Messages in this thread | | | Date | Mon, 31 Jan 2022 20:10:26 +0200 | Subject | Re: [PATCH net] net: e1000e: Recover at least in-memory copy of NVM checksum | From | "Neftin, Sasha" <> |
| |
On 1/31/2022 18:41, Thomas Bogendoerfer wrote: > On Mon, Jan 31, 2022 at 12:51:07PM +0200, Neftin, Sasha wrote: >> Hello Thomas, >> For security reasons starting from the TGL platform SPI controller will be >> locked for SW access. I've double-checked with our HW architect, not from >> SPT, from TGP. So, first, we can change the mac type e1000_pch_cnp to >> e1000_pch_tgp (as fix for initial patch) > > ok, that would fix the mentioned bug. Are you sending a patch for that ? Sure. I will send patch for this and inform you > >> Do we want (second) to allow HW initialization with the "wrong" NVM >> checksum? It could cause unexpected (HW) behavior in the future. Even if you >> will "recover" check in shadow RAM - there is no guarantee that NVM is good. > > sure. Out of curiosity why is the NVM fixup there in the first place ? It is legacy implementation (many years ago). I believe the 'main idea' was to allow SW to fix checksum when somehow it was computed wrongly. (probably recover checksum calculation bug in NVM release process) > > Thomas. > Sasha
| |