On Thu, Aug 13, 2026 at 6:27 AM Dan Cross <[email protected]> wrote:
> On Thu, Aug 13, 2026 at 4:15 AM ron minnich <[email protected]> wrote: > > you can not disable SMM no matter how early in the boot process you get. > > > > The memory it runs from is walled off by registers you can not set. > > It is enabled by write-once (sticky) bits. > > The interrupts vector to firmware. > > > > Nope, you can't just disable that code at boot time. It's been enabled > long before. > > Well, you kinda sorta maybe can, but only if you can intercede on the > boot process early enough. > When I say 'boot', in the plan 9 context, I mean at 9load time, which I suspect is what the OP meant by boot. At that point, as far as disabling SMM goes, there are no options. Yes, sure, if you get in early enough, by changing what's in SPI (FLASH back in the day), you can disable SMM. QED starting 1999, on several 10s of millions of $$$ supercomputers we built at Los Alamos. QED what Oxide has done. But very, very few people or organizations have the money, knowledge, leverage, and will to get there. In any event, for most people, changing firmware on x86 is not the option it was in 1999. Even in 1999, I had to spend about $100K on hardware just to debug. Nowadays, on many servers, if you change one bit of the firmware, the on-board infra will note that fact, and restore it to its original state -- and, if you've enabled that option, phone home and report a problem. Most customers love that, it's a big selling point. The amount of firmware-level work you can do on x86 is nothing like it used to be. Hence, for most people, disabling SMM is not an option. ron ------------------------------------------ 9fans: 9fans Permalink: https://9fans.topicbox.com/groups/9fans/T496d301e510c8cf1-M73c59d2da101e9ee00ff82e6 Delivery options: https://9fans.topicbox.com/groups/9fans/subscription
