Public bug reported:
---Problem Description---
Hugepages memory is not getting released even after destroying the guest
Machine Type = P10 Denali LPAR
---uname output---
Linux ubuntu2404lp3 6.8.0-22-generic #22-Ubuntu SMP Thu Apr 4 22:47:57 UTC
2024 ppc64le ppc64le ppc64le GNU/Linux
---Steps to Reproduce---
1. Create a guest which is backed by hugepages.
2. Destroy the guest
3. execute "free -h" or "cat /proc/meminfo" to see that Hugepage memory is
still getting held.
HugePages_Total: 20480
HugePages_Free:20419
HugePages_Rsvd:0
HugePages_Surp:0
Hugepagesize: 2048 kB
Hugetlb:41943040 kB
DirectMap4k: 0 kB
DirectMap64k:52428800 kB
DirectMap2M: 0 kB
DirectMap1G: 0 kB
root@ubuntu2404lp3:~# virsh list --all
Id Name State
---
-ramlp2g1 shut off
-ramlp2g2 shut off
-ramlp2g3 shut off
-ramlp3g3 shut off
root@ubuntu2404lp3:~# free -h
totalusedfree shared buff/cache available
Mem:48Gi43Gi 4.6Gi 2.6Mi 277Mi 4.6Gi
Swap: 8.0Gi 243Mi 7.8Gi
root@ubuntu2404lp3:~#
This is an issue created by commit 1b151e2435fc ("block: Remove special-casing
of compound pages") that moved the direct-io hugetlb handling from compound
pages to folios.
Following commit has been proposed and merged into 6.9-rc1 which fixes this
issue.
38b43539d64b2fa020b3b9a752a986769f87f7a6("block: Fix page refcounts for
unaligned buffers in __bio_release_pages()")
So the same needs to be backported to the Ubuntu24.04 kernel as well.
** Affects: linux (Ubuntu)
Importance: Undecided
Assignee: Ubuntu on IBM Power Systems Bug Triage (ubuntu-power-triage)
Status: New
** Tags: architecture-ppc64le bugnameltc-206058 severity-high
targetmilestone-inin2404
** Tags added: architecture-ppc64le bugnameltc-206058 severity-high
targetmilestone-inin2404
** Changed in: ubuntu
Assignee: (unassigned) => Ubuntu on IBM Power Systems Bug Triage
(ubuntu-power-triage)
** Package changed: ubuntu => linux (Ubuntu)
--
You received this bug notification because you are a member of Kernel
Packages, which is subscribed to linux in Ubuntu.
https://bugs.launchpad.net/bugs/2062556
Title:
[Ubuntu-24.04] Hugepage memory is not getting released even after
destroying the guest!
Status in linux package in Ubuntu:
New
Bug description:
---Problem Description---
Hugepages memory is not getting released even after destroying the guest
Machine Type = P10 Denali LPAR
---uname output---
Linux ubuntu2404lp3 6.8.0-22-generic #22-Ubuntu SMP Thu Apr 4 22:47:57 UTC
2024 ppc64le ppc64le ppc64le GNU/Linux
---Steps to Reproduce---
1. Create a guest which is backed by hugepages.
2. Destroy the guest
3. execute "free -h" or "cat /proc/meminfo" to see that Hugepage memory is
still getting held.
HugePages_Total: 20480
HugePages_Free:20419
HugePages_Rsvd:0
HugePages_Surp:0
Hugepagesize: 2048 kB
Hugetlb:41943040 kB
DirectMap4k: 0 kB
DirectMap64k:52428800 kB
DirectMap2M: 0 kB
DirectMap1G: 0 kB
root@ubuntu2404lp3:~# virsh list --all
Id Name State
---
-ramlp2g1 shut off
-ramlp2g2 shut off
-ramlp2g3 shut off
-ramlp3g3 shut off
root@ubuntu2404lp3:~# free -h
totalusedfree shared buff/cache
available
Mem:48Gi43Gi 4.6Gi 2.6Mi 277Mi
4.6Gi
Swap: 8.0Gi 243Mi 7.8Gi
root@ubuntu2404lp3:~#
This is an issue created by commit 1b151e2435fc ("block: Remove
special-casing of compound pages") that moved the direct-io hugetlb handling
from compound pages to folios.
Following commit has been proposed and merged into 6.9-rc1 which fixes this
issue.
38b43539d64b2fa020b3b9a752a986769f87f7a6("block: Fix page refcounts for
unaligned buffers in __bio_release_pages()")
So the same needs to be backported to the Ubuntu24.04 kernel as well.
To manage notifications about this bug go to:
https://bugs.launchpad.net/ubuntu/+source/linux/+bug/2062556/+subscriptions
--
Mailing list: https://launchpad.net/~kernel-packages
Post to : kernel-packages@lists.launchpad.net
Unsubscribe : https://launchpad.net/~kernel-packages
More help : https://help.launchpad.net/ListHelp