Re: [PATCH 1/3] hexdump: Implement macro for converting large buffers

[Paolo Abeni <pabeni@redhat.com>]

On 2/17/25 4:09 PM, Nick Child wrote:
> Thank you David and Simon for testing and review!
> 
> On Sun, Feb 16, 2025 at 11:24:30AM +0000, David Laight wrote:
>>
>> I just changed the prototypes (include/linux/printk.h) to make both
>> rowsize and groupsize 'unsigned int'.
>> The same change in lib/hexdump.c + changing the local 'i, linelen, remaining'
>> to unsigned int and it all compiled.
>>
>> FWIW that hexdump code is pretty horrid (especially if groupsize != 1).
>>
> 
> Given this discussion and my own head-scratching, I think I will take a
> closer look at hex_dump_to_buffer.
> 
> I was trying to avoid editing this function due the number of callers it
> has across the kernel. But I think we can get away with keeping the
> API (but change args to uint's) and editing the body of the function
> to always iterate byte-by-byte, addding space chars where necessary. At the
> cost of a few more cycles, this will allow for more dynamic values
> for rowsize and groupsize and shorten the code up a bit. This would also
> address the "Side question" in my cover letter. Will send a v3
> regardless if I can figure that out or not.
> 
> The return value of hex_dump_to_buffer on error still irks me a bit but
> I don't think that can easily be changed.

For the new versions, please:

- respect the 24h grace period:

https://elixir.bootlin.com/linux/v6.11.8/source/Documentation/process/maintainer-netdev.rst#L15

- add the target tree in the subj prefix (in this case 'net-next')

- ensure all the patches in the series have consistent subj prefix,
otherwise patchwork will be fooled.

I think it would be better to avoid modifying hex_dump_to_buffer(), if
not necessary, and I think you could do so either changing the type used
in patch 2 or even dropping such patch.

/P