I've written a network driver for ethernet , How to make the ethernet to use my driver code rather using the default driver in the kernel.
I'm trying to visualise the data transfer via OSI layers by adding prink at each layer of transmission (to be displayed in dmesg).
Related
I'm using Micropython on an ESP32-S3. I'm setting up DMA from an ADC by manually configuring the appropriate registers using mem32. I need to reserve some space in the SRAM1 region of Internal RAM for the linked list of descriptors. i.e. In address range 0x3FC88000 to 0x3FCEFFFF.
Is there something like array.aray() that lets me specify a range?
I assume either drivers or some other internal Micropython plumbing under the hood is already using some of that internal RAM address space, as I see non-zero values in it. I want to make sure my code doesn't stomp on anything being used (or vice versa).
I'm studying about network sockets, but unfortunately I couldn't understand if the so-called sockets are DATA or API...
I'm more tempted to believe that a network socket is a kernel-level object that serves as an in-memory register containing data that associates a process with another process to allow communication, and that register (socket network) has data that describes the type of communication that must be performed by the kernel. Some of the information that exists in this type of record that we call a network socket is an IP address, port number, an identifier and etc... So I believe that the kernel uses this dataset (socket) to make the connection.
Note that my definition is different from saying that "network sockets are an API", because in my view they are more for data than for a programming interface instance.
I would like to know if my definition makes sense or am I misunderstanding this concept?
I'm not sure if i understand well your question.
a network socket is more related to IPC (Inter-Process Communications) mechanisms.what is an IPC? (wikipedia). So it rely on the same principle than pipes, named pipes, shared memory etc... but act differently.
The way it is implemented is very dependant of the OS you are looking at. It mainly rely on kernel functions (C programming language for Linux -> see here) involving dedicated data spaces and buffers.
The data space and buffer allocation is made in the same way than when you declare a string=helloworld (or more complex structures) when writing an executable in C.
Modern operating systems does provides upper-level interfaces for manipulating those functions (i.e opening a socket). You may call it API, but the rest of the world is calling it: SYSTEM CALLS
Hope it helps. :)
EDIT (On linux systems):
In userland (ring 3), when process open a socket, it makes calls to linux API for network/sockets, and this API (system calls), rely on kernel functions (ring 0).
Those kernel functions are C classes, storing related data in a reserved memory (as C objects).
So for userland: the sockets are API.
for kernel: sockets are DATA structures.
there is also executable code coming from static functions in the TCP/IP stack and drivers executed by the kernel point of view...and relying on the data structures created from the API system calls (from the userland)
Is it clearer? ahah ;).
I am working on a project that requires me to use 5 TLC7524 DACs. They will all be receiving data from an 8 bit bus. For data transfer, the process is CS pin, WR pin, send Data. I was wondering if I could get away with grounding the CS pin on all of them and control them with just the WR pin.
Link to datasheet for TLC7524 -
http://www.ti.com/lit/ds/slas061d/slas061d.pdf
Usual way is select the DAC by /CS and the /WR signal is common for all of them.
In this way, you can connect it on FMC and create the /CS for each chip from the address (FMC is available on Nucleo144 boards, but not on Nucleo64 ones).
Also selected chips may have much higher current consumption (as they are active). For example SRAM chips are using much more current when enabled by /CS.
I am new to PCI express, I want to read/write into PCI Express configuration space via MMIO addresses. I know how port mapped IO read/write into PCI express config space via 0xCFC and 0xCF8 port addresses(on x86). I also wrote a sample linux kernel module to read pci config space via port mapped io which worked fine. I want to do the same via MMIO/MMCFG access.
I also did a search around but could not find a convincing answer. I am looking for the details and also some code sample to understand it better.
Any help is appreciated.
Hardware
The base address of the MMIO area for the configuration space of each PCIe devices in a PCI segment group is given in the ACPI table MCFG.
The MCFG table lists, for each PCI segment group, the first and last (inclusive) bus number of the PCI segment group and the base address of the extended configuration space.
The MCFG table is setup by the BIOS/UEFI based upon the value of the PCIEXBAR (for my processor is at offset 60h) in the Host Bridge/DRAM registers device located at 00:00.0.
This is the usual address, the device is integrated in the processor socket since the Nehalem architecture and never changed address.
You can google your processor generation datasheet to get the correct device address and register offset.
Also note that not all of the 256 MiB area may be mapped, my processor allows 256/128/64 MiB mapping with 128 MiB being the one selected by the BIOS/UEFI.
Linux
I don't know how to correctly handle this in Linux, there are the pci_{read|write}_config_XXX function function that seems to use the PCIe extended config space.
So accessing the config space should be very easy.
Alternatively, the pci_mcfg_lookup will give the physical address of extended configuration space for a PCI segment group and a bus range (you should be able to make it work by defining a resource structure with only the start and end fields set to the bus number).
In case you wanted a lower level approach.
Finally, you could get the address of the MCFG table and (re)parse it yourself - I don't know how to get such an address in Linux, exactly.
There is a acpi_tb_find_table where you can pass the signature of the table and null oem and table ids to get a table index.
At line 114 of the same file there is a piece of code that access a table by index, you can use it as documentation.
You probably have to import one or more symbols from the ACPI module.
Consider the following Ladder Program that checks if a connection is enabled (A202.00) then send a message from the PLC to the PC.
The documentation (Omron CX-Programmer) has a severe lack of explanation of the program convention. What I do not understand is:
To send a message from a node to a node. I should need to specify the receiver ID. It seems the function block does not have an option where I can insert an IP address. Am I supposed to MOV an IP address to a DM address (D300) then use it? If that's the case how (IP address has dots in between 4 bytes..)?
Can someone please explain what is S (First source word), D (First destination word) and C (First control word). Aren't they just memory address? E.g. sending content of a memory adress to another memory address?
[EDIT]
What am I trying to do?
I am trying to interface a measuring gauge (controlled through Ethernet by PC/C# application) to a robotic system (no RS232 or serial, no TCP/IP, only has the simplest I/O points) with an Omron PLC. When gauge completes a measurement, the C# app sends a command to the Omron PLC which, according to the command received, switch ON or OFF an output which triggers a voltage flow to the robot's I/O port.
Should I use FINS? What functions/protocol from the PLC I need to know to do this? I do not know so I am testing every function from the documentation. So far, zero progress.
1) All addressing information is encapsulated in the five control words (C -> C+4). C- "First Control Word" is the pointer to the first word in this table of five words you must have stored somewhere in your PLC to set up the communication.
2) First source word points to the first word in your PLC you wish to send. First destination word points to the first address in the PLC/device you wish to send to. In the example , the first control word specifies that 10 words should be sent. You point to the first one and it will send that one plus the next nine addresses as well.
To do this you have to use FINS communication - the PC stores a memory structure similar to the PLCs (CIO, DM, etc) called Event Memory and these are the addresses in the PC you are pointing to. The PC gets a FINS node number and address just like a PLC would - no IP addresses are involved. (see : FINS Manual) FINS is old, however, and has been superceded by things like Sysmac Gateway.
There are much better ways of communicating between PLC/PC, however, depending on what you are trying to do. Are you trying to write an HMI? If so, what language are you using?
Edit :
If you're using C#, I highly recommend you look into Sysmac Gateway and CX-Compolet. This is probably the most flexible, simple, and extensible way to get .NET working with Omron PLCs. If it is at all possible, however, a better way might even be to have the measurement unit communicate directly with the PLC via hardware I/O (relays, DIO, etc).
CX-Compolet, Sysmac Gateway link:
http://www.ia.omron.com/product/family/63/index_l_u.html