On 4/16/25 11:27, Jan Beulich wrote: > On 15.04.2025 18:54, Stewart Hildebrand wrote: >> In preparation for translating virtual PCI bus topology (where a pdev >> lookup will be needed in the vPCI I/O handlers), acquire d->pci_lock in >> the vPCI I/O handlers. > > I find it concerning that the locked regions (it's a domain-wide lock > after all) are further grown.
Hm. Right. Another approach may be to call
vpci_translate_virtual_device from within locked context of
vpci_{read,write}.
