Re: [PATCH V3] pci: exynos: split into two parts such as Synopsys part and Exynos part

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

 



On 7/23/2013 6:44 AM, Jingoo Han wrote:
+           if (restype == IORESOURCE_MEM) {
> >+                   of_pci_range_to_resource(&range, np, &pp->mem);
> >+                   pp->mem.name = "MEM";
> >+                   pp->config.mem_size = resource_size(&pp->mem);
> >+                   pp->config.mem_bus_addr = range.pci_addr;
> >+           }
> >+           if (restype == 0) {
> >+                   of_pci_range_to_resource(&range, np, &pp->cfg);
> >+                   pp->config.cfg0_size = resource_size(&pp->cfg)/2;
> >+                   pp->config.cfg1_size = resource_size(&pp->cfg)/2;
> >+           }
> >+   }
> >+
> >+   pp->dbi_base = devm_ioremap(pp->dev, pp->cfg.start,
> >+                           resource_size(&pp->cfg));
>
>Why is configuraion space divided into two?
Sorry, I don't know the exact reason.:(
Pratyush Anand may know about this.
Pratyush Anand, could you answer the question?


CfgRd1 and CfgWr1 transactions are needed when an EP is not directly connected to RC, rather connected through a bridge. For more detail see PCIe specs. Now, if we wish to generate a CfgRd/Wr1 transaction using SNPS controller then we can not use same programmed viewport as that for CfgRd/Wr0. However, a viewport can be reprogrammed using same physical cfg address to generate either cfg0 or cfg1. So, in that case we can do with only one cfg area and no reason to divide it into two. Current code assumes that there are only 4 viewports available (2 in outbound and 2 in inbound direction) and it always programs viewport dynamically for cfg0/1 as per need. But same can not be true for all implementations. There can be a case where hardware has sufficient number of viewports and software does not need to reprogram it dynamically. In that situation different physical memory area for each type of TLP would make the execution faster.

Also, if you find some problems, please let me know.


>Why should it be same as dbi_base?
>AFAIK, jacinto6 has a dedicated configuration/io/memory space that is entirely
>different from dbi_base.
According to the Synopsys designware PCIe datasheet,
in chapter 5.1.1 Register Space Layout,
'Port Logic Registers' are placed between (config space 0x0 + 0x700)
and (config space 0x0 + 0xFFF).
'dbi_base' is used for reading/writing 'Port Logic Registers'.
Exynos are using 'dbi_base' like this. Thus, the base addresses of
both 'dbi_base' and configuration/io/memory space are same.

Yes address space are same, but at any moment they work either as dbi space or as configuration/io/memory space depending on the dbi_enable bit of application register. Similar functionality was there in one older SPEAr implementation.


Just now, I looked at Spear PCIe driver.
However, in the case of Spear, the base address of configuration/io/memory
space is defined as 0x80000000. The base address of 'Port Logic Registers'
is defined as 0xb1000000.
I think that the situation of 'jacinto6' is similar with Spear, right?

In SPEAr1340 PCIe memory layout is as follows:

DBI base (0xB1000000 to 0xB1001FFF): This space is used to read/write own PCI Configuration Header, Capability and Port Logic(PL) Registers.

ELBI space (0xB1002000 to 0xB17FFFFF): These are completely SPEAr specific application registers.

configuration/io/memory space(0x80000000 to 0x8FFFFFFF): These can be used in viewport programming to generate different type of outbound transaction.

Regards
Pratyush



Then, I will move pp->dbi_base mapping code from pcie-designware.c
to pci-exynos.c.


--
To unsubscribe from this list: send the line "unsubscribe linux-pci" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html




[Index of Archives]     [DMA Engine]     [Linux Coverity]     [Linux USB]     [Video for Linux]     [Linux Audio Users]     [Yosemite News]     [Linux Kernel]     [Linux SCSI]     [Greybus]

  Powered by Linux