This patch fixes saving IV received in SNB to storage.
Previously after creating new node with IV 0 (or loading node with
IV > 0 but after long inactivity) first received SNB should update IV
(and reset sequence number to 0 if necessary).
The bug would prevent new IV being saved in storage
which resulted in sequence number being set to 0
on first SNB received after every daemon reset
but IV never being updated.
---
mesh/net.c | 20 +++++++-------------
1 file changed, 7 insertions(+), 13 deletions(-)
diff --git a/mesh/net.c b/mesh/net.c
index 5be95086d..1598933d4 100644
--- a/mesh/net.c
+++ b/mesh/net.c
@@ -2706,14 +2706,7 @@ static void update_kr_state(struct mesh_subnet *subnet, bool kr, uint32_t id)
static void update_iv_ivu_state(struct mesh_net *net, uint32_t iv_index,
bool ivu)
{
- uint32_t local_iv_index;
- bool local_ivu;
-
- /* Save original settings to differentiate what has changed */
- local_iv_index = net->iv_index;
- local_ivu = net->iv_update;
-
- if ((iv_index - ivu) > (local_iv_index - local_ivu)) {
+ if ((iv_index - ivu) > (net->iv_index - net->iv_update)) {
/* Don't accept IV_Index changes when performing SAR Out */
if (l_queue_length(net->sar_out))
return;
@@ -2737,26 +2730,27 @@ static void update_iv_ivu_state(struct mesh_net *net, uint32_t iv_index,
}
} else if (ivu) {
/* Ignore beacons with IVU if they come too soon */
- if (!local_ivu && net->iv_upd_state == IV_UPD_NORMAL_HOLD) {
+ if (!net->iv_update &&
+ net->iv_upd_state == IV_UPD_NORMAL_HOLD) {
l_error("Update attempted too soon");
return;
}
- if (!local_ivu) {
+ if (!net->iv_update) {
l_info("iv_upd_state = IV_UPD_UPDATING");
net->iv_upd_state = IV_UPD_UPDATING;
net->iv_update_timeout = l_timeout_create(
IV_IDX_UPD_MIN, iv_upd_to, net, NULL);
}
- } else if (local_ivu) {
+ } else if (net->iv_update) {
l_error("IVU clear attempted too soon");
return;
}
- if ((iv_index - ivu) > (local_iv_index - local_ivu))
+ if ((iv_index - ivu) > (net->iv_index - net->iv_update))
mesh_net_set_seq_num(net, 0);
- if (ivu != net->iv_update || local_iv_index != net->iv_index) {
+ if (ivu != net->iv_update || iv_index != net->iv_index) {
struct mesh_config *cfg = node_config_get(net->node);
mesh_config_write_iv_index(cfg, iv_index, ivu);
--
2.22.0
Hi Rafał,
On Wed, 2020-01-08 at 10:16 +0100, Rafał Gajda wrote:
> This patch fixes saving IV received in SNB to storage.
>
> Previously after creating new node with IV 0 (or loading node with
> IV > 0 but after long inactivity) first received SNB should update IV
> (and reset sequence number to 0 if necessary).
>
> The bug would prevent new IV being saved in storage
> which resulted in sequence number being set to 0
> on first SNB received after every daemon reset
> but IV never being updated.
> ---
> mesh/net.c | 20 +++++++-------------
> 1 file changed, 7 insertions(+), 13 deletions(-)
>
> diff --git a/mesh/net.c b/mesh/net.c
> index 5be95086d..1598933d4 100644
> --- a/mesh/net.c
> +++ b/mesh/net.c
>
<...>
> @@ -2737,26 +2730,27 @@ static void update_iv_ivu_state(struct mesh_net *net, uint32_t iv_index,
> }
> } else if (ivu) {
> /* Ignore beacons with IVU if they come too soon */
> - if (!local_ivu && net->iv_upd_state == IV_UPD_NORMAL_HOLD) {
> + if (!net->iv_update &&
> + net->iv_upd_state == IV_UPD_NORMAL_HOLD) {
> l_error("Update attempted too soon");
> return;
> }
>
> - if (!local_ivu) {
> + if (!net->iv_update) {
> l_info("iv_upd_state = IV_UPD_UPDATING");
> net->iv_upd_state = IV_UPD_UPDATING;
> net->iv_update_timeout = l_timeout_create(
> IV_IDX_UPD_MIN, iv_upd_to, net, NULL);
> }
> - } else if (local_ivu) {
> + } else if (net->iv_update) {
> l_error("IVU clear attempted too soon");
> return;
> }
>
> - if ((iv_index - ivu) > (local_iv_index - local_ivu))
> + if ((iv_index - ivu) > (net->iv_index - net->iv_update))
> mesh_net_set_seq_num(net, 0);
>
> - if (ivu != net->iv_update || local_iv_index != net->iv_index) {
> + if (ivu != net->iv_update || iv_index != net->iv_index) {
Would it be fair to say that this is the *only* change to the code that fixes the logic that was
malfunctioning?
If so, I would like to shrink this patch to just:
- if (ivu != net->iv_update || local_iv_index != net->iv_index) {
+ if (ivu != local_ivu || iv_index != local_iv_index) {
or, breaking this into a 2-patch patchset, where one of the patches removes the local caching of ivu and ivi,
and the second patch fixes the actual bug.
> struct mesh_config *cfg = node_config_get(net->node);
>
> mesh_config_write_iv_index(cfg, iv_index, ivu);
Brian, Rafał,
On 01/08, Gix, Brian wrote:
> > - if (ivu != net->iv_update || local_iv_index != net->iv_index) {
> > + if (ivu != net->iv_update || iv_index != net->iv_index) {
>
> Would it be fair to say that this is the *only* change to the code that fixes the logic that was
> malfunctioning?
That's correct.
>
> If so, I would like to shrink this patch to just:
>
> - if (ivu != net->iv_update || local_iv_index != net->iv_index) {
> + if (ivu != local_ivu || iv_index != local_iv_index) {
>
> or, breaking this into a 2-patch patchset, where one of the patches removes the local caching of ivu and ivi,
> and the second patch fixes the actual bug.
Ok, let's split this into 2 separate patches. I still think this "local
caching" was the underlying cause of the mistake.
regards
--
Michał Lowas-Rzechonek <[email protected]>
Silvair http://silvair.com
Jasnogórska 44, 31-358 Krakow, POLAND
On Wed, 2020-01-08 at 19:17 +0100, Michał Lowas-Rzechonek wrote:
> Brian, Rafał,
>
> On 01/08, Gix, Brian wrote:
> > > - if (ivu != net->iv_update || local_iv_index != net->iv_index) {
> > > + if (ivu != net->iv_update || iv_index != net->iv_index) {
> >
> > Would it be fair to say that this is the *only* change to the code that fixes the logic that was
> > malfunctioning?
>
> That's correct.
>
> > If so, I would like to shrink this patch to just:
> >
> > - if (ivu != net->iv_update || local_iv_index != net->iv_index) {
> > + if (ivu != local_ivu || iv_index != local_iv_index) {
> >
> > or, breaking this into a 2-patch patchset, where one of the patches removes the local caching of ivu and
> > ivi,
> > and the second patch fixes the actual bug.
>
> Ok, let's split this into 2 separate patches. I still think this "local
> caching" was the underlying cause of the mistake.
It probably makes more sense for the *first* patch of the set to fix the logic error,
and with the second patch, remove the local caching... Otherwise the cache removal
patch will include the absurdity of comparing net->iv_index against itself.
>
> regards