怎么使用AXI CDMA从OCM/DDR写入BRAM

我正在尝试在zc706板上获得一个基本的示例应用程序，以显示它是否值得使用Zynq用于我们的应用程序，但是我无法获得标准功能。
我正在寻找一个旨在与FMC卡接口的参考硬件设计，我正在尝试做的是添加一个可以从Zynq上运行的Linux控制的额外IP。
我要做的就是将一些配置数据写入PL中的BRAM。
要做到这一点，根据我在其他地方得到的建议，我提出了一个axi_cdma。
所以CDMA有一个S_AXI_LITE端口，连接到AXI_INTERCONNECT上的M08_AXI，然后转到PS-7上的GP0。
它还有一个M_AXI端口，它连接到AXI_INTERCONNECT的S00_AXI。
该互连具有M00_AXI，它连接到AXI_BRAM控制器。
它还有M01_AXI，它转到S_AXI_HP3。
AXI_BRAM控制器进入BLOCK MEMORY GENERATOR。
我想要做的就是在Linux中向DMA引擎提交一个事务，并将我提交的缓冲区中的数据写入块存储器生成器创建的BRAM中。
据我所知，我上面描述的设置应该允许这样做。
在我的地址编辑器中，在sys_ps7下，有一个AXI_CDMA_0条目，其偏移地址为0x43C0_0000，范围为64K。
在AXI_CDMA_0（sys_ps7的单独条目）下，有两个条目AXI_BRAM_CTRL_0，偏移量为0xC000_0000，范围为4K，sys_ps7，偏移量为0x0000_0000，范围为1G。
现在HSI不适合我，所以我不得不尝试自己填写一个条目。
我做了：
axibram @ 43C00000 {
＃address-cells =;
＃size-cells =;
#dma-cells =;
compatible =“cmp，axi-bram-ctrl-1.00”;
reg =;
中断=;
interrupt-parent =;
dma-channel @ 43C00000 {
buswidth =;
};
};
我不确定这是否正确，但似乎有些错误，因为我觉得它不起作用。
我看到的问题是
1. DMA事务每次都超时。
但是，我在测试中输入了一个读取命令，看起来我正在尝试编写的数据将被写入内存。
问题是我不知道在哪里。
我不能使用hexdump -C因为如果我使用-s用于超过2FFF0000的任何东西我会得到一个糟糕的地址失败。
但无论如何，DMA引擎从未承认传输已完成？
2.据我所知，没有数据写入BRAM。
我的第一个想法是，我可能会将错误的地址放入设备树中，因为我不知道43c00000是否是正确的使用方式。
如果有一个工具可以让我看到在执行过程中写入BRAM的数据，那会有所帮助
请帮忙，我一直试图解决这个问题



以下为原文

I'm trying to get a basic example application going on the zc706 board to show whether or not its worth using the Zynq for our application, but I'm having trouble getting standard functionality going. I am looking at a reference hardware design intended to interface with an FMC card, and what I'm trying to do is add an additional IP that can be controlled from Linux running on the Zynq.

All I'm trying to do is write some configuration data to a BRAM in PL. To do this, based on the recommendations I've gotten elsewhere, I put in an axi_cdma.

So the CDMA has an S_AXI_LITE port that goes to M08_AXI on the AXI_INTERCONNECT which goes to GP0 on the PS-7.

It also has a M_AXI port which goes to S00_AXI of an AXI_INTERCONNECT. That interconnect has M00_AXI, which goes to an AXI_BRAM CONTROLLER. It also has M01_AXI, which goes to S_AXI_HP3.

The AXI_BRAM CONTROLLER goes to a BLOCK MEMORY GENERATOR.

What I want to be able to do with this is submit a transaction to the DMA engine in Linux, and have the data in the buffer that I submit be written to the BRAM created by the block memory generator. As far as I understand it, the setup I described above should allow this.

In my Address Editor, under sys_ps7, there is an AXI_CDMA_0 entry, which has an offset address of 0x43C0_0000 and range of 64K.

under AXI_CDMA_0, a separate entry from sys_ps7, there are two entries,  AXI_BRAM_CTRL_0, with an offset of 0xC000_0000 and a range of 4K, and sys_ps7, with an offset of 0x0000_0000 and a range of 1G.

Now HSI isn't working for me, so I had to try making an entry myself. I did:

axibram@43C00000{
    #address-cells = <0x1>;
    #size-cells = <0x1>;
    #dma-cells = <0x1>;
    compatible = "cmp,axi-bram-ctrl-1.00";
    reg = <0x43C00000 0x1000>;
    interrupts = <0x0 0x37 0x4>;
    interrupt-parent = <0x1>;

    dma-channel@43C00000 {
        buswidth = <0x20>;
    };
};

Which I am not sure if that is correct or not, but something seems to be wrong as I don't think it is working.

The problem I am seeing is that

1. The DMA transaction times out every time. However, I put in a read comand in my test, and it looks like the data I am trying to write DOES get written into memory somewhere. The problem is I don't know where. I can't use hexdump -C because if I use a -s for anything past 2FFF0000 I get a bad address failure. But regardless, DMA engine never acknowledges that a transfer is finished?

2. As far as I can tell there is no data being written to the BRAM. My first thought is that maybe I put the wrong address into the device tree, becasue I dont know if 43c00000 was the right one to use. It would help if there was a tool that would actually let me see what the data written in the BRAM was during execution

Please help, I've been trying to figure this out for a while