This patchset fixes use-after-free bugs which show up if you put some of the hppa pci-host devices through an "init -> deinit" lifecycle, as the device-introspect-test does.
The problem in both cases is that we were calling pci_register_root_bus() in instance_init: we should only call this in realize, as all the other callers do. These bugs show up if you run 'make check' with an ASAN enabled build; they are also likely behind the intermittent segfaults on s390 that RTH has noticed recently. thanks -- PMM Peter Maydell (2): hw/pci-host/dino: Don't call pci_register_root_bus() in init hw/pci-host/astro: Don't call pci_regsiter_root_bus() in init hw/pci-host/astro.c | 27 +++++++------- hw/pci-host/dino.c | 90 +++++++++++++++++++++------------------------ 2 files changed, 55 insertions(+), 62 deletions(-) -- 2.43.0
