Re: [PATCH net-next] net: broadcom: use ethtool string helpers

[Rosen Penev]

On Tue, Oct 29, 2024 at 4:55 PM Jacob Keller <jacob.e.keller@xxxxxxxxx> wrote:
>
>
>
> On 10/29/2024 4:43 PM, Rosen Penev wrote:
> > On Tue, Oct 29, 2024 at 4:03 PM Jakub Kicinski <kuba@xxxxxxxxxx> wrote:
> >>
> >> On Tue, 22 Oct 2024 18:27:34 -0700 Rosen Penev wrote:
> >>> @@ -3220,13 +3212,13 @@ static void bnx2x_get_strings(struct net_device *dev, u32 stringset, u8 *buf)
> >>>                       start = 0;
> >>>               else
> >>>                       start = 4;
> >>> -             memcpy(buf, bnx2x_tests_str_arr + start,
> >>> -                    ETH_GSTRING_LEN * BNX2X_NUM_TESTS(bp));
> >>> +             for (i = start; i < BNX2X_NUM_TESTS(bp); i++)
> >>> +                     ethtool_puts(&buf, bnx2x_tests_str_arr[i]);
> >>
> >> I don't think this is equivalent.
> > What's wrong here?
>
> I was trying to figure that out too...
>
> I guess the memcpy does everything all at once and this does it via
> iteration...?
>
> memcpy would actually result in copying the padding between strings in
> the bnx2x_tests_str_arr, while the ethtool_puts turns into strscpy which
> doesn't pad the tail of the buffer with zeros?
I'll remove the change in the next version.

Still doesn't make much sense.
>
> >>
> >> Also, please split bnx2x to a separate patch, the other drivers in this
> >> patch IIUC are small embedded ones, the bnx2x is an "enterprise
> >> product".
> >> --
> >> pw-bot: cr
>