|
| ▲ | gkbrk 8 days ago | parent | next [-] |
| > You can physically do it with a microcode update. It's also anti-consumer that CPU vendors don't let customers who own the CPU perform whatever updates they want because they don't give out signing keys. |
| |
| ▲ | charcircuit 8 days ago | parent [-] | | If malware could install microcode it could break the security of the system. There is more consumer benefit than harm by locking it down to trusted updates. | | |
| ▲ | EvanAnderson 8 days ago | parent | next [-] | | The security model could allow the end user to install keys for the root of trust for the CPU, much like how UEFI Secure Boot allows you to install your own keys. That CPUs don't have this functionality may not be purposefully anti-consumer (and just laziness), but the net effect is anti-consumer. As it stands, besides preventing the user from making modifications to CPU functionality, the user is also forced to "trust" updates that might be created for specific anti-consumer purposes (say, compelled by government security services). | |
| ▲ | g-b-r 8 days ago | parent | prev [-] | | Trusted, sure |
|
|
|
| ▲ | cesarb 8 days ago | parent | prev | next [-] |
| > > Adding cpu instructions is something that you can't physically do > You can physically do it with a microcode update. Do these ARM CPUs even have microcode? Unlike on x86 CPUs where there are some very complex instructions which have to be microcoded, on ARM all instructions are simple enough that their decoding into micro-operations can be completely hard-coded in the decoder logic. |
| |
| ▲ | charcircuit 8 days ago | parent [-] | | Yes, it's too risky to make CPUs without microcode. Being able to fix bugs, or at least disable things so you don't have a complete paperweight is still important even for ARM. | | |
| ▲ | cesarb 8 days ago | parent [-] | | Disabling things can be done through "chicken bits" on configuration registers, no microcode necessary. Do you know of any ARM cores used on smarphones which actually have updatable microcode? I've never heard of any. All errata fixes I've seen are of the "set this bit in a specific register" kind. |
|
|
|
| ▲ | stavros 8 days ago | parent | prev [-] |
| I disagree. If they have to go out of their way to remove functionality the previous phones had, that's anti-consumer. |
| |
| ▲ | charcircuit 8 days ago | parent [-] | | It being your own device and removing a feature being anticonsumer can both be true. Every feature comes with trademarks off from the company providing them. It's up to consumers to validate products by buying them if they think the features offered is worth the price. If removing this feature doesn't hurt the sales of the device this feature may be more trouble than it's worth for them to provide. |
|
