| Message ID | 20240531132636.2637995-6-edumazet@google.com (mailing list archive) |
|---|---|
| State | Accepted |
| Commit | 2fe6fb36c781b50482b1c3323fb526bc07d1af59 |
| Delegated to: | Netdev Maintainers |
| Headers | show |
| Series | dst_cache: fix possible races | expand |
diff --git a/net/core/dst_cache.c b/net/core/dst_cache.c index 6a0482e676d379f1f9bffdda51c7535243b3ec38..70c634b9e7b02300188582a1634d5977838db132 100644 --- a/net/core/dst_cache.c +++ b/net/core/dst_cache.c @@ -27,6 +27,7 @@ struct dst_cache_pcpu { static void dst_cache_per_cpu_dst_set(struct dst_cache_pcpu *dst_cache, struct dst_entry *dst, u32 cookie) { + DEBUG_NET_WARN_ON_ONCE(!in_softirq()); dst_release(dst_cache->dst); if (dst) dst_hold(dst); @@ -40,6 +41,7 @@ static struct dst_entry *dst_cache_per_cpu_get(struct dst_cache *dst_cache, { struct dst_entry *dst; + DEBUG_NET_WARN_ON_ONCE(!in_softirq()); dst = idst->dst; if (!dst) goto fail;
After fixing four different bugs involving dst_cache users, it might be worth adding a check about BH being blocked by dst_cache callers. DEBUG_NET_WARN_ON_ONCE(!in_softirq()); It is not fatal, if we missed valid case where no BH deadlock is to be feared, we might change this. Signed-off-by: Eric Dumazet <edumazet@google.com> --- net/core/dst_cache.c | 2 ++ 1 file changed, 2 insertions(+)