Thread (2 messages) 2 messages, 2 authors, 4d ago

[PATCH 6.1.y v2] netfilter: nft_set_pipapo: restore cloned mapping table memcg accounting

flat view
COOLING4d

From: Karl Mehltretter <hidden>
Date: 2026-10-10 12:06:33
Also in: lkml, netfilter-devel, stable
Subsystem: netfilter, networking [general], the rest · Maintainers: Pablo Neira Ayuso, Florian Westphal, "David S. Miller", Eric Dumazet, Jakub Kicinski, Paolo Abeni, Linus Torvalds

The 6.1.y backport of commit 69e687cea79f
("netfilter: nf_tables: missing objects with no memcg accounting")
changed the pipapo_clone() mapping-table allocation to
GFP_KERNEL_ACCOUNT.

The later stable backport of commit 07ace0bbe03b ("netfilter:
nft_set_pipapo: do not rely on ZERO_SIZE_PTR") rewrote this allocation
and changed the flag back to GFP_KERNEL. Upstream applied the
ZERO_SIZE_PTR change before the memcg accounting change and retains
GFP_KERNEL_ACCOUNT after both.

nft_pipapo_deactivate() clones a populated set before looking up an
element to delete. If the element is absent, priv->dirty remains false
and the clone is retained. Its mapping tables therefore remain allocated
without being charged to the requesting memory cgroup.

Restore GFP_KERNEL_ACCOUNT for the cloned mapping table.

Fixes: a4983e89e3b1 ("netfilter: nft_set_pipapo: do not rely on ZERO_SIZE_PTR")
Assisted-by: LLM
Signed-off-by: Karl Mehltretter <redacted>
---
Changes in v2:
- Split the combined 6.1.y/6.6.y submission into one email per branch.
- Keep only the Fixes tag for this branch.
- Refresh the diff against 6.1.189; the code change is unchanged.
- Add fresh build, runtime and matched A/B results for this branch.

Original submission:
https://lore.kernel.org/r/20260831194905.45045-1-kmehltretter@gmail.com/ (local)

Testing on 6.1.189: built a minimal x86_64 kernel with GCC 15.2.0
and booted it in QEMU with CONFIG_MEMCG=y. A 64-element two-field
interval set passed creation, lookup, update, rollback, deletion, flush
and reuse checks.

Matched unpatched and patched boots used the same kernel configuration
and initramfs. During a normal update from a memory cgroup, both cloned
mapping-table allocations (512 bytes each) had accounted=false without
the patch. With the patch, both used GFP_KERNEL_ACCOUNT and had
accounted=true. Both variants passed normal set operations without
kernel warnings.

Earlier A/B testing on 6.1.186 and 6.6.155 measured 524,288 additional
charged bytes for the cloned mapping tables of a 32,768-element two-field
set. An unmodified 6.12.107 control already accounted that allocation.

 net/netfilter/nft_set_pipapo.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/net/netfilter/nft_set_pipapo.c b/net/netfilter/nft_set_pipapo.c
index 4b37c6684bff..456779003f58 100644
--- a/net/netfilter/nft_set_pipapo.c
+++ b/net/netfilter/nft_set_pipapo.c
@@ -1466,7 +1466,7 @@ static struct nft_pipapo_match *pipapo_clone(struct nft_pipapo_match *old)
 				goto out_mt;
 
 			dst->mt = kvmalloc_array(src->rules, sizeof(*src->mt),
-						 GFP_KERNEL);
+						 GFP_KERNEL_ACCOUNT);
 			if (!dst->mt)
 				goto out_mt;
 
Keyboard shortcuts
hback out one level
jnext message in thread
kprevious message in thread
ldrill in
Escclose help / fold thread tree
?toggle this help