Thadeu Lima de Souza Cascardo <cascardo@xxxxxxxxxx> wrote: > These fix different bugs when the probe fails. One of them is the addition > to a global list, which is not removed during the error path. That list has > been removed. > > Then, some memory leaks are fixed, which require a change in where the > workqueue is destroyed. > > Finally, the firmware completion is waited to prevent its callback from > accessing freed data. > > These were tested against an "emulated" rtl8192se. It was a changed rtl8139 > device under qemu with the rtl8192se PCI ID. Interesting. Does it mean qemu can support PCI pass-through to work with real hardware? > Memory allocation failures > were injected over 4 different places: init_sw_vars, rtl_pci_init, > rtl_init_core and ieee80211_register_hw. > For the Fixes tag of cleanup patches, I'm not sure if it should be or not. We can keep them and leave maintainers to decide taking to stable tree or not. If that happens, please carefully check the dependency of these patches.
