* [RFC PATCH 02/17] fib_trie: Make leaf and tnode more uniform
From: Alexander Duyck @ 2014-12-22 17:41 UTC (permalink / raw)
To: netdev
In-Reply-To: <20141222172632.1119.51469.stgit@ahduyck-vm-fedora20>
This change makes some fundamental changes to the way leaves and tnodes are
constructed. The big differences are:
1. Leaves now populate pos and bits indicating their full key size.
2. Trie nodes now mask out their lower bits to be consistent with the leaf
3. Both structures have been reordered so that rt_trie_node now consisists
of a much larger region including the pos, bits, and rcu portions of
the tnode structure.
On 32b systems this will result in the leaf being 4B larger as the pos and
bits values were added to a hole created by the key as it was only 4B in
length.
Signed-off-by: Alexander Duyck <alexander.h.duyck@redhat.com>
---
net/ipv4/fib_trie.c | 192 ++++++++++++++++++++++-----------------------------
1 file changed, 82 insertions(+), 110 deletions(-)
diff --git a/net/ipv4/fib_trie.c b/net/ipv4/fib_trie.c
index 861ef9b..934bfb0 100644
--- a/net/ipv4/fib_trie.c
+++ b/net/ipv4/fib_trie.c
@@ -87,24 +87,38 @@
typedef unsigned int t_key;
-#define T_TNODE 0
-#define T_LEAF 1
-#define NODE_TYPE_MASK 0x1UL
-#define NODE_TYPE(node) ((node)->parent & NODE_TYPE_MASK)
+#define IS_TNODE(n) ((n)->bits)
+#define IS_LEAF(n) (!(n)->bits)
-#define IS_TNODE(n) (!(n->parent & T_LEAF))
-#define IS_LEAF(n) (n->parent & T_LEAF)
+struct tnode {
+ t_key key;
+ unsigned char bits; /* 2log(KEYLENGTH) bits needed */
+ unsigned char pos; /* 2log(KEYLENGTH) bits needed */
+ struct tnode __rcu *parent;
+ union {
+ struct rcu_head rcu;
+ struct tnode *tnode_free;
+ };
+ unsigned int full_children; /* KEYLENGTH bits needed */
+ unsigned int empty_children; /* KEYLENGTH bits needed */
+ struct rt_trie_node __rcu *child[0];
+};
struct rt_trie_node {
- unsigned long parent;
t_key key;
+ unsigned char bits;
+ unsigned char pos;
+ struct tnode __rcu *parent;
+ struct rcu_head rcu;
};
struct leaf {
- unsigned long parent;
t_key key;
- struct hlist_head list;
+ unsigned char bits;
+ unsigned char pos;
+ struct tnode __rcu *parent;
struct rcu_head rcu;
+ struct hlist_head list;
};
struct leaf_info {
@@ -115,20 +129,6 @@ struct leaf_info {
struct rcu_head rcu;
};
-struct tnode {
- unsigned long parent;
- t_key key;
- unsigned char pos; /* 2log(KEYLENGTH) bits needed */
- unsigned char bits; /* 2log(KEYLENGTH) bits needed */
- unsigned int full_children; /* KEYLENGTH bits needed */
- unsigned int empty_children; /* KEYLENGTH bits needed */
- union {
- struct rcu_head rcu;
- struct tnode *tnode_free;
- };
- struct rt_trie_node __rcu *child[0];
-};
-
#ifdef CONFIG_IP_FIB_TRIE_STATS
struct trie_use_stats {
unsigned int gets;
@@ -176,38 +176,27 @@ static const int sync_pages = 128;
static struct kmem_cache *fn_alias_kmem __read_mostly;
static struct kmem_cache *trie_leaf_kmem __read_mostly;
-/*
- * caller must hold RTNL
- */
-static inline struct tnode *node_parent(const struct rt_trie_node *node)
-{
- unsigned long parent;
+/* caller must hold RTNL */
+#define node_parent(n) rtnl_dereference((n)->parent)
- parent = rcu_dereference_index_check(node->parent, lockdep_rtnl_is_held());
+/* caller must hold RCU read lock or RTNL */
+#define node_parent_rcu(n) rcu_dereference_rtnl((n)->parent)
- return (struct tnode *)(parent & ~NODE_TYPE_MASK);
-}
-
-/*
- * caller must hold RCU read lock or RTNL
- */
-static inline struct tnode *node_parent_rcu(const struct rt_trie_node *node)
+/* wrapper for rcu_assign_pointer */
+static inline void node_set_parent(struct rt_trie_node *node, struct tnode *ptr)
{
- unsigned long parent;
-
- parent = rcu_dereference_index_check(node->parent, rcu_read_lock_held() ||
- lockdep_rtnl_is_held());
-
- return (struct tnode *)(parent & ~NODE_TYPE_MASK);
+ if (node)
+ rcu_assign_pointer(node->parent, ptr);
}
-/* Same as rcu_assign_pointer
- * but that macro() assumes that value is a pointer.
+#define NODE_INIT_PARENT(n, p) RCU_INIT_POINTER((n)->parent, p)
+
+/* This provides us with the number of children in this node, in the case of a
+ * leaf this will return 0 meaning none of the children are accessible.
*/
-static inline void node_set_parent(struct rt_trie_node *node, struct tnode *ptr)
+static inline int tnode_child_length(const struct tnode *tn)
{
- smp_wmb();
- node->parent = (unsigned long)ptr | NODE_TYPE(node);
+ return (1ul << tn->bits) & ~(1ul);
}
/*
@@ -215,7 +204,7 @@ static inline void node_set_parent(struct rt_trie_node *node, struct tnode *ptr)
*/
static inline struct rt_trie_node *tnode_get_child(const struct tnode *tn, unsigned int i)
{
- BUG_ON(i >= 1U << tn->bits);
+ BUG_ON(i >= tnode_child_length(tn));
return rtnl_dereference(tn->child[i]);
}
@@ -225,16 +214,11 @@ static inline struct rt_trie_node *tnode_get_child(const struct tnode *tn, unsig
*/
static inline struct rt_trie_node *tnode_get_child_rcu(const struct tnode *tn, unsigned int i)
{
- BUG_ON(i >= 1U << tn->bits);
+ BUG_ON(i >= tnode_child_length(tn));
return rcu_dereference_rtnl(tn->child[i]);
}
-static inline int tnode_child_length(const struct tnode *tn)
-{
- return 1 << tn->bits;
-}
-
static inline t_key mask_pfx(t_key k, unsigned int l)
{
return (l == 0) ? 0 : k >> (KEYLENGTH-l) << (KEYLENGTH-l);
@@ -336,11 +320,6 @@ static inline int tkey_mismatch(t_key a, int offset, t_key b)
*/
-static inline void check_tnode(const struct tnode *tn)
-{
- WARN_ON(tn && tn->pos+tn->bits > 32);
-}
-
static const int halve_threshold = 25;
static const int inflate_threshold = 50;
static const int halve_threshold_root = 15;
@@ -426,11 +405,20 @@ static void tnode_free_flush(void)
}
}
-static struct leaf *leaf_new(void)
+static struct leaf *leaf_new(t_key key)
{
struct leaf *l = kmem_cache_alloc(trie_leaf_kmem, GFP_KERNEL);
if (l) {
- l->parent = T_LEAF;
+ l->parent = NULL;
+ /* set key and pos to reflect full key value
+ * any trailing zeros in the key should be ignored
+ * as the nodes are searched
+ */
+ l->key = key;
+ l->pos = KEYLENGTH;
+ /* set bits to 0 indicating we are not a tnode */
+ l->bits = 0;
+
INIT_HLIST_HEAD(&l->list);
}
return l;
@@ -451,12 +439,16 @@ static struct tnode *tnode_new(t_key key, int pos, int bits)
{
size_t sz = sizeof(struct tnode) + (sizeof(struct rt_trie_node *) << bits);
struct tnode *tn = tnode_alloc(sz);
+ unsigned int shift = pos + bits;
+
+ /* verify bits and pos their msb bits clear and values are valid */
+ BUG_ON(!bits || (shift > KEYLENGTH));
if (tn) {
- tn->parent = T_TNODE;
+ tn->parent = NULL;
tn->pos = pos;
tn->bits = bits;
- tn->key = key;
+ tn->key = mask_pfx(key, pos);
tn->full_children = 0;
tn->empty_children = 1<<bits;
}
@@ -473,10 +465,7 @@ static struct tnode *tnode_new(t_key key, int pos, int bits)
static inline int tnode_full(const struct tnode *tn, const struct rt_trie_node *n)
{
- if (n == NULL || IS_LEAF(n))
- return 0;
-
- return ((struct tnode *) n)->pos == tn->pos + tn->bits;
+ return n && IS_TNODE(n) && (n->pos == (tn->pos + tn->bits));
}
static inline void put_child(struct tnode *tn, int i,
@@ -514,8 +503,7 @@ static void tnode_put_child_reorg(struct tnode *tn, int i, struct rt_trie_node *
else if (!wasfull && isfull)
tn->full_children++;
- if (n)
- node_set_parent(n, tn);
+ node_set_parent(n, tn);
rcu_assign_pointer(tn->child[i], n);
}
@@ -523,7 +511,7 @@ static void tnode_put_child_reorg(struct tnode *tn, int i, struct rt_trie_node *
#define MAX_WORK 10
static struct rt_trie_node *resize(struct trie *t, struct tnode *tn)
{
- int i;
+ struct rt_trie_node *n = NULL;
struct tnode *old_tn;
int inflate_threshold_use;
int halve_threshold_use;
@@ -536,12 +524,11 @@ static struct rt_trie_node *resize(struct trie *t, struct tnode *tn)
tn, inflate_threshold, halve_threshold);
/* No children */
- if (tn->empty_children == tnode_child_length(tn)) {
- tnode_free_safe(tn);
- return NULL;
- }
+ if (tn->empty_children > (tnode_child_length(tn) - 1))
+ goto no_children;
+
/* One child */
- if (tn->empty_children == tnode_child_length(tn) - 1)
+ if (tn->empty_children == (tnode_child_length(tn) - 1))
goto one_child;
/*
* Double as long as the resulting node has a number of
@@ -607,11 +594,9 @@ static struct rt_trie_node *resize(struct trie *t, struct tnode *tn)
*
*/
- check_tnode(tn);
-
/* Keep root node larger */
- if (!node_parent((struct rt_trie_node *)tn)) {
+ if (!node_parent(tn)) {
inflate_threshold_use = inflate_threshold_root;
halve_threshold_use = halve_threshold_root;
} else {
@@ -637,8 +622,6 @@ static struct rt_trie_node *resize(struct trie *t, struct tnode *tn)
}
}
- check_tnode(tn);
-
/* Return if at least one inflate is run */
if (max_work != MAX_WORK)
return (struct rt_trie_node *) tn;
@@ -666,21 +649,16 @@ static struct rt_trie_node *resize(struct trie *t, struct tnode *tn)
/* Only one child remains */
- if (tn->empty_children == tnode_child_length(tn) - 1) {
+ if (tn->empty_children == (tnode_child_length(tn) - 1)) {
+ unsigned long i;
one_child:
- for (i = 0; i < tnode_child_length(tn); i++) {
- struct rt_trie_node *n;
-
- n = rtnl_dereference(tn->child[i]);
- if (!n)
- continue;
-
- /* compress one level */
-
- node_set_parent(n, NULL);
- tnode_free_safe(tn);
- return n;
- }
+ for (i = tnode_child_length(tn); !n && i;)
+ n = tnode_get_child(tn, --i);
+no_children:
+ /* compress one level */
+ node_set_parent(n, NULL);
+ tnode_free_safe(tn);
+ return n;
}
return (struct rt_trie_node *) tn;
}
@@ -760,8 +738,7 @@ static struct tnode *inflate(struct trie *t, struct tnode *tn)
/* A leaf or an internal node with skipped bits */
- if (IS_LEAF(node) || ((struct tnode *) node)->pos >
- tn->pos + tn->bits - 1) {
+ if (IS_LEAF(node) || (node->pos > (tn->pos + tn->bits - 1))) {
put_child(tn,
tkey_extract_bits(node->key, oldtnode->pos, oldtnode->bits + 1),
node);
@@ -958,11 +935,9 @@ fib_find_node(struct trie *t, u32 key)
pos = 0;
n = rcu_dereference_rtnl(t->trie);
- while (n != NULL && NODE_TYPE(n) == T_TNODE) {
+ while (n && IS_TNODE(n)) {
tn = (struct tnode *) n;
- check_tnode(tn);
-
if (tkey_sub_equals(tn->key, pos, tn->pos-pos, key)) {
pos = tn->pos + tn->bits;
n = tnode_get_child_rcu(tn,
@@ -988,7 +963,7 @@ static void trie_rebalance(struct trie *t, struct tnode *tn)
key = tn->key;
- while (tn != NULL && (tp = node_parent((struct rt_trie_node *)tn)) != NULL) {
+ while (tn != NULL && (tp = node_parent(tn)) != NULL) {
cindex = tkey_extract_bits(key, tp->pos, tp->bits);
wasfull = tnode_full(tp, tnode_get_child(tp, cindex));
tn = (struct tnode *)resize(t, tn);
@@ -996,7 +971,7 @@ static void trie_rebalance(struct trie *t, struct tnode *tn)
tnode_put_child_reorg(tp, cindex,
(struct rt_trie_node *)tn, wasfull);
- tp = node_parent((struct rt_trie_node *) tn);
+ tp = node_parent(tn);
if (!tp)
rcu_assign_pointer(t->trie, (struct rt_trie_node *)tn);
@@ -1048,11 +1023,9 @@ static struct list_head *fib_insert_node(struct trie *t, u32 key, int plen)
* If it doesn't, we need to replace it with a T_TNODE.
*/
- while (n != NULL && NODE_TYPE(n) == T_TNODE) {
+ while (n && IS_TNODE(n)) {
tn = (struct tnode *) n;
- check_tnode(tn);
-
if (tkey_sub_equals(tn->key, pos, tn->pos-pos, key)) {
tp = tn;
pos = tn->pos + tn->bits;
@@ -1087,12 +1060,11 @@ static struct list_head *fib_insert_node(struct trie *t, u32 key, int plen)
insert_leaf_info(&l->list, li);
goto done;
}
- l = leaf_new();
+ l = leaf_new(key);
if (!l)
return NULL;
- l->key = key;
li = leaf_info_new(plen);
if (!li) {
@@ -1569,7 +1541,7 @@ backtrace:
if (chopped_off <= pn->bits) {
cindex &= ~(1 << (chopped_off-1));
} else {
- struct tnode *parent = node_parent_rcu((struct rt_trie_node *) pn);
+ struct tnode *parent = node_parent_rcu(pn);
if (!parent)
goto failed;
@@ -1597,7 +1569,7 @@ EXPORT_SYMBOL_GPL(fib_table_lookup);
*/
static void trie_leaf_remove(struct trie *t, struct leaf *l)
{
- struct tnode *tp = node_parent((struct rt_trie_node *) l);
+ struct tnode *tp = node_parent(l);
pr_debug("entering trie_leaf_remove(%p)\n", l);
@@ -2374,7 +2346,7 @@ static int fib_trie_seq_show(struct seq_file *seq, void *v)
if (IS_TNODE(n)) {
struct tnode *tn = (struct tnode *) n;
- __be32 prf = htonl(mask_pfx(tn->key, tn->pos));
+ __be32 prf = htonl(tn->key);
seq_indent(seq, iter->depth-1);
seq_printf(seq, " +-- %pI4/%d %d %d %d\n",
^ permalink raw reply related
* [RFC PATCH 01/17] fib_trie: Update usage stats to be percpu instead of global variables
From: Alexander Duyck @ 2014-12-22 17:40 UTC (permalink / raw)
To: netdev
In-Reply-To: <20141222172632.1119.51469.stgit@ahduyck-vm-fedora20>
The trie usage stats were currently being shared by all threads that were
calling fib_table_lookup. As a result when multiple threads were
performing lookups simultaneously the trie would begin to cache bounce
between those threads.
In order to prevent this I have updated the usage stats to use a set of
percpu variables. By doing this we should be able to avoid the cache
bouncing and still make use of these stats.
Signed-off-by: Alexander Duyck <alexander.h.duyck@redhat.com>
---
net/ipv4/fib_frontend.c | 2 +
net/ipv4/fib_trie.c | 68 +++++++++++++++++++++++++++++++++--------------
2 files changed, 49 insertions(+), 21 deletions(-)
diff --git a/net/ipv4/fib_frontend.c b/net/ipv4/fib_frontend.c
index 23104a3..6689020 100644
--- a/net/ipv4/fib_frontend.c
+++ b/net/ipv4/fib_frontend.c
@@ -67,7 +67,7 @@ static int __net_init fib4_rules_init(struct net *net)
return 0;
fail:
- kfree(local_table);
+ fib_free_table(local_table);
return -ENOMEM;
}
#else
diff --git a/net/ipv4/fib_trie.c b/net/ipv4/fib_trie.c
index 18bcaf2..861ef9b 100644
--- a/net/ipv4/fib_trie.c
+++ b/net/ipv4/fib_trie.c
@@ -153,7 +153,7 @@ struct trie_stat {
struct trie {
struct rt_trie_node __rcu *trie;
#ifdef CONFIG_IP_FIB_TRIE_STATS
- struct trie_use_stats stats;
+ struct trie_use_stats __percpu *stats;
#endif
};
@@ -631,7 +631,7 @@ static struct rt_trie_node *resize(struct trie *t, struct tnode *tn)
if (IS_ERR(tn)) {
tn = old_tn;
#ifdef CONFIG_IP_FIB_TRIE_STATS
- t->stats.resize_node_skipped++;
+ this_cpu_inc(t->stats->resize_node_skipped);
#endif
break;
}
@@ -658,7 +658,7 @@ static struct rt_trie_node *resize(struct trie *t, struct tnode *tn)
if (IS_ERR(tn)) {
tn = old_tn;
#ifdef CONFIG_IP_FIB_TRIE_STATS
- t->stats.resize_node_skipped++;
+ this_cpu_inc(t->stats->resize_node_skipped);
#endif
break;
}
@@ -1357,7 +1357,7 @@ static int check_leaf(struct fib_table *tb, struct trie *t, struct leaf *l,
err = fib_props[fa->fa_type].error;
if (err) {
#ifdef CONFIG_IP_FIB_TRIE_STATS
- t->stats.semantic_match_passed++;
+ this_cpu_inc(t->stats->semantic_match_passed);
#endif
return err;
}
@@ -1372,7 +1372,7 @@ static int check_leaf(struct fib_table *tb, struct trie *t, struct leaf *l,
continue;
#ifdef CONFIG_IP_FIB_TRIE_STATS
- t->stats.semantic_match_passed++;
+ this_cpu_inc(t->stats->semantic_match_passed);
#endif
res->prefixlen = li->plen;
res->nh_sel = nhsel;
@@ -1388,7 +1388,7 @@ static int check_leaf(struct fib_table *tb, struct trie *t, struct leaf *l,
}
#ifdef CONFIG_IP_FIB_TRIE_STATS
- t->stats.semantic_match_miss++;
+ this_cpu_ptr(t->stats->semantic_match_miss);
#endif
}
@@ -1399,6 +1399,9 @@ int fib_table_lookup(struct fib_table *tb, const struct flowi4 *flp,
struct fib_result *res, int fib_flags)
{
struct trie *t = (struct trie *) tb->tb_data;
+#ifdef CONFIG_IP_FIB_TRIE_STATS
+ struct trie_use_stats __percpu *stats = t->stats;
+#endif
int ret;
struct rt_trie_node *n;
struct tnode *pn;
@@ -1417,7 +1420,7 @@ int fib_table_lookup(struct fib_table *tb, const struct flowi4 *flp,
goto failed;
#ifdef CONFIG_IP_FIB_TRIE_STATS
- t->stats.gets++;
+ this_cpu_inc(stats->gets);
#endif
/* Just a leaf? */
@@ -1441,7 +1444,7 @@ int fib_table_lookup(struct fib_table *tb, const struct flowi4 *flp,
if (n == NULL) {
#ifdef CONFIG_IP_FIB_TRIE_STATS
- t->stats.null_node_hit++;
+ this_cpu_inc(stats->null_node_hit);
#endif
goto backtrace;
}
@@ -1576,7 +1579,7 @@ backtrace:
chopped_off = 0;
#ifdef CONFIG_IP_FIB_TRIE_STATS
- t->stats.backtrack++;
+ this_cpu_inc(stats->backtrack);
#endif
goto backtrace;
}
@@ -1830,6 +1833,11 @@ int fib_table_flush(struct fib_table *tb)
void fib_free_table(struct fib_table *tb)
{
+#ifdef CONFIG_IP_FIB_TRIE_STATS
+ struct trie *t = (struct trie *)tb->tb_data;
+
+ free_percpu(t->stats);
+#endif /* CONFIG_IP_FIB_TRIE_STATS */
kfree(tb);
}
@@ -1973,7 +1981,14 @@ struct fib_table *fib_trie_table(u32 id)
tb->tb_num_default = 0;
t = (struct trie *) tb->tb_data;
- memset(t, 0, sizeof(*t));
+ RCU_INIT_POINTER(t->trie, NULL);
+#ifdef CONFIG_IP_FIB_TRIE_STATS
+ t->stats = alloc_percpu(struct trie_use_stats);
+ if (!t->stats) {
+ kfree(tb);
+ tb = NULL;
+ }
+#endif
return tb;
}
@@ -2139,18 +2154,31 @@ static void trie_show_stats(struct seq_file *seq, struct trie_stat *stat)
#ifdef CONFIG_IP_FIB_TRIE_STATS
static void trie_show_usage(struct seq_file *seq,
- const struct trie_use_stats *stats)
+ const struct trie_use_stats __percpu *stats)
{
+ struct trie_use_stats s = { 0 };
+ int cpu;
+
+ /* loop through all of the CPUs and gather up the stats */
+ for_each_possible_cpu(cpu) {
+ const struct trie_use_stats *pcpu = per_cpu_ptr(stats, cpu);
+
+ s.gets += pcpu->gets;
+ s.backtrack += pcpu->backtrack;
+ s.semantic_match_passed += pcpu->semantic_match_passed;
+ s.semantic_match_miss += pcpu->semantic_match_miss;
+ s.null_node_hit += pcpu->null_node_hit;
+ s.resize_node_skipped += pcpu->resize_node_skipped;
+ }
+
seq_printf(seq, "\nCounters:\n---------\n");
- seq_printf(seq, "gets = %u\n", stats->gets);
- seq_printf(seq, "backtracks = %u\n", stats->backtrack);
+ seq_printf(seq, "gets = %u\n", s.gets);
+ seq_printf(seq, "backtracks = %u\n", s.backtrack);
seq_printf(seq, "semantic match passed = %u\n",
- stats->semantic_match_passed);
- seq_printf(seq, "semantic match miss = %u\n",
- stats->semantic_match_miss);
- seq_printf(seq, "null node hit= %u\n", stats->null_node_hit);
- seq_printf(seq, "skipped node resize = %u\n\n",
- stats->resize_node_skipped);
+ s.semantic_match_passed);
+ seq_printf(seq, "semantic match miss = %u\n", s.semantic_match_miss);
+ seq_printf(seq, "null node hit= %u\n", s.null_node_hit);
+ seq_printf(seq, "skipped node resize = %u\n\n", s.resize_node_skipped);
}
#endif /* CONFIG_IP_FIB_TRIE_STATS */
@@ -2191,7 +2219,7 @@ static int fib_triestat_seq_show(struct seq_file *seq, void *v)
trie_collect_stats(t, &stat);
trie_show_stats(seq, &stat);
#ifdef CONFIG_IP_FIB_TRIE_STATS
- trie_show_usage(seq, &t->stats);
+ trie_show_usage(seq, t->stats);
#endif
}
}
^ permalink raw reply related
* [RFC PATCH 00/17] fib_trie: Reduce time spent in fib_table_lookup by 35 to 75%
From: Alexander Duyck @ 2014-12-22 17:40 UTC (permalink / raw)
To: netdev
These patches are meant to address several performance issues I have seen
in the fib_trie implementation, and fib_table_lookup specifically. With
these changes in place I have seen a reduction of up to 35 to 75% for the
total time spent in fib_table_lookup depending on the type of search being
performed.
On a VM running in my Corei7-4930K system with a trie of maximum depth of 7
this resulted in a reduction of over 370ns per packet in the total time to
process packets received from an ixgbe interface and route them to a dummy
interface. This represents a failed lookup in the local trie followed by
a successful search in the main trie.
Baseline Refactor
ixgbe->dummy routing 1.20Mpps 2.21Mpps
------------------------------------------------------------
processing time per packet 835ns 453ns
fib_table_lookup 50.1% 418ns 25.0% 113ns
check_leaf.isra.9 7.9% 66ns -- --
ixgbe_clean_rx_irq 5.3% 44ns 9.8% 44ns
ip_route_input_noref 2.9% 25ns 4.6% 21ns
pvclock_clocksource_read 2.6% 21ns 4.6% 21ns
ip_rcv 2.6% 22ns 4.0% 18ns
In the simple case of receiving a frame and dropping it before it can reach
the socket layer I saw a reduction of 40ns per packet. This represents a
trip through the local trie with the correct leaf found with no need for
any backtracing.
Baseline Refactor
ixgbe->local receive 2.65Mpps 2.96Mpps
------------------------------------------------------------
processing time per packet 377ns 337ns
fib_table_lookup 25.1% 95ns 25.8% 87ns
ixgbe_clean_rx_irq 8.7% 33ns 9.0% 30ns
check_leaf.isra.9 7.2% 27ns -- --
ip_rcv 5.7% 21ns 6.5% 22ns
These changes have resulted in several functions being inlined such as
check_leaf and fib_find_node, but due to the code simplification the
overall size of the code has been reduced.
text data bss dec hex filename
16932 376 16 17324 43ac net/ipv4/fib_trie.o - before
15259 376 8 15643 3d1b net/ipv4/fib_trie.o - after
---
Alexander Duyck (17):
fib_trie: Update usage stats to be percpu instead of global variables
fib_trie: Make leaf and tnode more uniform
fib_trie: Merge tnode_free and leaf_free into node_free
fib_trie: Merge leaf into tnode
fib_trie: Optimize fib_table_lookup to avoid wasting time on loops/variables
fib_trie: Optimize fib_find_node
fib_trie: Optimize fib_table_insert
fib_trie: Update meaning of pos to represent unchecked bits
fib_trie: Use unsigned long for anything dealing with a shift by bits
fib_trie: Push rcu_read_lock/unlock to callers
fib_trie: Move resize to after inflate/halve
fib_trie: Add functions should_inflate and should_halve
fib_trie: Push assignment of child to parent down into inflate/halve
fib_trie: Push tnode flushing down to inflate/halve
fib_trie: inflate/halve nodes in a more RCU friendly way
fib_trie: Remove checks for index >= tnode_child_length from tnode_get_child
fib_trie: Add tracking value for suffix length
include/net/ip_fib.h | 50 +
net/ipv4/fib_frontend.c | 29 -
net/ipv4/fib_rules.c | 22 -
net/ipv4/fib_trie.c | 1915 ++++++++++++++++++++++-------------------------
4 files changed, 945 insertions(+), 1071 deletions(-)
--
^ permalink raw reply
* [PATCH for 3.19] rtlwifi: Fix error when accessing unmapped memory in skb
From: Larry Finger @ 2014-12-22 17:37 UTC (permalink / raw)
To: kvalo; +Cc: linux-wireless, Larry Finger, netdev, Eric Biggers, Stable
Under heavy memory pressure, it is possible for the allocation of a
new skb to fail. When this happens, the kernel gets a memory access
violation. Previous versions of the drivers would drop the read request;
however, this logic was missed in the 3.18 update. This patch restores
the previous behavior.
Reported-by: Eric Biggers <ebiggers3@gmail.com>
Signed-off-by: Larry Finger <Larry.Finger@lwfinger.net>
Cc: Eric Biggers <ebiggers3@gmail.com>
Cc: Stable <stable@vger.kernel.org> [3.18]
---
drivers/net/wireless/rtlwifi/pci.c | 14 ++++++++------
1 file changed, 8 insertions(+), 6 deletions(-)
diff --git a/drivers/net/wireless/rtlwifi/pci.c b/drivers/net/wireless/rtlwifi/pci.c
index 846a2e6..55334ca 100644
--- a/drivers/net/wireless/rtlwifi/pci.c
+++ b/drivers/net/wireless/rtlwifi/pci.c
@@ -912,13 +912,15 @@ static void _rtl_pci_rx_interrupt(struct ieee80211_hw *hw)
}
end:
if (rtlpriv->use_new_trx_flow) {
- _rtl_pci_init_one_rxdesc(hw, (u8 *)buffer_desc,
- rxring_idx,
- rtlpci->rx_ring[rxring_idx].idx);
+ if (!_rtl_pci_init_one_rxdesc(hw, (u8 *)buffer_desc,
+ rxring_idx,
+ rtlpci->rx_ring[rxring_idx].idx))
+ return;
} else {
- _rtl_pci_init_one_rxdesc(hw, (u8 *)pdesc, rxring_idx,
- rtlpci->rx_ring[rxring_idx].idx);
-
+ if (!_rtl_pci_init_one_rxdesc(hw, (u8 *)pdesc,
+ rxring_idx,
+ rtlpci->rx_ring[rxring_idx].idx))
+ return;
if (rtlpci->rx_ring[rxring_idx].idx ==
rtlpci->rxringcount - 1)
rtlpriv->cfg->ops->set_desc(hw, (u8 *)pdesc,
--
2.1.2
^ permalink raw reply related
* [PATCH net] tcp6: don't move IP6CB before xfrm6_policy_check()
From: Nicolas Dichtel @ 2014-12-22 17:22 UTC (permalink / raw)
To: davem, eric.dumazet; +Cc: netdev, Nicolas Dichtel
In-Reply-To: <1419264939.11185.18.camel@edumazet-glaptop2.roam.corp.google.com>
When xfrm6_policy_check() is used, _decode_session6() is called after some
intermediate functions. This function uses IP6CB(), thus TCP_SKB_CB() must be
prepared after the call of xfrm6_policy_check().
Before this patch, scenarii with IPv6 + TCP + IPsec Transport are broken.
Fixes: 971f10eca186 ("tcp: better TCP_SKB_CB layout to reduce cache line misses")
Reported-by: Huaibin Wang <huaibin.wang@6wind.com>
Suggested-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: Nicolas Dichtel <nicolas.dichtel@6wind.com>
---
net/ipv6/tcp_ipv6.c | 45 +++++++++++++++++++++++++++++----------------
1 file changed, 29 insertions(+), 16 deletions(-)
diff --git a/net/ipv6/tcp_ipv6.c b/net/ipv6/tcp_ipv6.c
index 5ff87805258e..9c0b54e87b47 100644
--- a/net/ipv6/tcp_ipv6.c
+++ b/net/ipv6/tcp_ipv6.c
@@ -1387,6 +1387,28 @@ ipv6_pktoptions:
return 0;
}
+static void tcp_v6_fill_cb(struct sk_buff *skb, const struct ipv6hdr *hdr,
+ const struct tcphdr *th)
+{
+ /* This is tricky: we move IP6CB at its correct location into
+ * TCP_SKB_CB(). It must be done after xfrm6_policy_check(), because
+ * _decode_session6() uses IP6CB().
+ * barrier() makes sure compiler won't play aliasing games.
+ */
+ memmove(&TCP_SKB_CB(skb)->header.h6, IP6CB(skb),
+ sizeof(struct inet6_skb_parm));
+ barrier();
+
+ TCP_SKB_CB(skb)->seq = ntohl(th->seq);
+ TCP_SKB_CB(skb)->end_seq = (TCP_SKB_CB(skb)->seq + th->syn + th->fin +
+ skb->len - th->doff*4);
+ TCP_SKB_CB(skb)->ack_seq = ntohl(th->ack_seq);
+ TCP_SKB_CB(skb)->tcp_flags = tcp_flag_byte(th);
+ TCP_SKB_CB(skb)->tcp_tw_isn = 0;
+ TCP_SKB_CB(skb)->ip_dsfield = ipv6_get_dsfield(hdr);
+ TCP_SKB_CB(skb)->sacked = 0;
+}
+
static int tcp_v6_rcv(struct sk_buff *skb)
{
const struct tcphdr *th;
@@ -1418,24 +1440,9 @@ static int tcp_v6_rcv(struct sk_buff *skb)
th = tcp_hdr(skb);
hdr = ipv6_hdr(skb);
- /* This is tricky : We move IPCB at its correct location into TCP_SKB_CB()
- * barrier() makes sure compiler wont play fool^Waliasing games.
- */
- memmove(&TCP_SKB_CB(skb)->header.h6, IP6CB(skb),
- sizeof(struct inet6_skb_parm));
- barrier();
-
- TCP_SKB_CB(skb)->seq = ntohl(th->seq);
- TCP_SKB_CB(skb)->end_seq = (TCP_SKB_CB(skb)->seq + th->syn + th->fin +
- skb->len - th->doff*4);
- TCP_SKB_CB(skb)->ack_seq = ntohl(th->ack_seq);
- TCP_SKB_CB(skb)->tcp_flags = tcp_flag_byte(th);
- TCP_SKB_CB(skb)->tcp_tw_isn = 0;
- TCP_SKB_CB(skb)->ip_dsfield = ipv6_get_dsfield(hdr);
- TCP_SKB_CB(skb)->sacked = 0;
sk = __inet6_lookup_skb(&tcp_hashinfo, skb, th->source, th->dest,
- tcp_v6_iif(skb));
+ inet6_iif(skb));
if (!sk)
goto no_tcp_socket;
@@ -1451,6 +1458,8 @@ process:
if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb))
goto discard_and_relse;
+ tcp_v6_fill_cb(skb, hdr, th);
+
#ifdef CONFIG_TCP_MD5SIG
if (tcp_v6_inbound_md5_hash(sk, skb))
goto discard_and_relse;
@@ -1482,6 +1491,8 @@ no_tcp_socket:
if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb))
goto discard_it;
+ tcp_v6_fill_cb(skb, hdr, th);
+
if (skb->len < (th->doff<<2) || tcp_checksum_complete(skb)) {
csum_error:
TCP_INC_STATS_BH(net, TCP_MIB_CSUMERRORS);
@@ -1505,6 +1516,8 @@ do_time_wait:
goto discard_it;
}
+ tcp_v6_fill_cb(skb, hdr, th);
+
if (skb->len < (th->doff<<2)) {
inet_twsk_put(inet_twsk(sk));
goto bad_packet;
--
2.1.0
^ permalink raw reply related
* Re: Regression with commit 971f10eca186 ("tcp: better TCP_SKB_CB layout to reduce cache line misses")
From: Nicolas Dichtel @ 2014-12-22 16:54 UTC (permalink / raw)
To: Eric Dumazet; +Cc: netdev, Eric Dumazet
In-Reply-To: <1419264939.11185.18.camel@edumazet-glaptop2.roam.corp.google.com>
Le 22/12/2014 17:15, Eric Dumazet a écrit :
> On Mon, 2014-12-22 at 16:17 +0100, Nicolas Dichtel wrote:
>> One of our engineer (Huaibin Wang <huaibin.wang@6wind.com>) has reported and
>> analysed this bug:
>>
>> This commit introduces a regression with IPv6 + IPsec transport + TCP.
>>
>> In TCP (net/ipv6/tcp_ipv6.c), xfrm6_policy_check() is called and thus, after
>> some intermediate functions, _decode_session6() is also called.
>>
>> This function uses IP6CB() (u8 nexthdr = nh[IP6CB(skb)->nhoff]), which is wrong
>> becauses it has been moved to the end of TCP_SKB_CB().
>>
>> Not sure what is the best way to fix this, any suggestion?
>
> Thanks for the report
>
> Presumably tcp_v6_rcv() needs to be reordered so that
> xfrm6_policy_check() calls are done before the CB swap.
>
> swap should probably be done right before bh_lock_sock_nested()
>
> I am currently traveling and I am not sure if I can get Internet access
> to post a patch soon.
Ok, thank you for the tip. I will try to cook a patch.
^ permalink raw reply
* Re: virtio_net: Fix napi poll list corruption
From: Marcelo Ricardo Leitner @ 2014-12-22 16:19 UTC (permalink / raw)
To: Herbert Xu, David Vrabel
Cc: netdev, xen-devel, konrad.wilk, boris.ostrovsky, edumazet,
David S. Miller
In-Reply-To: <20141220002327.GA31975@gondor.apana.org.au>
On 19-12-2014 22:23, Herbert Xu wrote:
> David Vrabel <david.vrabel@citrix.com> wrote:
>> After d75b1ade567ffab085e8adbbdacf0092d10cd09c (net: less interrupt
>> masking in NAPI) the napi instance is removed from the per-cpu list
>> prior to calling the n->poll(), and is only requeued if all of the
>> budget was used. This inadvertently broke netfront because netfront
>> does not use NAPI correctly.
>
> A similar bug exists in virtio_net.
>
> -- >8 --
> The commit d75b1ade567ffab085e8adbbdacf0092d10cd09c (net: less
> interrupt masking in NAPI) breaks virtio_net in an insidious way.
>
> It is now required that if the entire budget is consumed when poll
> returns, the napi poll_list must remain empty. However, like some
> other drivers virtio_net tries to do a last-ditch check and if
> there is more work it will call napi_schedule and then immediately
> process some of this new work. Should the entire budget be consumed
> while processing such new work then we will violate the new caller
> contract.
>
> This patch fixes this by not touching any work when we reschedule
> in virtio_net.
>
> The worst part of this bug is that the list corruption causes other
> napi users to be moved off-list. In my case I was chasing a stall
> in IPsec (IPsec uses netif_rx) and I only belatedly realised that it
> was virtio_net which caused the stall even though the virtio_net
> poll was still functioning perfectly after IPsec stalled.
Thanks for finding/fixing this, Herbert. I was debugging this one too. In my
case, vxlan interface was getting stuck.
Marcelo
^ permalink raw reply
* Re: Announce: follow #netdev01 on tweeter
From: Jesper Dangaard Brouer @ 2014-12-22 16:17 UTC (permalink / raw)
To: Jamal Hadi Salim; +Cc: brouer, netdev@vger.kernel.org, Richard Guy Briggs
In-Reply-To: <54970E46.1070902@mojatatu.com>
On Sun, 21 Dec 2014 13:15:34 -0500
Jamal Hadi Salim <jhs@mojatatu.com> wrote:
> Please help us advertise netdev01.
> For folks with twitter accounts, please follow #netdev01
> and retweet the announcements when they come in.
Link to the account:
https://twitter.com/netdev01/
--
Best regards,
Jesper Dangaard Brouer
MSc.CS, Sr. Network Kernel Developer at Red Hat
Author of http://www.iptv-analyzer.org
LinkedIn: http://www.linkedin.com/in/brouer
^ permalink raw reply
* Re: Regression with commit 971f10eca186 ("tcp: better TCP_SKB_CB layout to reduce cache line misses")
From: Eric Dumazet @ 2014-12-22 16:15 UTC (permalink / raw)
To: nicolas.dichtel; +Cc: netdev, Eric Dumazet
In-Reply-To: <54983612.8000600@6wind.com>
On Mon, 2014-12-22 at 16:17 +0100, Nicolas Dichtel wrote:
> One of our engineer (Huaibin Wang <huaibin.wang@6wind.com>) has reported and
> analysed this bug:
>
> This commit introduces a regression with IPv6 + IPsec transport + TCP.
>
> In TCP (net/ipv6/tcp_ipv6.c), xfrm6_policy_check() is called and thus, after
> some intermediate functions, _decode_session6() is also called.
>
> This function uses IP6CB() (u8 nexthdr = nh[IP6CB(skb)->nhoff]), which is wrong
> becauses it has been moved to the end of TCP_SKB_CB().
>
> Not sure what is the best way to fix this, any suggestion?
Thanks for the report
Presumably tcp_v6_rcv() needs to be reordered so that
xfrm6_policy_check() calls are done before the CB swap.
swap should probably be done right before bh_lock_sock_nested()
I am currently traveling and I am not sure if I can get Internet access
to post a patch soon.
^ permalink raw reply
* [PATCH net] net: Generalize ndo_gso_check to ndo_features_check
From: Jesse Gross @ 2014-12-22 16:03 UTC (permalink / raw)
To: David Miller; +Cc: netdev, Tom Herbert, Joe Stringer, Eric Dumazet
GSO isn't the only offload feature with restrictions that
potentially can't be expressed with the current features mechanism.
Checksum is another although it's a general issue that could in
theory apply to anything. Even if it may be possible to
implement these restrictions in other ways, it can result in
duplicate code or inefficient per-packet behavior.
This generalizes ndo_gso_check so that drivers can remove any
features that don't make sense for a given packet, similar to
netif_skb_features(). It also converts existing driver
restrictions to the new format, completing the work that was
done to support tunnel protocols since the issues apply to
checksums as well.
CC: Tom Herbert <therbert@google.com>
CC: Joe Stringer <joestringer@nicira.com>
CC: Eric Dumazet <edumazet@google.com>
Signed-off-by: Jesse Gross <jesse@nicira.com>
Fixes: 04ffcb255f22 ("net: Add ndo_gso_check")
---
drivers/net/ethernet/broadcom/bnx2x/bnx2x_main.c | 8 ++++---
drivers/net/ethernet/emulex/benet/be_main.c | 8 ++++---
drivers/net/ethernet/mellanox/mlx4/en_netdev.c | 10 +++++----
drivers/net/ethernet/qlogic/qlcnic/qlcnic_main.c | 8 ++++---
include/linux/netdevice.h | 20 +++++++++--------
include/net/vxlan.h | 28 ++++++++++++++++++++----
net/core/dev.c | 28 ++++++++++++++++--------
7 files changed, 75 insertions(+), 35 deletions(-)
diff --git a/drivers/net/ethernet/broadcom/bnx2x/bnx2x_main.c b/drivers/net/ethernet/broadcom/bnx2x/bnx2x_main.c
index 9f5e387..72eef9f 100644
--- a/drivers/net/ethernet/broadcom/bnx2x/bnx2x_main.c
+++ b/drivers/net/ethernet/broadcom/bnx2x/bnx2x_main.c
@@ -12553,9 +12553,11 @@ static int bnx2x_get_phys_port_id(struct net_device *netdev,
return 0;
}
-static bool bnx2x_gso_check(struct sk_buff *skb, struct net_device *dev)
+static netdev_features_t bnx2x_features_check(struct sk_buff *skb,
+ struct net_device *dev,
+ netdev_features_t features)
{
- return vxlan_gso_check(skb);
+ return vxlan_features_check(skb, features);
}
static const struct net_device_ops bnx2x_netdev_ops = {
@@ -12589,7 +12591,7 @@ static const struct net_device_ops bnx2x_netdev_ops = {
#endif
.ndo_get_phys_port_id = bnx2x_get_phys_port_id,
.ndo_set_vf_link_state = bnx2x_set_vf_link_state,
- .ndo_gso_check = bnx2x_gso_check,
+ .ndo_features_check = bnx2x_features_check,
};
static int bnx2x_set_coherency_mask(struct bnx2x *bp)
diff --git a/drivers/net/ethernet/emulex/benet/be_main.c b/drivers/net/ethernet/emulex/benet/be_main.c
index 1960731..41a0a54 100644
--- a/drivers/net/ethernet/emulex/benet/be_main.c
+++ b/drivers/net/ethernet/emulex/benet/be_main.c
@@ -4459,9 +4459,11 @@ done:
adapter->vxlan_port_count--;
}
-static bool be_gso_check(struct sk_buff *skb, struct net_device *dev)
+static netdev_features_t be_features_check(struct sk_buff *skb,
+ struct net_device *dev,
+ netdev_features_t features)
{
- return vxlan_gso_check(skb);
+ return vxlan_features_check(skb, features);
}
#endif
@@ -4492,7 +4494,7 @@ static const struct net_device_ops be_netdev_ops = {
#ifdef CONFIG_BE2NET_VXLAN
.ndo_add_vxlan_port = be_add_vxlan_port,
.ndo_del_vxlan_port = be_del_vxlan_port,
- .ndo_gso_check = be_gso_check,
+ .ndo_features_check = be_features_check,
#endif
};
diff --git a/drivers/net/ethernet/mellanox/mlx4/en_netdev.c b/drivers/net/ethernet/mellanox/mlx4/en_netdev.c
index 190cbd9..d0d6dc1 100644
--- a/drivers/net/ethernet/mellanox/mlx4/en_netdev.c
+++ b/drivers/net/ethernet/mellanox/mlx4/en_netdev.c
@@ -2365,9 +2365,11 @@ static void mlx4_en_del_vxlan_port(struct net_device *dev,
queue_work(priv->mdev->workqueue, &priv->vxlan_del_task);
}
-static bool mlx4_en_gso_check(struct sk_buff *skb, struct net_device *dev)
+static netdev_features_t mlx4_en_features_check(struct sk_buff *skb,
+ struct net_device *dev,
+ netdev_features_t features)
{
- return vxlan_gso_check(skb);
+ return vxlan_features_check(skb, features);
}
#endif
@@ -2400,7 +2402,7 @@ static const struct net_device_ops mlx4_netdev_ops = {
#ifdef CONFIG_MLX4_EN_VXLAN
.ndo_add_vxlan_port = mlx4_en_add_vxlan_port,
.ndo_del_vxlan_port = mlx4_en_del_vxlan_port,
- .ndo_gso_check = mlx4_en_gso_check,
+ .ndo_features_check = mlx4_en_features_check,
#endif
};
@@ -2434,7 +2436,7 @@ static const struct net_device_ops mlx4_netdev_ops_master = {
#ifdef CONFIG_MLX4_EN_VXLAN
.ndo_add_vxlan_port = mlx4_en_add_vxlan_port,
.ndo_del_vxlan_port = mlx4_en_del_vxlan_port,
- .ndo_gso_check = mlx4_en_gso_check,
+ .ndo_features_check = mlx4_en_features_check,
#endif
};
diff --git a/drivers/net/ethernet/qlogic/qlcnic/qlcnic_main.c b/drivers/net/ethernet/qlogic/qlcnic/qlcnic_main.c
index 1aa25b1..9929b97 100644
--- a/drivers/net/ethernet/qlogic/qlcnic/qlcnic_main.c
+++ b/drivers/net/ethernet/qlogic/qlcnic/qlcnic_main.c
@@ -505,9 +505,11 @@ static void qlcnic_del_vxlan_port(struct net_device *netdev,
adapter->flags |= QLCNIC_DEL_VXLAN_PORT;
}
-static bool qlcnic_gso_check(struct sk_buff *skb, struct net_device *dev)
+static netdev_features_t qlcnic_features_check(struct sk_buff *skb,
+ struct net_device *dev,
+ netdev_features_t features)
{
- return vxlan_gso_check(skb);
+ return vxlan_features_check(skb, features);
}
#endif
@@ -532,7 +534,7 @@ static const struct net_device_ops qlcnic_netdev_ops = {
#ifdef CONFIG_QLCNIC_VXLAN
.ndo_add_vxlan_port = qlcnic_add_vxlan_port,
.ndo_del_vxlan_port = qlcnic_del_vxlan_port,
- .ndo_gso_check = qlcnic_gso_check,
+ .ndo_features_check = qlcnic_features_check,
#endif
#ifdef CONFIG_NET_POLL_CONTROLLER
.ndo_poll_controller = qlcnic_poll_controller,
diff --git a/include/linux/netdevice.h b/include/linux/netdevice.h
index c31f74d..679e6e9 100644
--- a/include/linux/netdevice.h
+++ b/include/linux/netdevice.h
@@ -1012,12 +1012,15 @@ typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
* Callback to use for xmit over the accelerated station. This
* is used in place of ndo_start_xmit on accelerated net
* devices.
- * bool (*ndo_gso_check) (struct sk_buff *skb,
- * struct net_device *dev);
+ * netdev_features_t (*ndo_features_check) (struct sk_buff *skb,
+ * struct net_device *dev
+ * netdev_features_t features);
* Called by core transmit path to determine if device is capable of
- * performing GSO on a packet. The device returns true if it is
- * able to GSO the packet, false otherwise. If the return value is
- * false the stack will do software GSO.
+ * performing offload operations on a given packet. This is to give
+ * the device an opportunity to implement any restrictions that cannot
+ * be otherwise expressed by feature flags. The check is called with
+ * the set of features that the stack has calculated and it returns
+ * those the driver believes to be appropriate.
*
* int (*ndo_switch_parent_id_get)(struct net_device *dev,
* struct netdev_phys_item_id *psid);
@@ -1178,8 +1181,9 @@ struct net_device_ops {
struct net_device *dev,
void *priv);
int (*ndo_get_lock_subclass)(struct net_device *dev);
- bool (*ndo_gso_check) (struct sk_buff *skb,
- struct net_device *dev);
+ netdev_features_t (*ndo_features_check) (struct sk_buff *skb,
+ struct net_device *dev,
+ netdev_features_t features);
#ifdef CONFIG_NET_SWITCHDEV
int (*ndo_switch_parent_id_get)(struct net_device *dev,
struct netdev_phys_item_id *psid);
@@ -3611,8 +3615,6 @@ static inline bool netif_needs_gso(struct net_device *dev, struct sk_buff *skb,
netdev_features_t features)
{
return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
- (dev->netdev_ops->ndo_gso_check &&
- !dev->netdev_ops->ndo_gso_check(skb, dev)) ||
unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
(skb->ip_summed != CHECKSUM_UNNECESSARY)));
}
diff --git a/include/net/vxlan.h b/include/net/vxlan.h
index 57cccd0..903461a 100644
--- a/include/net/vxlan.h
+++ b/include/net/vxlan.h
@@ -1,6 +1,9 @@
#ifndef __NET_VXLAN_H
#define __NET_VXLAN_H 1
+#include <linux/ip.h>
+#include <linux/ipv6.h>
+#include <linux/if_vlan.h>
#include <linux/skbuff.h>
#include <linux/netdevice.h>
#include <linux/udp.h>
@@ -51,16 +54,33 @@ int vxlan_xmit_skb(struct vxlan_sock *vs,
__be32 src, __be32 dst, __u8 tos, __u8 ttl, __be16 df,
__be16 src_port, __be16 dst_port, __be32 vni, bool xnet);
-static inline bool vxlan_gso_check(struct sk_buff *skb)
+static inline netdev_features_t vxlan_features_check(struct sk_buff *skb,
+ netdev_features_t features)
{
- if ((skb_shinfo(skb)->gso_type & SKB_GSO_UDP_TUNNEL) &&
+ u8 l4_hdr = 0;
+
+ if (!skb->encapsulation)
+ return features;
+
+ switch (vlan_get_protocol(skb)) {
+ case htons(ETH_P_IP):
+ l4_hdr = ip_hdr(skb)->protocol;
+ break;
+ case htons(ETH_P_IPV6):
+ l4_hdr = ipv6_hdr(skb)->nexthdr;
+ break;
+ default:
+ return features;;
+ }
+
+ if ((l4_hdr == IPPROTO_UDP) &&
(skb->inner_protocol_type != ENCAP_TYPE_ETHER ||
skb->inner_protocol != htons(ETH_P_TEB) ||
(skb_inner_mac_header(skb) - skb_transport_header(skb) !=
sizeof(struct udphdr) + sizeof(struct vxlanhdr))))
- return false;
+ return features & ~(NETIF_F_ALL_CSUM | NETIF_F_GSO_MASK);
- return true;
+ return features;
}
/* IP header + UDP + VXLAN + Ethernet header */
diff --git a/net/core/dev.c b/net/core/dev.c
index f411c28..fc13f72 100644
--- a/net/core/dev.c
+++ b/net/core/dev.c
@@ -2562,7 +2562,7 @@ static netdev_features_t harmonize_features(struct sk_buff *skb,
netdev_features_t netif_skb_features(struct sk_buff *skb)
{
- const struct net_device *dev = skb->dev;
+ struct net_device *dev = skb->dev;
netdev_features_t features = dev->features;
u16 gso_segs = skb_shinfo(skb)->gso_segs;
__be16 protocol = skb->protocol;
@@ -2570,11 +2570,19 @@ netdev_features_t netif_skb_features(struct sk_buff *skb)
if (gso_segs > dev->gso_max_segs || gso_segs < dev->gso_min_segs)
features &= ~NETIF_F_GSO_MASK;
+ /* If encapsulation offload request, verify we are testing
+ * hardware encapsulation features instead of standard
+ * features for the netdev
+ */
+ if (skb->encapsulation)
+ features = netdev_intersect_features(features,
+ dev->hw_enc_features);
+
if (protocol == htons(ETH_P_8021Q) || protocol == htons(ETH_P_8021AD)) {
struct vlan_ethhdr *veh = (struct vlan_ethhdr *)skb->data;
protocol = veh->h_vlan_encapsulated_proto;
} else if (!vlan_tx_tag_present(skb)) {
- return harmonize_features(skb, features);
+ goto finalize;
}
features = netdev_intersect_features(features,
@@ -2591,6 +2599,15 @@ netdev_features_t netif_skb_features(struct sk_buff *skb)
NETIF_F_HW_VLAN_CTAG_TX |
NETIF_F_HW_VLAN_STAG_TX);
+finalize:
+ if (dev->netdev_ops->ndo_features_check) {
+ netdev_features_t dev_features;
+
+ dev_features = dev->netdev_ops->ndo_features_check(skb, dev,
+ features);
+ features = netdev_intersect_features(features, dev_features);
+ }
+
return harmonize_features(skb, features);
}
EXPORT_SYMBOL(netif_skb_features);
@@ -2661,13 +2678,6 @@ static struct sk_buff *validate_xmit_skb(struct sk_buff *skb, struct net_device
if (unlikely(!skb))
goto out_null;
- /* If encapsulation offload request, verify we are testing
- * hardware encapsulation features instead of standard
- * features for the netdev
- */
- if (skb->encapsulation)
- features &= dev->hw_enc_features;
-
if (netif_needs_gso(dev, skb, features)) {
struct sk_buff *segs;
--
1.9.1
^ permalink raw reply related
* [PATCH iproute2 v2] tc: Show classes in tree view
From: Vadim Kochan @ 2014-12-22 15:51 UTC (permalink / raw)
To: netdev; +Cc: Vadim Kochan
From: Vadim Kochan <vadim4j@gmail.com>
Added new '-t[ree]' which shows classes dependency
in the tree view. Meanwhile only generic stats info
is supported.
e.g.:
$ tc/tc -t class show dev tap0
+---(1:2) htb rate 6Mbit ceil 6Mbit burst 15Kb cburst 1599b
| +---(1:40) htb prio 0 rate 5Mbit ceil 5Mbit burst 15Kb cburst 1600b
| +---(1:50) htb rate 3Mbit ceil 6Mbit burst 15Kb cburst 1599b
| | +---(1:51) htb prio 0 rate 1Kbit ceil 6Mbit burst 15Kb cburst 1599b
| |
| +---(1:60) htb prio 0 rate 1Kbit ceil 6Mbit burst 15Kb cburst 1599b
|
+---(1:1) htb rate 6Mbit ceil 6Mbit burst 15Kb cburst 1599b
+---(1:10) htb prio 0 rate 5Mbit ceil 5Mbit burst 15Kb cburst 1600b
+---(1:20) htb prio 0 rate 3Mbit ceil 6Mbit burst 15Kb cburst 1599b
+---(1:30) htb prio 0 rate 1Kbit ceil 6Mbit burst 15Kb cburst 1599b
$ tc/tc -t -s class show dev tap0
+---(1:2) htb rate 6Mbit ceil 6Mbit burst 15Kb cburst 1599b
| | Sent 0 bytes 0 pkt (dropped 0, overlimits 0 requeues 0)
| | rate 0bit 0pps backlog 0b 0p requeues 0
| |
| +---(1:40) htb prio 0 rate 5Mbit ceil 5Mbit burst 15Kb cburst 1600b
| | Sent 0 bytes 0 pkt (dropped 0, overlimits 0 requeues 0)
| | rate 0bit 0pps backlog 0b 0p requeues 0
| |
| +---(1:50) htb rate 3Mbit ceil 6Mbit burst 15Kb cburst 1599b
| | | Sent 0 bytes 0 pkt (dropped 0, overlimits 0 requeues 0)
| | | rate 0bit 0pps backlog 0b 0p requeues 0
| | |
| | +---(1:51) htb prio 0 rate 1Kbit ceil 6Mbit burst 15Kb cburst 1599b
| | Sent 0 bytes 0 pkt (dropped 0, overlimits 0 requeues 0)
| | rate 0bit 0pps backlog 0b 0p requeues 0
| |
| +---(1:60) htb prio 0 rate 1Kbit ceil 6Mbit burst 15Kb cburst 1599b
| Sent 0 bytes 0 pkt (dropped 0, overlimits 0 requeues 0)
| rate 0bit 0pps backlog 0b 0p requeues 0
|
+---(1:1) htb rate 6Mbit ceil 6Mbit burst 15Kb cburst 1599b
| Sent 0 bytes 0 pkt (dropped 0, overlimits 0 requeues 0)
| rate 0bit 0pps backlog 0b 0p requeues 0
|
+---(1:10) htb prio 0 rate 5Mbit ceil 5Mbit burst 15Kb cburst 1600b
| Sent 0 bytes 0 pkt (dropped 0, overlimits 0 requeues 0)
| rate 0bit 0pps backlog 0b 0p requeues 0
|
+---(1:20) htb prio 0 rate 3Mbit ceil 6Mbit burst 15Kb cburst 1599b
| Sent 0 bytes 0 pkt (dropped 0, overlimits 0 requeues 0)
| rate 0bit 0pps backlog 0b 0p requeues 0
|
+---(1:30) htb prio 0 rate 1Kbit ceil 6Mbit burst 15Kb cburst 1599b
Sent 0 bytes 0 pkt (dropped 0, overlimits 0 requeues 0)
rate 0bit 0pps backlog 0b 0p requeues 0
Signed-off-by: Vadim Kochan <vadim4j@gmail.com>
---
Changes v2:
Removed "Date:" from commit message which was added by mistake.
Changes RFC -> PATCH:
#1 get rid of INIT_HLIST_NODE
#2 added sample output to commit message
#3 use "show_tree=1" instead of "show_tree++"
#4 no need update include/hlist.h (because of #1)
#5 changed a little tree output: parentheses around class id instead of qdisc name
tc/tc.c | 5 +-
tc/tc_class.c | 165 ++++++++++++++++++++++++++++++++++++++++++++++++++++++++-
tc/tc_common.h | 2 +
3 files changed, 169 insertions(+), 3 deletions(-)
diff --git a/tc/tc.c b/tc/tc.c
index 9b50e74..30950a6 100644
--- a/tc/tc.c
+++ b/tc/tc.c
@@ -34,8 +34,9 @@ int show_stats = 0;
int show_details = 0;
int show_raw = 0;
int show_pretty = 0;
-int batch_mode = 0;
+int show_tree = 0;
+int batch_mode = 0;
int resolve_hosts = 0;
int use_iec = 0;
int force = 0;
@@ -278,6 +279,8 @@ int main(int argc, char **argv)
++show_raw;
} else if (matches(argv[1], "-pretty") == 0) {
++show_pretty;
+ } else if (matches(argv[1], "-tree") == 0) {
+ show_tree = 1;
} else if (matches(argv[1], "-Version") == 0) {
printf("tc utility, iproute2-ss%s\n", SNAPSHOT);
return 0;
diff --git a/tc/tc_class.c b/tc/tc_class.c
index e56bf07..a7b3ecd 100644
--- a/tc/tc_class.c
+++ b/tc/tc_class.c
@@ -24,6 +24,22 @@
#include "utils.h"
#include "tc_util.h"
#include "tc_common.h"
+#include "hlist.h"
+
+struct cls_node {
+ struct hlist_node hlist;
+ __u32 handle;
+ __u32 parent;
+ int level;
+ struct cls_node *cls_parent;
+ struct cls_node *cls_right;
+ struct rtattr *attr;
+ int attr_len;
+ int childs_count;
+};
+
+static struct hlist_head cls_list = {};
+static struct hlist_head root_cls_list = {};
static void usage(void);
@@ -148,13 +164,149 @@ int filter_ifindex;
__u32 filter_qdisc;
__u32 filter_classid;
+static void tree_cls_add(__u32 parent, __u32 handle, struct rtattr *attr, int len)
+{
+ struct cls_node *cls = malloc(sizeof(struct cls_node));
+
+ memset(cls, 0, sizeof(*cls));
+ cls->handle = handle;
+ cls->parent = parent;
+ cls->attr = malloc(len);
+ cls->attr_len = len;
+
+ memcpy(cls->attr, attr, len);
+
+ if (parent == TC_H_ROOT)
+ hlist_add_head(&cls->hlist, &root_cls_list);
+ else
+ hlist_add_head(&cls->hlist, &cls_list);
+}
+
+static void tree_cls_indent(char *buf, struct cls_node *cls, int is_newline,
+ int add_spaces)
+{
+ char spaces[100] = {0};
+
+ while (cls && cls->cls_parent) {
+ cls->cls_parent->cls_right = cls;
+ cls = cls->cls_parent;
+ }
+ while (cls && cls->cls_right)
+ {
+ if (cls->hlist.next)
+ strcat(buf, "| ");
+ else
+ strcat(buf, " ");
+
+ cls = cls->cls_right;
+ }
+
+ if (is_newline) {
+ if (cls->hlist.next && cls->childs_count)
+ strcat(buf, "| |");
+ else if (cls->hlist.next)
+ strcat(buf, "| ");
+ else if (cls->childs_count)
+ strcat(buf, " |");
+ else if (!cls->hlist.next)
+ strcat(buf, " ");
+ }
+ if (add_spaces > 0)
+ {
+ sprintf(spaces, "%-*s", add_spaces, "");
+ strcat(buf, spaces);
+ }
+}
+
+static void tree_cls_show(FILE *fp, char *buf, struct hlist_head *root_list, int level)
+{
+ struct hlist_node *n, *tmp_cls;
+ char cls_id_str[256] = {};
+ struct rtattr * tb[TCA_MAX+1] = {};
+ struct qdisc_util *q;
+ char str[100] = {};
+
+ hlist_for_each_safe(n, tmp_cls, root_list) {
+ struct hlist_node *c, *tmp_chld;
+ struct hlist_head childs = {};
+ struct cls_node *cls = container_of(n, struct cls_node, hlist);
+
+ hlist_for_each_safe(c, tmp_chld, &cls_list) {
+ struct cls_node *child = container_of(c, struct cls_node, hlist);
+
+ if (cls->handle == child->parent) {
+ hlist_del(c);
+ hlist_add_head(c, &childs);
+ cls->childs_count++;
+ child->cls_parent = cls;
+ }
+ }
+
+ tree_cls_indent(buf, cls, 0, 0);
+
+ print_tc_classid(cls_id_str, sizeof(cls_id_str), cls->handle);
+ sprintf(str, "+---(%s)", cls_id_str);
+ strcat(buf, str);
+
+ parse_rtattr(tb, TCA_MAX, cls->attr, cls->attr_len);
+
+ if (tb[TCA_KIND] == NULL) {
+ strcat(buf, " [unknown qdisc kind] ");
+ } else {
+ const char *kind = rta_getattr_str(tb[TCA_KIND]);
+
+ sprintf(str, " %s ", kind);
+ strcat(buf, str);
+ fprintf(fp, "%s", buf);
+ buf[0] = '\0';
+
+ q = get_qdisc_kind(kind);
+ if (q && q->print_copt) {
+ q->print_copt(q, fp, tb[TCA_OPTIONS]);
+ }
+ if (q && show_stats)
+ {
+ int cls_indent = strlen(q->id) - 2 +
+ strlen(cls_id_str);
+ struct rtattr *xstats = NULL;
+
+ tree_cls_indent(buf, cls, 1, cls_indent);
+
+ if (tb[TCA_STATS] || tb[TCA_STATS2]) {
+ fprintf(fp, "\n");
+ print_tcstats_attr(fp, tb, buf, &xstats);
+ buf[0] = '\0';
+ }
+ if (cls->hlist.next || cls->childs_count)
+ {
+ strcat(buf, "\n");
+ tree_cls_indent(buf, cls, 1, 0);
+ }
+ }
+ }
+ free(cls->attr);
+ fprintf(fp, "%s\n", buf);
+ buf[0] = '\0';
+
+ tree_cls_show(fp, buf, &childs, level + 1);
+ if (!cls->hlist.next) {
+ tree_cls_indent(buf, cls, 0, 0);
+ strcat(buf, "\n");
+ }
+
+ fprintf(fp, "%s", buf);
+ buf[0] = '\0';
+ free(cls);
+ }
+}
+
int print_class(const struct sockaddr_nl *who,
struct nlmsghdr *n, void *arg)
{
FILE *fp = (FILE*)arg;
struct tcmsg *t = NLMSG_DATA(n);
int len = n->nlmsg_len;
- struct rtattr * tb[TCA_MAX+1];
+ struct rtattr * tb[TCA_MAX+1] = {};
struct qdisc_util *q;
char abuf[256];
@@ -167,13 +319,18 @@ int print_class(const struct sockaddr_nl *who,
fprintf(stderr, "Wrong len %d\n", len);
return -1;
}
+
+ if (show_tree) {
+ tree_cls_add(t->tcm_parent, t->tcm_handle, TCA_RTA(t), len);
+ return 0;
+ }
+
if (filter_qdisc && TC_H_MAJ(t->tcm_handle^filter_qdisc))
return 0;
if (filter_classid && t->tcm_handle != filter_classid)
return 0;
- memset(tb, 0, sizeof(tb));
parse_rtattr(tb, TCA_MAX, TCA_RTA(t), len);
if (tb[TCA_KIND] == NULL) {
@@ -236,6 +393,7 @@ static int tc_class_list(int argc, char **argv)
{
struct tcmsg t;
char d[16];
+ char buf[1024] = {0};
memset(&t, 0, sizeof(t));
t.tcm_family = AF_UNSPEC;
@@ -306,6 +464,9 @@ static int tc_class_list(int argc, char **argv)
return 1;
}
+ if (show_tree)
+ tree_cls_show(stdout, &buf[0], &root_cls_list, 0);
+
return 0;
}
diff --git a/tc/tc_common.h b/tc/tc_common.h
index 4f88856..0ee009b 100644
--- a/tc/tc_common.h
+++ b/tc/tc_common.h
@@ -19,3 +19,5 @@ extern int parse_estimator(int *p_argc, char ***p_argv, struct tc_estimator *est
struct tc_sizespec;
extern int parse_size_table(int *p_argc, char ***p_argv, struct tc_sizespec *s);
extern int check_size_table_opts(struct tc_sizespec *s);
+
+extern int show_tree;
--
2.1.3
^ permalink raw reply related
* Re: [PATCH net] net: ndo_gso_check() must force segmentation
From: Jesse Gross @ 2014-12-22 15:59 UTC (permalink / raw)
To: Eric Dumazet
Cc: Hayes Wang, Tom Herbert, David Miller, netdev@vger.kernel.org,
nic_swsd
In-Reply-To: <1419261697.11185.15.camel@edumazet-glaptop2.roam.corp.google.com>
On Mon, Dec 22, 2014 at 10:21 AM, Eric Dumazet <eric.dumazet@gmail.com> wrote:
> From: Eric Dumazet <edumazet@google.com>
>
> If ndo_gso_check() returns true, it means a driver wants the stack
> to perform software segmentation, even if device features initially
> claimed hardware was able handle a TSO packet.
>
> This means netif_needs_gso() needs to modify the features.
>
> Signed-off-by: Eric Dumazet <edumazet@google.com>
> Reported-by: Hayes Wang <hayeswang@realtek.com>
> Tested-by: Hayes Wang <hayeswang@realtek.com>
> Fixes: 04ffcb255f22 ("net: Add ndo_gso_check")
There's actually another problem with ndo_gso_check() - it doesn't
deal with other features like checksum offloading which usually have
similar constraints. As a result, the GSO code generates segments with
checksums offloading that immediately fails.
I have a patch that I was working on last night that behaves similarly
to yours but is also generalized to handle both cases. Let me send it
out now.
^ permalink raw reply
* Re: [PATCH] bonding: avoid re-entry of bond_release
From: Andy Gospodarek @ 2014-12-22 15:45 UTC (permalink / raw)
To: Wengang; +Cc: Ding Tianhong, netdev
In-Reply-To: <5497D6B0.2040402@oracle.com>
On Mon, Dec 22, 2014 at 04:30:40PM +0800, Wengang wrote:
> OK. Will change as suggested and re-post.
Sounds great. Thanks for your work on this.
>
> thanks,
> wengang
>
> 于 2014年12月22日 10:05, Ding Tianhong 写道:
> >On 2014/12/22 9:09, Wengang wrote:
> >>Hi Andy and Ding,
> >>
> >>Thanks for your reviews!
> >>In the ioctl path, removing a interface that is not currently actually a slave
> >>can happen from user space(by mistake), we should avoid the noisy message.
> >>
> >>While, __bond_release_one() has another call path which is from bond_uninit().
> >>In the later case, it should be treated as an error if the interface is not with
> >>IFF_SLAVE flag. To notice that error occurred, the message is printed. I think
> >>the message is needed for this path.
> >>
> >>How do you think?
> >>
> >Just like the bond_enslave(), it is only a warning.
> >
> >Ding
> >
> >>thanks,
> >>wengang
> >>
> >>于 2014年12月21日 10:01, Ding Tianhong 写道:
> >>>On 2014/12/19 23:11, Andy Gospodarek wrote:
> >>>>On Fri, Dec 19, 2014 at 04:56:57PM +0800, Wengang Wang wrote:
> >>>>>If bond_release is run against an interface which is already detached from
> >>>>>it's master, then there is an error message shown like
> >>>>> "<master name> cannot release <slave name>".
> >>>>>
> >>>>>The call path is:
> >>>>> bond_do_ioctl()
> >>>>> bond_release()
> >>>>> __bond_release_one()
> >>>>>
> >>>>>Though it does not really harm, the message the message is misleading.
> >>>>>This patch tries to avoid the message.
> >>>>>
> >>>>>Signed-off-by: Wengang Wang <wen.gang.wang@oracle.com>
> >>>>>---
> >>>>> drivers/net/bonding/bond_main.c | 5 ++++-
> >>>>> 1 file changed, 4 insertions(+), 1 deletion(-)
> >>>>>
> >>>>>diff --git a/drivers/net/bonding/bond_main.c b/drivers/net/bonding/bond_main.c
> >>>>>index 184c434..4a71bbd 100644
> >>>>>--- a/drivers/net/bonding/bond_main.c
> >>>>>+++ b/drivers/net/bonding/bond_main.c
> >>>>>@@ -3256,7 +3256,10 @@ static int bond_do_ioctl(struct net_device *bond_dev, struct ifreq *ifr, int cmd
> >>>>> break;
> >>>>> case BOND_RELEASE_OLD:
> >>>>> case SIOCBONDRELEASE:
> >>>>>- res = bond_release(bond_dev, slave_dev);
> >>>>>+ if (slave_dev->flags & IFF_SLAVE)
> >>>>>+ res = bond_release(bond_dev, slave_dev);
> >>>>>+ else
> >>>>>+ res = 0;
> >>>>Functionally this patch is fine, but I would prefer that you simply
> >>>>change the check in __bond_release_one to not be so noisy. There is a
> >>>>check[1] in bond_enslave to see if a slave is already in a bond and that
> >>>>just prints a message of netdev_dbg (rather than netdev_err) and it
> >>>>seems that would be appropriate for this type of message.
> >>>>
> >>>>[1] from bond_enslave():
> >>>>
> >>>> /* already enslaved */
> >>>> if (slave_dev->flags & IFF_SLAVE) {
> >>>> netdev_dbg(bond_dev, "Error: Device was already enslaved\n");
> >>>> return -EBUSY;
> >>>> }
> >>>>
> >>>>
> >>>>> break;
> >>>>> case BOND_SETHWADDR_OLD:
> >>>>> case SIOCBONDSETHWADDR:
> >>>>>--
> >>>agree ,use netdev_dbg looks more better and enough.
> >>>
> >>>Ding
> >>>
> >>>
> >>
>
^ permalink raw reply
* Re: e1000_netpoll(): disable_irq() triggers might_sleep() on linux-next
From: Bart Van Assche @ 2014-12-22 15:28 UTC (permalink / raw)
To: Sabrina Dubroca, Peter Zijlstra
Cc: Thomas Gleixner, netdev, linux-kernel, jeffrey.t.kirsher
In-Reply-To: <20141202163530.GA19420@kria>
On 12/02/14 17:35, Sabrina Dubroca wrote:
> Hello, sorry for the delay.
>
> 2014-10-29, 20:36:03 +0100, Peter Zijlstra wrote:
>> On Wed, Oct 29, 2014 at 07:33:00PM +0100, Thomas Gleixner wrote:
>>> Yuck. No. You are just papering over the problem.
>>>
>>> What happens if you add 'threadirqs' to the kernel command line? Or if
>>> the interrupt line is shared with a real threaded interrupt user?
>>>
>>> The proper solution is to have a poll_lock for e1000 which serializes
>>> the hardware interrupt against netpoll instead of using
>>> disable/enable_irq().
>>>
>>> In fact that's less expensive than the disable/enable_irq() dance and
>>> the chance of contention is pretty low. If done right it will be a
>>> NOOP for the CONFIG_NET_POLL_CONTROLLER=n case.
>>>
>>
>> OK a little something like so then I suppose.. But I suspect most all
>> the network drivers will need this and maybe more, disable_irq() is a
>> popular little thing and we 'just' changed semantics on them.
>>
>> ---
>> drivers/net/ethernet/intel/e1000/e1000.h | 2 ++
>> drivers/net/ethernet/intel/e1000/e1000_main.c | 22 +++++++++++++++++-----
>> kernel/irq/manage.c | 2 +-
>> 3 files changed, 20 insertions(+), 6 deletions(-)
>
> I've been running with variants of this patch, things seem ok.
>
> As noted earlier, there are a lot of drivers doing this disable_irq +
> irq_handler + enable_irq sequence. I found about 60.
> Many already take a lock in the interrupt handler, and look like we
> could just remove the call to disable_irq (example: cp_interrupt,
> drivers/net/ethernet/realtek/8139cp.c).
>
> Here's how I modified your patch. The locking compiles away if
> CONFIG_NET_POLL_CONTROLLER=n.
>
> I can work on converting all the drivers from disable_irq to
> netpoll_irq_lock, if that's okay with you.
>
> In igb there's also a synchronize_irq() called from the netpoll
> controller (in igb_irq_disable()), I think a similar locking scheme
> would work.
> I also saw a few disable_irq_nosync and disable_percpu_irq. These are
> okay?
>
> [ ... ]
Hello,
Earlier today I ran into the bug mentioned at the start of this thread
with kernel 3.19-rc1 and the e1000e driver. Can anyone tell me what the
latest status is ?
Thanks,
Bart.
^ permalink raw reply
* [PATCH net] net: ndo_gso_check() must force segmentation
From: Eric Dumazet @ 2014-12-22 15:21 UTC (permalink / raw)
To: Hayes Wang; +Cc: Tom Herbert, David Miller, netdev@vger.kernel.org, nic_swsd
In-Reply-To: <0835B3720019904CB8F7AA43166CEEB2ED5B01@RTITMBSV03.realtek.com.tw>
From: Eric Dumazet <edumazet@google.com>
If ndo_gso_check() returns true, it means a driver wants the stack
to perform software segmentation, even if device features initially
claimed hardware was able handle a TSO packet.
This means netif_needs_gso() needs to modify the features.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Reported-by: Hayes Wang <hayeswang@realtek.com>
Tested-by: Hayes Wang <hayeswang@realtek.com>
Fixes: 04ffcb255f22 ("net: Add ndo_gso_check")
---
drivers/net/macvtap.c | 2 +-
drivers/net/xen-netfront.c | 4 +++-
include/linux/netdevice.h | 18 ++++++++++++------
net/core/dev.c | 2 +-
4 files changed, 17 insertions(+), 9 deletions(-)
diff --git a/drivers/net/macvtap.c b/drivers/net/macvtap.c
index 7df221788cd4..0346bcfe72a5 100644
--- a/drivers/net/macvtap.c
+++ b/drivers/net/macvtap.c
@@ -314,7 +314,7 @@ static rx_handler_result_t macvtap_handle_frame(struct sk_buff **pskb)
*/
if (q->flags & IFF_VNET_HDR)
features |= vlan->tap_features;
- if (netif_needs_gso(dev, skb, features)) {
+ if (netif_needs_gso(dev, skb, &features)) {
struct sk_buff *segs = __skb_gso_segment(skb, features, false);
if (IS_ERR(segs))
diff --git a/drivers/net/xen-netfront.c b/drivers/net/xen-netfront.c
index 22bcb4e12e2a..9cacabaea175 100644
--- a/drivers/net/xen-netfront.c
+++ b/drivers/net/xen-netfront.c
@@ -578,6 +578,7 @@ static int xennet_start_xmit(struct sk_buff *skb, struct net_device *dev)
unsigned long flags;
struct netfront_queue *queue = NULL;
unsigned int num_queues = dev->real_num_tx_queues;
+ netdev_features_t features;
u16 queue_index;
/* Drop the packet if no queues are set up */
@@ -611,9 +612,10 @@ static int xennet_start_xmit(struct sk_buff *skb, struct net_device *dev)
spin_lock_irqsave(&queue->tx_lock, flags);
+ features = netif_skb_features(skb);
if (unlikely(!netif_carrier_ok(dev) ||
(slots > 1 && !xennet_can_sg(dev)) ||
- netif_needs_gso(dev, skb, netif_skb_features(skb)))) {
+ netif_needs_gso(dev, skb, &features))) {
spin_unlock_irqrestore(&queue->tx_lock, flags);
goto drop;
}
diff --git a/include/linux/netdevice.h b/include/linux/netdevice.h
index c31f74d76ebd..fb1f8c900df9 100644
--- a/include/linux/netdevice.h
+++ b/include/linux/netdevice.h
@@ -3608,13 +3608,19 @@ static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
}
static inline bool netif_needs_gso(struct net_device *dev, struct sk_buff *skb,
- netdev_features_t features)
+ netdev_features_t *features)
{
- return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
- (dev->netdev_ops->ndo_gso_check &&
- !dev->netdev_ops->ndo_gso_check(skb, dev)) ||
- unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
- (skb->ip_summed != CHECKSUM_UNNECESSARY)));
+ if (!skb_is_gso(skb))
+ return false;
+ if (!skb_gso_ok(skb, *features))
+ return true;
+ if (dev->netdev_ops->ndo_gso_check &&
+ !dev->netdev_ops->ndo_gso_check(skb, dev)) {
+ *features &= ~NETIF_F_GSO_MASK;
+ return true;
+ }
+ return skb->ip_summed != CHECKSUM_PARTIAL &&
+ skb->ip_summed != CHECKSUM_UNNECESSARY;
}
static inline void netif_set_gso_max_size(struct net_device *dev,
diff --git a/net/core/dev.c b/net/core/dev.c
index f411c28d0a66..b61c26b45bb7 100644
--- a/net/core/dev.c
+++ b/net/core/dev.c
@@ -2668,7 +2668,7 @@ static struct sk_buff *validate_xmit_skb(struct sk_buff *skb, struct net_device
if (skb->encapsulation)
features &= dev->hw_enc_features;
- if (netif_needs_gso(dev, skb, features)) {
+ if (netif_needs_gso(dev, skb, &features)) {
struct sk_buff *segs;
segs = skb_gso_segment(skb, features);
^ permalink raw reply related
* Regression with commit 971f10eca186 ("tcp: better TCP_SKB_CB layout to reduce cache line misses")
From: Nicolas Dichtel @ 2014-12-22 15:17 UTC (permalink / raw)
To: netdev, Eric Dumazet
One of our engineer (Huaibin Wang <huaibin.wang@6wind.com>) has reported and
analysed this bug:
This commit introduces a regression with IPv6 + IPsec transport + TCP.
In TCP (net/ipv6/tcp_ipv6.c), xfrm6_policy_check() is called and thus, after
some intermediate functions, _decode_session6() is also called.
This function uses IP6CB() (u8 nexthdr = nh[IP6CB(skb)->nhoff]), which is wrong
becauses it has been moved to the end of TCP_SKB_CB().
Not sure what is the best way to fix this, any suggestion?
Regards,
Nicolas
^ permalink raw reply
* Re: [PATCH net]tg3: tg3_disable_ints using uninitialized mailbox value to disable interrupts
From: Marcelo Ricardo Leitner @ 2014-12-22 14:06 UTC (permalink / raw)
To: Nils Holland, Prashant Sreedharan
Cc: davem, netdev, linux-pci, bhelgaas, rajatxjain, Michael Chan
In-Reply-To: <20141220221606.GA2591@teela.fritz.box>
On 20-12-2014 20:16, Nils Holland wrote:
> On Sat, Dec 20, 2014 at 12:16:17PM -0800, Prashant Sreedharan wrote:
>>
>> This driver bug was exposed because of the commit a7877b17a667 (PCI: Check only
>> the Vendor ID to identify Configuration Request Retry). Also this issue is only
>> seen in older generation chipsets like 5722 because config space write to offset
>> 0 from driver is possible.
>>
>> Fixed by initializing the interrupt mailbox registers before calling tg3_halt.
>>
>> Please queue for -stable.
>
> I gave this patch a try and can confirm what was to be expected: It
> fixes the issue and the network interface is once again working
> properly on my system. Thus, I guess the issue is adequately solved.
>
> Thanks to everyone involved, especially to Marcelo for additional
> debugging and the guys at Broadcom for the quick fix!
>
> Greetings,
> Nils
Same here, it works again.
Thanks!
Marcelo
^ permalink raw reply
* [PATCH iproute2] tc: Show classes in tree view
From: Vadim Kochan @ 2014-12-22 12:57 UTC (permalink / raw)
To: netdev; +Cc: Vadim Kochan
From: Vadim Kochan <vadim4j@gmail.com>
Date: Tue, 16 Dec 2014 22:59:30 +0200
Added new '-t[ree]' which shows classes dependency
in the tree view. Meanwhile only generic stats info
is supported.
e.g.:
$ tc/tc -t class show dev tap0
+---(1:2) htb rate 6Mbit ceil 6Mbit burst 15Kb cburst 1599b
| +---(1:40) htb prio 0 rate 5Mbit ceil 5Mbit burst 15Kb cburst 1600b
| +---(1:50) htb rate 3Mbit ceil 6Mbit burst 15Kb cburst 1599b
| | +---(1:51) htb prio 0 rate 1Kbit ceil 6Mbit burst 15Kb cburst 1599b
| |
| +---(1:60) htb prio 0 rate 1Kbit ceil 6Mbit burst 15Kb cburst 1599b
|
+---(1:1) htb rate 6Mbit ceil 6Mbit burst 15Kb cburst 1599b
+---(1:10) htb prio 0 rate 5Mbit ceil 5Mbit burst 15Kb cburst 1600b
+---(1:20) htb prio 0 rate 3Mbit ceil 6Mbit burst 15Kb cburst 1599b
+---(1:30) htb prio 0 rate 1Kbit ceil 6Mbit burst 15Kb cburst 1599b
$ tc/tc -t -s class show dev tap0
+---(1:2) htb rate 6Mbit ceil 6Mbit burst 15Kb cburst 1599b
| | Sent 0 bytes 0 pkt (dropped 0, overlimits 0 requeues 0)
| | rate 0bit 0pps backlog 0b 0p requeues 0
| |
| +---(1:40) htb prio 0 rate 5Mbit ceil 5Mbit burst 15Kb cburst 1600b
| | Sent 0 bytes 0 pkt (dropped 0, overlimits 0 requeues 0)
| | rate 0bit 0pps backlog 0b 0p requeues 0
| |
| +---(1:50) htb rate 3Mbit ceil 6Mbit burst 15Kb cburst 1599b
| | | Sent 0 bytes 0 pkt (dropped 0, overlimits 0 requeues 0)
| | | rate 0bit 0pps backlog 0b 0p requeues 0
| | |
| | +---(1:51) htb prio 0 rate 1Kbit ceil 6Mbit burst 15Kb cburst 1599b
| | Sent 0 bytes 0 pkt (dropped 0, overlimits 0 requeues 0)
| | rate 0bit 0pps backlog 0b 0p requeues 0
| |
| +---(1:60) htb prio 0 rate 1Kbit ceil 6Mbit burst 15Kb cburst 1599b
| Sent 0 bytes 0 pkt (dropped 0, overlimits 0 requeues 0)
| rate 0bit 0pps backlog 0b 0p requeues 0
|
+---(1:1) htb rate 6Mbit ceil 6Mbit burst 15Kb cburst 1599b
| Sent 0 bytes 0 pkt (dropped 0, overlimits 0 requeues 0)
| rate 0bit 0pps backlog 0b 0p requeues 0
|
+---(1:10) htb prio 0 rate 5Mbit ceil 5Mbit burst 15Kb cburst 1600b
| Sent 0 bytes 0 pkt (dropped 0, overlimits 0 requeues 0)
| rate 0bit 0pps backlog 0b 0p requeues 0
|
+---(1:20) htb prio 0 rate 3Mbit ceil 6Mbit burst 15Kb cburst 1599b
| Sent 0 bytes 0 pkt (dropped 0, overlimits 0 requeues 0)
| rate 0bit 0pps backlog 0b 0p requeues 0
|
+---(1:30) htb prio 0 rate 1Kbit ceil 6Mbit burst 15Kb cburst 1599b
Sent 0 bytes 0 pkt (dropped 0, overlimits 0 requeues 0)
rate 0bit 0pps backlog 0b 0p requeues 0
Signed-off-by: Vadim Kochan <vadim4j@gmail.com>
---
Changes RFC -> PATCH:
#1 get rid of INIT_HLIST_NODE
#2 added sample output to commit message
#3 use "show_tree=1" instead of "show_tree++"
#4 no need update include/hlist.h (because of #1)
#5 changed a little tree output: parentheses around class id instead of qdisc name
tc/tc.c | 5 +-
tc/tc_class.c | 165 ++++++++++++++++++++++++++++++++++++++++++++++++++++++++-
tc/tc_common.h | 2 +
3 files changed, 169 insertions(+), 3 deletions(-)
diff --git a/tc/tc.c b/tc/tc.c
index 9b50e74..30950a6 100644
--- a/tc/tc.c
+++ b/tc/tc.c
@@ -34,8 +34,9 @@ int show_stats = 0;
int show_details = 0;
int show_raw = 0;
int show_pretty = 0;
-int batch_mode = 0;
+int show_tree = 0;
+int batch_mode = 0;
int resolve_hosts = 0;
int use_iec = 0;
int force = 0;
@@ -278,6 +279,8 @@ int main(int argc, char **argv)
++show_raw;
} else if (matches(argv[1], "-pretty") == 0) {
++show_pretty;
+ } else if (matches(argv[1], "-tree") == 0) {
+ show_tree = 1;
} else if (matches(argv[1], "-Version") == 0) {
printf("tc utility, iproute2-ss%s\n", SNAPSHOT);
return 0;
diff --git a/tc/tc_class.c b/tc/tc_class.c
index e56bf07..a7b3ecd 100644
--- a/tc/tc_class.c
+++ b/tc/tc_class.c
@@ -24,6 +24,22 @@
#include "utils.h"
#include "tc_util.h"
#include "tc_common.h"
+#include "hlist.h"
+
+struct cls_node {
+ struct hlist_node hlist;
+ __u32 handle;
+ __u32 parent;
+ int level;
+ struct cls_node *cls_parent;
+ struct cls_node *cls_right;
+ struct rtattr *attr;
+ int attr_len;
+ int childs_count;
+};
+
+static struct hlist_head cls_list = {};
+static struct hlist_head root_cls_list = {};
static void usage(void);
@@ -148,13 +164,149 @@ int filter_ifindex;
__u32 filter_qdisc;
__u32 filter_classid;
+static void tree_cls_add(__u32 parent, __u32 handle, struct rtattr *attr, int len)
+{
+ struct cls_node *cls = malloc(sizeof(struct cls_node));
+
+ memset(cls, 0, sizeof(*cls));
+ cls->handle = handle;
+ cls->parent = parent;
+ cls->attr = malloc(len);
+ cls->attr_len = len;
+
+ memcpy(cls->attr, attr, len);
+
+ if (parent == TC_H_ROOT)
+ hlist_add_head(&cls->hlist, &root_cls_list);
+ else
+ hlist_add_head(&cls->hlist, &cls_list);
+}
+
+static void tree_cls_indent(char *buf, struct cls_node *cls, int is_newline,
+ int add_spaces)
+{
+ char spaces[100] = {0};
+
+ while (cls && cls->cls_parent) {
+ cls->cls_parent->cls_right = cls;
+ cls = cls->cls_parent;
+ }
+ while (cls && cls->cls_right)
+ {
+ if (cls->hlist.next)
+ strcat(buf, "| ");
+ else
+ strcat(buf, " ");
+
+ cls = cls->cls_right;
+ }
+
+ if (is_newline) {
+ if (cls->hlist.next && cls->childs_count)
+ strcat(buf, "| |");
+ else if (cls->hlist.next)
+ strcat(buf, "| ");
+ else if (cls->childs_count)
+ strcat(buf, " |");
+ else if (!cls->hlist.next)
+ strcat(buf, " ");
+ }
+ if (add_spaces > 0)
+ {
+ sprintf(spaces, "%-*s", add_spaces, "");
+ strcat(buf, spaces);
+ }
+}
+
+static void tree_cls_show(FILE *fp, char *buf, struct hlist_head *root_list, int level)
+{
+ struct hlist_node *n, *tmp_cls;
+ char cls_id_str[256] = {};
+ struct rtattr * tb[TCA_MAX+1] = {};
+ struct qdisc_util *q;
+ char str[100] = {};
+
+ hlist_for_each_safe(n, tmp_cls, root_list) {
+ struct hlist_node *c, *tmp_chld;
+ struct hlist_head childs = {};
+ struct cls_node *cls = container_of(n, struct cls_node, hlist);
+
+ hlist_for_each_safe(c, tmp_chld, &cls_list) {
+ struct cls_node *child = container_of(c, struct cls_node, hlist);
+
+ if (cls->handle == child->parent) {
+ hlist_del(c);
+ hlist_add_head(c, &childs);
+ cls->childs_count++;
+ child->cls_parent = cls;
+ }
+ }
+
+ tree_cls_indent(buf, cls, 0, 0);
+
+ print_tc_classid(cls_id_str, sizeof(cls_id_str), cls->handle);
+ sprintf(str, "+---(%s)", cls_id_str);
+ strcat(buf, str);
+
+ parse_rtattr(tb, TCA_MAX, cls->attr, cls->attr_len);
+
+ if (tb[TCA_KIND] == NULL) {
+ strcat(buf, " [unknown qdisc kind] ");
+ } else {
+ const char *kind = rta_getattr_str(tb[TCA_KIND]);
+
+ sprintf(str, " %s ", kind);
+ strcat(buf, str);
+ fprintf(fp, "%s", buf);
+ buf[0] = '\0';
+
+ q = get_qdisc_kind(kind);
+ if (q && q->print_copt) {
+ q->print_copt(q, fp, tb[TCA_OPTIONS]);
+ }
+ if (q && show_stats)
+ {
+ int cls_indent = strlen(q->id) - 2 +
+ strlen(cls_id_str);
+ struct rtattr *xstats = NULL;
+
+ tree_cls_indent(buf, cls, 1, cls_indent);
+
+ if (tb[TCA_STATS] || tb[TCA_STATS2]) {
+ fprintf(fp, "\n");
+ print_tcstats_attr(fp, tb, buf, &xstats);
+ buf[0] = '\0';
+ }
+ if (cls->hlist.next || cls->childs_count)
+ {
+ strcat(buf, "\n");
+ tree_cls_indent(buf, cls, 1, 0);
+ }
+ }
+ }
+ free(cls->attr);
+ fprintf(fp, "%s\n", buf);
+ buf[0] = '\0';
+
+ tree_cls_show(fp, buf, &childs, level + 1);
+ if (!cls->hlist.next) {
+ tree_cls_indent(buf, cls, 0, 0);
+ strcat(buf, "\n");
+ }
+
+ fprintf(fp, "%s", buf);
+ buf[0] = '\0';
+ free(cls);
+ }
+}
+
int print_class(const struct sockaddr_nl *who,
struct nlmsghdr *n, void *arg)
{
FILE *fp = (FILE*)arg;
struct tcmsg *t = NLMSG_DATA(n);
int len = n->nlmsg_len;
- struct rtattr * tb[TCA_MAX+1];
+ struct rtattr * tb[TCA_MAX+1] = {};
struct qdisc_util *q;
char abuf[256];
@@ -167,13 +319,18 @@ int print_class(const struct sockaddr_nl *who,
fprintf(stderr, "Wrong len %d\n", len);
return -1;
}
+
+ if (show_tree) {
+ tree_cls_add(t->tcm_parent, t->tcm_handle, TCA_RTA(t), len);
+ return 0;
+ }
+
if (filter_qdisc && TC_H_MAJ(t->tcm_handle^filter_qdisc))
return 0;
if (filter_classid && t->tcm_handle != filter_classid)
return 0;
- memset(tb, 0, sizeof(tb));
parse_rtattr(tb, TCA_MAX, TCA_RTA(t), len);
if (tb[TCA_KIND] == NULL) {
@@ -236,6 +393,7 @@ static int tc_class_list(int argc, char **argv)
{
struct tcmsg t;
char d[16];
+ char buf[1024] = {0};
memset(&t, 0, sizeof(t));
t.tcm_family = AF_UNSPEC;
@@ -306,6 +464,9 @@ static int tc_class_list(int argc, char **argv)
return 1;
}
+ if (show_tree)
+ tree_cls_show(stdout, &buf[0], &root_cls_list, 0);
+
return 0;
}
diff --git a/tc/tc_common.h b/tc/tc_common.h
index 4f88856..0ee009b 100644
--- a/tc/tc_common.h
+++ b/tc/tc_common.h
@@ -19,3 +19,5 @@ extern int parse_estimator(int *p_argc, char ***p_argv, struct tc_estimator *est
struct tc_sizespec;
extern int parse_size_table(int *p_argc, char ***p_argv, struct tc_sizespec *s);
extern int check_size_table_opts(struct tc_sizespec *s);
+
+extern int show_tree;
--
2.1.3
^ permalink raw reply related
* Re: SRIOV as bridge Re: [PATCH net-next RESEND] net: Do not call ndo_dflt_fdb_dump if ndo_fdb_dump is defined.
From: Jamal Hadi Salim @ 2014-12-22 13:04 UTC (permalink / raw)
To: Roopa Prabhu
Cc: John Fastabend, Hubert Sokolowski, netdev@vger.kernel.org,
Vlad Yasevich, Shrijeet Mukherjee
In-Reply-To: <54980A33.6000801@mojatatu.com>
On 12/22/14 07:10, Jamal Hadi Salim wrote:
> Actually you cant list them as netdevs (please someone
> correct me if i am wrong). What you see are PCI devices.
I have been corrected.
It turns out i was wrong. The VFs infact show up as ethx devices.
So your idea of having the devices with "bridge .. self" would work
except when the device moves to a VM and you cant see them anymore...
and you want to add an fdb entry to the embedded switch..
cheers,
jamal
^ permalink raw reply
* Re: SRIOV as bridge Re: [PATCH net-next RESEND] net: Do not call ndo_dflt_fdb_dump if ndo_fdb_dump is defined.
From: Jamal Hadi Salim @ 2014-12-22 12:10 UTC (permalink / raw)
To: Roopa Prabhu
Cc: John Fastabend, Hubert Sokolowski, netdev@vger.kernel.org,
Vlad Yasevich, Shrijeet Mukherjee
In-Reply-To: <5497B916.1020605@cumulusnetworks.com>
On 12/22/14 01:24, Roopa Prabhu wrote:
> On 12/21/14, 7:13 PM, Jamal Hadi Salim wrote:
>> For SRIOV it is. Example to add via pf eth10 an
>> fdb entry to the igb hardware fdb to point to vf1:
>> ip link set eth10 vf 1 mac aa:bb:cc:dd:ee:ff vlan 10
>> That last part "vf 1 mac aa:bb:cc:dd:ee:ff vlan 10" is typically
>> part of an "fdb add semantic" - but we explicitly call out
>> eth10, the parent. The PF has control of the hardware fdb.
>
> Ah......i did not know this syntax with 'ip link set'. thanks for
> pointing out.
> I always thought that you can still use 'bridge fdb add' for vfs.
> Curious why its not that way.
>
It was there before bridge command showed up. I think thats why
John feels it can be repaired. The main challenge is that most
of the time the VFs are "migrated" from host to VMs, so you cant
even list them on the host.
Actually you cant list them as netdevs (please someone
correct me if i am wrong). What you see are PCI devices.
From my notes:
---
modprobe igb max_vfs=6
# lspci | grep 82576
0b:00.0 Ethernet controller: Intel Corporation 82576 Gigabit Network
Connection (rev 01)
0b:00.1 Ethernet controller: Intel Corporation 82576 Gigabit Network
Connection(rev 01)
0b:10.0 Ethernet controller: Intel Corporation 82576 Virtual Function
(rev 01)
0b:10.1 Ethernet controller: Intel Corporation 82576 Virtual Function
(rev 01)
..
...
....
---
And then you can tell qemu to use one of these devices as a network
device and holla it disappears from the host.
What you can do from the host side is tell the embedded hardware switch
to add an fdb which sends packets to a VF using the above syntax (via
PF). I think PF as control interface to the VFs is a convinience issue.
It just happened to be there and could be used as the anchor.
This is why i was mapping it instead to theclassthingy which i think
is more generic.
>
>> So what do you do if the user sets either one of master/self and it
>> doesnt make sense?
>
> Am guessing it will continue to do what it does today. If there is no
> master or if there is master and the master does not support the op, it
> will return -EOPNOTSUPP. And, self does not make sense in cases where
> the port driver does not support the op. In which case again you will
> get a -EOPNOTSUPP. Have not thought through all the other cases yet.
>
So master is only valid if there is a software device parent?
Again note - none of these VFs are attached to a bridge today.
It doesnt seem to me like it even makes sense people want to
do that.
The PF could be and in such a case its master is the bridge.
cheers,
jamal
^ permalink raw reply
* Re: [RFC PATCH net-next] tun: support retrieving multiple packets in a single read with IFF_MULTI_READ
From: Herbert Xu @ 2014-12-22 12:09 UTC (permalink / raw)
To: Alex Gartrell
Cc: jasonwang, davem, netdev, linux-kernel, mst, herbert, kernel-team,
agartrell
In-Reply-To: <1417752000-27171-1-git-send-email-agartrell@fb.com>
Alex Gartrell <agartrell@fb.com> wrote:
> This patch adds the IFF_MULTI_READ flag. This has the following behavior.
>
> 1) If a read is too short for a packet, a single stripped packet will be read
>
> 2) If a read is long enough for multiple packets, as many *full* packets
> will be read as possible. We will not return a stripped packet, so even if
> there are many, many packets, we may get a short read.
>
> In casual performance testing with a simple test program that simply reads
> and counts packets, IFF_MULTI_READ conservatively yielded a 30% CPU win, as
> measured by top. Load was being driven by a bunch of hpings running on a
> server on the same L2 network (single hop through a top-of-rack switch).
>
> Signed-off-by: Alex Gartrell <agartrell@fb.com>
As tun already has a socket interface can we do this through
recvmmsg?
Thanks,
--
Email: Herbert Xu <herbert@gondor.apana.org.au>
Home Page: http://gondor.apana.org.au/~herbert/
PGP Key: http://gondor.apana.org.au/~herbert/pubkey.txt
^ permalink raw reply
* Re: REGRESSION in nfnetlink on 3.18.x (bisected)
From: Pablo Neira Ayuso @ 2014-12-22 11:56 UTC (permalink / raw)
To: Andre Tomt; +Cc: netfilter-devel, netdev
In-Reply-To: <5496075F.3060204@tomt.net>
[-- Attachment #1: Type: text/plain, Size: 816 bytes --]
On Sun, Dec 21, 2014 at 12:33:51AM +0100, Andre Tomt wrote:
> On at least Ubuntu 14.04 LTS and Ubuntu 14.10 "conntrack -E" has
> started failing with Linux 3.18.x. conntrack -L still works.
>
> 14.04 and 14.10 ships conntrack-utils version 1.4.1, but 1.4.2 does
> not work either.
>
> It fails with:
> ># conntrack -E
> >conntrack v1.4.2 (conntrack-tools): Can't open handler
>
> strace shows:
> >bind(3, {sa_family=AF_NETLINK, pid=0, groups=00000000}, 12) = 0
> >getsockname(3, {sa_family=AF_NETLINK, pid=14092, groups=00000000}, [12]) = 0
> >bind(3, {sa_family=AF_NETLINK, pid=14092, groups=00000007}, 12) = -1 EINVAL (Invalid argument)
>
> Reverting 97840cb67ff5ac8add836684f011fd838518d698 - netfilter:
> nfnetlink: fix insufficient validation in nfnetlink_bind
Could you give a test to this patch? Thanks.
[-- Attachment #2: 0001-netlink-fix-wrong-subscription-bitmask-to-group-mapp.patch --]
[-- Type: text/x-diff, Size: 1827 bytes --]
>From f4f65150fd2129607a7bd25f007c258045237c8c Mon Sep 17 00:00:00 2001
From: Pablo Neira Ayuso <pablo@netfilter.org>
Date: Sun, 21 Dec 2014 21:48:36 +0100
Subject: [PATCH nf] netlink: fix wrong subscription bitmask to group mapping in
binding callbacks
The subscription bitmask passed via struct sockaddr_nl is converted to
the group number when calling the netlink_bind() and netlink_unbind()
callbacks.
The conversion is however incorrect since bitmask (1 << 0) needs to be
mapped to group number 1. Note that you cannot specify the group number 0
(usually known as _NONE) from setsockopt() using NETLINK_ADD_MEMBERSHIP
since this is rejected through -EINVAL.
This problem became noticeable since 97840cb ("netfilter: nfnetlink:
fix insufficient validation in nfnetlink_bind") when binding to bitmask
(1 << 0) in ctnetlink.
Reported-by: Andre Tomt <andre@tomt.net>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
---
net/netlink/af_netlink.c | 4 ++--
1 file changed, 2 insertions(+), 2 deletions(-)
diff --git a/net/netlink/af_netlink.c b/net/netlink/af_netlink.c
index 074cf3e..cbcf73b 100644
--- a/net/netlink/af_netlink.c
+++ b/net/netlink/af_netlink.c
@@ -1420,7 +1420,7 @@ static void netlink_unbind(int group, long unsigned int groups,
for (undo = 0; undo < group; undo++)
if (test_bit(undo, &groups))
- nlk->netlink_unbind(undo);
+ nlk->netlink_unbind(undo + 1);
}
static int netlink_bind(struct socket *sock, struct sockaddr *addr,
@@ -1458,7 +1458,7 @@ static int netlink_bind(struct socket *sock, struct sockaddr *addr,
for (group = 0; group < nlk->ngroups; group++) {
if (!test_bit(group, &groups))
continue;
- err = nlk->netlink_bind(group);
+ err = nlk->netlink_bind(group + 1);
if (!err)
continue;
netlink_unbind(group, groups, nlk);
--
1.7.10.4
^ permalink raw reply related
* RE: [PATCH net] r8152: drop the tx packet with invalid length
From: Hayes Wang @ 2014-12-22 11:34 UTC (permalink / raw)
To: Eric Dumazet
Cc: Tom Herbert, David Miller, netdev@vger.kernel.org, nic_swsd,
linux-kernel@vger.kernel.org, linux-usb@vger.kernel.org
In-Reply-To: <1419012837.9773.85.camel@edumazet-glaptop2.roam.corp.google.com>
> -----Original Message-----
> From: Hayes Wang
> Sent: Monday, December 22, 2014 10:23 AM
> To: 'Eric Dumazet'
> Cc: Tom Herbert; David Miller; netdev@vger.kernel.org;
> nic_swsd; linux-kernel@vger.kernel.org; linux-usb@vger.kernel.org
> Subject: RE: [PATCH net] r8152: drop the tx packet with invalid length
>
> Eric Dumazet [mailto:eric.dumazet@gmail.com]
> > Sent: Saturday, December 20, 2014 2:14 AM
> [...]
> > Could you try following patch ?
>
> Thank you. I would test it.
It works for me. Thanks.
Best Regards,
Hayes
^ permalink raw reply
* Re: OOPS in nf_ct_unlink_expect_report using Polycom RealPresence Mobile
From: zhuyj @ 2014-12-22 10:34 UTC (permalink / raw)
To: Mike Galbraith, astx; +Cc: linux-kernel, netdev, zyjzyj2000
In-Reply-To: <1391174223.6395.3.camel@marge.simpson.net>
Please check the number of iptables rule. Maybe it results from the big
number of iptables rules.
Best Regards!
Zhu Yanjun
On 01/31/2014 09:17 PM, Mike Galbraith wrote:
> (CC netdev)
>
> On Fri, 2014-01-31 at 12:05 +0100, astx wrote:
>> Using Polycom video conferencing software my homebrew linux NAT router
>> crashes with attached kernel oops message.
>> This error can be reproduced also using kernel 3.2.54. Kernel 2.6.35
>> seems to be stable.
>>
>> Disabling nf_nat_h323 and nf_conntrack_h323 avoids crash - but video
>> conferencing software is no more usable.
>>
>>
>> ===================================================================================
>> BUG: unable to handle kernel paging request at 00100104
>> IP: [<f8214f07>] nf_ct_unlink_expect_report+0x57/0xf0 [nf_conntrack]
>> *pdpt = 00000000359aa001 *pde = 0000000000000000
>> Oops: 0002 [#1] SMP
>> Modules linked in: nf_conntrack_netlink nfnetlink xt_mac xt_TCPMSS
>> ipt_MASQUERADE
>> xt_pkttype xt_multiport xt_REDIRECT xt_nat iptable_mangle xt_LOG
>> xt_limit af_packet
>> act_mirred cls_u32 sch_ingress sch_hfsc ifb xt_tcpudp ip6t_REJECT ipt_REJECT
>> ip6table_raw iptable_raw xt_CT iptable_filter nf_nat_pptp nf_nat_proto_gre
>> nf_conntrack_proto_udplite nf_conntrack_proto_dccp ip6table_mangle
>> iptable_nat
>> nf_nat_ipv4 nf_nat_sip nf_nat_irc nf_nat_snmp_basic nf_conntrack_snmp
>> nf_conntrack_broadcast nf_nat_h323 nf_nat_tftp nf_nat_ftp nf_nat
>> nf_conntrack_h323
>> nf_conntrack_tftp nf_conntrack_proto_sctp nf_conntrack_sip nf_conntrack_irc
>> nf_conntrack_pptp nf_conntrack_proto_gre nf_conntrack_ftp nf_conntrack_ipv4
>> nf_defrag_ipv4 ip_tables xt_conntrack nf_conntrack ip6table_filter ip6_tables
>> x_tables padlock_sha padlock_aes e_powersaver freq_table mperf via_cputemp
>> hwmon_vid serio_raw pcspkr i2c_viapro ehci_pci fan thermal processor 8139too
>> sg thermal_sys button shpchp 8139cp pci_hotplug mii via_agp ext4 crc16 jbd2
>> pata_via sata_via libata sd_mod scsi_mod ohci_hcd uhci_hcd ehci_hcd
>> CPU: 0 PID: 0 Comm: swapper/0 Not tainted 3.10.28-9500-smp_m #1
>> Hardware name: /CN700-8237, BIOS 6.00 PG 08/30/2007
>> task: c07ce180 ti: f6408000 task.ti: c07c2000
>> EIP: 0060:[<f8214f07>] EFLAGS: 00210206 CPU: 0
>> EIP is at nf_ct_unlink_expect_report+0x57/0xf0 [nf_conntrack]
>> EAX: 00100100 EBX: eb636bc0 ECX: 00000000 EDX: eb461540
>> ESI: c0804e00 EDI: eb461544 EBP: f6409f08 ESP: f6409eec
>> DS: 007b ES: 007b FS: 00d8 GS: 0000 SS: 0068
>> CR0: 8005003b CR2: 00100104 CR3: 359d4000 CR4: 000006b0
>> DR0: 00000000 DR1: 00000000 DR2: 00000000 DR3: 00000000
>> DR6: ffff0ff0 DR7: 00000400
>> Stack:
>> 00000000 00200286 f6409f08 c0244bd8 eb636bc0 00100100 00000000 f6409f18
>> f8215687 f598ede8 c0804e00 f6409f28 f8211c99 f598ede8 f598ee50 f6409f5c
>> f8212e5e 00000003 00000000 00000000 00000004 eb461514 f598ede8 00000000
>> Call Trace:
>> [<c0244bd8>] ? del_timer+0x48/0x70
>> [<f8215687>] nf_ct_remove_expectations+0x47/0x60 [nf_conntrack]
>> [<f8211c99>] nf_ct_delete_from_lists+0x59/0x90 [nf_conntrack]
>> [<f8212e5e>] death_by_timeout+0x14e/0x1c0 [nf_conntrack]
>> [<f8212d10>] ? nf_conntrack_set_hashsize+0x190/0x190 [nf_conntrack]
>> [<c024442d>] call_timer_fn+0x1d/0x80
>> [<c024461e>] run_timer_softirq+0x18e/0x1a0
>> [<f8212d10>] ? nf_conntrack_set_hashsize+0x190/0x190 [nf_conntrack]
>> [<c023e6f3>] __do_softirq+0xa3/0x170
>> [<c023e650>] ? __local_bh_enable+0x70/0x70
>> <IRQ>
>> [<c023e587>] ? irq_exit+0x67/0xa0
>> [<c0202af6>] ? do_IRQ+0x46/0xb0
>> [<c027ad05>] ? clockevents_notify+0x35/0x110
>> [<c066ac6c>] ? common_interrupt+0x2c/0x40
>> [<c056e3c1>] ? cpuidle_enter_state+0x41/0xf0
>> [<c056e6fb>] ? cpuidle_idle_call+0x8b/0x100
>> [<c02085f8>] ? arch_cpu_idle+0x8/0x30
>> [<c027314b>] ? cpu_idle_loop+0x4b/0x140
>> [<c0273258>] ? cpu_startup_entry+0x18/0x20
>> [<c066056d>] ? rest_init+0x5d/0x70
>> [<c0813ac8>] ? start_kernel+0x2ec/0x2f2
>> [<c081364f>] ? repair_env_string+0x5b/0x5b
>> [<c0813269>] ? i386_start_kernel+0x33/0x35
>> Code: 8b 7b 0c 8b b6 98 00 00 00 85 c0 89 07 74 03 89 78 04 c7 43 0c 00
>> 02 20 00 83 ae ec 05 00 00 01 8b 03 8b 7b 04 85 c0 89 07 74 03 <89> 78
>> 04 8b 43 7c c7 03 00 01 10 00 c7 43 04 00 02 20 00 80 6c
>> EIP: [<f8214f07>] nf_ct_unlink_expect_report+0x57/0xf0 [nf_conntrack]
>> SS:ESP 0068:f6409eec
>> CR2: 0000000000100104
>> ---[ end trace 79fe2e6b81f54dee ]---
>> Kernel panic - not syncing: Fatal exception in interrupt
>> Rebooting in 300 seconds..
>> ===================================================================================
>>
>>
>> Polycom Version: 3.1-44477
>> running on device: Apple iPad Mini
>> using operating system: iOS Version: 7.0.4
>>
>>
>> Attached also my kernel config. Hopefully someone could help...
>>
>> BR, Toni
>
> --
> To unsubscribe from this list: send the line "unsubscribe netdev" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at http://vger.kernel.org/majordomo-info.html
>
^ permalink raw reply
* [PATCH net] net: Fix stacked vlan offload features computation
From: Toshiaki Makita @ 2014-12-22 10:04 UTC (permalink / raw)
To: David S . Miller; +Cc: Toshiaki Makita, Jesse Gross, netdev
When vlan tags are stacked, it is very likely that the outer tag is stored
in skb->vlan_tci and skb->protocol shows the inner tag's vlan_proto.
Currently netif_skb_features() first looks at skb->protocol even if there
is the outer tag in vlan_tci, thus it incorrectly retrieves the protocol
encapsulated by the inner vlan instead of the inner vlan protocol.
This allows GSO packets to be passed to HW and they end up being
corrupted.
Fixes: 58e998c6d239 ("offloading: Force software GSO for multiple vlan tags.")
Signed-off-by: Toshiaki Makita <makita.toshiaki@lab.ntt.co.jp>
---
net/core/dev.c | 13 ++++++++-----
1 file changed, 8 insertions(+), 5 deletions(-)
diff --git a/net/core/dev.c b/net/core/dev.c
index f411c28..a6afd70 100644
--- a/net/core/dev.c
+++ b/net/core/dev.c
@@ -2570,11 +2570,14 @@ netdev_features_t netif_skb_features(struct sk_buff *skb)
if (gso_segs > dev->gso_max_segs || gso_segs < dev->gso_min_segs)
features &= ~NETIF_F_GSO_MASK;
- if (protocol == htons(ETH_P_8021Q) || protocol == htons(ETH_P_8021AD)) {
- struct vlan_ethhdr *veh = (struct vlan_ethhdr *)skb->data;
- protocol = veh->h_vlan_encapsulated_proto;
- } else if (!vlan_tx_tag_present(skb)) {
- return harmonize_features(skb, features);
+ if (!vlan_tx_tag_present(skb)) {
+ if (unlikely(protocol == htons(ETH_P_8021Q) ||
+ protocol == htons(ETH_P_8021AD))) {
+ struct vlan_ethhdr *veh = (struct vlan_ethhdr *)skb->data;
+ protocol = veh->h_vlan_encapsulated_proto;
+ } else {
+ return harmonize_features(skb, features);
+ }
}
features = netdev_intersect_features(features,
--
1.8.1.2
^ permalink raw reply related
page: next (older) | prev (newer) | latest
- recent:[subjects (threaded)|topics (new)|topics (active)]
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox