diff options
| author | Jack Wang <jinpu.wang@ionos.com> | 2023-11-21 14:03:16 +0100 |
|---|---|---|
| committer | Leon Romanovsky <leon@kernel.org> | 2023-11-26 11:32:00 +0200 |
| commit | 50af5d12f7e24b85fc10270d7700f4aa1b20b8e4 (patch) | |
| tree | ebb4a1b8a506224c61f663f8e84cbecfe1bfe461 /drivers/infiniband/ulp/ipoib/ipoib_ib.c | |
| parent | 753fff78f430704548f45eda52d6d55371a52c0f (diff) | |
| download | linux-next-50af5d12f7e24b85fc10270d7700f4aa1b20b8e4.tar.gz linux-next-50af5d12f7e24b85fc10270d7700f4aa1b20b8e4.zip | |
RDMA/IPoIB: Add tx timeout work to recover queue stop situation
As we sometime run into TX timeout from IPoIB, queue seems stopped
and can't recover. Diff with Mellanox OFED show Mellanox driver
has timeout work to recover in such case.
Add TX timeout work/NAPI work to recover such case.
Also increase the watchdog_timeo to 10 seconds, so more tolerant to
error.
Signed-off-by: Jack Wang <jinpu.wang@ionos.com>
Link: https://lore.kernel.org/r/20231121130316.126364-3-jinpu.wang@ionos.com
Signed-off-by: Leon Romanovsky <leon@kernel.org>
Diffstat (limited to 'drivers/infiniband/ulp/ipoib/ipoib_ib.c')
| -rw-r--r-- | drivers/infiniband/ulp/ipoib/ipoib_ib.c | 26 |
1 files changed, 25 insertions, 1 deletions
diff --git a/drivers/infiniband/ulp/ipoib/ipoib_ib.c b/drivers/infiniband/ulp/ipoib/ipoib_ib.c index 7f84d9866cef..5cde275daa94 100644 --- a/drivers/infiniband/ulp/ipoib/ipoib_ib.c +++ b/drivers/infiniband/ulp/ipoib/ipoib_ib.c @@ -531,11 +531,35 @@ void ipoib_ib_rx_completion(struct ib_cq *cq, void *ctx_ptr) napi_schedule(&priv->recv_napi); } +/* The function will force napi_schedule */ +void ipoib_napi_schedule_work(struct work_struct *work) +{ + struct ipoib_dev_priv *priv = + container_of(work, struct ipoib_dev_priv, reschedule_napi_work); + bool ret; + + do { + ret = napi_schedule(&priv->send_napi); + if (!ret) + msleep(3); + } while (!ret && netif_queue_stopped(priv->dev) && + test_bit(IPOIB_FLAG_INITIALIZED, &priv->flags)); +} + void ipoib_ib_tx_completion(struct ib_cq *cq, void *ctx_ptr) { struct ipoib_dev_priv *priv = ctx_ptr; + bool ret; - napi_schedule(&priv->send_napi); + ret = napi_schedule(&priv->send_napi); + /* + * if the queue is closed the driver must be able to schedule napi, + * otherwise we can end with closed queue forever, because no new + * packets to send and napi callback might not get new event after + * its re-arm of the napi. + */ + if (!ret && netif_queue_stopped(priv->dev)) + schedule_work(&priv->reschedule_napi_work); } static inline int post_send(struct ipoib_dev_priv *priv, |