Skip to contentSkip to conversation

hash: fix gcc 10 maybe-uninitialized warning

PATCH Applied Kevin Traynor 3 participants 5 messages May 15, 2020 – May 18, 2020

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 hash: fix gcc 10 maybe-uninitialized warning

gcc 10.1.1 reports a warning for the ext_bkt_id variable:

../lib/librte_hash/rte_cuckoo_hash.c:
In function ‘__rte_hash_add_key_with_hash’:
../lib/librte_hash/rte_cuckoo_hash.c:1104:29:
warning: ‘ext_bkt_id’ may be used uninitialized in this function [-Wmaybe-uninitialized]

 1104 |  (h->buckets_ext[ext_bkt_id - 1]).sig_current[0] = short_sig;
      |                  ~~~~~~~~~~~^~~

The return value of rte_ring_sc_dequeue_elem() is already checked, but also initialize ext_bkt_id to zero (invalid value) and check that it also overwritten.

Fixes fbfe568103b0 ("hash: use 32-bit elements rings to save memory")
---
 lib/librte_hash/rte_cuckoo_hash.c | 5 +++--
 1 file changed, 3 insertions(+), 2 deletions(-)
lib/librte_hash/rte_cuckoo_hash.c +3 -2
index 38767a8a1..90cb99b0e 100644
@@ -940,6 +940,6 @@ __rte_hash_add_key_with_hash(const struct rte_hash *h, const void *key,
 	struct rte_hash_bucket *prim_bkt, *sec_bkt, *cur_bkt;
 	struct rte_hash_key *new_k, *keys = h->key_store;
+	uint32_t ext_bkt_id = 0;
 	uint32_t slot_id;
-	uint32_t ext_bkt_id;
 	int ret;
 	unsigned n_slots;
@@ -1096,5 +1096,6 @@ __rte_hash_add_key_with_hash(const struct rte_hash *h, const void *key,
 	 */
 	if (rte_ring_sc_dequeue_elem(h->free_ext_bkts, &ext_bkt_id,
-						sizeof(uint32_t)) != 0) {
+						sizeof(uint32_t)) != 0 ||
+					ext_bkt_id == 0) {
 		ret = -ENOSPC;
 		goto failure;
Signature · 1 line
2.21.3
Acked-by on v1 1/1
#1 lore raw
Quoted 48 lines

-----Original Message-----
From: Kevin Traynor <[email protected]>
Sent: Friday, May 15, 2020 7:28 AM
To: [email protected]; Wang, Yipeng1 <[email protected]>; Gobriel, Sameh <[email protected]>; Richardson, Bruce <[email protected]>
Cc: [email protected]; Yigit, Ferruh <[email protected]>; [email protected]; Kevin Traynor <[email protected]>; [email protected]
Subject: [PATCH] hash: fix gcc 10 maybe-uninitialized warning

gcc 10.1.1 reports a warning for the ext_bkt_id variable:

../lib/librte_hash/rte_cuckoo_hash.c:
In function ‘__rte_hash_add_key_with_hash’:
../lib/librte_hash/rte_cuckoo_hash.c:1104:29:
warning: ‘ext_bkt_id’ may be used uninitialized in this function [-Wmaybe- uninitialized]

 1104 |  (h->buckets_ext[ext_bkt_id - 1]).sig_current[0] = short_sig;
      |                  ~~~~~~~~~~~^~~

The return value of rte_ring_sc_dequeue_elem() is already checked, but also initialize ext_bkt_id to zero (invalid value) and check that it also overwritten.

Fixes fbfe568103b0 ("hash: use 32-bit elements rings to save memory")
---
 lib/librte_hash/rte_cuckoo_hash.c | 5 +++--
 1 file changed, 3 insertions(+), 2 deletions(-)
lib/librte_hash/rte_cuckoo_hash.c +1 -1
index 38767a8a1..90cb99b0e 100644
@@ -940,6 +940,6 @@ __rte_hash_add_key_with_hash(const struct
 rte_hash *h, const void *key,
 	struct rte_hash_bucket *prim_bkt, *sec_bkt, *cur_bkt;
 	struct rte_hash_key *new_k, *keys = h->key_store;
+	uint32_t ext_bkt_id = 0;
 	uint32_t slot_id;
-	uint32_t ext_bkt_id;
 	int ret;
lib/librte_hash/rte_cuckoo_hash.c +0 -0
 	unsigned n_slots;

lib/librte_hash/rte_cuckoo_hash.c __rte_hash_add_key_with_hash(const struct · line 1100

Quoted 6 lines
lib/librte_hash/rte_cuckoo_hash.c +2 -1
@@ -1096,5 +1096,6 @@ __rte_hash_add_key_with_hash(const struct
 rte_hash *h, const void *key,
 	 */
 	if (rte_ring_sc_dequeue_elem(h->free_ext_bkts, &ext_bkt_id,
-						sizeof(uint32_t)) != 0) {
+						sizeof(uint32_t)) != 0 ||
+					ext_bkt_id == 0) {

[Wang, Yipeng]
If convenient, it would be better to make the two lines aligned with same indent...

Quoted 4 lines
lib/librte_hash/rte_cuckoo_hash.c +0 -0
 		ret = -ENOSPC;

goto failure;
--
2.21.3

[Wang, Yipeng]
Thanks for the fix. I think It is also better code in general.

#2 lore raw
Quoted 48 lines · On 15/05/2020 20:06, Wang, Yipeng1 wrote:
Quoted 48 lines

-----Original Message-----
From: Kevin Traynor <[email protected]>
Sent: Friday, May 15, 2020 7:28 AM
To: [email protected]; Wang, Yipeng1 <[email protected]>; Gobriel, Sameh <[email protected]>; Richardson, Bruce <[email protected]>
Cc: [email protected]; Yigit, Ferruh <[email protected]>; [email protected]; Kevin Traynor <[email protected]>; [email protected]
Subject: [PATCH] hash: fix gcc 10 maybe-uninitialized warning

gcc 10.1.1 reports a warning for the ext_bkt_id variable:

../lib/librte_hash/rte_cuckoo_hash.c:
In function ‘__rte_hash_add_key_with_hash’:
../lib/librte_hash/rte_cuckoo_hash.c:1104:29:
warning: ‘ext_bkt_id’ may be used uninitialized in this function [-Wmaybe- uninitialized]

 1104 |  (h->buckets_ext[ext_bkt_id - 1]).sig_current[0] = short_sig;
      |                  ~~~~~~~~~~~^~~

The return value of rte_ring_sc_dequeue_elem() is already checked, but also initialize ext_bkt_id to zero (invalid value) and check that it also overwritten.

Fixes fbfe568103b0 ("hash: use 32-bit elements rings to save memory")
---
 lib/librte_hash/rte_cuckoo_hash.c | 5 +++--
 1 file changed, 3 insertions(+), 2 deletions(-)
lib/librte_hash/rte_cuckoo_hash.c +1 -1
index 38767a8a1..90cb99b0e 100644
@@ -940,6 +940,6 @@ __rte_hash_add_key_with_hash(const struct
 rte_hash *h, const void *key,
 	struct rte_hash_bucket *prim_bkt, *sec_bkt, *cur_bkt;
 	struct rte_hash_key *new_k, *keys = h->key_store;
+	uint32_t ext_bkt_id = 0;
 	uint32_t slot_id;
-	uint32_t ext_bkt_id;
 	int ret;
lib/librte_hash/rte_cuckoo_hash.c +0 -0
 	unsigned n_slots;
@@ -1096,5 +1096,6 @@ __rte_hash_add_key_with_hash(const struct
Quoted 9 lines

lib/librte_hash/rte_cuckoo_hash.c __rte_hash_add_key_with_hash(const struct rte_hash *h, const void *key, · line 1099

Quoted 6 lines
lib/librte_hash/rte_cuckoo_hash.c +2 -1
 rte_hash *h, const void *key,
 	 */
 	if (rte_ring_sc_dequeue_elem(h->free_ext_bkts, &ext_bkt_id,
-						sizeof(uint32_t)) != 0) {
+						sizeof(uint32_t)) != 0 ||
+					ext_bkt_id == 0) {

[Wang, Yipeng]
If convenient, it would be better to make the two lines aligned with same indent...

Hi Yipeng, I had checked the coding style [1] about this and I think it's correct as 'sizeof..' is a wrap from the first condition so gets a second tab to indicate that, whereas 'ext_bkt_id..' is the second condition with no wrap. Fine to change it, if I interpret incorrectly.

[1] third bullet,
http://doc.dpdk.org/guides/contributing/coding_style.html#general

Quoted 9 lines
Quoted 4 lines
lib/librte_hash/rte_cuckoo_hash.c +0 -0
 		ret = -ENOSPC;

goto failure;
--
2.21.3

[Wang, Yipeng]
Thanks for the fix. I think It is also better code in general.

#3 lore raw
Quoted 71 lines

-----Original Message-----
From: Kevin Traynor <[email protected]>
Sent: Friday, May 15, 2020 12:18 PM
To: Wang, Yipeng1 <[email protected]>; [email protected]; Gobriel, Sameh <[email protected]>; Richardson, Bruce <[email protected]>
Cc: [email protected]; Yigit, Ferruh <[email protected]>; [email protected]; [email protected]
Subject: Re: [PATCH] hash: fix gcc 10 maybe-uninitialized warning

Quoted 49 lines · On 15/05/2020 20:06, Wang, Yipeng1 wrote:
Quoted 49 lines

-----Original Message-----
From: Kevin Traynor <[email protected]>
Sent: Friday, May 15, 2020 7:28 AM
To: [email protected]; Wang, Yipeng1 <[email protected]>; Gobriel, Sameh <[email protected]>; Richardson, Bruce <[email protected]>
Cc: [email protected]; Yigit, Ferruh <[email protected]>; [email protected]; Kevin Traynor <[email protected]>; [email protected]
Subject: [PATCH] hash: fix gcc 10 maybe-uninitialized warning

gcc 10.1.1 reports a warning for the ext_bkt_id variable:

../lib/librte_hash/rte_cuckoo_hash.c:
In function ‘__rte_hash_add_key_with_hash’:
../lib/librte_hash/rte_cuckoo_hash.c:1104:29:
warning: ‘ext_bkt_id’ may be used uninitialized in this function [-Wmaybe- uninitialized]

 1104 |  (h->buckets_ext[ext_bkt_id - 1]).sig_current[0] = short_sig;
      |                  ~~~~~~~~~~~^~~

The return value of rte_ring_sc_dequeue_elem() is already checked, but also initialize ext_bkt_id to zero (invalid value) and check that it also overwritten.

Fixes fbfe568103b0 ("hash: use 32-bit elements rings to save

memory")
Cc: [email protected]

---
 lib/librte_hash/rte_cuckoo_hash.c | 5 +++--
 1 file changed, 3 insertions(+), 2 deletions(-)
lib/librte_hash/rte_cuckoo_hash.c +1 -1
index 38767a8a1..90cb99b0e 100644
@@ -940,6 +940,6 @@ __rte_hash_add_key_with_hash(const struct
 rte_hash *h, const void *key,
 	struct rte_hash_bucket *prim_bkt, *sec_bkt, *cur_bkt;
 	struct rte_hash_key *new_k, *keys = h->key_store;
+	uint32_t ext_bkt_id = 0;
 	uint32_t slot_id;
-	uint32_t ext_bkt_id;
 	int ret;
lib/librte_hash/rte_cuckoo_hash.c +0 -0
 	unsigned n_slots;
@@ -1096,5 +1096,6 @@ __rte_hash_add_key_with_hash(const struct
Quoted 8 lines

lib/librte_hash/rte_cuckoo_hash.c __rte_hash_add_key_with_hash(const struct rte_hash *h, const void *key, · line 1099

Quoted 6 lines
lib/librte_hash/rte_cuckoo_hash.c +2 -1
 rte_hash *h, const void *key,
 	 */
 	if (rte_ring_sc_dequeue_elem(h->free_ext_bkts, &ext_bkt_id,
-						sizeof(uint32_t)) != 0) {
+						sizeof(uint32_t)) != 0 ||
+					ext_bkt_id == 0) {

[Wang, Yipeng]
If convenient, it would be better to make the two lines aligned with same

indent...

Quoted 1 line
Quoted 7 lines

Hi Yipeng, I had checked the coding style [1] about this and I think it's correct as 'sizeof..' is a wrap from the first condition so gets a second tab to indicate that, whereas 'ext_bkt_id..' is the second condition with no wrap. Fine to change it, if I interpret incorrectly.

[1] third bullet,
http://doc.dpdk.org/guides/contributing/coding_style.html#general

[Wang, Yipeng] I see your point now. Thanks for explaining and I think it makes sense to distinguish these two. I guess it just came from my own aesthetic preference.
and I don’t have the best coding style : )

Quoted 10 lines
Quoted 10 lines
Quoted 5 lines
lib/librte_hash/rte_cuckoo_hash.c +0 -0
 		ret = -ENOSPC;

goto failure;
--
2.21.3

[Wang, Yipeng]
Thanks for the fix. I think It is also better code in general.

Acked-by Yipeng Wang on v1 1/1 Applied
#4 lore raw
In reply to Kevin Traynor
Quoted 11 lines · Kevin Traynor

gcc 10.1.1 reports a warning for the ext_bkt_id variable:

../lib/librte_hash/rte_cuckoo_hash.c:
In function ‘__rte_hash_add_key_with_hash’:
../lib/librte_hash/rte_cuckoo_hash.c:1104:29:
warning: ‘ext_bkt_id’ may be used uninitialized in this function [-Wmaybe-uninitialized]

 1104 |  (h->buckets_ext[ext_bkt_id - 1]).sig_current[0] = short_sig;
      |                  ~~~~~~~~~~~^~~
Quoted 8 lines

The return value of rte_ring_sc_dequeue_elem() is already checked, but also initialize ext_bkt_id to zero (invalid value) and check that it also overwritten.

Fixes fbfe568103b0 ("hash: use 32-bit elements rings to save memory")

Applied, thanks.

Signature · 1 line
David Marchand