NACK: [SRU][Zesty] fix soft lockup on overcommited hugepages

Seth Forshee seth.forshee at
Fri Jun 16 20:48:22 UTC 2017

On Thu, Jun 15, 2017 at 09:13:47AM -0500, Manoj Iyer wrote:
> Please review and consider this pull request for fixing bug
> This patch fixes a case where if
> hugepage migration is not enabled in the kernel, on failing to migrate a
> page ref-counts are left in an inconsistent state leading to soft lockups.
> Please refer to bug report for detailed test reports.
> The patch was cherry-picked from upstream with minor backport to Zesty. The
> kernel is available in the PPA:
> and tested on
> ARM64, and boot tested on AMD64 and Power8 for any regressions.
> The following changes since commit 357b492ed1cc4beb9d46e3bb017116a362dca5f0:
>  arm64: kconfig: allow support for memory failure handling (2017-06-13
> 15:45:24 -0500)
> are available in the git repository at:
>  git+ssh:// lp1696165
> for you to fetch changes up to 080d471941f3c814aa64ba0c3c622dd712adc75d:
>  mm/migrate: fix refcount handling when !hugepage_migration_supported()
> (2017-06-14 13:57:18 -0500)

The upstream patch looks like a reasonable one to backport, but there
are two problems. First, you're commit says "cherry picked from ..."
even though you say above you had to backport it, so it ought to say
"backported from ...". Second, I don't see that there's any reason to
include unrelated changes from other commits in your backport, even if
they're harmless.

Fix those issues and I'll be happy to ack the patch.

Also I'll note that this patch was already applied to artful/master-next
as part of the 4.11.4 stable update.


More information about the kernel-team mailing list