On Wed, Sep 04, 2024 at 02:56:46PM +0900, Takamitsu Iwai wrote:
> > So you have confirmed with the datsheet that the write is not needed?
> >
> > As i said, this is a hardware register, not memory. Writes are not
> > always idempotent. It might be necessary to write it twice.
>
> I have checked following datasheets and I can not find that we need to write
> RDH, RDT, TDH, TDT registers twice at initialization.
>
> https://www.intel.com/content/dam/www/public/us/en/documents/datasheets/82577-gbe-phy-datasheet.pdf
> https://www.intel.com/content/www/us/en/content-details/613460/intel-82583v-gbe-controller-datasheet.html
>
> Write happened once before commit 0845d45e900c, so just out of curiosity,
> have you seen such a device?
This is just risk minimisation. I don't want e1000e to be broken
because you removed a write. I'm trying to ensure you fully understand
what you are changing, and have verified it is a safe change. I don't
have this hardware, so i cannot test it.
> My colleague, Kohei, tested the patch with a real hardware and will provide
> his
> Tested-by shortly.
Please resend the patch, adding his Tested-by: and update the commit
message to summarise this discussion. Explain how you determined this
is safe.
Thanks
Andrew
