Comment 0 for bug 1968749

Revision history for this message
Tim Gardner (timg-tpi) wrote :

SRU Justification

[Impact]

focal/linux-azure is missing a number of config options compared to impish/linux-azure.

From Microsoft:
We tested this kernel. The P0P1 test cases are finished. Below are we found. P2 test cases are still running. I will update you when they are finished.

1. From dmesg, we can see following lines. Could you check why it has these messages and if the messages impact the system?
[ 3.341802] squashfs: SQUASHFS error: Xattrs in filesystem, these will be ignored
[ 3.346056] unable to read xattr id index table

2. The kernel hasn't backport Michael's SRIOV performance improvement patches. The network performance is not expected.
Here are the patches:
https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?h=v5.18-rc2&id=37200078ed6aa2ac3c88a01a64996133dccfdd34
https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?h=v5.18-rc2&id=8d21732475c637c7efcdb91dc927a4c594e97898

3. It is failed to create raid 0 using below command.
# mdadm --create /dev/md0 --level 0 --raid-devices 16 /dev/sd[abcdghijklmnopqr][1-5]
mdadm: Defaulting to version 1.2 metadata
mdadm: RUN_ARRAY failed: Invalid argument

From dmesg, it has following lines:
[ 235.830712] md: personality for level 0 is not loaded!
[ 235.830733] md: md0 stopped.

I see many raid related configuration is not set. Could you check the raid related and other configuration?
# cat /boot/config-5.4.0-1076-azure | grep RAID
CONFIG_RAID_ATTRS=m
# CONFIG_BLK_DEV_3W_XXXX_RAID is not set
# CONFIG_SCSI_AACRAID is not set
# CONFIG_MEGARAID_NEWGEN is not set
# CONFIG_MEGARAID_LEGACY is not set
CONFIG_MEGARAID_SAS=y
# CONFIG_SCSI_PMCRAID is not set
# CONFIG_MD_RAID0 is not set
# CONFIG_MD_RAID1 is not set
# CONFIG_MD_RAID10 is not set
# CONFIG_MD_RAID456 is not set
# CONFIG_DM_RAID is not set
CONFIG_BCM_SBA_RAID=m
CONFIG_DMA_ENGINE_RAID=y
CONFIG_RAID6_PQ=m
CONFIG_RAID6_PQ_BENCHMARK=y

[Test Case]

Microsoft tested

[Where things could go wrong]

There should be little chance of regression given that these are config options.