Message ID | 20210311080155.47562-1-andrea.righi@canonical.com |
---|---|
Headers | show |
Series | aws: revert change to support instance types > 255 cpu cores | expand |
On 11.03.21 09:01, Andrea Righi wrote: > BugLink: https://bugs.launchpad.net/bugs/1918590 > > [Impact] > > In LP: #1913739 we applied a fix to properly detect the right amount of > cores in AWS large instance types (> 255). However, this fix seems to > work properly only with kernels > 4.4. Even if it doesn't seem to > introduce any failure or regression, it is probably just safer to revert > this commit in the Xenial kernel, and keep it on the other kernels (> 4.4). > > [Test plan] > > Boot an instance type with > 255 cpus and check the amount of cores > detected. > > [Fix] > > Revert commit: > > 561ef22a7d396289a1c5a1d18057ccdf9bb59826 ("iommu/vt-d: Gracefully handle DMAR units with no supported address widths") > > [Regression potential] > > Reverting the commit simply restores the old behavior, so we won't be > able to detect the right amount of CPUs in large instances, but we are > sure that we won't also introduce any potential failure or regressions > by restoring the previous behavior. And this is affecting only the > Xenial kernel, that is approaching to EOL. > > ChangeLog (v1 -> v2): > - create a new bug instead of re-using the old one when the original > fix was applied > > Applied to xenial:linux-aws/master. I took the ACKs from v1 since there was no difference in the content other than the bug report. Thanks. -Stefan
On 11/03/2021 05:01, Andrea Righi wrote: > BugLink: https://bugs.launchpad.net/bugs/1918590 > > [Impact] > > In LP: #1913739 we applied a fix to properly detect the right amount of > cores in AWS large instance types (> 255). However, this fix seems to > work properly only with kernels > 4.4. Even if it doesn't seem to > introduce any failure or regression, it is probably just safer to revert > this commit in the Xenial kernel, and keep it on the other kernels (> 4.4). > > [Test plan] > > Boot an instance type with > 255 cpus and check the amount of cores > detected. > > [Fix] > > Revert commit: > > 561ef22a7d396289a1c5a1d18057ccdf9bb59826 ("iommu/vt-d: Gracefully handle DMAR units with no supported address widths") > > [Regression potential] > > Reverting the commit simply restores the old behavior, so we won't be > able to detect the right amount of CPUs in large instances, but we are > sure that we won't also introduce any potential failure or regressions > by restoring the previous behavior. And this is affecting only the > Xenial kernel, that is approaching to EOL. > > ChangeLog (v1 -> v2): > - create a new bug instead of re-using the old one when the original > fix was applied > > Thanks Andrea, makes sense! Acked-by: Guilherme G. Piccoli <gpiccoli@canonical.com>