On Wed, 24 Jun 2026 at 08:51, David Marchand <[email protected]> wrote: > > > diff --git a/drivers/bus/pci/pci_common.c b/drivers/bus/pci/pci_common.c > > > index fd18b8772b..791e9a7b49 100644 > > > --- a/drivers/bus/pci/pci_common.c > > > +++ b/drivers/bus/pci/pci_common.c > > > @@ -344,6 +344,10 @@ pci_cleanup(void) > > > rte_errno = errno; > > > error = -1; > > > } > > > + > > > + if (drv->drv_flags & RTE_PCI_DRV_NEED_MAPPING) > > > + rte_pci_unmap_device(dev); > > > > rte_pci_unmap_device() also invoked in some drivers remove callback, e.g. > > eth_virtio_pci_uninit() > > ifcvf_pci_remove() > > ... > > > > In this case it may doulbe-free > > $ git grep -l rte_pci_unmap_device drivers :^drivers/bus/pci | cut -d > / -f 2,3 | sort -u > net/nbl > net/virtio > net/xsc > vdpa/ifc > vdpa/nfp > vdpa/sfc > > Those drivers do not set RTE_PCI_DRV_NEED_MAPPING and have their own > logic when it comes to mapping/unmapping resources. > > $ for drv in $(git grep -l rte_pci_unmap_device drivers > :^drivers/bus/pci | cut -d / -f 2,3 | sort -u); do git grep -l > RTE_PCI_DRV_NEED_MAPPING drivers/$drv; done > drivers/vdpa/nfp/nfp_vdpa.c > > I suspect calling rte_pci_unmap_device twice will end up with some > error before a double free. > Worth a fix on vdpa/nfp regardless, I'll work on it quickly.
Just a note. On the principle, the vdpa/nfp driver has a double unmap bug on device unplug, regardless of this change of mine. -- David Marchand
