Date
21 - 22 of 22
Watchdog Spec Questions
|
Phil McCoy <pnm@...>
Several modifications to the WDT spec were proposed, and seemingly agreed in this discussion thread. Is there any plan to actually update the spec? Should I try to learn enough git-foo to generate a pull request?
Any plans to try to get this ratified in the foreseeable future? I'm currently trying to finalize a WDT implementation spec, and would like to align with the official spec as much as possible. Thanks, Phil
|
|
|
|
Greg Favor
On Tue, Apr 19, 2022 at 4:48 PM Phil McCoy <pnm@...> wrote: Several modifications to the WDT spec were proposed, and seemingly agreed in this discussion thread. Is there any plan to actually update the spec? Should I try to learn enough git-foo to generate a pull request? I wouldn't say there was true convergence on a near-final hardware WDT spec. To start with there have also been arguments to abstract any hardware away from what S/HS-mode sees and to instead have an OS/hypervisor do SBI calls to set up its watchdog and to "refresh" the watchdog, and it could receive some form of callback on a first-stage watchdog timeout. This is based on the frequency of watchdog refreshes being relatively very low and not anywhere close to a performance overhead issue. In any case, proper discussion and resolution of this issue should happen in the OS-A SEE TG (whether that results in a RISC-V hardware standard like in ARM architecture or in a RISC-V SBI standard extension). Greg
|
|
|