kernel/linux.git/net/mac802154/tx.c, branch v6.6.134

net: mac802154: Fix racy device stats updates by DEV_STATS_INC() and DEV_STATS_ADD()

2024-07-25T07:50:52+00:00

[ Upstream commit b8ec0dc3845f6c9089573cb5c2c4b05f7fc10728 ] mac802154 devices update their dev->stats fields locklessly. Therefore these counters should be updated atomically. Adopt SMP safe DEV_STATS_INC() and DEV_STATS_ADD() to achieve this. Signed-off-by: Yunshui Jiang Message-ID: <20240531080739.2608969-1-jiangyunshui@kylinos.cn> Signed-off-by: Stefan Schmidt Signed-off-by: Sasha Levin

mac802154: Add MLME Tx locked helpers

2023-01-03T18:45:44+00:00

These have the exact same behavior as before, except they expect the rtnl to be already taken (and will complain otherwise). This allows performing MLME transmissions from different contexts. Signed-off-by: Miquel Raynal Acked-by: Alexander Aring Link: https://lore.kernel.org/r/20230103165644.432209-6-miquel.raynal@bootlin.com Signed-off-by: Stefan Schmidt

net: mac802154: Fix a Tx warning check

2022-06-27T08:57:32+00:00

The purpose of the netif_is_down() helper was to ensure that the network interface used was still up when performing the transmission. What it actually did was to check if _all_ interfaces were up. This was not noticed at that time because I did not use interfaces at all before discussing with Alexander Aring about how to handle coordinators properly. Drop the helper and call netif_running() on the right sub interface object directly. Fixes: 4f790184139b ("net: mac802154: Add a warning in the slow path") Signed-off-by: Miquel Raynal Acked-by: Alexander Aring Link: https://lore.kernel.org/r/20220617192914.1275611-1-miquel.raynal@bootlin.com Signed-off-by: Stefan Schmidt

mac802154: fix atomic_dec_and_test checks

2022-06-14T08:23:51+00:00

We need to call wake_up() when hold_txs reaches zero. The semantic of atomic_dec_and_test() is that it returns true when it's zero. Fixes: f0feb3490473 ("net: mac802154: Introduce a tx queue flushing mechanism") Signed-off-by: Alexander Aring Reviewed-by: Miquel Raynal Link: https://lore.kernel.org/r/20220613043735.1039895-3-aahringo@redhat.com Signed-off-by: Stefan Schmidt

net: mac802154: Add a warning in the slow path

2022-06-10T07:48:41+00:00

In order to be able to detect possible conflicts between the net interface core and the ieee802154 core, let's add a warning in the slow path: we want to be sure that whenever we start an asynchronous MLME transmission (which can be fully asynchronous) the net core somehow agrees that this transmission is possible, ie. the device was not stopped. Warning in this case would allow us to track down more easily possible issues with the MLME logic if we ever get reports. Unlike in the hot path, such a situation cannot be handled. Signed-off-by: Miquel Raynal Acked-by: Alexander Aring Link: https://lore.kernel.org/r/20220519150516.443078-12-miquel.raynal@bootlin.com Signed-off-by: Stefan Schmidt

net: mac802154: Add a warning in the hot path

2022-06-10T07:48:41+00:00

We should never start a transmission after the queue has been stopped. But because it might work we don't kill the function here but rather warn loudly the user that something is wrong. Set a flag when the queue should remain stopped. Reset this flag when the queue actually gets restarded. Just check this value to know if a transmission is legitimate, warn if it is not. Turn the flags variable into an unsigned long to allow the use of atomic helpers on it. Signed-off-by: Miquel Raynal Acked-by: Alexander Aring Link: https://lore.kernel.org/r/20220519150516.443078-11-miquel.raynal@bootlin.com Signed-off-by: Stefan Schmidt

net: mac802154: Introduce a synchronous API for MLME commands

2022-06-10T07:48:40+00:00

This is the slow path, we need to wait for each command to be processed before continuing so let's introduce an helper which does the transmission and blocks until it gets notified of its asynchronous completion. This helper is going to be used when introducing scan support. Signed-off-by: Miquel Raynal Acked-by: Alexander Aring Link: https://lore.kernel.org/r/20220519150516.443078-10-miquel.raynal@bootlin.com Signed-off-by: Stefan Schmidt

net: mac802154: Introduce a tx queue flushing mechanism

2022-06-10T07:48:40+00:00

Right now we are able to stop a queue but we have no indication if a transmission is ongoing or not. Thanks to recent additions, we can track the number of ongoing transmissions so we know if the last transmission is over. Adding on top of it an internal wait queue also allows to be woken up asynchronously when this happens. If, beforehands, we marked the queue to be held and stopped it, we end up flushing and stopping the tx queue. Thanks to this feature, we will soon be able to introduce a synchronous transmit API. Signed-off-by: Miquel Raynal Acked-by: Alexander Aring Link: https://lore.kernel.org/r/20220519150516.443078-9-miquel.raynal@bootlin.com Signed-off-by: Stefan Schmidt

net: mac802154: Create a hot tx path

2022-06-10T07:48:40+00:00

Let's rename the current Tx path to show that this is the "hot" Tx path. We will soon introduce a slower Tx path for MLME commands. Signed-off-by: Miquel Raynal Acked-by: Alexander Aring Link: https://lore.kernel.org/r/20220519150516.443078-7-miquel.raynal@bootlin.com Signed-off-by: Stefan Schmidt

net: mac802154: Bring the ability to hold the transmit queue

2022-06-10T07:48:40+00:00

Create a hold_txs atomic variable and increment/decrement it when relevant, ie. when we want to hold the queue or release it: currently all the "stopped" situations are suitable, but very soon we will more extensively use this feature for MLME purposes. Upon release, the atomic counter is decremented and checked. If it is back to 0, then the netif queue gets woken up. This makes the whole process fully transparent, provided that all the users of ieee802154_wake/stop_queue() now call ieee802154_hold/release_queue() instead. In no situation individual drivers should call any of these helpers manually in order to avoid messing with the counters. There are other functions more suited for this purpose which have been introduced, such as the _xmit_complete() and _xmit_error() helpers which will handle all that for them. One advantage is that, as no more drivers call the stop/wake helpers directly, we can safely stop exporting them and only declare the hold/release ones in a header only accessible to the core. Signed-off-by: Miquel Raynal Acked-by: Alexander Aring Link: https://lore.kernel.org/r/20220519150516.443078-6-miquel.raynal@bootlin.com Signed-off-by: Stefan Schmidt