Alex Williamson <alex.williamson@xxxxxxxxxx> 于2019年6月5日周三 上午1:29写道： > > On Mon, 20 May 2019 21:17:24 +0800 > Hao Zheng <mowendugu@xxxxxxxxx> wrote: > > > Set the pcie_cap field before getting the config space size for > > other VFs. Otherwise, the config space size of other VFs are error > > set to 256, while the size of VF0 is 4096. > > > > Signed-off-by: Hao Zheng <mowendugu@xxxxxxxxx> > > Signed-off-by: Quan Xu <quan.xu0@xxxxxxxxx> > > --- > > drivers/pci/iov.c | 1 + > > 1 file changed, 1 insertion(+) > > > > diff --git a/drivers/pci/iov.c b/drivers/pci/iov.c > > index 3aa115e..239fad1 100644 > > --- a/drivers/pci/iov.c > > +++ b/drivers/pci/iov.c > > @@ -133,6 +133,7 @@ static void pci_read_vf_config_common(struct pci_dev *virtfn) > > pci_read_config_word(virtfn, PCI_SUBSYSTEM_ID, > > &physfn->sriov->subsystem_device); > > > > + set_pcie_port_type(virtfn); > > physfn->sriov->cfg_size = pci_cfg_space_size(virtfn); > > } > > > > This results in set_pci_port_type() being called multiple times on > VF0. Why not simply delay calling pci_read_vf_config_common() until > after pci_setup_device()? Here's the alternate approach: > > https://lore.kernel.org/linux-pci/155966918965.10361.16228304474160813310.stgit@xxxxxxxxxx/ I get it. This is a better approach. Thank you for your advice! > > Thanks, > Alex -- Best Regards!
