kernel: add driver support for Cisco UCS platforms

[foersleo]

Enable networking and storage drivers for Cisco UCS platforms for
bare metal variants.

Signed-off-by: Leonard Foerster <[email protected]>

Issue number: #2265

Description of changes:

This commit enabled all the different *NIC drivers that Cisco provides for their UCS platforms, specifically:

• ENIC: CONFIG_ENIC - Cisco VIC Ethernet NIC Support
• USNIC: CONFIG_INFINIBAND_USNIC - Verbs support for Cisco VIC
• FNIC: CONFIG_FCOE_FNIC - Cisco FNIC driver
• SNIC: CONFIG_SCSI_SNIC - Cisco SNIC driver

Testing done:

We unfortunately can not test these changes on a UCS machine given availability of the hardware. Checking the diff for the resulting configs shows the enablement we expect:

kernel-5.10:

--- config-bottlerocket-x86_64-kernel-5.10-5.10.118-1.x86_64_develop	2022-07-11 09:50:28.000000000 +0000
+++ config-bottlerocket-x86_64-kernel-5.10-5.10.118-1.x86_64_cisco	2022-07-11 09:24:51.000000000 +0000
@@ -2031,8 +2031,9 @@
 CONFIG_LIBFC=m
 CONFIG_LIBFCOE=m
 # CONFIG_FCOE is not set
-# CONFIG_FCOE_FNIC is not set
-# CONFIG_SCSI_SNIC is not set
+CONFIG_FCOE_FNIC=m
+CONFIG_SCSI_SNIC=y
+# CONFIG_SCSI_SNIC_DEBUG_FS is not set
 # CONFIG_SCSI_DMX3191D is not set
 # CONFIG_SCSI_FDOMAIN_PCI is not set
 # CONFIG_SCSI_GDTH is not set
@@ -2312,7 +2313,8 @@
 # CONFIG_MACB is not set
 # CONFIG_NET_VENDOR_CAVIUM is not set
 # CONFIG_NET_VENDOR_CHELSIO is not set
-# CONFIG_NET_VENDOR_CISCO is not set
+CONFIG_NET_VENDOR_CISCO=y
+CONFIG_ENIC=m
 # CONFIG_NET_VENDOR_CORTINA is not set
 # CONFIG_CX_ECAT is not set
 # CONFIG_DNET is not set
@@ -3981,7 +3983,7 @@
 CONFIG_MLX5_INFINIBAND=m
 # CONFIG_INFINIBAND_OCRDMA is not set
 # CONFIG_INFINIBAND_VMWARE_PVRDMA is not set
-# CONFIG_INFINIBAND_USNIC is not set
+CONFIG_INFINIBAND_USNIC=m
 # CONFIG_INFINIBAND_BNXT_RE is not set
 CONFIG_INFINIBAND_QEDR=m
 # CONFIG_INFINIBAND_RDMAVT is not set

kernel-5.15:

--- config-bottlerocket-x86_64-kernel-5.15-5.15.43-1.x86_64_develop	2022-07-11 09:56:21.000000000 +0000
+++ config-bottlerocket-x86_64-kernel-5.15-5.15.43-1.x86_64_cisco	2022-07-11 09:35:45.000000000 +0000
@@ -2086,8 +2086,9 @@
 CONFIG_LIBFC=m
 CONFIG_LIBFCOE=m
 # CONFIG_FCOE is not set
-# CONFIG_FCOE_FNIC is not set
-# CONFIG_SCSI_SNIC is not set
+CONFIG_FCOE_FNIC=m
+CONFIG_SCSI_SNIC=y
+# CONFIG_SCSI_SNIC_DEBUG_FS is not set
 # CONFIG_SCSI_DMX3191D is not set
 # CONFIG_SCSI_FDOMAIN_PCI is not set
 # CONFIG_SCSI_ISCI is not set
@@ -2361,7 +2362,8 @@
 # CONFIG_MACB is not set
 # CONFIG_NET_VENDOR_CAVIUM is not set
 # CONFIG_NET_VENDOR_CHELSIO is not set
-# CONFIG_NET_VENDOR_CISCO is not set
+CONFIG_NET_VENDOR_CISCO=y
+CONFIG_ENIC=m
 # CONFIG_NET_VENDOR_CORTINA is not set
 # CONFIG_DNET is not set
 # CONFIG_NET_VENDOR_DEC is not set
@@ -4079,7 +4081,7 @@
 CONFIG_MLX5_INFINIBAND=m
 # CONFIG_INFINIBAND_OCRDMA is not set
 # CONFIG_INFINIBAND_VMWARE_PVRDMA is not set
-# CONFIG_INFINIBAND_USNIC is not set
+CONFIG_INFINIBAND_USNIC=m
 # CONFIG_INFINIBAND_BNXT_RE is not set
 CONFIG_INFINIBAND_QEDR=m
 # CONFIG_INFINIBAND_RDMAVT is not set

Terms of contribution:

By submitting this pull request, I agree that this contribution is dual-licensed under the terms of both the Apache License, version 2.0, and the MIT license.

[markusboehme]

Building libxcrypt for metal-k8s-1.23, x86_64, false, false failed with checking for archiver @FILE support... unexpected EOF. I triggered a rerun for now since this cannot be related to the changes. I have seen the "unexpected EOF" error before (particularly remember other cases during configure). Has anyone already looked into the cause of those failures?

[bcressey]

I triggered a rerun for now since this cannot be related to the changes. I have seen the "unexpected EOF" error before (particularly remember other cases during configure). Has anyone already looked into the cause of those failures?

From what I can tell, this happens if a CI job gets scheduled onto a worker that's already finished a job and ends up getting shut down. We're still using runners with the --once flag which has known failure modes like this. Once we switch to ephemeral runners this problem should go away.

[zmrow]

🍔