The bug is here: "} else if (item) {".
The list iterator value will *always* be set and non-NULL by
list_for_each_entry(), so it is incorrect to assume that the iterator
value will be NULL if the list is empty or no element is found in list.
Use a new value 'iter' as the list iterator, while use the old value
'item' as a dedicated pointer to point to the found element, which
1. can fix this bug, due to now 'item' is NULL only if it's not found.
2. do not need to change all the uses of 'item' after the loop.
3. can also limit the scope of the list iterator 'iter' *only inside*
the traversal loop by simply declaring 'iter' inside the loop in the
future, as usage of the iterator outside of the list_for_each_entry
is considered harmful. https://lkml.org/lkml/2022/2/17/1032
Fixes: a910e4a94f692 ("cw1200: add driver for the ST-E CW1100 & CW1200 WLAN chipsets")
Signed-off-by: Xiaomeng Tong <[email protected]>
---
changes since v2:
- rebase on latest wireless-next (Kalle Valo)
changes since v1:
- fix incorrect check to item (Jakob Koschel)
v2: https://lore.kernel.org/lkml/[email protected]/
v1: https://lore.kernel.org/all/[email protected]/
---
drivers/net/wireless/st/cw1200/queue.c | 18 ++++++++++--------
1 file changed, 10 insertions(+), 8 deletions(-)
diff --git a/drivers/net/wireless/st/cw1200/queue.c b/drivers/net/wireless/st/cw1200/queue.c
index e06da4b3b0d4..805a3c1bf8fe 100644
--- a/drivers/net/wireless/st/cw1200/queue.c
+++ b/drivers/net/wireless/st/cw1200/queue.c
@@ -91,23 +91,25 @@ static void __cw1200_queue_gc(struct cw1200_queue *queue,
bool unlock)
{
struct cw1200_queue_stats *stats = queue->stats;
- struct cw1200_queue_item *item = NULL, *tmp;
+ struct cw1200_queue_item *item = NULL, *iter, *tmp;
bool wakeup_stats = false;
- list_for_each_entry_safe(item, tmp, &queue->queue, head) {
- if (time_is_after_jiffies(item->queue_timestamp + queue->ttl))
+ list_for_each_entry_safe(iter, tmp, &queue->queue, head) {
+ if (time_is_after_jiffies(iter->queue_timestamp + queue->ttl)) {
+ item = iter;
break;
+ }
--queue->num_queued;
- --queue->link_map_cache[item->txpriv.link_id];
+ --queue->link_map_cache[iter->txpriv.link_id];
spin_lock_bh(&stats->lock);
--stats->num_queued;
- if (!--stats->link_map_cache[item->txpriv.link_id])
+ if (!--stats->link_map_cache[iter->txpriv.link_id])
wakeup_stats = true;
spin_unlock_bh(&stats->lock);
cw1200_debug_tx_ttl(stats->priv);
- cw1200_queue_register_post_gc(head, item);
- item->skb = NULL;
- list_move_tail(&item->head, &queue->free_pool);
+ cw1200_queue_register_post_gc(head, iter);
+ iter->skb = NULL;
+ list_move_tail(&iter->head, &queue->free_pool);
}
if (wakeup_stats)
--
2.17.1
Xiaomeng Tong <[email protected]> writes:
> The bug is here: "} else if (item) {".
>
> The list iterator value will *always* be set and non-NULL by
> list_for_each_entry(), so it is incorrect to assume that the iterator
> value will be NULL if the list is empty or no element is found in list.
>
> Use a new value 'iter' as the list iterator, while use the old value
> 'item' as a dedicated pointer to point to the found element, which
> 1. can fix this bug, due to now 'item' is NULL only if it's not found.
> 2. do not need to change all the uses of 'item' after the loop.
> 3. can also limit the scope of the list iterator 'iter' *only inside*
> the traversal loop by simply declaring 'iter' inside the loop in the
> future, as usage of the iterator outside of the list_for_each_entry
> is considered harmful. https://lkml.org/lkml/2022/2/17/1032
>
> Fixes: a910e4a94f692 ("cw1200: add driver for the ST-E CW1100 & CW1200 WLAN chipsets")
> Signed-off-by: Xiaomeng Tong <[email protected]>
> ---
> changes since v2:
> - rebase on latest wireless-next (Kalle Valo)
> changes since v1:
> - fix incorrect check to item (Jakob Koschel)
>
> v2: https://lore.kernel.org/lkml/[email protected]/
> v1: https://lore.kernel.org/all/[email protected]/
> ---
> drivers/net/wireless/st/cw1200/queue.c | 18 ++++++++++--------
> 1 file changed, 10 insertions(+), 8 deletions(-)
>
> diff --git a/drivers/net/wireless/st/cw1200/queue.c b/drivers/net/wireless/st/cw1200/queue.c
> index e06da4b3b0d4..805a3c1bf8fe 100644
> --- a/drivers/net/wireless/st/cw1200/queue.c
> +++ b/drivers/net/wireless/st/cw1200/queue.c
> @@ -91,23 +91,25 @@ static void __cw1200_queue_gc(struct cw1200_queue *queue,
> bool unlock)
> {
> struct cw1200_queue_stats *stats = queue->stats;
> - struct cw1200_queue_item *item = NULL, *tmp;
> + struct cw1200_queue_item *item = NULL, *iter, *tmp;
> bool wakeup_stats = false;
>
> - list_for_each_entry_safe(item, tmp, &queue->queue, head) {
> - if (time_is_after_jiffies(item->queue_timestamp + queue->ttl))
> + list_for_each_entry_safe(iter, tmp, &queue->queue, head) {
> + if (time_is_after_jiffies(iter->queue_timestamp + queue->ttl)) {
> + item = iter;
> break;
> + }
> --queue->num_queued;
> - --queue->link_map_cache[item->txpriv.link_id];
> + --queue->link_map_cache[iter->txpriv.link_id];
> spin_lock_bh(&stats->lock);
> --stats->num_queued;
> - if (!--stats->link_map_cache[item->txpriv.link_id])
> + if (!--stats->link_map_cache[iter->txpriv.link_id])
> wakeup_stats = true;
> spin_unlock_bh(&stats->lock);
> cw1200_debug_tx_ttl(stats->priv);
> - cw1200_queue_register_post_gc(head, item);
> - item->skb = NULL;
> - list_move_tail(&item->head, &queue->free_pool);
> + cw1200_queue_register_post_gc(head, iter);
> + iter->skb = NULL;
> + list_move_tail(&iter->head, &queue->free_pool);
> }
>
> if (wakeup_stats)
I started to look at this myself. I don't know if I'm missing something,
but is the time_is_after_jiffies() really correct? This was added by
Wang in commit 8cbc3d51b4ae ("cw1200: use time_is_after_jiffies()
instead of open coding it"):
- if (jiffies - item->queue_timestamp < queue->ttl)
+ if (time_is_after_jiffies(item->queue_timestamp + queue->ttl))
https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=8cbc3d51b4ae
--
https://patchwork.kernel.org/project/linux-wireless/list/
https://wireless.wiki.kernel.org/en/developers/documentation/submittingpatches
Xiaomeng Tong <[email protected]> wrote:
> The bug is here: "} else if (item) {".
>
> The list iterator value will *always* be set and non-NULL by
> list_for_each_entry(), so it is incorrect to assume that the iterator
> value will be NULL if the list is empty or no element is found in list.
>
> Use a new value 'iter' as the list iterator, while use the old value
> 'item' as a dedicated pointer to point to the found element, which
> 1. can fix this bug, due to now 'item' is NULL only if it's not found.
> 2. do not need to change all the uses of 'item' after the loop.
> 3. can also limit the scope of the list iterator 'iter' *only inside*
> the traversal loop by simply declaring 'iter' inside the loop in the
> future, as usage of the iterator outside of the list_for_each_entry
> is considered harmful. https://lkml.org/lkml/2022/2/17/1032
>
> Fixes: a910e4a94f692 ("cw1200: add driver for the ST-E CW1100 & CW1200 WLAN chipsets")
> Signed-off-by: Xiaomeng Tong <[email protected]>
Patch applied to wireless-next.git, thanks.
86df5de5c632 cw1200: fix incorrect check to determine if no element is found in list
--
https://patchwork.kernel.org/project/linux-wireless/patch/[email protected]/
https://wireless.wiki.kernel.org/en/developers/documentation/submittingpatches