Skip to contentSkip to conversation

How to extract opcode, addr, dummy and buffer in transfer_one_message()?

DISCUSSION Stefan Schaeckeler 2 participants 2 messages Jun 11, 2020 – Jun 12, 2020 SPI

lore
Share & tools
lore
#0 lore raw

Hello linux-spi,

I can implement mem_ops() a/o transfer_one_message().

spi-mem.c:spi_mem_exec_op() will directly pass op to mem_ops->exec_op(). There I can easily extract opcode, addr, dummy and data buffer. It would probably look roughly like

exec_op(struct spi_mem *mem, const struct spi_mem_op *op){
        dummy = addr = -1;

        opcode = op->cmd.opcode;

        addrlen=op->addr.nbytes;
        if (addrlen)
            addr = op->addr.val;

        if (op->dummy.nbytes)
            dummy = QSPI_DUMMY_CYCLE;

        len=op->data.nbytes;
        if (len)
            if (op->data.dir == SPI_MEM_DATA_IN) {
                rxbuf = op->data.buf.in;
                is_wr = false;
            } else {
                txbuf = op->data.buf.out;
                is_wr = true;
            }
}

That's great. What about transfer_one_message()?

spi-mem.c:spi_mem_exec_op() builds a message consisting of multiple transfers. If there is no address, dummy or data buffer, then they will not be included in the transfer array as transfers are created like this

        /* op is the first transfer */
        tmpbuf[0] = op->cmd.opcode;
        xfers[xferpos].tx_buf = tmpbuf;
        ...

        if (op->addr.nbytes) {
        /* add addr transfer */
        ...
        }

        if (op->dummy.nbytes) {
        /* add dummy transfer */
        ...
        }

        if (op->data.nbytes) {
        /* add data transfer */
        ...
        }

These transfers are then passed as a message consisting of 1 to 4 transfers to transfer_one_message().

In transfer_one_message(), the first transfer is the opcode. How do I now what the next transfers represent? Like in the exec_op() example above, I would like to create and initialize my variables opcode, addr, dummy and read or write buffer.

Related question: what are the disadvantages of always using mem_ops() and never transfer_one_message()?

Thanks,
Stefan

#1 lore raw

Hi Stefan,

Quoted 55 lines · Stefan Schaeckeler

I can implement mem_ops() a/o transfer_one_message().

spi-mem.c:spi_mem_exec_op() will directly pass op to mem_ops->exec_op(). There I can easily extract opcode, addr, dummy and data buffer. It would probably look roughly like

exec_op(struct spi_mem *mem, const struct spi_mem_op *op){
        dummy = addr = -1;

        opcode = op->cmd.opcode;

        addrlen=op->addr.nbytes;
        if (addrlen)
            addr = op->addr.val;

        if (op->dummy.nbytes)
            dummy = QSPI_DUMMY_CYCLE;

        len=op->data.nbytes;
        if (len)
            if (op->data.dir == SPI_MEM_DATA_IN) {
                rxbuf = op->data.buf.in;
                is_wr = false;
            } else {
                txbuf = op->data.buf.out;
                is_wr = true;
            }
}

That's great. What about transfer_one_message()?

spi-mem.c:spi_mem_exec_op() builds a message consisting of multiple transfers. If there is no address, dummy or data buffer, then they will not be included in the transfer array as transfers are created like this

        /* op is the first transfer */
        tmpbuf[0] = op->cmd.opcode;
        xfers[xferpos].tx_buf = tmpbuf;
        ...

        if (op->addr.nbytes) {
        /* add addr transfer */
        ...
        }

        if (op->dummy.nbytes) {
        /* add dummy transfer */
        ...
        }

        if (op->data.nbytes) {
        /* add data transfer */
        ...
        }
Quoted 7 lines

These transfers are then passed as a message consisting of 1 to 4 transfers to transfer_one_message().

In transfer_one_message(), the first transfer is the opcode. How do I now what the next transfers represent? Like in the exec_op() example above, I would like to create and initialize my variables opcode, addr, dummy and read or write buffer.

The meaning of the transfers depends on the protocol used to talk to the SPI slave.
The mem_ops() are intended for memory devices. transfer_one_message() (and transfer_one()) are meant for communicating with generic SPI slaves.

Quoted 2 lines

Related question: what are the disadvantages of always using mem_ops() and never transfer_one_message()?

If you do that, your driver cannot talk to SPI slaves that are not supported by spi_mem, i.e. anything except memory devices.

Gr{oetje,eeting}s,

Geert

Signature · 5 lines
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- [email protected]

In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
                                -- Linus Torvalds