Intel I217LM defaults to legacy interrupts

For Developers
1

I have an X86 board with an Intel® I217LM gig-e onborad.

During boot the logs state:

e1000e 0000:00:19.0 eth0: MSI interrupt test failed, using legacy interrupt.

Some forum reading suggests this will cause a performance impact, but there doesn't seem to be any work-around. There are also some mailing-list posts to suggest that the MSI test can produce false-negatives in certain scenarios.

Can the e1000e driver be forced to use MSI/MSI-X interrupts somehow? The docs suggest it's possible to specify a 'best case' interrupt mode, but this doesn't override the test and it will drop to Legacy if the test fails.

2

Infact it's worse than performance... if any significant load is put through this adaptor I repeatedly get these:

Sun Sep 15 13:19:48 2019 kern.err kernel: [  915.996324] e1000e 0000:00:19.0 eth0: Detected Hardware Unit Hang:
Sun Sep 15 13:19:48 2019 kern.err kernel: [  915.996324]   TDH                  <db>
Sun Sep 15 13:19:48 2019 kern.err kernel: [  915.996324]   TDT                  <f1>
Sun Sep 15 13:19:48 2019 kern.err kernel: [  915.996324]   next_to_use          <f1>
Sun Sep 15 13:19:48 2019 kern.err kernel: [  915.996324]   next_to_clean        <db>
Sun Sep 15 13:19:48 2019 kern.err kernel: [  915.996324] buffer_info[next_to_clean]:
Sun Sep 15 13:19:48 2019 kern.err kernel: [  915.996324]   time_stamp           <10000efce>
Sun Sep 15 13:19:48 2019 kern.err kernel: [  915.996324]   next_to_watch        <db>
Sun Sep 15 13:19:48 2019 kern.err kernel: [  915.996324]   jiffies              <10000f0a0>
Sun Sep 15 13:19:48 2019 kern.err kernel: [  915.996324]   next_to_watch.status <0>
Sun Sep 15 13:19:48 2019 kern.err kernel: [  915.996324] MAC Status             <80083>
Sun Sep 15 13:19:48 2019 kern.err kernel: [  915.996324] PHY Status             <796d>
Sun Sep 15 13:19:48 2019 kern.err kernel: [  915.996324] PHY 1000BASE-T Status  <3800>
Sun Sep 15 13:19:48 2019 kern.err kernel: [  915.996324] PHY Extended Status    <3000>
Sun Sep 15 13:19:48 2019 kern.err kernel: [  915.996324] PCI Status             <10>
Sun Sep 15 13:19:50 2019 kern.err kernel: [  917.996439] e1000e 0000:00:19.0 eth0: Detected Hardware Unit Hang:
Sun Sep 15 13:19:50 2019 kern.err kernel: [  917.996439]   TDH                  <db>
Sun Sep 15 13:19:50 2019 kern.err kernel: [  917.996439]   TDT                  <f1>
Sun Sep 15 13:19:50 2019 kern.err kernel: [  917.996439]   next_to_use          <f1>
Sun Sep 15 13:19:50 2019 kern.err kernel: [  917.996439]   next_to_clean        <db>
Sun Sep 15 13:19:50 2019 kern.err kernel: [  917.996439] buffer_info[next_to_clean]:
Sun Sep 15 13:19:50 2019 kern.err kernel: [  917.996439]   time_stamp           <10000efce>
Sun Sep 15 13:19:50 2019 kern.err kernel: [  917.996439]   next_to_watch        <db>
Sun Sep 15 13:19:50 2019 kern.err kernel: [  917.996439]   jiffies              <10000f168>
Sun Sep 15 13:19:50 2019 kern.err kernel: [  917.996439]   next_to_watch.status <0>
Sun Sep 15 13:19:50 2019 kern.err kernel: [  917.996439] MAC Status             <80083>
Sun Sep 15 13:19:50 2019 kern.err kernel: [  917.996439] PHY Status             <796d>
Sun Sep 15 13:19:50 2019 kern.err kernel: [  917.996439] PHY 1000BASE-T Status  <3800>
Sun Sep 15 13:19:50 2019 kern.err kernel: [  917.996439] PHY Extended Status    <3000>
Sun Sep 15 13:19:50 2019 kern.err kernel: [  917.996439] PCI Status             <10>