[PATCH net-next] net/ipv6: stop leaking percpu memory in fib6 info

The fib6_info_alloc() function allocates percpu memory to hold per CPU
pointers to rt6_info, but this memory is never freed. Fix it.

Fixes: a64efe142f5e ("net/ipv6: introduce fib6_info struct and helpers")

Signed-off-by: Mike Rapoport <redacted>
Cc: stable@vger.kernel.org
---
 net/ipv6/ip6_fib.c | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/net/ipv6/ip6_fib.c b/net/ipv6/ip6_fib.c
index cf709eadc932..cc7de7eb8b9c 100644
--- a/net/ipv6/ip6_fib.c
+++ b/net/ipv6/ip6_fib.c

@@ -194,6 +194,8 @@ void fib6_info_destroy_rcu(struct rcu_head *head)
 				*ppcpu_rt = NULL;
 			}
 		}
+
+		free_percpu(f6i->rt6i_pcpu);
 	}
 
 	lwtstate_put(f6i->fib6_nh.nh_lwtstate);

-- 
2.7.4

On 10/8/18 6:06 AM, Mike Rapoport wrote:

Odd that KMEMLEAK is not detecting this. Thanks for the fix.

Reviewed-by: David Ahern <redacted>

From: Mike Rapoport <redacted>
Date: Mon,  8 Oct 2018 15:06:03 +0300

The fib6_info_alloc() function allocates percpu memory to hold per CPU
pointers to rt6_info, but this memory is never freed. Fix it.

Fixes: a64efe142f5e ("net/ipv6: introduce fib6_info struct and helpers")

Please do not put empty lines between Fixes: and other tags.  They
belong together as a group.

Signed-off-by: Mike Rapoport <redacted>
Cc: stable@vger.kernel.org

Several problems here.

You cannot submit patches to net-next and expect them to go onward to
-stable.  That is not appropriate.  If it is going to stable, you
must target 'net' not 'net-next'.

Furthermore, for networking changes, explicit CC: of stable is not
appropriate.  Instead you must explicitly ask me to queue the patch
up for -stable as I handle networking stable submissions myself.

Thank you.

On Mon, Oct 08, 2018 at 12:15:54PM -0600, David Ahern wrote:

On 10/8/18 6:06 AM, Mike Rapoport wrote:

quoted

The fib6_info_alloc() function allocates percpu memory to hold per CPU
pointers to rt6_info, but this memory is never freed. Fix it.

Fixes: a64efe142f5e ("net/ipv6: introduce fib6_info struct and helpers")

Signed-off-by: Mike Rapoport <redacted>
Cc: stable@vger.kernel.org
---
 net/ipv6/ip6_fib.c | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/net/ipv6/ip6_fib.c b/net/ipv6/ip6_fib.c
index cf709eadc932..cc7de7eb8b9c 100644
--- a/net/ipv6/ip6_fib.c
+++ b/net/ipv6/ip6_fib.c

@@ -194,6 +194,8 @@ void fib6_info_destroy_rcu(struct rcu_head *head)
 				*ppcpu_rt = NULL;
 			}
 		}
+
+		free_percpu(f6i->rt6i_pcpu);
 	}
 
 	lwtstate_put(f6i->fib6_nh.nh_lwtstate);

Odd that KMEMLEAK is not detecting this. Thanks for the fix.

There's a comment in kmemleak that says:

	/*
	 * Percpu allocations are only scanned and not reported as leaks
	 * (min_count is set to 0).
	 */

No idea why, though...
 

Reviewed-by: David Ahern <redacted>

-- 
Sincerely yours,
Mike.