diff options
author | Mike Marciniszyn <mike.marciniszyn@intel.com> | 2020-06-23 23:43:28 +0300 |
---|---|---|
committer | Jason Gunthorpe <jgg@nvidia.com> | 2020-06-24 22:13:38 +0300 |
commit | 38fd98afeeb79d3b148db49f81f2ec6a37a4ee00 (patch) | |
tree | 8c3b8047aacb3a84a6253e505461199c40407fb0 /drivers/infiniband/hw/hfi1/ipoib.h | |
parent | 82172b765530f84b4b9da929f2dcf46f2b7b232b (diff) | |
download | linux-38fd98afeeb79d3b148db49f81f2ec6a37a4ee00.tar.xz |
IB/hfi1: Add atomic triggered sleep/wakeup
When running iperf in a two host configuration the following trace can
occur:
[ 319.728730] NETDEV WATCHDOG: ib0 (hfi1): transmit queue 0 timed out
The issue happens because the current implementation relies on the netif
txq being stopped to control the flushing of the tx list.
There are two resources that the transmit logic can wait on and stop the
txq:
- SDMA descriptors
- Ring space to hold completions
The ring space is tested on the sending side and relieved when the ring is
consumed in the napi tx reaping.
Unfortunately, that reaping can run conncurrently with the workqueue
flushing of the txlist. If the txq is started just before the workitem
executes, the txlist will never be flushed, leading to the txq being
stuck.
Fix by:
- Adding sleep/wakeup wrappers
* Use an atomic to control the call to the netif routines inside the
wrappers
- Use another atomic to record ring space exhaustion
* Only wakeup when the a ring space exhaustion has happened and it
relieved
Add additional wrappers to clarify the ring space resource handling.
Fixes: d99dc602e2a5 ("IB/hfi1: Add functions to transmit datagram ipoib packets")
Link: https://lore.kernel.org/r/20200623204327.108092.4024.stgit@awfm-01.aw.intel.com
Reviewed-by: Kaike Wan <kaike.wan@intel.com>
Signed-off-by: Mike Marciniszyn <mike.marciniszyn@intel.com>
Signed-off-by: Dennis Dalessandro <dennis.dalessandro@intel.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
Diffstat (limited to 'drivers/infiniband/hw/hfi1/ipoib.h')
-rw-r--r-- | drivers/infiniband/hw/hfi1/ipoib.h | 6 |
1 files changed, 6 insertions, 0 deletions
diff --git a/drivers/infiniband/hw/hfi1/ipoib.h b/drivers/infiniband/hw/hfi1/ipoib.h index 185c9b02c974..b8c9d0a003fb 100644 --- a/drivers/infiniband/hw/hfi1/ipoib.h +++ b/drivers/infiniband/hw/hfi1/ipoib.h @@ -67,6 +67,9 @@ struct hfi1_ipoib_circ_buf { * @sde: sdma engine * @tx_list: tx request list * @sent_txreqs: count of txreqs posted to sdma + * @stops: count of stops of queue + * @ring_full: ring has been filled + * @no_desc: descriptor shortage seen * @flow: tracks when list needs to be flushed for a flow change * @q_idx: ipoib Tx queue index * @pkts_sent: indicator packets have been sent from this queue @@ -80,6 +83,9 @@ struct hfi1_ipoib_txq { struct sdma_engine *sde; struct list_head tx_list; u64 sent_txreqs; + atomic_t stops; + atomic_t ring_full; + atomic_t no_desc; union hfi1_ipoib_flow flow; u8 q_idx; bool pkts_sent; |