On Mon, 21 Aug 2023 at 21:44, Brian Norris <computersforpeace@xxxxxxxxx> wrote: > > On Mon, Aug 21, 2023 at 12:35 PM Robert Marko <robimarko@xxxxxxxxx> wrote: > > On Mon, 21 Aug 2023 at 21:31, Rob Herring <robh@xxxxxxxxxx> wrote: > > > Why can't you just disable SDI unconditionally when going into debug > > > mode? Is doing that when not enabled going to crash the system or > > > something? > > I asked the same, to resounding silence: > > https://lore.kernel.org/all/20200721080054.2803881-1-computersforpeace@xxxxxxxxx/ > https://lore.kernel.org/all/ZNlhSdh0qDMieTAS@localhost/ > > > Because if not disabled you will enter the Secure Debug mode even on a > > regular reboot and then you have to pull the power in order to boot again. > > Even according to QCA docs they intended for the Linux to disable SDI as > > TZ/QSEE will always enable it as part of booting. > > NB: I've never read such docs. Presumably they're internal/private to > Qualcomm and/or its direct partners? I'd love to see them. Sadly they are all behind the NDA. > > But, I think you (robinmarko) are not really answering the same > question that Rob (robh) is asking. Rob is asking why you ever *don't* > want to disable SDI. You're answering why we ever need to disable it > at all. I don't think the latter question is controversial. I understood his question differently, hence my answer. > > FWIW, your description of those docs sounds like we should > unconditionally *disable* SDI (like my first RFC above), which would > answer Rob's question, and would agree with my RFC above :) And as a > bonus, no Device Tree change would be required. Well, the thing is that I only have docs for some of the IPQ chips, and with the insane variety of SoC-s that use SCM and TZ/QSEE but completely different FW base or version something would break for sure so I would prefer to opt-in if its really required as SDI was something that was until IPQ5018 came along, always disabled by default, except for the weird Google WiFI board. Regards, Robert > > Brian
