Re: Regression caused by commit bc976233a872, ethernet r8169 stops working after system S3

From: Kai-Heng Feng
Date: Sat Mar 31 2018 - 11:35:28 EST


Thomas Gleixner <tglx@xxxxxxxxxxxxx> wrote:

On Sat, 31 Mar 2018, Kai-Heng Feng wrote:
A user reported [1] that the Realtek ethernet r8169 stops working after S3
since v4.15-rc6. The issue still exists in Linus' tree:

[ 150.877998] IPv6: ADDRCONF(NETDEV_UP): enp1s0: link is not ready
[ 150.944101] do_IRQ: 3.37 No irq handler for vector
[ 150.944105] r8169 0000:01:00.0 enp1s0: link down
[ 150.944180] IPv6: ADDRCONF(NETDEV_UP): enp1s0: link is not ready

My desktop also has this device, which is also affected by this regression. So
I did a bisect, and this is the bad commit:
bc976233a872 genirq/msi x86/vector: Prevent reservation mode for non maskable
MSI

After this commit gets reverted, the issue is gone.

Sigh. Can you please apply the debug patch below and provide the output of

# cat /proc/interrupts

and full dmesg after resume.

Weird. I don't see this issue with this debug patch.

dmesg after resume attached.


Thanks,

tglx

8<------------------
--- a/kernel/irq/cpuhotplug.c
+++ b/kernel/irq/cpuhotplug.c
@@ -134,6 +134,10 @@ static bool migrate_one_irq(struct irq_d
brokeaff = false;
}

+ pr_info("IRQ%u: New affinity: %*pbl effective: %*pbl\n",
+ cpumask_pr_args(affinity),
+ irq_data_get_effective_affinity_mask(d));
+
if (maskchip && chip->irq_unmask)
chip->irq_unmask(d);

Attachment: dmesg
Description: Binary data