On 10/15/2026 2:08 AM, Anton Johansson wrote: > On 15/10/26, Pierrick Bouvier wrote: >> This series removes duplication of cpu_get_note_size and cpu_get_dump_info >> per >> target. For this, we define a new set of Dump operations for a given cpu, and >> implement it per target. >> >> Existing implementations were using first_cpu, which is what we continue to >> do. >> When we'll work on heterogeneous later, we can revisit this to either forbid >> a >> dump, or perform a dump for each target active. >> >> Based-on: [email protected] >> "[PATCH v3 00/12] multi-target: add hexagon" >> This base is only needed due to conflicts with current wip, it's not >> important >> for this series implementation. > > I have a very similar series going locally here > > https://gitlab.com/AntonJohansson/qemu/-/commits/b4/arch-dump > > where I added get_dump_info()/get_note_size() to SysemuCPUOps given > that we already have write_elf[32|64]_note() there. Any reason it would > be preferable to have a new struct of function pointers? > > Also I saw some uses of first_cpu in the write_elf[32|64]_note() > functions on i386 at least that we should eliminate. >
It didn't cross my mind to be honest, but taking a look, yes it seems like a good place to hold this. It even has monitor related content already. I'll respin current series with this. Thanks, Pierrick
