From: Borislav Petkov <[email protected]>
Booting v5.1-rc1 causes these new messages in dmesg here:
genirq: Threaded irq requested with handler=NULL and !ONESHOT for irq 24
pcie_bw_notification: probe of 0000:00:02.0:pcie010 failed with error -22
because requesting a threaded IRQ with NULL handler needs to be oneshot.
Otherwise a level-triggered interrupt endless loop can happen, see the
detailed explanation above the "Threaded irq... " error message in
kernel/irq/manage.c.
And PCI-MSI type interrupts are one-shot safe but not in the IOAPIC
case:
24: ... IO-APIC 28-fasteoi PCIe BW notif, PCIe BW notif, PCIe BW notif
25: ... IO-APIC 29-fasteoi PCIe BW notif, PCIe BW notif
Make the BW notification handler oneshot too.
Fixes: e8303bb7a75c ("PCI/LINK: Report degraded links via link bandwidth notification")
Suggested-by: Thomas Gleixner <[email protected]>
Signed-off-by: Borislav Petkov <[email protected]>
Cc: Bjorn Helgaas <[email protected]>
Cc: Lukas Wunner <[email protected]>
Cc: Alexandru Gagniuc <[email protected]>
Cc: [email protected]
---
drivers/pci/pcie/bw_notification.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/drivers/pci/pcie/bw_notification.c b/drivers/pci/pcie/bw_notification.c
index d2eae3b7cc0f..16e11c09f6bd 100644
--- a/drivers/pci/pcie/bw_notification.c
+++ b/drivers/pci/pcie/bw_notification.c
@@ -81,7 +81,7 @@ static int pcie_bandwidth_notification_probe(struct pcie_device *srv)
return -ENODEV;
ret = request_threaded_irq(srv->irq, NULL, pcie_bw_notification_handler,
- IRQF_SHARED, "PCIe BW notif", srv);
+ IRQF_SHARED | IRQF_ONESHOT, "PCIe BW notif", srv);
if (ret)
return ret;
--
2.21.0
Hi Borislav,
Thanks for the update. We've settled on a different fix [1], since Lukas
was not happy with IRQF_ONESHOT [2].
Alex
[1]
https://lore.kernel.org/linux-pci/[email protected]/
[2]
https://lore.kernel.org/linux-pci/[email protected]/
On 3/25/19 1:23 PM, Borislav Petkov wrote:
> From: Borislav Petkov <[email protected]>
>
> Booting v5.1-rc1 causes these new messages in dmesg here:
>
> genirq: Threaded irq requested with handler=NULL and !ONESHOT for irq 24
> pcie_bw_notification: probe of 0000:00:02.0:pcie010 failed with error -22
>
> because requesting a threaded IRQ with NULL handler needs to be oneshot.
> Otherwise a level-triggered interrupt endless loop can happen, see the
> detailed explanation above the "Threaded irq... " error message in
> kernel/irq/manage.c.
>
> And PCI-MSI type interrupts are one-shot safe but not in the IOAPIC
> case:
>
> 24: ... IO-APIC 28-fasteoi PCIe BW notif, PCIe BW notif, PCIe BW notif
> 25: ... IO-APIC 29-fasteoi PCIe BW notif, PCIe BW notif
>
> Make the BW notification handler oneshot too.
>
> Fixes: e8303bb7a75c ("PCI/LINK: Report degraded links via link bandwidth notification")
> Suggested-by: Thomas Gleixner <[email protected]>
> Signed-off-by: Borislav Petkov <[email protected]>
> Cc: Bjorn Helgaas <[email protected]>
> Cc: Lukas Wunner <[email protected]>
> Cc: Alexandru Gagniuc <[email protected]>
> Cc: [email protected]
> ---
> drivers/pci/pcie/bw_notification.c | 2 +-
> 1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/drivers/pci/pcie/bw_notification.c b/drivers/pci/pcie/bw_notification.c
> index d2eae3b7cc0f..16e11c09f6bd 100644
> --- a/drivers/pci/pcie/bw_notification.c
> +++ b/drivers/pci/pcie/bw_notification.c
> @@ -81,7 +81,7 @@ static int pcie_bandwidth_notification_probe(struct pcie_device *srv)
> return -ENODEV;
>
> ret = request_threaded_irq(srv->irq, NULL, pcie_bw_notification_handler,
> - IRQF_SHARED, "PCIe BW notif", srv);
> + IRQF_SHARED | IRQF_ONESHOT, "PCIe BW notif", srv);
> if (ret)
> return ret;
>
>