[PATCH v2 intel-net 0/2] ice XDP fixes

[PATCH v2 intel-net 0/2] ice XDP fixes

[Maciej Fijalkowski]

[resending as v2, vim session got broken while editing patch]

Hi,

here are two small fixes around XDP support in ice driver.

Jamal reported that ice driver does not support XDP on his side. This
got me really puzzling and I had no clue what was going on. Turned that
this is the case when device is in 'safe mode', so let's add a dedicated
ndo_bpf for safe mode ops and make it clear to user what needs to be
fixed. I've described that in the commit message of patch 1 more
thoroughly.

Second issue was found during implementing XDP Tx fallback path for
unsufficient queue count case, which I will send on next week once I'm
back from woods. Hopefully.

Thanks!

Maciej Fijalkowski (2):
  ice: add ndo_bpf callback for safe mode netdev ops
  ice: parametrize functions responsible for Tx ring management

 drivers/net/ethernet/intel/ice/ice_lib.c  | 18 ++++++++++--------
 drivers/net/ethernet/intel/ice/ice_main.c | 15 +++++++++++++++
 2 files changed, 25 insertions(+), 8 deletions(-)

-- 
2.20.1

[PATCH v2 intel-net 1/2] ice: add ndo_bpf callback for safe mode netdev ops

[Maciej Fijalkowski]

ice driver requires a programmable pipeline firmware package in order to
have a support for advanced features. Otherwise, driver falls back to so
called 'safe mode'. For that mode, ndo_bpf callback is not exposed and
when user tries to load XDP program, the following happens:

$ sudo ./xdp1 enp179s0f1
libbpf: Kernel error message: Underlying driver does not support XDP in native mode
link set xdp fd failed

which is sort of confusing, as there is a native XDP support, but not in
the current mode. Improve the user experience by providing the specific
ndo_bpf callback dedicated for safe mode which will make use of extack
to explicitly let the user know that the DDP package is missing and
that's the reason that the XDP can't be loaded onto interface currently.

Cc: Jamal Hadi Salim <jhs@mojatatu.com>
Fixes: efc2214b6047 ("ice: Add support for XDP")
Signed-off-by: Maciej Fijalkowski <maciej.fijalkowski@intel.com>
---
 drivers/net/ethernet/intel/ice/ice_main.c | 15 +++++++++++++++
 1 file changed, 15 insertions(+)

diff --git a/drivers/net/ethernet/intel/ice/ice_main.c b/drivers/net/ethernet/intel/ice/ice_main.c
index 1033d7836891..dfd4f07837cf 100644
--- a/drivers/net/ethernet/intel/ice/ice_main.c
+++ b/drivers/net/ethernet/intel/ice/ice_main.c
@@ -2571,6 +2571,20 @@ ice_xdp_setup_prog(struct ice_vsi *vsi, struct bpf_prog *prog,
 	return (ret || xdp_ring_err) ? -ENOMEM : 0;
 }
 
+/**
+ * ice_xdp_safe_mode - XDP handler for safe mode
+ * @dev: netdevice
+ * @xdp: XDP command
+ */
+static int ice_xdp_safe_mode(struct net_device __always_unused *dev,
+			     struct netdev_bpf *xdp)
+{
+	NL_SET_ERR_MSG_MOD(xdp->extack,
+			   "Please provide working DDP firmware package in order to use XDP\n"
+			   "Refer to Documentation/networking/device_drivers/ethernet/intel/ice.rst");
+	return -EOPNOTSUPP;
+}
+
 /**
  * ice_xdp - implements XDP handler
  * @dev: netdevice
@@ -6953,6 +6967,7 @@ static const struct net_device_ops ice_netdev_safe_mode_ops = {
 	.ndo_change_mtu = ice_change_mtu,
 	.ndo_get_stats64 = ice_get_stats64,
 	.ndo_tx_timeout = ice_tx_timeout,
+	.ndo_bpf = ice_xdp_safe_mode,
 };
 
 static const struct net_device_ops ice_netdev_ops = {
-- 
2.20.1

[PATCH v2 intel-net 2/2] ice: parametrize functions responsible for Tx ring management

[Maciej Fijalkowski]

Commit ae15e0ba1b33 ("ice: Change number of XDP Tx queues to match
number of Rx queues") tried to address the incorrect setting of XDP
queue count that was based on the Tx queue count, whereas in theory we
should provide the XDP queue per Rx queue. However, the routines that
setup and destroy the set of Tx resources are still based on the
vsi->num_txq.

Ice supports the asynchronous Tx/Rx queue count, so for a setup where
vsi->num_txq > vsi->num_rxq, ice_vsi_stop_tx_rings and ice_vsi_cfg_txqs
will be accessing the vsi->xdp_rings out of the bounds.

Parametrize two mentioned functions so they get the size of Tx resources
array as the input.

Fixes: ae15e0ba1b33 ("ice: Change number of XDP Tx queues to match number of Rx queues")
Signed-off-by: Maciej Fijalkowski <maciej.fijalkowski@intel.com>
---
 drivers/net/ethernet/intel/ice/ice_lib.c | 18 ++++++++++--------
 1 file changed, 10 insertions(+), 8 deletions(-)

diff --git a/drivers/net/ethernet/intel/ice/ice_lib.c b/drivers/net/ethernet/intel/ice/ice_lib.c
index 4f8b29d6dcfd..3c8668d8b964 100644
--- a/drivers/net/ethernet/intel/ice/ice_lib.c
+++ b/drivers/net/ethernet/intel/ice/ice_lib.c
@@ -1705,12 +1705,13 @@ int ice_vsi_cfg_rxqs(struct ice_vsi *vsi)
  * ice_vsi_cfg_txqs - Configure the VSI for Tx
  * @vsi: the VSI being configured
  * @rings: Tx ring array to be configured
+ * @count: number of Tx ring array elements
  *
  * Return 0 on success and a negative value on error
  * Configure the Tx VSI for operation.
  */
 static int
-ice_vsi_cfg_txqs(struct ice_vsi *vsi, struct ice_ring **rings)
+ice_vsi_cfg_txqs(struct ice_vsi *vsi, struct ice_ring **rings, u16 count)
 {
 	struct ice_aqc_add_tx_qgrp *qg_buf;
 	u16 q_idx = 0;
@@ -1722,7 +1723,7 @@ ice_vsi_cfg_txqs(struct ice_vsi *vsi, struct ice_ring **rings)
 
 	qg_buf->num_txqs = 1;
 
-	for (q_idx = 0; q_idx < vsi->num_txq; q_idx++) {
+	for (q_idx = 0; q_idx < count; q_idx++) {
 		err = ice_vsi_cfg_txq(vsi, rings[q_idx], qg_buf);
 		if (err)
 			goto err_cfg_txqs;
@@ -1742,7 +1743,7 @@ ice_vsi_cfg_txqs(struct ice_vsi *vsi, struct ice_ring **rings)
  */
 int ice_vsi_cfg_lan_txqs(struct ice_vsi *vsi)
 {
-	return ice_vsi_cfg_txqs(vsi, vsi->tx_rings);
+	return ice_vsi_cfg_txqs(vsi, vsi->tx_rings, vsi->num_txq);
 }
 
 /**
@@ -1757,7 +1758,7 @@ int ice_vsi_cfg_xdp_txqs(struct ice_vsi *vsi)
 	int ret;
 	int i;
 
-	ret = ice_vsi_cfg_txqs(vsi, vsi->xdp_rings);
+	ret = ice_vsi_cfg_txqs(vsi, vsi->xdp_rings, vsi->num_xdp_txq);
 	if (ret)
 		return ret;
 
@@ -1997,17 +1998,18 @@ int ice_vsi_stop_all_rx_rings(struct ice_vsi *vsi)
  * @rst_src: reset source
  * @rel_vmvf_num: Relative ID of VF/VM
  * @rings: Tx ring array to be stopped
+ * @count: number of Tx ring array elements
  */
 static int
 ice_vsi_stop_tx_rings(struct ice_vsi *vsi, enum ice_disq_rst_src rst_src,
-		      u16 rel_vmvf_num, struct ice_ring **rings)
+		      u16 rel_vmvf_num, struct ice_ring **rings, u16 count)
 {
 	u16 q_idx;
 
 	if (vsi->num_txq > ICE_LAN_TXQ_MAX_QDIS)
 		return -EINVAL;
 
-	for (q_idx = 0; q_idx < vsi->num_txq; q_idx++) {
+	for (q_idx = 0; q_idx < count; q_idx++) {
 		struct ice_txq_meta txq_meta = { };
 		int status;
 
@@ -2035,7 +2037,7 @@ int
 ice_vsi_stop_lan_tx_rings(struct ice_vsi *vsi, enum ice_disq_rst_src rst_src,
 			  u16 rel_vmvf_num)
 {
-	return ice_vsi_stop_tx_rings(vsi, rst_src, rel_vmvf_num, vsi->tx_rings);
+	return ice_vsi_stop_tx_rings(vsi, rst_src, rel_vmvf_num, vsi->tx_rings, vsi->num_txq);
 }
 
 /**
@@ -2044,7 +2046,7 @@ ice_vsi_stop_lan_tx_rings(struct ice_vsi *vsi, enum ice_disq_rst_src rst_src,
  */
 int ice_vsi_stop_xdp_tx_rings(struct ice_vsi *vsi)
 {
-	return ice_vsi_stop_tx_rings(vsi, ICE_NO_RESET, 0, vsi->xdp_rings);
+	return ice_vsi_stop_tx_rings(vsi, ICE_NO_RESET, 0, vsi->xdp_rings, vsi->num_xdp_txq);
 }
 
 /**
-- 
2.20.1

RE: [Intel-wired-lan] [PATCH v2 intel-net 1/2] ice: add ndo_bpf callback for safe mode netdev ops

[Bhandare, KiranX]

> -----Original Message-----
> From: Intel-wired-lan <intel-wired-lan-bounces@osuosl.org> On Behalf Of
> Maciej Fijalkowski
> Sent: Thursday, May 20, 2021 12:05 PM
> To: intel-wired-lan@lists.osuosl.org
> Cc: netdev@vger.kernel.org; Jamal Hadi Salim <jhs@mojatatu.com>;
> bjorn@kernel.org; kuba@kernel.org; bpf@vger.kernel.org;
> davem@davemloft.net; Karlsson, Magnus <magnus.karlsson@intel.com>
> Subject: [Intel-wired-lan] [PATCH v2 intel-net 1/2] ice: add ndo_bpf callback
> for safe mode netdev ops
> 
> ice driver requires a programmable pipeline firmware package in order to
> have a support for advanced features. Otherwise, driver falls back to so
> called 'safe mode'. For that mode, ndo_bpf callback is not exposed and when
> user tries to load XDP program, the following happens:
> 
> $ sudo ./xdp1 enp179s0f1
> libbpf: Kernel error message: Underlying driver does not support XDP in
> native mode link set xdp fd failed
> 
> which is sort of confusing, as there is a native XDP support, but not in the
> current mode. Improve the user experience by providing the specific
> ndo_bpf callback dedicated for safe mode which will make use of extack to
> explicitly let the user know that the DDP package is missing and that's the
> reason that the XDP can't be loaded onto interface currently.
> 
> Cc: Jamal Hadi Salim <jhs@mojatatu.com>
> Fixes: efc2214b6047 ("ice: Add support for XDP")
> Signed-off-by: Maciej Fijalkowski <maciej.fijalkowski@intel.com>
> ---
>  drivers/net/ethernet/intel/ice/ice_main.c | 15 +++++++++++++++
>  1 file changed, 15 insertions(+)
> 

Tested-by: Kiran Bhandare <kiranx.bhandare@intel.com>  A Contingent Worker at Intel

RE: [Intel-wired-lan] [PATCH v2 intel-net 2/2] ice: parametrize functions responsible for Tx ring management

[Bhandare, KiranX]

> -----Original Message-----
> From: Intel-wired-lan <intel-wired-lan-bounces@osuosl.org> On Behalf Of
> Maciej Fijalkowski
> Sent: Thursday, May 20, 2021 12:05 PM
> To: intel-wired-lan@lists.osuosl.org
> Cc: netdev@vger.kernel.org; bjorn@kernel.org; kuba@kernel.org;
> bpf@vger.kernel.org; davem@davemloft.net; Karlsson, Magnus
> <magnus.karlsson@intel.com>
> Subject: [Intel-wired-lan] [PATCH v2 intel-net 2/2] ice: parametrize functions
> responsible for Tx ring management
> 
> Commit ae15e0ba1b33 ("ice: Change number of XDP Tx queues to match
> number of Rx queues") tried to address the incorrect setting of XDP queue
> count that was based on the Tx queue count, whereas in theory we should
> provide the XDP queue per Rx queue. However, the routines that setup and
> destroy the set of Tx resources are still based on the
> vsi->num_txq.
> 
> Ice supports the asynchronous Tx/Rx queue count, so for a setup where
> vsi->num_txq > vsi->num_rxq, ice_vsi_stop_tx_rings and ice_vsi_cfg_txqs
> will be accessing the vsi->xdp_rings out of the bounds.
> 
> Parametrize two mentioned functions so they get the size of Tx resources
> array as the input.
> 
> Fixes: ae15e0ba1b33 ("ice: Change number of XDP Tx queues to match
> number of Rx queues")
> Signed-off-by: Maciej Fijalkowski <maciej.fijalkowski@intel.com>
> ---
>  drivers/net/ethernet/intel/ice/ice_lib.c | 18 ++++++++++--------
>  1 file changed, 10 insertions(+), 8 deletions(-)
> 

Tested-by: Kiran Bhandare <kiranx.bhandare@intel.com>  A Contingent Worker at Intel