Hi Eric, >-----Original Message----- >From: Eric Auger <eric.au...@redhat.com> >Sent: Wednesday, September 27, 2023 9:14 PM >Subject: Re: [PATCH v2 11/12] vfio/common: Introduce two kinds of VFIO device >lists > >Hi Zhenzhong, >On 9/26/23 13:32, Zhenzhong Duan wrote: >> In VFIO subsystem, there are different VFIO device iteration requirements. >> One requirement is to iterate all VFIO devices, the other is to iterate >> VFIO device in same container. >> >> Currently VFIO device is iterated through VFIO group list which is group >> perceivable and less efficient. >> >> Introduce two kinds of VFIO device lists, one is a global list, the other >> is per container list. With the two lists added, we can make some migration >> and reset related functions group agnostic. >> >> For example, vfio_device_list is used in below functions: >> vfio_mig_active >> vfio_reset_handler >> vfio_multiple_devices_migration_is_supported >> >> Per container list is used in below functions: >> vfio_devices_all_dirty_tracking >> vfio_devices_all_device_dirty_tracking >> vfio_devices_all_running_and_mig_active >> vfio_devices_dma_logging_stop >> vfio_devices_dma_logging_start >> vfio_devices_query_dirty_bitmap >> >> This is a prerequisite for future IOMMUFD backend support which >> has same kind of iteration requirement. >> >> vfio_group_list is preserved for some functions which honor group >> iteration, those functions are all related to legacy backend. >> >> Suggested-by: Alex Williamson <alex.william...@redhat.com> >> Signed-off-by: Zhenzhong Duan <zhenzhong.d...@intel.com> > >This may be split into 3 patches >1. creation of container->device_list >2. creation of global device list >3. addition of container field in vbasedev (which is not described in >the commit msg by the way) and looks somehow unrelated to me?
Ah, ok, I just thought adding pointer to container is trivial which is counterpart to container->device_list. And yes, I'm fine to split it into 3 patches. Thanks Zhenzhong
