On Fri, Feb 11, 2022 at 09:07:32PM +1300, Paulo Miguel Almeida wrote:
> dev_<level> functions don't support printing hex dumps and the
> alternative available (print_hex_dump_debug) doesn't print the device
> information such as device's driver name and device name. That type of
> information which comes in handy for situations in which you can more
> than 1 device attached at the same type.
>
> this patch adds a utility function that can obtain the same result as
> print_hex_dump_debug while being able to honour all possible flags that
> one may be interested in when dynamic debug is used.
>
> Signed-off-by: Paulo Miguel Almeida <[email protected]>
> ---
> Meta-comments:
>
> the initial discussion to use print_hex_dump_debug started in this patch
> but the original idea got merged into the brach.
>
> https://lore.kernel.org/lkml/[email protected]/#t
>
> ---
> drivers/staging/pi433/rf69.c | 39 ++++++++++++++++++++++++++++--------
> 1 file changed, 31 insertions(+), 8 deletions(-)
>
> diff --git a/drivers/staging/pi433/rf69.c b/drivers/staging/pi433/rf69.c
> index 901f8db3e3ce..82d4ba24c35f 100644
> --- a/drivers/staging/pi433/rf69.c
> +++ b/drivers/staging/pi433/rf69.c
> @@ -822,9 +822,37 @@ int rf69_set_dagc(struct spi_device *spi, enum dagc dagc)
>
> /*-------------------------------------------------------------------------*/
>
> +static void rf69_dbg_hex(struct spi_device *spi, u8 *buf, unsigned int size,
> + const char *fmt, ...)
> +{
> + va_list args;
> + char textbuf[512] = {};
> + char *text = textbuf;
> + int text_pos;
> +
> + int rowsize = 16;
> + int i, linelen, remaining = size;
> +
> + va_start(args, fmt);
> + text_pos = vscnprintf(text, sizeof(textbuf), fmt, args);
> + text += text_pos;
> + va_end(args);
> +
> + for (i = 0; i < size; i += rowsize) {
> + linelen = min(remaining, rowsize);
> + remaining -= rowsize;
> +
> + hex_dump_to_buffer(buf + i, linelen, rowsize, 1,
> + text, sizeof(textbuf) - text_pos, false);
> +
> + dev_dbg(&spi->dev, "%s\n", textbuf);
> +
> + memset(text, 0, sizeof(textbuf) - text_pos);
> + }
> +}
This is a lot of additional complexity for almost no real benefit.
> +
> int rf69_read_fifo(struct spi_device *spi, u8 *buffer, unsigned int size)
> {
> - int i;
> struct spi_transfer transfer;
> u8 local_buffer[FIFO_SIZE + 1];
> int retval;
> @@ -844,9 +872,7 @@ int rf69_read_fifo(struct spi_device *spi, u8 *buffer, unsigned int size)
>
> retval = spi_sync_transfer(spi, &transfer, 1);
>
> - /* print content read from fifo for debugging purposes */
> - for (i = 0; i < size; i++)
> - dev_dbg(&spi->dev, "%d - 0x%x\n", i, local_buffer[i + 1]);
What is wrong with this simple line?
> + rf69_dbg_hex(spi, local_buffer + 1, size, "%s - ", __func__);
>
> memcpy(buffer, &local_buffer[1], size);
>
> @@ -855,7 +881,6 @@ int rf69_read_fifo(struct spi_device *spi, u8 *buffer, unsigned int size)
>
> int rf69_write_fifo(struct spi_device *spi, u8 *buffer, unsigned int size)
> {
> - int i;
> u8 local_buffer[FIFO_SIZE + 1];
>
> if (size > FIFO_SIZE) {
> @@ -867,9 +892,7 @@ int rf69_write_fifo(struct spi_device *spi, u8 *buffer, unsigned int size)
> local_buffer[0] = REG_FIFO | WRITE_BIT;
> memcpy(&local_buffer[1], buffer, size);
>
> - /* print content written from fifo for debugging purposes */
> - for (i = 0; i < size; i++)
> - dev_dbg(&spi->dev, "0x%x\n", buffer[i]);
> + rf69_dbg_hex(spi, local_buffer + 1, size, "%s - ", __func__);
Again, the original is fine here, why make this so complex?
Also, you are using local_buffer here, not buffer, why?
I think the original is just fine, no need to polish something as tiny
as a hex dump for debugging only.
thanks,
greg k-h
On Fri, Feb 11, 2022 at 09:14:40AM +0100, Greg KH wrote:
>
> This is a lot of additional complexity for almost no real benefit.
>
you're right. I will no longer pursue this approach.
> > - /* print content read from fifo for debugging purposes */
> > - for (i = 0; i < size; i++)
> > - dev_dbg(&spi->dev, "%d - 0x%x\n", i, local_buffer[i + 1]);
>
> What is wrong with this simple line?
>
to be honest, I think that 1 register per line isn't the easiest way to
read them. Given that print_hex_dump_debug existed and had this
horizontal-style priting format, I thought that it would be a better
way of visualizing the fifo data.
the only problems with print_hex_dump_debug was the absense of device
name and string format... so I saw a couple of drivers implementing
alternative hex_dump-like functions and thought that pi433 would benefit
from similar approach.
> > - /* print content written from fifo for debugging purposes */
> > - for (i = 0; i < size; i++)
> > - dev_dbg(&spi->dev, "0x%x\n", buffer[i]);
if we are keeping this format, I may need to add the register idx to
dev_dbg:
dev_dbg(&spi->dev, "%d - 0x%x\n", i, buffer[i]);
> Again, the original is fine here, why make this so complex?
[thinking out loud/brainstorm]
I do agree that, for just a single driver, having a method like that
seemed unnecessary but do you think it would be a good idea having
something like dev_dbg_hex_dump or similar?
print_hex_dump_debug has the following limitation:
1) lacks string format
2) doesn't honor dynamic debug flags (other then 'p')
3) doesn't print device driver name and device name
> > + rf69_dbg_hex(spi, local_buffer + 1, size, "%s - ", __func__);
>
> Also, you are using local_buffer here, not buffer, why?
>
That was a mistake, good catch.
> I think the original is just fine, no need to polish something as tiny
> as a hex dump for debugging only.
>
> thanks,
>
> greg k-h
thanks for taking the time to review the patch, I won't pursue this
approach anymore.
thanks,
Paulo Almeida
On Sat, 2022-02-12 at 08:39 +1300, Paulo Miguel Almeida wrote:
> On Fri, Feb 11, 2022 at 09:14:40AM +0100, Greg KH wrote:
> >
> > This is a lot of additional complexity for almost no real benefit.
> >
>
> you're right. I will no longer pursue this approach.
>
> > > - /* print content read from fifo for debugging purposes */
> > > - for (i = 0; i < size; i++)
> > > - dev_dbg(&spi->dev, "%d - 0x%x\n", i, local_buffer[i + 1]);
> >
> > What is wrong with this simple line?
> >
>
> to be honest, I think that 1 register per line isn't the easiest way to
> read them. Given that print_hex_dump_debug existed and had this
> horizontal-style priting format, I thought that it would be a better
> way of visualizing the fifo data.
>
> the only problems with print_hex_dump_debug was the absense of device
> name and string format... so I saw a couple of drivers implementing
> alternative hex_dump-like functions and thought that pi433 would benefit
> from similar approach.
>
> > > - /* print content written from fifo for debugging purposes */
> > > - for (i = 0; i < size; i++)
> > > - dev_dbg(&spi->dev, "0x%x\n", buffer[i]);
>
> if we are keeping this format, I may need to add the register idx to
> dev_dbg:
> dev_dbg(&spi->dev, "%d - 0x%x\n", i, buffer[i]);
You could use %*ph with a buffer length up to 64 bytes.
Multiple times if necessary.
Something like:
for (i = 0; i < size; i += 64)
dev_dbg(&spi->dev, "FIFO buffer: %d: %*ph\n",
i, min_t(int, 64, size - i), buffer + i);
Or for a more realistic length like 16
for (i = 0; i < size; i += 16)
dev_dbg(&spi->dev, "FIFO buffer: %d: %*ph\n",
i, min_t(int, 16, size - i), buffer + i);