kernel: Enable virtio-scsi #202
Conversation
amshinde
force-pushed
the
enable-virtio-scsi
branch
from
35f370d to
bf2db1a
Compare
kernel/kernel-config-4.9.x
Outdated
| @@ -1,6 +1,6 @@ | |||
| # | |||
| # Automatically generated file; DO NOT EDIT. | |||
| # Linux/x86 4.9.54 Kernel Configuration | |||
| # Linux/x86 4.9.60 Kernel Configuration | |||
There was a problem hiding this comment.
This change should be on a different commit I think?
| # CONFIG_SCSI is not set | ||
| # CONFIG_SCSI_DMA is not set | ||
| CONFIG_SCSI=y | ||
| CONFIG_SCSI_DMA=y |
There was a problem hiding this comment.
Can I ask - do we know if we need the SCSI_DMA set?
I was going to ask the same of the SG, but the virtio scsi driver does have SG code, and I can see how that could be used. Having DMA for a purely virtual device feels slightly odd though (but maybe it too is tied into how SG is passed/used?).
There was a problem hiding this comment.
@grahamwhaley This is enabled by default with virtio-scsi, maybe pulled in with SG as well.
kernel/kernel-config-4.9.x
Outdated
| @@ -603,7 +603,7 @@ CONFIG_PCI_MSI_IRQ_DOMAIN=y | |||
| # CONFIG_HT_IRQ is not set | |||
| CONFIG_PCI_ATS=y | |||
| # CONFIG_PCI_IOV is not set | |||
| # CONFIG_PCI_PRI is not set | |||
| CONFIG_PCI_PRI=y | |||
There was a problem hiding this comment.
Do we know we need this? I looked it up - it allows PCI devices behind an iommu to recover from page faults - sounds a little 'fringe' for our virtio case?
There was a problem hiding this comment.
@grahamwhaley Yes you are right, we dont typically need this for the virtio case, taking this out.
kernel/kernel-config-4.9.x
Outdated
| @@ -1,6 +1,6 @@ | |||
| # | |||
| # Automatically generated file; DO NOT EDIT. | |||
| # Linux/x86 4.9.54 Kernel Configuration | |||
| # Linux/x86 4.9.60 Kernel Configuration | |||
There was a problem hiding this comment.
I agree with @jodh-intel - if you did this upgrade against a new kernel (and took a defconfig or allyes or similar), and/or you need to have the latest kernel for a feature, can you do the version upgrade as a separate commit (and possibly a separate PR?).
There was a problem hiding this comment.
We are currently using kernel 4.9.60, but our config file still contains 4.9.54!
I can put this change in a separate PR itself.
There was a problem hiding this comment.
This sounds like a bit of a "process gap".
@grahamwhaley, @jcvenegas, @erick0z - the packaging scripts run make oldconfig, but shouldn't we be committing the changes to avoid this sort of issue? Ideally, we'd build both .deb and .rpm packages, diff the two make oldconfig-updated .config files (they should be the same we'd hope!) then commit one of them as https://github.com/clearcontainers/packaging/blob/master/kernel/kernel-config-4.9.x so the comment line above is always in sync with the kernel version we're building?
kernel/kernel-config-4.9.x
Outdated
| # CONFIG_SCSI_NETLINK is not set | ||
| # CONFIG_SCSI_MQ_DEFAULT is not set |
There was a problem hiding this comment.
@amshinde can we enable mq and see if we can leverage it in the virtio case? mq typically gives us clear performance and scaling benefits.
There was a problem hiding this comment.
@mcastelino I have enabled this option
amshinde
force-pushed
the
enable-virtio-scsi
branch
from
bf2db1a to
42bd9e8
Compare
|
I agree. At the very least this should be done when the kernel is updated to a new dot release. |
|
@grahamwhaley Can you take another look? |
|
ping @grahamwhaley |
Need this config to move to virtio-scsi. Fixes clearcontainers#201 Signed-off-by: Archana Shinde <archana.m.shinde@intel.com>
amshinde
force-pushed
the
enable-virtio-scsi
branch
from
42bd9e8 to
f5af055
Compare
There was a problem hiding this comment.
lgtm
I know @egernst noted we could land this with a kernel minor upgrade - but I'm not sure how we'd co-ordinate that.
Merging...
create-repo-branch.sh: Fix a typo
Need this config to move to virtio-scsi.
Fixes #201
Signed-off-by: Archana Shinde archana.m.shinde@intel.com