,
>>>>>>>>>               <0x1 0x0 0x0 0x80000000 0x0 0x40000000>,
>>>>>>>>>               <0x1 0x40000000 0x0 0xc0000000 0x0 0x40000000>;
>>>>>>>>>
>>>>>>>>> not working for you? I haven't tried this myself, but since DT permits
>>>>>>>>> describing the inbound mappings this way, we should fix the code if it
>>>>>>>>> doesn't work at the moment.
>>>>>>>>
>>>>>>>> You mean encoding the memory controller index in the first cell? If that
>>>>>>>> works, that's indeed a much cleaner solution, though is it standard
>>>>>>>> compliant in any form?
>>>>>>>
>>>>>>> No those are just memory addresses (although I may have screwed up the
>>>>>>> order). From Documentation/devicetree/booting-without-of.txt:
>>>>>>>
>>>>>>> """
>>>>>>> Optional property:
>>>>>>> - dma-ranges: <prop-encoded-array> encoded as arbitrary number of triplets of
>>>>>>>          (child-bus-address, parent-bus-address, length). Each triplet specified
>>>>>>>          describes a contiguous DMA address range.
>>>>>>> """
>>>>>>>
>>>>>>
>>>>>> Then I am confused by your comment, that's what this patch does, it adds
>>>>>> support for reading "dma-ranges" from Device Tree and setting up inbound
>>>>>> windows using that. The only caveat is that because the PCIe root
>>>>>> complex has some ties with the memory bus architecture it is connected
>>>>>> to (SCB in our case) there is still a requirement to know the
>>>>>> translation between a given physical address and its backing memory
>>>>>> controller/aperture.
>>>>>>
>>>>>
>>>>> Ah ok, apologies for the noise then.
>>>>>
>>>>> I was hoping that having working support for dma-ranges would remove
>>>>> the need for the special phys<->dma conversion routines.
>>>>
>>>> What you describe definitively works with platform devices, but I am not
>>>> sure this is working for PCIe devices, although, conceptually it should,
>>>> yes.
>>> Sorry for my delay in responding.  One problem is that
>>> of_dma_configure() only looks at the first dma-range given and then
>>> converts it to dev->dma_pfn_offset which is respected by the DMA API.
>>> However, we often have multiple dma-ranges, not just one.  This is the
>>> big issue.
>>>
>>
>> Given the recent attention to getting these APIs in shape, this may be
>> something Robin or Christoph may care to look into?
> 
> It looks like this has been brought up before in the "[RFC PATCH] of:
> Fix DMA configuration for non-DT masters" thread aka
> 
> https://lists.linuxfoundation.org/pipermail/iommu/2017-April/021325.html
> 
> In the thread "Oza Oza", a Broadcom coworker probably dealing with the
> same exact problem as I,  enumerates three problems.   #1 and #2 are
> the exact same ones I've just given: the "dma-ranges" prop of the RC
> DT node is "skipped", and of_dma_get_range() only considers the first
> entry in any "dma-ranges".

Robin, is that something that is expected or should the "dma-ranges" 
somehow propagate from host bridge down the PCIe end-point drivers?

> 
> Thanks, Jim
> 
>>
>> In any case, the description of dma-ranges should be in sync with the
>> way Linux interprets it, so this is either a documentation bug or a
>> DMA layer bug.
>>
>>> There is another issue with of_dma_configure() being invoked by the EP
>>> driver on "bridge->parent->of_node", which is our RC device,
>>> Of_dma_configure() calls of_dma_range() on the of_get_next_parent() of
>>> our RC's device node and this misses the dma-ranges property which is
>>> contained within the RC.  I think I could workaround this but there is
>>> no getting around the first problem.
>>>
>>
>> IIUC dma-ranges should be added to the parent bus of a device, which I
>> guess is slightly ambiguous for a root complex that incorporates a
>> host bridge.

Humm, why is that ambiguous for a host bridge/root complex?
Ard Biesheuvel Sept. 26, 2018, 7:59 a.m. | #11
On Wed, 26 Sep 2018 at 04:59, Florian Fainelli <f.fainelli@gmail.com> wrote:
>
>
>
> On 9/24/2018 8:01 AM, Jim Quinlan wrote:
> > On Mon, Sep 24, 2018 at 4:25 AM Ard Biesheuvel
> > <ard.biesheuvel@linaro.org> wrote:
> >>
> >> On Fri, 21 Sep 2018 at 19:41, Jim Quinlan <jim2101024@gmail.com> wrote:
> >>>
> >>> On Thu, Sep 20, 2018 at 5:39 PM Florian Fainelli <f.fainelli@gmail.com> wrote:
> >>>>
> >>>> On 09/20/2018 02:33 PM, Ard Biesheuvel wrote:
> >>>>> On 20 September 2018 at 14:31, Florian Fainelli <f.fainelli@gmail.com> wrote:
> >>>>>> On 09/20/2018 02:04 PM, Ard Biesheuvel wrote:
> >>>>>>> On 20 September 2018 at 13:55, Florian Fainelli <f.fainelli@gmail.com> wrote:
> >>>>>>>> On 09/19/2018 07:19 PM, Ard Biesheuvel wrote:
> >>>>>>>>> On 19 September 2018 at 07:31, Jim Quinlan <jim2101024@gmail.com> wrote:
> >>>>>>>>>> The Broadcom STB PCIe host controller is intimately related to the
> >>>>>>>>>> memory subsystem.  This close relationship adds complexity to how cpu
> >>>>>>>>>> system memory is mapped to PCIe memory.  Ideally, this mapping is an
> >>>>>>>>>> identity mapping, or an identity mapping off by a constant.  Not so in
> >>>>>>>>>> this case.
> >>>>>>>>>>
> >>>>>>>>>> Consider the Broadcom reference board BCM97445LCC_4X8 which has 6 GB
> >>>>>>>>>> of system memory.  Here is how the PCIe controller maps the
> >>>>>>>>>> system memory to PCIe memory:
> >>>>>>>>>>
> >>>>>>>>>>    memc0-a@[        0....3fffffff] <=> pci@[        0....3fffffff]
> >>>>>>>>>>    memc0-b@[100000000...13fffffff] <=> pci@[ 40000000....7fffffff]
> >>>>>>>>>>    memc1-a@[ 40000000....7fffffff] <=> pci@[ 80000000....bfffffff]
> >>>>>>>>>>    memc1-b@[300000000...33fffffff] <=> pci@[ c0000000....ffffffff]
> >>>>>>>>>>    memc2-a@[ 80000000....bfffffff] <=> pci@[100000000...13fffffff]
> >>>>>>>>>>    memc2-b@[c00000000...c3fffffff] <=> pci@[140000000...17fffffff]
> >>>>>>>>>>
> >>>>>>>>>
> >>>>>>>>> So is describing this as
> >>>>>>>>>
> >>>>>>>>> dma-ranges = <0x0 0x0 0x0 0x0 0x0 0x40000000>,
> >>>>>>>>>               <0x0 0x40000000 0x1 0x0 0x0 0x40000000>,
> >>>>>>>>>               <0x0 0x80000000 0x0 0x40000000 0x0 0x40000000>,
> >>>>>>>>>               <0x0 0xc0000000 0x3 0x0 0x0 0x40000000>,
> >>>>>>>>>               <0x1 0x0 0x0 0x80000000 0x0 0x40000000>,
> >>>>>>>>>               <0x1 0x40000000 0x0 0xc0000000 0x0 0x40000000>;
> >>>>>>>>>
> >>>>>>>>> not working for you? I haven't tried this myself, but since DT permits
> >>>>>>>>> describing the inbound mappings this way, we should fix the code if it
> >>>>>>>>> doesn't work at the moment.
> >>>>>>>>
> >>>>>>>> You mean encoding the memory controller index in the first cell? If that
> >>>>>>>> works, that's indeed a much cleaner solution, though is it standard
> >>>>>>>> compliant in any form?
> >>>>>>>
> >>>>>>> No those are just memory addresses (although I may have screwed up the
> >>>>>>> order). From Documentation/devicetree/booting-without-of.txt:
> >>>>>>>
> >>>>>>> """
> >>>>>>> Optional property:
> >>>>>>> - dma-ranges: <prop-encoded-array> encoded as arbitrary number of triplets of
> >>>>>>>          (child-bus-address, parent-bus-address, length). Each triplet specified
> >>>>>>>          describes a contiguous DMA address range.
> >>>>>>> """
> >>>>>>>
> >>>>>>
> >>>>>> Then I am confused by your comment, that's what this patch does, it adds
> >>>>>> support for reading "dma-ranges" from Device Tree and setting up inbound
> >>>>>> windows using that. The only caveat is that because the PCIe root
> >>>>>> complex has some ties with the memory bus architecture it is connected
> >>>>>> to (SCB in our case) there is still a requirement to know the
> >>>>>> translation between a given physical address and its backing memory
> >>>>>> controller/aperture.
> >>>>>>
> >>>>>
> >>>>> Ah ok, apologies for the noise then.
> >>>>>
> >>>>> I was hoping that having working support for dma-ranges would remove
> >>>>> the need for the special phys<->dma conversion routines.
> >>>>
> >>>> What you describe definitively works with platform devices, but I am not
> >>>> sure this is working for PCIe devices, although, conceptually it should,
> >>>> yes.
> >>> Sorry for my delay in responding.  One problem is that
> >>> of_dma_configure() only looks at the first dma-range given and then
> >>> converts it to dev->dma_pfn_offset which is respected by the DMA API.
> >>> However, we often have multiple dma-ranges, not just one.  This is the
> >>> big issue.
> >>>
> >>
> >> Given the recent attention to getting these APIs in shape, this may be
> >> something Robin or Christoph may care to look into?
> >
> > It looks like this has been brought up before in the "[RFC PATCH] of:
> > Fix DMA configuration for non-DT masters" thread aka
> >
> > https://lists.linuxfoundation.org/pipermail/iommu/2017-April/021325.html
> >
> > In the thread "Oza Oza", a Broadcom coworker probably dealing with the
> > same exact problem as I,  enumerates three problems.   #1 and #2 are
> > the exact same ones I've just given: the "dma-ranges" prop of the RC
> > DT node is "skipped", and of_dma_get_range() only considers the first
> > entry in any "dma-ranges".
>
> Robin, is that something that is expected or should the "dma-ranges"
> somehow propagate from host bridge down the PCIe end-point drivers?
>
> >
> > Thanks, Jim
> >
> >>
> >> In any case, the description of dma-ranges should be in sync with the
> >> way Linux interprets it, so this is either a documentation bug or a
> >> DMA layer bug.
> >>
> >>> There is another issue with of_dma_configure() being invoked by the EP
> >>> driver on "bridge->parent->of_node", which is our RC device,
> >>> Of_dma_configure() calls of_dma_range() on the of_get_next_parent() of
> >>> our RC's device node and this misses the dma-ranges property which is
> >>> contained within the RC.  I think I could workaround this but there is
> >>> no getting around the first problem.
> >>>
> >>
> >> IIUC dma-ranges should be added to the parent bus of a device, which I
> >> guess is slightly ambiguous for a root complex that incorporates a
> >> host bridge.
>
> Humm, why is that ambiguous for a host bridge/root complex?

Well, perhaps I am the only one who is slightly confused by this, but
the fact that the device node describes both the host bridge and the
root port/bridge (which is arguably a device behind the bus) makes it
somewhat non-intuitive to me.
Robin Murphy Sept. 26, 2018, 10:56 a.m. | #12
On 26/09/18 03:59, Florian Fainelli wrote:
> 
> 
> On 9/24/2018 8:01 AM, Jim Quinlan wrote:
>> On Mon, Sep 24, 2018 at 4:25 AM Ard Biesheuvel
>> <ard.biesheuvel@linaro.org> wrote:
>>>
>>> On Fri, 21 Sep 2018 at 19:41, Jim Quinlan <jim2101024@gmail.com> wrote:
>>>>
>>>> On Thu, Sep 20, 2018 at 5:39 PM Florian Fainelli 
>>>> <f.fainelli@gmail.com> wrote:
>>>>>
>>>>> On 09/20/2018 02:33 PM, Ard Biesheuvel wrote:
>>>>>> On 20 September 2018 at 14:31, Florian Fainelli 
>>>>>> <f.fainelli@gmail.com> wrote:
>>>>>>> On 09/20/2018 02:04 PM, Ard Biesheuvel wrote:
>>>>>>>> On 20 September 2018 at 13:55, Florian Fainelli 
>>>>>>>> <f.fainelli@gmail.com> wrote:
>>>>>>>>> On 09/19/2018 07:19 PM, Ard Biesheuvel wrote:
>>>>>>>>>> On 19 September 2018 at 07:31, Jim Quinlan 
>>>>>>>>>> <jim2101024@gmail.com> wrote:
>>>>>>>>>>> The Broadcom STB PCIe host controller is intimately related 
>>>>>>>>>>> to the
>>>>>>>>>>> memory subsystem.  This close relationship adds complexity to 
>>>>>>>>>>> how cpu
>>>>>>>>>>> system memory is mapped to PCIe memory.  Ideally, this 
>>>>>>>>>>> mapping is an
>>>>>>>>>>> identity mapping, or an identity mapping off by a constant.  
>>>>>>>>>>> Not so in
>>>>>>>>>>> this case.
>>>>>>>>>>>
>>>>>>>>>>> Consider the Broadcom reference board BCM97445LCC_4X8 which 
>>>>>>>>>>> has 6 GB
>>>>>>>>>>> of system memory.  Here is how the PCIe controller maps the
>>>>>>>>>>> system memory to PCIe memory:
>>>>>>>>>>>
>>>>>>>>>>>    memc0-a@[        0....3fffffff] <=> pci@[        
>>>>>>>>>>> 0....3fffffff]
>>>>>>>>>>>    memc0-b@[100000000...13fffffff] <=> pci@[ 
>>>>>>>>>>> 40000000....7fffffff]
>>>>>>>>>>>    memc1-a@[ 40000000....7fffffff] <=> pci@[ 
>>>>>>>>>>> 80000000....bfffffff]
>>>>>>>>>>>    memc1-b@[300000000...33fffffff] <=> pci@[ 
>>>>>>>>>>> c0000000....ffffffff]
>>>>>>>>>>>    memc2-a@[ 80000000....bfffffff] <=> 
>>>>>>>>>>> pci@[100000000...13fffffff]
>>>>>>>>>>>    memc2-b@[c00000000...c3fffffff] <=> 
>>>>>>>>>>> pci@[140000000...17fffffff]
>>>>>>>>>>>
>>>>>>>>>>
>>>>>>>>>> So is describing this as
>>>>>>>>>>
>>>>>>>>>> dma-ranges = <0x0 0x0 0x0 0x0 0x0 0x40000000>,
>>>>>>>>>>               <0x0 0x40000000 0x1 0x0 0x0 0x40000000>,
>>>>>>>>>>               <0x0 0x80000000 0x0 0x40000000 0x0 0x40000000>,
>>>>>>>>>>               <0x0 0xc0000000 0x3 0x0 0x0 0x40000000>,
>>>>>>>>>>               <0x1 0x0 0x0 0x80000000 0x0 0x40000000>,
>>>>>>>>>>               <0x1 0x40000000 0x0 0xc0000000 0x0 0x40000000>;
>>>>>>>>>>
>>>>>>>>>> not working for you? I haven't tried this myself, but since DT 
>>>>>>>>>> permits
>>>>>>>>>> describing the inbound mappings this way, we should fix the 
>>>>>>>>>> code if it
>>>>>>>>>> doesn't work at the moment.
>>>>>>>>>
>>>>>>>>> You mean encoding the memory controller index in the first 
>>>>>>>>> cell? If that
>>>>>>>>> works, that's indeed a much cleaner solution, though is it 
>>>>>>>>> standard
>>>>>>>>> compliant in any form?
>>>>>>>>
>>>>>>>> No those are just memory addresses (although I may have screwed 
>>>>>>>> up the
>>>>>>>> order). From Documentation/devicetree/booting-without-of.txt:
>>>>>>>>
>>>>>>>> """
>>>>>>>> Optional property:
>>>>>>>> - dma-ranges: <prop-encoded-array> encoded as arbitrary number 
>>>>>>>> of triplets of
>>>>>>>>          (child-bus-address, parent-bus-address, length). Each 
>>>>>>>> triplet specified
>>>>>>>>          describes a contiguous DMA address range.
>>>>>>>> """
>>>>>>>>
>>>>>>>
>>>>>>> Then I am confused by your comment, that's what this patch does, 
>>>>>>> it adds
>>>>>>> support for reading "dma-ranges" from Device Tree and setting up 
>>>>>>> inbound
>>>>>>> windows using that. The only caveat is that because the PCIe root
>>>>>>> complex has some ties with the memory bus architecture it is 
>>>>>>> connected
>>>>>>> to (SCB in our case) there is still a requirement to know the
>>>>>>> translation between a given physical address and its backing memory
>>>>>>> controller/aperture.
>>>>>>>
>>>>>>
>>>>>> Ah ok, apologies for the noise then.
>>>>>>
>>>>>> I was hoping that having working support for dma-ranges would remove
>>>>>> the need for the special phys<->dma conversion routines.
>>>>>
>>>>> What you describe definitively works with platform devices, but I 
>>>>> am not
>>>>> sure this is working for PCIe devices, although, conceptually it 
>>>>> should,
>>>>> yes.
>>>> Sorry for my delay in responding.  One problem is that
>>>> of_dma_configure() only looks at the first dma-range given and then
>>>> converts it to dev->dma_pfn_offset which is respected by the DMA API.
>>>> However, we often have multiple dma-ranges, not just one.  This is the
>>>> big issue.
>>>>
>>>
>>> Given the recent attention to getting these APIs in shape, this may be
>>> something Robin or Christoph may care to look into?
>>
>> It looks like this has been brought up before in the "[RFC PATCH] of:
>> Fix DMA configuration for non-DT masters" thread aka
>>
>> https://lists.linuxfoundation.org/pipermail/iommu/2017-April/021325.html
>>
>> In the thread "Oza Oza", a Broadcom coworker probably dealing with the
>> same exact problem as I,  enumerates three problems.   #1 and #2 are
>> the exact same ones I've just given: the "dma-ranges" prop of the RC
>> DT node is "skipped", and of_dma_get_range() only considers the first
>> entry in any "dma-ranges".
> 
> Robin, is that something that is expected or should the "dma-ranges" 
> somehow propagate from host bridge down the PCIe end-point drivers?

Nope, the code is most definitely incomplete - it's sufficient to 
support the system it was originally needed for (i.e. platform devices 
with a single range), but can by no means even pretend to support the 
binding fully. Furthermore, the way that PCI support was later grafted 
into of_dma_configure() was *only* in support of dma-coherent without 
consideration for dma-ranges. Hence the current mess.

>>
>> Thanks, Jim
>>
>>>
>>> In any case, the description of dma-ranges should be in sync with the
>>> way Linux interprets it, so this is either a documentation bug or a
>>> DMA layer bug.
>>>
>>>> There is another issue with of_dma_configure() being invoked by the EP
>>>> driver on "bridge->parent->of_node", which is our RC device,
>>>> Of_dma_configure() calls of_dma_range() on the of_get_next_parent() of
>>>> our RC's device node and this misses the dma-ranges property which is
>>>> contained within the RC.  I think I could workaround this but there is
>>>> no getting around the first problem.
>>>>
>>>
>>> IIUC dma-ranges should be added to the parent bus of a device, which I
>>> guess is slightly ambiguous for a root complex that incorporates a
>>> host bridge.
> 
> Humm, why is that ambiguous for a host bridge/root complex?

The real problem is that FDT machines don't describe the PCI hierarchy 
in DT as proper OF does, so we have this awkward crossing between the DT 
model and the Linux device model where the devices have no DT 
representation and the "parent bus" is a DT leaf node, which cocks up 
the way the current code is expecting to work.

Robin.
Florian Fainelli Oct. 1, 2018, 6:13 p.m. | #13
On 09/26/2018 03:56 AM, Robin Murphy wrote:
> On 26/09/18 03:59, Florian Fainelli wrote:
>>
>>
>> On 9/24/2018 8:01 AM, Jim Quinlan wrote:
>>> On Mon, Sep 24, 2018 at 4:25 AM Ard Biesheuvel
>>> <ard.biesheuvel@linaro.org> wrote:
>>>>
>>>> On Fri, 21 Sep 2018 at 19:41, Jim Quinlan <jim2101024@gmail.com> wrote:
>>>>>
>>>>> On Thu, Sep 20, 2018 at 5:39 PM Florian Fainelli
>>>>> <f.fainelli@gmail.com> wrote:
>>>>>>
>>>>>> On 09/20/2018 02:33 PM, Ard Biesheuvel wrote:
>>>>>>> On 20 September 2018 at 14:31, Florian Fainelli
>>>>>>> <f.fainelli@gmail.com> wrote:
>>>>>>>> On 09/20/2018 02:04 PM, Ard Biesheuvel wrote:
>>>>>>>>> On 20 September 2018 at 13:55, Florian Fainelli
>>>>>>>>> <f.fainelli@gmail.com> wrote:
>>>>>>>>>> On 09/19/2018 07:19 PM, Ard Biesheuvel wrote:
>>>>>>>>>>> On 19 September 2018 at 07:31, Jim Quinlan
>>>>>>>>>>> <jim2101024@gmail.com> wrote:
>>>>>>>>>>>> The Broadcom STB PCIe host controller is intimately related
>>>>>>>>>>>> to the
>>>>>>>>>>>> memory subsystem.  This close relationship adds complexity
>>>>>>>>>>>> to how cpu
>>>>>>>>>>>> system memory is mapped to PCIe memory.  Ideally, this
>>>>>>>>>>>> mapping is an
>>>>>>>>>>>> identity mapping, or an identity mapping off by a constant. 
>>>>>>>>>>>> Not so in
>>>>>>>>>>>> this case.
>>>>>>>>>>>>
>>>>>>>>>>>> Consider the Broadcom reference board BCM97445LCC_4X8 which
>>>>>>>>>>>> has 6 GB
>>>>>>>>>>>> of system memory.  Here is how the PCIe controller maps the
>>>>>>>>>>>> system memory to PCIe memory:
>>>>>>>>>>>>
>>>>>>>>>>>>    memc0-a@[        0....3fffffff] <=> pci@[       
>>>>>>>>>>>> 0....3fffffff]
>>>>>>>>>>>>    memc0-b@[100000000...13fffffff] <=> pci@[
>>>>>>>>>>>> 40000000....7fffffff]
>>>>>>>>>>>>    memc1-a@[ 40000000....7fffffff] <=> pci@[
>>>>>>>>>>>> 80000000....bfffffff]
>>>>>>>>>>>>    memc1-b@[300000000...33fffffff] <=> pci@[
>>>>>>>>>>>> c0000000....ffffffff]
>>>>>>>>>>>>    memc2-a@[ 80000000....bfffffff] <=>
>>>>>>>>>>>> pci@[100000000...13fffffff]
>>>>>>>>>>>>    memc2-b@[c00000000...c3fffffff] <=>
>>>>>>>>>>>> pci@[140000000...17fffffff]
>>>>>>>>>>>>
>>>>>>>>>>>
>>>>>>>>>>> So is describing this as
>>>>>>>>>>>
>>>>>>>>>>> dma-ranges = <0x0 0x0 0x0 0x0 0x0 0x40000000>,
>>>>>>>>>>>               <0x0 0x40000000 0x1 0x0 0x0 0x40000000>,
>>>>>>>>>>>               <0x0 0x80000000 0x0 0x40000000 0x0 0x40000000>,
>>>>>>>>>>>               <0x0 0xc0000000 0x3 0x0 0x0 0x40000000>,
>>>>>>>>>>>               <0x1 0x0 0x0 0x80000000 0x0 0x40000000>,
>>>>>>>>>>>               <0x1 0x40000000 0x0 0xc0000000 0x0 0x40000000>;
>>>>>>>>>>>
>>>>>>>>>>> not working for you? I haven't tried this myself, but since
>>>>>>>>>>> DT permits
>>>>>>>>>>> describing the inbound mappings this way, we should fix the
>>>>>>>>>>> code if it
>>>>>>>>>>> doesn't work at the moment.
>>>>>>>>>>
>>>>>>>>>> You mean encoding the memory controller index in the first
>>>>>>>>>> cell? If that
>>>>>>>>>> works, that's indeed a much cleaner solution, though is it
>>>>>>>>>> standard
>>>>>>>>>> compliant in any form?
>>>>>>>>>
>>>>>>>>> No those are just memory addresses (although I may have screwed
>>>>>>>>> up the
>>>>>>>>> order). From Documentation/devicetree/booting-without-of.txt:
>>>>>>>>>
>>>>>>>>> """
>>>>>>>>> Optional property:
>>>>>>>>> - dma-ranges: <prop-encoded-array> encoded as arbitrary number
>>>>>>>>> of triplets of
>>>>>>>>>          (child-bus-address, parent-bus-address, length). Each
>>>>>>>>> triplet specified
>>>>>>>>>          describes a contiguous DMA address range.
>>>>>>>>> """
>>>>>>>>>
>>>>>>>>
>>>>>>>> Then I am confused by your comment, that's what this patch does,
>>>>>>>> it adds
>>>>>>>> support for reading "dma-ranges" from Device Tree and setting up
>>>>>>>> inbound
>>>>>>>> windows using that. The only caveat is that because the PCIe root
>>>>>>>> complex has some ties with the memory bus architecture it is
>>>>>>>> connected
>>>>>>>> to (SCB in our case) there is still a requirement to know the
>>>>>>>> translation between a given physical address and its backing memory
>>>>>>>> controller/aperture.
>>>>>>>>
>>>>>>>
>>>>>>> Ah ok, apologies for the noise then.
>>>>>>>
>>>>>>> I was hoping that having working support for dma-ranges would remove
>>>>>>> the need for the special phys<->dma conversion routines.
>>>>>>
>>>>>> What you describe definitively works with platform devices, but I
>>>>>> am not
>>>>>> sure this is working for PCIe devices, although, conceptually it
>>>>>> should,
>>>>>> yes.
>>>>> Sorry for my delay in responding.  One problem is that
>>>>> of_dma_configure() only looks at the first dma-range given and then
>>>>> converts it to dev->dma_pfn_offset which is respected by the DMA API.
>>>>> However, we often have multiple dma-ranges, not just one.  This is the
>>>>> big issue.
>>>>>
>>>>
>>>> Given the recent attention to getting these APIs in shape, this may be
>>>> something Robin or Christoph may care to look into?
>>>
>>> It looks like this has been brought up before in the "[RFC PATCH] of:
>>> Fix DMA configuration for non-DT masters" thread aka
>>>
>>> https://lists.linuxfoundation.org/pipermail/iommu/2017-April/021325.html
>>>
>>> In the thread "Oza Oza", a Broadcom coworker probably dealing with the
>>> same exact problem as I,  enumerates three problems.   #1 and #2 are
>>> the exact same ones I've just given: the "dma-ranges" prop of the RC
>>> DT node is "skipped", and of_dma_get_range() only considers the first
>>> entry in any "dma-ranges".
>>
>> Robin, is that something that is expected or should the "dma-ranges"
>> somehow propagate from host bridge down the PCIe end-point drivers?
> 
> Nope, the code is most definitely incomplete - it's sufficient to
> support the system it was originally needed for (i.e. platform devices
> with a single range), but can by no means even pretend to support the
> binding fully. Furthermore, the way that PCI support was later grafted
> into of_dma_configure() was *only* in support of dma-coherent without
> consideration for dma-ranges. Hence the current mess.

Is this the way we should go? If the PCIe root complex has a
"dma-ranges" property, we need to make sure that all child devices
behind it get that translation applied as well? Right now we tried to
follow Christoph's recommendations from the v1 through v4 submissions,
but that does not really seem to be working well.

I am really not a fan of having to go audit every architecture, and
possibly insert the right translation in there, while introducing a
possible layering problem too...

> 
>>>
>>> Thanks, Jim
>>>
>>>>
>>>> In any case, the description of dma-ranges should be in sync with the
>>>> way Linux interprets it, so this is either a documentation bug or a
>>>> DMA layer bug.
>>>>
>>>>> There is another issue with of_dma_configure() being invoked by the EP
>>>>> driver on "bridge->parent->of_node", which is our RC device,
>>>>> Of_dma_configure() calls of_dma_range() on the of_get_next_parent() of
>>>>> our RC's device node and this misses the dma-ranges property which is
>>>>> contained within the RC.  I think I could workaround this but there is
>>>>> no getting around the first problem.
>>>>>
>>>>
>>>> IIUC dma-ranges should be added to the parent bus of a device, which I
>>>> guess is slightly ambiguous for a root complex that incorporates a
>>>> host bridge.
>>
>> Humm, why is that ambiguous for a host bridge/root complex?
> 
> The real problem is that FDT machines don't describe the PCI hierarchy
> in DT as proper OF does, so we have this awkward crossing between the DT
> model and the Linux device model where the devices have no DT
> representation and the "parent bus" is a DT leaf node, which cocks up
> the way the current code is expecting to work.

OK, I see what you mean, we still have to work with what we have here,
which is FDT...

Patch

diff --git a/drivers/pci/controller/pcie-brcmstb.c b/drivers/pci/controller/pcie-brcmstb.c
index 9c87d10..abfa429 100644
--- a/drivers/pci/controller/pcie-brcmstb.c
+++ b/drivers/pci/controller/pcie-brcmstb.c
@@ -21,6 +21,7 @@ 
 #include <linux/printk.h>
 #include <linux/sizes.h>
 #include <linux/slab.h>
+#include <soc/brcmstb/common.h>
 #include <soc/brcmstb/memory_api.h>
 #include <linux/string.h>
 #include <linux/types.h>
@@ -321,6 +322,7 @@  static void __iomem *brcm_pcie_map_conf(struct pci_bus *bus, unsigned int devfn,
 	(((val) & ~reg##_##field##_MASK) | \
 	 (reg##_##field##_MASK & (field_val << reg##_##field##_SHIFT)))
 
+static struct of_pci_range *brcm_dma_ranges;
 static phys_addr_t scb_size[BRCM_MAX_SCB];
 static int num_memc;
 static int num_pcie;
@@ -599,6 +601,79 @@  static inline void brcm_pcie_perst_set(struct brcm_pcie *pcie,
 		WR_FLD_RB(pcie->base, PCIE_MISC_PCIE_CTRL, PCIE_PERSTB, !val);
 }
 
+static int brcm_pcie_parse_map_dma_ranges(struct brcm_pcie *pcie)
+{
+	int i;
+	struct of_pci_range_parser parser;
+	struct device_node *dn = pcie->dn;
+
+	/*
+	 * Parse dma-ranges property if present.  If there are multiple
+	 * PCIe controllers, we only have to parse from one of them since
+	 * the others will have an identical mapping.
+	 */
+	if (!of_pci_dma_range_parser_init(&parser, dn)) {
+		struct of_pci_range *p;
+		unsigned int max_ranges = (parser.end - parser.range)
+			/ parser.np;
+
+		/* Add a null entry to indicate the end of the array */
+		brcm_dma_ranges = kcalloc(max_ranges + 1,
+					  sizeof(struct of_pci_range),
+					  GFP_KERNEL);
+		if (!brcm_dma_ranges)
+			return -ENOMEM;
+
+		p = brcm_dma_ranges;
+		while (of_pci_range_parser_one(&parser, p))
+			p++;
+	}
+
+	for (i = 0, num_memc = 0; i < BRCM_MAX_SCB; i++) {
+		u64 size = brcmstb_memory_memc_size(i);
+
+		if (size == (u64)-1) {
+			dev_err(pcie->dev, "cannot get memc%d size", i);
+			return -EINVAL;
+		} else if (size) {
+			scb_size[i] = roundup_pow_of_two_64(size);
+			num_memc++;
+		} else {
+			break;
+		}
+	}
+
+	return 0;
+}
+
+dma_addr_t brcm_phys_to_dma(struct device *dev, phys_addr_t paddr)
+{
+	struct of_pci_range *p;
+
+	if (!dev || !dev_is_pci(dev))
+		return (dma_addr_t)paddr;
+	for (p = brcm_dma_ranges; p && p->size; p++)
+		if (paddr >= p->cpu_addr && paddr < (p->cpu_addr + p->size))
+			return (dma_addr_t)(paddr - p->cpu_addr + p->pci_addr);
+
+	return (dma_addr_t)paddr;
+}
+
+phys_addr_t brcm_dma_to_phys(struct device *dev, dma_addr_t dev_addr)
+{
+	struct of_pci_range *p;
+
+	if (!dev || !dev_is_pci(dev))
+		return (phys_addr_t)dev_addr;
+	for (p = brcm_dma_ranges; p && p->size; p++)
+		if (dev_addr >= p->pci_addr
+		    && dev_addr < (p->pci_addr + p->size))
+			return (phys_addr_t)
+				(dev_addr - p->pci_addr + p->cpu_addr);
+
+	return (phys_addr_t)dev_addr;
+}
+
 static int brcm_pcie_add_controller(struct brcm_pcie *pcie)
 {
 	int i, ret = 0;
@@ -610,6 +685,10 @@  static int brcm_pcie_add_controller(struct brcm_pcie *pcie)
 		goto done;
 	}
 
+	ret = brcm_pcie_parse_map_dma_ranges(pcie);
+	if (ret)
+		goto done;
+
 	/* Determine num_memc and their sizes */
 	for (i = 0, num_memc = 0; i < BRCM_MAX_SCB; i++) {
 		u64 size = brcmstb_memory_memc_size(i);
@@ -639,8 +718,13 @@  static int brcm_pcie_add_controller(struct brcm_pcie *pcie)
 static void brcm_pcie_remove_controller(struct brcm_pcie *pcie)
 {
 	mutex_lock(&brcm_pcie_lock);
-	if (--num_pcie == 0)
-		num_memc = 0;
+	if (--num_pcie > 0)
+		goto out;
+
+	kfree(brcm_dma_ranges);
+	brcm_dma_ranges = NULL;
+	num_memc = 0;
+out:
 	mutex_unlock(&brcm_pcie_lock);
 }
 
@@ -747,11 +831,37 @@  static int brcm_pcie_setup(struct brcm_pcie *pcie)
 	 */
 	rc_bar2_size = roundup_pow_of_two_64(total_mem_size);
 
-	/*
-	 * Set simple configuration based on memory sizes
-	 * only.  We always start the viewport at address 0.
-	 */
-	rc_bar2_offset = 0;
+	if (brcm_dma_ranges) {
+		/*
+		 * The best-case scenario is to place the inbound
+		 * region in the first 4GB of pcie-space, as some
+		 * legacy devices can only address 32bits.
+		 * We would also like to put the MSI under 4GB
+		 * as well, since some devices require a 32bit
+		 * MSI target address.
+		 */
+		if (total_mem_size <= 0xc0000000ULL &&
+		    rc_bar2_size <= 0x100000000ULL) {
+			rc_bar2_offset = 0;
+		} else {
+			/*
+			 * The system memory is 4GB or larger so we
+			 * cannot start the inbound region at location
+			 * 0 (since we have to allow some space for
+			 * outbound memory @ 3GB).  So instead we
+			 * start it at the 1x multiple of its size
+			 */
+			rc_bar2_offset = rc_bar2_size;
+		}
+
+	} else {
+		/*
+		 * Set simple configuration based on memory sizes
+		 * only.  We always start the viewport at address 0,
+		 * and set the MSI target address accordingly.
+		 */
+		rc_bar2_offset = 0;
+	}
 
 	tmp = lower_32_bits(rc_bar2_offset);
 	tmp = INSERT_FIELD(tmp, PCIE_MISC_RC_BAR2_CONFIG_LO, SIZE,
@@ -969,7 +1079,6 @@  static int brcm_pcie_probe(struct platform_device *pdev)
 	struct brcm_pcie *pcie;
 	struct resource *res;
 	void __iomem *base;
-	u32 tmp;
 	struct pci_host_bridge *bridge;
 	struct pci_bus *child;
 
@@ -986,11 +1095,6 @@  static int brcm_pcie_probe(struct platform_device *pdev)
 		return -EINVAL;
 	}
 
-	if (of_property_read_u32(dn, "dma-ranges", &tmp) == 0) {
-		dev_err(&pdev->dev, "cannot yet handle dma-ranges\n");
-		return -EINVAL;
-	}
-
 	data = of_id->data;
 	pcie->reg_offsets = data->offsets;
 	pcie->reg_field_info = data->reg_field_info;
diff --git a/include/soc/brcmstb/common.h b/include/soc/brcmstb/common.h
index cfb5335..a7f19e0 100644
--- a/include/soc/brcmstb/common.h
+++ b/include/soc/brcmstb/common.h
@@ -12,4 +12,20 @@ 
 
 bool soc_is_brcmstb(void);
 
+#if defined(CONFIG_PCIE_BRCMSTB)
+dma_addr_t brcm_phys_to_dma(struct device *dev, phys_addr_t paddr);
+phys_addr_t brcm_dma_to_phys(struct device *dev, dma_addr_t dev_addr);
+#else
+static inline dma_addr_t brcm_phys_to_dma(struct device *dev, phys_addr_t paddr)
+{
+	return (dma_addr_t)paddr;
+}
+
+static inline phys_addr_t brcm_dma_to_phys(struct device *dev,
+					   dma_addr_t dev_addr)
+{
+	return (phys_addr_t)dev_addr;
+}
+#endif
+
 #endif /* __SOC_BRCMSTB_COMMON_H__ */