diff options
| author | Christian Marangi | 2023-06-10 16:56:11 +0000 |
|---|---|---|
| committer | Christian Marangi | 2023-06-10 19:08:23 +0000 |
| commit | bf1d5fdf623473419f8f738fc9071fee84cb13e5 (patch) | |
| tree | ec578500355d3363c47f8de6f2b102dc5f6ea759 | |
| parent | ffba75c9cd8f96ef34b7b3ab35315ce66ba8f37d (diff) | |
| download | firewall3-bf1d5fdf623473419f8f738fc9071fee84cb13e5.tar.gz | |
iptables: fix regression with unintended free in need_protomatch
xtables_find_match memory allocation is funny. It can return something
allocated in a linked list or return a just allocated match clone and is
never freed.
This caused confusion and made a broken patch where an unintended free
is done in the case of entry not cloned.
xtables_find_match have a way to comunicate that the entry is cloned by
returning a looping linked list where the next entry is the same entry.
We can use this to understand where the entry has to be freed.
Fixes: ffba75c9cd8f ("iptables: free xtables_match if found in need_protomatch")
Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
Tested-by: Rui Salvaterra <rsalvaterra@gmail.com>
| -rw-r--r-- | iptables.c | 4 |
1 files changed, 3 insertions, 1 deletions
@@ -718,7 +718,9 @@ need_protomatch(struct fw3_ipt_rule *r, const char *pname) if (!match) return true; - free(match); + /* Free any kind of clone from xtables_find_match */ + if (match == match->next) + free(match); return !r->protocol_loaded; } |