> -----Original Message-----
> From: Dexuan Cui <decui@xxxxxxxxxxxxx>
> Sent: Friday, September 4, 2020 7:55 PM
> To: wei.liu@xxxxxxxxxx; KY Srinivasan <kys@xxxxxxxxxxxxx>; Haiyang Zhang
> <haiyangz@xxxxxxxxxxxxx>; Stephen Hemminger <sthemmin@xxxxxxxxxxxxx>;
> lorenzo.pieralisi@xxxxxxx; bhelgaas@xxxxxxxxxx; linux-hyperv@xxxxxxxxxxxxxxx;
> linux-pci@xxxxxxxxxxxxxxx; linux-kernel@xxxxxxxxxxxxxxx; Michael Kelley
> <mikelley@xxxxxxxxxxxxx>
> Cc: Dexuan Cui <decui@xxxxxxxxxxxxx>; Jake Oshins <jakeo@xxxxxxxxxxxxx>
> Subject: [PATCH] PCI: hv: Fix hibernation in case interrupts are not re-created
>
> Hyper-V doesn't trap and emulate the accesses to the MSI/MSI-X registers, and we
> must use hv_compose_msi_msg() to ask Hyper-V to create the IOMMU Interrupt
> Remapping Table Entries. This is not an issue for a lot of PCI device drivers (e.g.
> NVMe driver, Mellanox NIC drivers), which destroy and re-create the interrupts
> across hibernation, so
> hv_compose_msi_msg() is called automatically. However, some other PCI device
> drivers (e.g. the Nvidia driver) may not destroy and re-create the interrupts across
> hibernation, so hv_pci_resume() has to call hv_compose_msi_msg(), otherwise the
> PCI device drivers can no longer receive MSI/MSI-X interrupts after hibernation.
>
> Fixes: ac82fc832708 ("PCI: hv: Add hibernation support")
> Cc: Jake Oshins <jakeo@xxxxxxxxxxxxx>
> Signed-off-by: Dexuan Cui <decui@xxxxxxxxxxxxx>
> ---
> drivers/pci/controller/pci-hyperv.c | 44 +++++++++++++++++++++++++++++
> 1 file changed, 44 insertions(+)
>
> diff --git a/drivers/pci/controller/pci-hyperv.c b/drivers/pci/controller/pci-hyperv.c
> index fc4c3a15e570..abefff9a20e1 100644
> --- a/drivers/pci/controller/pci-hyperv.c
> +++ b/drivers/pci/controller/pci-hyperv.c
> @@ -1211,6 +1211,21 @@ static void hv_irq_unmask(struct irq_data *data)
> pbus = pdev->bus;
> hbus = container_of(pbus->sysdata, struct hv_pcibus_device, sysdata);
>
> + if (hbus->state == hv_pcibus_removing) {
> + /*
> + * During hibernatin, when a CPU is offlined, the kernel tries
> + * to move the interrupt to the remaining CPUs that haven't
> + * been offlined yet. In this case, the below hv_do_hypercall()
> + * always fails since the vmbus channel has been closed, so we
> + * should not call the hypercall, but we still need
> + * pci_msi_unmask_irq() to reset the mask bit in desc->masked:
> + * see cpu_disable_common() -> fixup_irqs() ->
> + * irq_migrate_all_off_this_cpu() -> migrate_one_irq().
> + */
> + pci_msi_unmask_irq(data);
> + return;
> + }
> +
> spin_lock_irqsave(&hbus->retarget_msi_interrupt_lock, flags);
>
> params = &hbus->retarget_msi_interrupt_params;
> @@ -3372,6 +3387,33 @@ static int hv_pci_suspend(struct hv_device *hdev)
> return 0;
> }
>
> +static int hv_pci_restore_msi_msg(struct pci_dev *pdev, void *arg) {
> + struct msi_desc *entry;
> + struct irq_data *irq_data;
> +
> + for_each_pci_msi_entry(entry, pdev) {
> + irq_data = irq_get_irq_data(entry->irq);
> + if (WARN_ON_ONCE(!irq_data))
> + return -EINVAL;
> +
> + hv_compose_msi_msg(irq_data, &entry->msg);
> + }
> +
> + return 0;
> +}
> +
> +/*
> + * Upon resume, pci_restore_msi_state() -> ... ->
> +__pci_write_msi_msg()
> + * re-writes the MSI/MSI-X registers, but since Hyper-V doesn't trap
> +and
> + * emulate the accesses, we have to call hv_compose_msi_msg() to ask
> + * Hyper-V to re-create the IOMMU Interrupt Remapping Table Entries.
> + */
> +static void hv_pci_restore_msi_state(struct hv_pcibus_device *hbus) {
> + pci_walk_bus(hbus->pci_bus, hv_pci_restore_msi_msg, NULL); }
> +
> static int hv_pci_resume(struct hv_device *hdev) {
> struct hv_pcibus_device *hbus = hv_get_drvdata(hdev); @@ -3405,6
> +3447,8 @@ static int hv_pci_resume(struct hv_device *hdev)
>
> prepopulate_bars(hbus);
>
> + hv_pci_restore_msi_state(hbus);
> +
> hbus->state = hv_pcibus_installed;
> return 0;
> out:
> --
> 2.19.1
Reviewed-by: Jake Oshins (jakeo@xxxxxxxxxxxxx)
