Hi Breno,

On Mon, Jul 08, 2024 at 12:08:05PM -0700, Breno Leitao wrote:

I thought about a patch like the following (compile tested only). What
do you think?

To be honest, there are several things I don't really like about this
patch.

- I really struggled with applying it in the current format. Could you
  please post the output of git format-patch in the future?
- You addressed dpaa_set_coalesce() but not also dpaa_fq_setup()
- You misrepresented the patch content by saying you only allocate size
  for online CPUs in the commit message. But you allocate for all
  possible CPUs.
- You only kfree(needs_revert) in the error (revert_values) case, but
  not in the normal (return 0) case.
- The netdev coding style is to sort the lines with variable
  declarations in reverse order of line length (they call this "reverse
  Christmas tree"). Your patch broke that order.
- You should use kcalloc() instead of kmalloc_array() + memset()

I have prepared and tested the attached alternative patch on a board and
I am preparing to submit it myself, if you don't have any objection.

Thanks,
Vladimir