If the Root/Downstream Port supports active link reporting we can check if the link is trained before waiting for the device to respond. If the link is not trained, there is no point waiting for the whole ~60s so bail out early in that case. Signed-off-by: Mika Westerberg <mika.westerberg@xxxxxxxxxxxxxxx> --- As discussed in the email thread of the previous version here: https://lore.kernel.org/linux-pci/20230404052714.51315-1-mika.westerberg@xxxxxxxxxxxxxxx/ This adds the last change on top of https://git.kernel.org/pub/scm/linux/kernel/git/pci/pci.git/log/?h=reset drivers/pci/pci.c | 16 ++++++++++++++++ 1 file changed, 16 insertions(+) diff --git a/drivers/pci/pci.c b/drivers/pci/pci.c index 0b4f3b08f780..61bf8a4b2099 100644 --- a/drivers/pci/pci.c +++ b/drivers/pci/pci.c @@ -5037,6 +5037,22 @@ int pci_bridge_wait_for_secondary_bus(struct pci_dev *dev, char *reset_type) } } + /* + * Everything above is handling the delays mandated by the PCIe r6.0 + * sec 6.6.1. + * + * If the port supports active link reporting we now check one more + * time if the link is active and if not bail out early with the + * assumption that the device is not present anymore. + */ + if (dev->link_active_reporting) { + u16 status; + + pcie_capability_read_word(dev, PCI_EXP_LNKSTA, &status); + if (!(status & PCI_EXP_LNKSTA_DLLLA)) + return -ENOTTY; + } + return pci_dev_wait(child, reset_type, PCIE_RESET_READY_POLL_MS - delay); } -- 2.39.2
