On Wed, Jul 15, 2026 at 7:19 AM Michael Tokarev <[email protected]> wrote: > > On 5/26/26 00:53, Matt Turner wrote: > > Define HAVE_ELF_CORE_DUMP and target_elf_gregset_t in target_elf.h, > > mirroring the kernel's elf_gregset_t (ELF_NGREG = 66): r0-r31 > > [0..31], f0-f31 [32..63], pc [64], ps [65]. Implement > > elf_core_copy_regs() in elfload.c to populate the gregset from > > CPUAlphaState. > > > > Without this, bprm->core_dump is NULL for Alpha targets. When a > > guest signal goes unhandled, dump_core_and_abort() skips the core > > write and falls through to die_with_signal(), which re-raises the > > signal to the host. The host kernel then writes an x86-64 core file > > for the qemu-alpha process instead of an Alpha guest core. > > I'm picking this change up for currently active qemu stable series > (10.0.x and 11.0.x). While it is not a bugfix per se, but a lack > of feature which can be very useful for debugging. > > For 11.0.x, picking up this patch makes subsequent patches in this area > to apply cleanly (e.g. 7a2e863f9dd "populate AT_HWCAP from env->amask"). > > For 10.0.x, where things are quite a bit different, both mentioned > changes require manual apply, since in there, things are in different > place (in linux-user/elfload.c) and organized a bit differently > (HAVE_ELF_CORE_DUMP vs USE_ELF_CORE_DUMP). But since I picked this > one for 11.0.x, and since 10.0.x is an LTS series, I'm picking it for > 10.0.x anyway. > > The result can be seen here: > > https://gitlab.com/mjt0k/qemu/-/commit/bb9f5a3bbc2f06fba918432b40b56e5aeb2b3c76 > > Hopefully I did the back-port correctly. It would be nice to have > another view on the result.
Looks good to me. Thank you!
