i40e: fix regression that enables AF_XDP ZC unconditionally
Subsystems (3)
cover i40e: fix regression that enables AF_XDP ZC unconditionally
This series addresses a recent AF_XDP zero-copy regression.
In commit f3fef2b6e1cc ("i40e: Remove umem from VSI") a regression was introduced; When the VSI was reset, the setup code would try to enable AF_XDP ZC unconditionally (as long as there was a umem placed in the netdev._rx struct). Here, we add a bitmap to the VSI that tracks if a certain queue pair has been "zero-copy enabled" via the ndo_bpf. The bitmap is used in i40e_xsk_umem, and enables zero-copy if and only if XDP is enabled, the corresponding qid in the bitmap is set and the umem is non-NULL.
Thanks,
Bj?rn
Bj?rn T?pel (2):
i40e: move i40e_xsk_umem function
i40e: add tracking of AF_XDP ZC state for each queue pair
drivers/net/ethernet/intel/i40e/i40e.h | 16 ++---------- drivers/net/ethernet/intel/i40e/i40e_main.c | 28 +++++++++++++++++++++ drivers/net/ethernet/intel/i40e/i40e_xsk.c | 3 +++ 3 files changed, 33 insertions(+), 14 deletions(-)
Signature · 1 line
2.19.1
PATCH 1/2 i40e: move i40e_xsk_umem function
From: Bj?rn T?pel <[email protected]>
The i40e_xsk_umem function was explicitly inlined in i40e.h. There is no reason for that, so move it to i40e_main.c instead.
--- drivers/net/ethernet/intel/i40e/i40e.h | 14 -------------- drivers/net/ethernet/intel/i40e/i40e_main.c | 20 ++++++++++++++++++++ 2 files changed, 20 insertions(+), 14 deletions(-)
drivers/net/ethernet/intel/i40e/i40e.h +0 -14
@@ -1096,20 +1096,6 @@ static inline bool i40e_enabled_xdp_vsi(struct i40e_vsi *vsi)return !!vsi->xdp_prog; } -static inline struct xdp_umem *i40e_xsk_umem (struct i40e_ring *ring) -{ -bool xdp_on =i40e_enabled_xdp_vsi (ring->vsi); -int qid = ring->queue_index; - -if (ring_is_xdp (ring)) - qid -= ring->vsi->alloc_queue_pairs; - -if (!xdp_on) -return NULL ; - -return xdp_get_umem_from_qid (ring->vsi->netdev, qid); -} -int i40e_create_queue_channel (struct i40e_vsi *vsi,struct i40e_channel *ch);int i40e_set_bw_limit (struct i40e_vsi *vsi,u16 seid,u64 max_tx_rate); int i40e_add_del_cloud_filter(struct i40e_vsi *vsi,
drivers/net/ethernet/intel/i40e/i40e_main.c +20 -0
@@ -3063,6 +3063,26 @@ static void i40e_config_xps_tx_ring(struct i40e_ring *ring) ring->queue_index); } +/** +* i40e_xsk_umem - Retrieve the AF_XDP ZC if XDP and ZC is enabled +* @ring: The Tx or Rx ring +* +* Returns the UMEM or NULL. +**/ +static struct xdp_umem *i40e_xsk_umem (struct i40e_ring *ring) +{ +bool xdp_on =i40e_enabled_xdp_vsi (ring->vsi); +int qid = ring->queue_index; + +if (ring_is_xdp (ring)) + qid -= ring->vsi->alloc_queue_pairs; + +if (!xdp_on) +return NULL ; + +return xdp_get_umem_from_qid (ring->vsi->netdev, qid); +} +/** * i40e_configure_tx_ring - Configure a transmit ring context and rest * @ring: The Tx ring to configure
Signature · 1 line
2.19.1
RE: [PATCH 1/2] i40e: move i40e_xsk_umem function
Quoted 11 lines
From: netdev-owner at vger.kernel.org [mailto:netdev- owner at vger.kernel.org] On Behalf Of Bj?rn T?pel
Sent: Tuesday, February 12, 2019 12:52 AM
To: intel-wired-lan at lists.osuosl.org
Cc: Topel, Bjorn <[email protected]>; Karlsson, Magnus <[email protected]>; magnus.karlsson at gmail.com; netdev at vger.kernel.org; Sokolowski, Jan <[email protected]>
Subject: [PATCH 1/2] i40e: movei40e_xsk_umemfunctionFrom: Bj?rn T?pel <[email protected]>
Quoted 9 lines
The
i40e_xsk_umemfunction was explicitly inlined ini40e.h. There is no reason for that, so move it toi40e_main.cinstead.Signed-off-by Bj?rn T?pel <[email protected]>--- drivers/net/ethernet/intel/i40e/i40e.h | 14 -------------- drivers/net/ethernet/intel/i40e/i40e_main.c | 20 ++++++++++++++++++++ 2 files changed, 20 insertions(+), 14 deletions(-)
PATCH 2/2 i40e: add tracking of AF_XDP ZC state for each queue pair
In reply to Björn TöpelFrom: Bj?rn T?pel <[email protected]>
In commit f3fef2b6e1cc ("i40e: Remove umem from VSI") a regression was introduced; When the VSI was reset, the setup code would try to enable AF_XDP ZC unconditionally (as long as there was a umem placed in the netdev._rx struct). Here, we add a bitmap to the VSI that tracks if a certain queue pair has been "zero-copy enabled" via the ndo_bpf. The bitmap is used in i40e_xsk_umem, and enables zero-copy if and only if XDP is enabled, the corresponding qid in the bitmap is set and the umem is non-NULL.
--- drivers/net/ethernet/intel/i40e/i40e.h | 2 ++ drivers/net/ethernet/intel/i40e/i40e_main.c | 10 +++++++++- drivers/net/ethernet/intel/i40e/i40e_xsk.c | 3 +++ 3 files changed, 14 insertions(+), 1 deletion(-)
drivers/net/ethernet/intel/i40e/i40e.h +2 -0
@@ -790,6 +790,8 @@ struct i40e_vsi {/* VSI specific handlers */ irqreturn_t (*irq_handler)(int irq,void *data); + +unsigned long *af_xdp_zc_qps;/* tracks AF_XDP ZC enabled qps */ }____cacheline_internodealigned_in_smp ;struct i40e_netdev_priv {
drivers/net/ethernet/intel/i40e/i40e_main.c +9 -1
@@ -3077,7 +3077,7 @@ static struct xdp_umem *i40e_xsk_umem(struct i40e_ring *ring)if (ring_is_xdp (ring)) qid -= ring->vsi->alloc_queue_pairs; -if (!xdp_on) +if (!xdp_on || !test_bit (qid, ring->vsi->af_xdp_zc_qps))return NULL ;return xdp_get_umem_from_qid (ring->vsi->netdev, qid);
@@ -10076,6 +10076,12 @@ static int i40e_vsi_mem_alloc(struct i40e_pf *pf, enum i40e_vsi_type type)hash_init (vsi->mac_filter_hash); vsi->irqs_ready = false; +if (type ==I40E_VSI_MAIN ) { + vsi->af_xdp_zc_qps =bitmap_zalloc (pf->num_lan_qps,GFP_KERNEL ); +if (!vsi->af_xdp_zc_qps) + goto err_rings; + } + ret =i40e_set_num_rings_in_vsi (vsi);if (ret) goto err_rings;
@@ -10094,6 +10100,7 @@ static int i40e_vsi_mem_alloc(struct i40e_pf *pf, enum i40e_vsi_type type) goto unlock_pf; err_rings: +bitmap_free (vsi->af_xdp_zc_qps); pf->next_vsi = i -1 ;kfree (vsi); unlock_pf:
@@ -10174,6 +10181,7 @@ static int i40e_vsi_clear(struct i40e_vsi *vsi)i40e_put_lump (pf->qp_pile, vsi->base_queue, vsi->idx);i40e_put_lump (pf->irq_pile, vsi->base_vector, vsi->idx); +bitmap_free (vsi->af_xdp_zc_qps);i40e_vsi_free_arrays (vsi, true);i40e_clear_rss_config_user (vsi);
drivers/net/ethernet/intel/i40e/i40e_xsk.c +3 -0
@@ -102,6 +102,8 @@ static int i40e_xsk_umem_enable(struct i40e_vsi *vsi, struct xdp_umem *umem,if (err)return err; +set_bit (qid, vsi->af_xdp_zc_qps); + if_running =netif_running (vsi->netdev) &&i40e_enabled_xdp_vsi (vsi);if (if_running) {
@@ -143,6 +145,7 @@ static int i40e_xsk_umem_disable(struct i40e_vsi *vsi, u16 qid)return err; } +clear_bit (qid, vsi->af_xdp_zc_qps);i40e_xsk_umem_dma_unmap (vsi, umem);if (if_running) {
Signature · 1 line
2.19.1
RE: [PATCH 2/2] i40e: add tracking of AF_XDP ZC state for each queue pair
Quoted 21 lines
From: netdev-owner at vger.kernel.org [mailto:netdev-
owner at vger.kernel.org] On Behalf Of Bj?rn T?pel
Sent: Tuesday, February 12, 2019 12:52 AM
To: intel-wired-lan at lists.osuosl.org
Cc: Topel, Bjorn <[email protected]>; Karlsson, Magnus <[email protected]>; magnus.karlsson at gmail.com; netdev at vger.kernel.org; Sokolowski, Jan <[email protected]>
Subject: [PATCH 2/2] i40e: add tracking ofAF_XDPZC state for each queue pairFrom: Bj?rn T?pel <[email protected]>
In commit
f3fef2b6e1cc("i40e: Remove umem from VSI") a regression was introduced; When the VSI was reset, the setup code would try to enableAF_XDPZC unconditionally (as long as there was a umem placed in the netdev._rx struct). Here, we add a bitmap to the VSI that tracks if a certain queue pair has been "zero-copy enabled" via thendo_bpf. The bitmap is used ini40e_xsk_umem, and enables zero-copy if and only if XDP is enabled, the corresponding qid in the bitmap is set and the umem is non-NULL.
Quoted 8 lines
Fixes f3fef2b6e1cc ("i40e: Remove umem from VSI")Signed-off-by Bj?rn T?pel <[email protected]>--- drivers/net/ethernet/intel/i40e/i40e.h | 2 ++ drivers/net/ethernet/intel/i40e/i40e_main.c | 10 +++++++++- drivers/net/ethernet/intel/i40e/i40e_xsk.c | 3 +++ 3 files changed, 14 insertions(+), 1 deletion(-)