mbox series

[b:azure-4.15,f:azure,0/2] Keep Chrony time in sync

Message ID 20201216001223.463961-1-kelsey.skunberg@canonical.com
Headers show
Series Keep Chrony time in sync | expand

Message

Kelsey Skunberg Dec. 16, 2020, 12:12 a.m. UTC
BugLink: https://bugs.launchpad.net/bugs/1896784

SRU Justification:

[Impact]

Microsoft would like to request a change to the time keeping method in
the Azure images.

We would like to request that an Azure image uses the Azure host for
time sync by default.  This will require the use of the Chrony package.

The Chrony package would need to be installed by default in all images
and be configured to use /dev/ptp0 for timekeeping.

There is a request for two patches to be backported so Chrony time can
stay in sync before being set as the default.

Patches are already in the 5.8 kernel. Only needed for azure 5.4 and
4.15


[Fix]

backport these two patches:
b46b4a8a57c3 ("hv_utils: drain the timesync packets on onchannelcallback")
90b125f4cd26 ("hv_utils: return error if host timesysnc update is stale")

[Test Case]

Desired test cases:
 - verify Chrony time can stay in sync
 - verify caller is notified if host timesync update is stale
 - verify all timesync packets are drained on onchannelcallback to avoid
   stale packets

Microsoft holds test cases for the above. They verified a 4.15 and 5.4
test kernels with the above patches backported to work as expected.

[Regression Potential]

Caller may get an error now if host timesync messages were not processed
by the guest. This wasn't getting reported to the caller before and may
appear to be a regression if it occurs. 

Regression potential may also lay in still incorrectly managing timesync
packets on onchannelcallback.

Vineeth Pillai (2):
  hv_utils: return error if host timesysnc update is stale
  hv_utils: drain the timesync packets on onchannelcallback

 drivers/hv/hv_util.c | 64 ++++++++++++++++++++++++++++++++++----------
 1 file changed, 50 insertions(+), 14 deletions(-)

Comments

Kleber Sacilotto de Souza Dec. 16, 2020, 9:11 a.m. UTC | #1
On 16.12.20 01:12, Kelsey Skunberg wrote:
> BugLink: https://bugs.launchpad.net/bugs/1896784
> 
> SRU Justification:
> 
> [Impact]
> 
> Microsoft would like to request a change to the time keeping method in
> the Azure images.
> 
> We would like to request that an Azure image uses the Azure host for
> time sync by default.  This will require the use of the Chrony package.
> 
> The Chrony package would need to be installed by default in all images
> and be configured to use /dev/ptp0 for timekeeping.
> 
> There is a request for two patches to be backported so Chrony time can
> stay in sync before being set as the default.
> 
> Patches are already in the 5.8 kernel. Only needed for azure 5.4 and
> 4.15
> 
> 
> [Fix]
> 
> backport these two patches:
> b46b4a8a57c3 ("hv_utils: drain the timesync packets on onchannelcallback")
> 90b125f4cd26 ("hv_utils: return error if host timesysnc update is stale")
> 
> [Test Case]
> 
> Desired test cases:
>   - verify Chrony time can stay in sync
>   - verify caller is notified if host timesync update is stale
>   - verify all timesync packets are drained on onchannelcallback to avoid
>     stale packets
> 
> Microsoft holds test cases for the above. They verified a 4.15 and 5.4
> test kernels with the above patches backported to work as expected.
> 
> [Regression Potential]
> 
> Caller may get an error now if host timesync messages were not processed
> by the guest. This wasn't getting reported to the caller before and may
> appear to be a regression if it occurs.
> 
> Regression potential may also lay in still incorrectly managing timesync
> packets on onchannelcallback.
> 
> Vineeth Pillai (2):
>    hv_utils: return error if host timesysnc update is stale
>    hv_utils: drain the timesync packets on onchannelcallback
> 
>   drivers/hv/hv_util.c | 64 ++++++++++++++++++++++++++++++++++----------
>   1 file changed, 50 insertions(+), 14 deletions(-)
> 

The backports look good to me.

Acked-by: Kleber Sacilotto de Souza <kleber.souza@canonical.com>
Stefan Bader Dec. 17, 2020, 7:39 a.m. UTC | #2
On 16.12.20 01:12, Kelsey Skunberg wrote:
> BugLink: https://bugs.launchpad.net/bugs/1896784
> 
> SRU Justification:
> 
> [Impact]
> 
> Microsoft would like to request a change to the time keeping method in
> the Azure images.
> 
> We would like to request that an Azure image uses the Azure host for
> time sync by default.  This will require the use of the Chrony package.
> 
> The Chrony package would need to be installed by default in all images
> and be configured to use /dev/ptp0 for timekeeping.
> 
> There is a request for two patches to be backported so Chrony time can
> stay in sync before being set as the default.
> 
> Patches are already in the 5.8 kernel. Only needed for azure 5.4 and
> 4.15
> 
> 
> [Fix]
> 
> backport these two patches:
> b46b4a8a57c3 ("hv_utils: drain the timesync packets on onchannelcallback")
> 90b125f4cd26 ("hv_utils: return error if host timesysnc update is stale")
> 
> [Test Case]
> 
> Desired test cases:
>  - verify Chrony time can stay in sync
>  - verify caller is notified if host timesync update is stale
>  - verify all timesync packets are drained on onchannelcallback to avoid
>    stale packets
> 
> Microsoft holds test cases for the above. They verified a 4.15 and 5.4
> test kernels with the above patches backported to work as expected.
> 
> [Regression Potential]
> 
> Caller may get an error now if host timesync messages were not processed
> by the guest. This wasn't getting reported to the caller before and may
> appear to be a regression if it occurs. 
> 
> Regression potential may also lay in still incorrectly managing timesync
> packets on onchannelcallback.
> 
> Vineeth Pillai (2):
>   hv_utils: return error if host timesysnc update is stale
>   hv_utils: drain the timesync packets on onchannelcallback
> 
>  drivers/hv/hv_util.c | 64 ++++++++++++++++++++++++++++++++++----------
>  1 file changed, 50 insertions(+), 14 deletions(-)
> 
Acked-by: Stefan Bader <stefan.bader@canonical.com>
Ian May Dec. 17, 2020, 8:32 p.m. UTC | #3
Applied to:
  bionic/linux-azure-4.15
  focal/linux-azure

Thanks,
Ian

On 2020-12-15 17:12:21 , Kelsey Skunberg wrote:
> BugLink: https://bugs.launchpad.net/bugs/1896784
> 
> SRU Justification:
> 
> [Impact]
> 
> Microsoft would like to request a change to the time keeping method in
> the Azure images.
> 
> We would like to request that an Azure image uses the Azure host for
> time sync by default.  This will require the use of the Chrony package.
> 
> The Chrony package would need to be installed by default in all images
> and be configured to use /dev/ptp0 for timekeeping.
> 
> There is a request for two patches to be backported so Chrony time can
> stay in sync before being set as the default.
> 
> Patches are already in the 5.8 kernel. Only needed for azure 5.4 and
> 4.15
> 
> 
> [Fix]
> 
> backport these two patches:
> b46b4a8a57c3 ("hv_utils: drain the timesync packets on onchannelcallback")
> 90b125f4cd26 ("hv_utils: return error if host timesysnc update is stale")
> 
> [Test Case]
> 
> Desired test cases:
>  - verify Chrony time can stay in sync
>  - verify caller is notified if host timesync update is stale
>  - verify all timesync packets are drained on onchannelcallback to avoid
>    stale packets
> 
> Microsoft holds test cases for the above. They verified a 4.15 and 5.4
> test kernels with the above patches backported to work as expected.
> 
> [Regression Potential]
> 
> Caller may get an error now if host timesync messages were not processed
> by the guest. This wasn't getting reported to the caller before and may
> appear to be a regression if it occurs. 
> 
> Regression potential may also lay in still incorrectly managing timesync
> packets on onchannelcallback.
> 
> Vineeth Pillai (2):
>   hv_utils: return error if host timesysnc update is stale
>   hv_utils: drain the timesync packets on onchannelcallback
> 
>  drivers/hv/hv_util.c | 64 ++++++++++++++++++++++++++++++++++----------
>  1 file changed, 50 insertions(+), 14 deletions(-)
> 
> -- 
> 2.25.1
> 
> 
> -- 
> kernel-team mailing list
> kernel-team@lists.ubuntu.com
> https://lists.ubuntu.com/mailman/listinfo/kernel-team