Hi Alexander,

aahringo@redhat.com wrote on Sat, 3 Jun 2023 07:30:05 -0400:

Hi,

On Thu, Jun 1, 2023 at 11:50 AM Miquel Raynal [off-list ref] wrote:

quoted

Devices may decide to disassociate from their coordinator for different
reasons (device turning off, coordinator signal strength too low, etc),
the MAC layer just has to send a disassociation notification.

If the ack of the disassociation notification is not received, the
device may consider itself disassociated anyway.

Signed-off-by: Miquel Raynal <miquel.raynal@bootlin.com>
---
 net/ieee802154/pan.c         |   2 +
 net/mac802154/cfg.c          | 102 +++++++++++++++++++++++++++++++++++
 net/mac802154/ieee802154_i.h |   4 ++
 net/mac802154/scan.c         |  60 +++++++++++++++++++++
 4 files changed, 168 insertions(+)

diff --git a/net/ieee802154/pan.c b/net/ieee802154/pan.c
index e2a12a42ba2b..477e8dad0cf0 100644
--- a/net/ieee802154/pan.c
+++ b/net/ieee802154/pan.c

@@ -49,6 +49,7 @@ bool cfg802154_device_is_parent(struct wpan_dev *wpan_dev,

        return false;
 }
+EXPORT_SYMBOL_GPL(cfg802154_device_is_parent);

 struct ieee802154_pan_device *
 cfg802154_device_is_child(struct wpan_dev *wpan_dev,

@@ -64,3 +65,4 @@ cfg802154_device_is_child(struct wpan_dev *wpan_dev,

        return NULL;
 }
+EXPORT_SYMBOL_GPL(cfg802154_device_is_child);

diff --git a/net/mac802154/cfg.c b/net/mac802154/cfg.c
index 89112d2bcee7..c27c05e825ff 100644
--- a/net/mac802154/cfg.c
+++ b/net/mac802154/cfg.c

@@ -386,6 +386,107 @@ static int mac802154_associate(struct wpan_phy *wpan_phy,
        return ret;
 }

+static int mac802154_disassociate_from_parent(struct wpan_phy *wpan_phy,
+                                             struct wpan_dev *wpan_dev)
+{
+       struct ieee802154_local *local = wpan_phy_priv(wpan_phy);
+       struct ieee802154_pan_device *child, *tmp;
+       struct ieee802154_sub_if_data *sdata;
+       u64 eaddr;
+       int ret;
+
+       sdata = IEEE802154_WPAN_DEV_TO_SUB_IF(wpan_dev);
+
+       /* Start by disassociating all the children and preventing new ones to
+        * attempt associations.
+        */
+       list_for_each_entry_safe(child, tmp, &wpan_dev->children, node) {
+               ret = mac802154_send_disassociation_notif(sdata, child,
+                                                         IEEE802154_COORD_WISHES_DEVICE_TO_LEAVE);
+               if (ret) {
+                       eaddr = swab64((__force u64)child->extended_addr);  

Does this pass sparse? I think this needs to be le64_to_cpu()?

I never feel comfortable with sparse given the dozens (hundreds) of
lines it outputs, but I think yes, parse does not seem to complain. To
be honest I think we should keep it this way just because I copy-pasted
it from other locations in the core:

$ git grep -c "swab64((__force u64)" -- net/ieee802154/ net/mac802154/
net/ieee802154/nl-mac.c:1
net/mac802154/cfg.c:4
net/mac802154/llsec.c:1
net/mac802154/rx.c:2
net/mac802154/scan.c:6

So if we ever want to change that, we could easily find them all and
replace them all in one go?

Thanks,
Miquèl