On Wed, 2022-11-02 at 09:36 +0000, Christophe Leroy wrote: > Le 25/10/2022 à 06:44, Benjamin Gray a écrit : > > Detect and abort __do_patch_instruction() when there is no > > text_poke_area, > > which implies there is no patching address. This allows > > patch_instruction() > > to fail gracefully and let the caller decide what to do, as opposed > > to > > the current behaviour of kernel panicking when the null pointer is > > dereferenced. > > Is there any reason at all to have no text_poke_area ? > > If that's the boot CPU, then it means we are really early in the boot > process and will use raw_patch_instruction() directly. Or it means we > don't have enough memory for the boot CPU, in which case you are > going > to have so many problems that it is not worth any effort. > > If it is not the boot CPU, isn't there a way to not add a CPU for > which > the allocation has failed ? If text_area_cpu_up() returns an error, > isn't the CPU deactivated ?
Right, I hadn't seen that the CPU isn't onlined if the startup fails. That would make the check redundant then.
