From: Jacob Keller Date: Thu, 9 Jun 2016 13:28:17 -0700 Subject: [Intel-wired-lan] [PATCH] fm10k: use software values when checking for Tx hangs in hot path Message-ID: <20160609202817.6083-1-jacob.e.keller@intel.com> MIME-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit To: intel-wired-lan@osuosl.org List-ID: A previous patch added support to check for hardware tx pending in the fm10k_down routine. This support was intended to ensure that we accurately check what the hardware state is. However, checking for Tx hangs in this manor during the hotpath results in a large performance hit. Avoid this by making the hotpath check use the SW counters instead. Fixes: a0f53cf49cb0 ("fm10k: use actual hardware registers when checking for pending Tx", 2016-06-08) Signed-off-by: Jacob Keller --- drivers/net/ethernet/intel/fm10k/fm10k_main.c | 19 +++++++++++++++---- drivers/net/ethernet/intel/fm10k/fm10k_pci.c | 2 +- 2 files changed, 16 insertions(+), 5 deletions(-) Jeff, This patch could be squashed into the one mentioned in "fixes" which is currently on your queue. However, it will create conflicts with another patch as well which adds a use of the function. Please let me know if you want me to squash these changes in? If you do it by hand you have to squash this part in and also fix the follow-on patch that uses the function. Regards, Jake diff --git a/drivers/net/ethernet/intel/fm10k/fm10k_main.c b/drivers/net/ethernet/intel/fm10k/fm10k_main.c index e9767b6366a8..a4ea84273ea1 100644 --- a/drivers/net/ethernet/intel/fm10k/fm10k_main.c +++ b/drivers/net/ethernet/intel/fm10k/fm10k_main.c @@ -1128,13 +1128,24 @@ static u64 fm10k_get_tx_completed(struct fm10k_ring *ring) return ring->stats.packets; } -u64 fm10k_get_tx_pending(struct fm10k_ring *ring) +/** + * fm10k_get_tx_pending - how many tx descriptors not processed + * @ring: the ring structure + * @in_sw: is tx_pending being checked in SW or in HW? + */ +u64 fm10k_get_tx_pending(struct fm10k_ring *ring, bool in_sw) { struct fm10k_intfc *interface = ring->q_vector->interface; struct fm10k_hw *hw = &interface->hw; + u32 head, tail; - u32 head = fm10k_read_reg(hw, FM10K_TDH(ring->reg_idx)); - u32 tail = fm10k_read_reg(hw, FM10K_TDT(ring->reg_idx)); + if (likely(in_sw)) { + head = ring->next_to_clean; + tail = ring->next_to_use; + } else { + head = fm10k_read_reg(hw, FM10K_TDH(ring->reg_idx)); + tail = fm10k_read_reg(hw, FM10K_TDT(ring->reg_idx)); + } return ((head <= tail) ? tail : tail + ring->count) - head; } @@ -1143,7 +1154,7 @@ bool fm10k_check_tx_hang(struct fm10k_ring *tx_ring) { u32 tx_done = fm10k_get_tx_completed(tx_ring); u32 tx_done_old = tx_ring->tx_stats.tx_done_old; - u32 tx_pending = fm10k_get_tx_pending(tx_ring); + u32 tx_pending = fm10k_get_tx_pending(tx_ring, true); clear_check_for_tx_hang(tx_ring); diff --git a/drivers/net/ethernet/intel/fm10k/fm10k_pci.c b/drivers/net/ethernet/intel/fm10k/fm10k_pci.c index 559cda999f49..b0c3d7b3a34c 100644 --- a/drivers/net/ethernet/intel/fm10k/fm10k_pci.c +++ b/drivers/net/ethernet/intel/fm10k/fm10k_pci.c @@ -1699,7 +1699,7 @@ void fm10k_down(struct fm10k_intfc *interface) /* start checking at the last ring to have pending Tx */ for (; i < interface->num_tx_queues; i++) - if (fm10k_get_tx_pending(interface->tx_ring[i])) + if (fm10k_get_tx_pending(interface->tx_ring[i], false)) break; /* if all the queues are drained, we can break now */ -- 2.9.0.rc1.405.g81f467e