AF_XDP ZC selftests
Subsystems (8)
Maciej Fijalkowski
maintainerv2 cover AF_XDP ZC selftests
v1->v2:
* collect acks from Magnus
* drop redundant 'ret' variable in patch 4 (Magnus)
* drop redundant assignments to ifobject->xdp_flags in patch 10 (Magnus)
* use NETIF_F_LOOPBACK instead of introducing priv-flag (Alexandr)
Hi!
This set makes it possible to test ice driver with test suite that xdpxceiver provides. In order to do it, device needs to be turned on to loopback mode, so in first patch NETIF_F_LOOPBACK support is added to ice_set_features(). Furthermore, ethtool's loopback test is fixed for ice (patch 2 and 3). Besides turning device into loopback mode, I was limiting queue count to 1, so that none of the steering filters are needed.
Rest of the work is related to xdpxceiver. Currently it is assumed that underlying device supports native XDP. It is true for veth, but might not be for other device that might be used with xdpxceiver once this patch set land. So, patch 4 adds a logic to find out if underlying device supports XDP so that TEST_MODE_DRV can be used for test suite. Patch 5 restores close() on netns fd that was removed by accident.
In patch 6, default Rx pkt stream is added so physical device testing will be able to use shared UMEM in a way that Tx will utilize first half of buffer space and Rx second one. Then, patch 7 adds support for running xdpxceiver on physical devices.
Patch 8 and 9 prepares xdpxceiver to address ZC drivers that clean lazily Tx descriptors (and therefore produce XSK descs to CQ) so that pacing algorithm works fine.
Patch 10 finally adds new TEST_MODE_ZC for testing zero copy AF_XDP driver support.
This work already allowed us to spot and fix two bugs in AF_XDP kernel side ([0], [1]).
v1 is here [2].
[0]: https://lore.kernel.org/bpf/[email protected]/
[1]: https://lore.kernel.org/bpf/[email protected]/
[2]: https://lore.kernel.org/bpf/[email protected]/
Thank you.
Maciej Fijalkowski (10):
ice: allow toggling loopback mode via ndo_set_features callback
ice: check DD bit on Rx descriptor rather than (EOP | RS)
ice: do not setup vlan for loopback VSI
selftests: xsk: query for native XDP support
selftests: xsk: add missing close() on netns fd
selftests: xsk: introduce default Rx pkt stream
selftests: xsk: add support for executing tests on physical device
selftests: xsk: rely on pkts_in_flight in wait_for_tx_completion()
selftests: xsk: remove struct xsk_socket_info::outstanding_tx
selftests: xsk: add support for zero copy testing
drivers/net/ethernet/intel/ice/ice_ethtool.c | 2 +- drivers/net/ethernet/intel/ice/ice_main.c | 62 +-- tools/testing/selftests/bpf/test_xsk.sh | 52 ++- tools/testing/selftests/bpf/xdpxceiver.c | 400 +++++++++++++------ tools/testing/selftests/bpf/xdpxceiver.h | 10 +- 5 files changed, 364 insertions(+), 162 deletions(-)
Signature · 1 line
2.27.0
Maciej Fijalkowski
maintainerPATCH v2 1/10 ice: allow toggling loopback mode via ndo_set_features callback
Add support for NETIF_F_LOOPBACK. Also, simplify checks throughout whole ice_set_features().
--- drivers/net/ethernet/intel/ice/ice_main.c | 54 ++++++++++++++--------- 1 file changed, 34 insertions(+), 20 deletions(-)
drivers/net/ethernet/intel/ice/ice_main.c +34 -20
@@ -3358,6 +3358,7 @@ static void ice_set_netdev_features(struct net_device *netdev) netdev->features |= netdev->hw_features; netdev->hw_features |=NETIF_F_HW_TC ; + netdev->hw_features |=NETIF_F_LOOPBACK ;/* encap and VLAN devices inherit default, csumo and tso features */ netdev->hw_enc_features |= dflt_features | csumo_features |
@@ -5902,6 +5903,18 @@ ice_set_vlan_features(struct net_device *netdev, netdev_features_t features)return 0 ; } +static void ice_set_loopback (struct ice_pf *pf,struct net_device *netdev,bool ena) +{ +bool if_running =netif_running (netdev); + +if (if_running) +ice_stop (netdev); +if (ice_aq_set_mac_loopback (&pf->hw, ena,NULL )) +dev_err (ice_pf_to_dev (pf),"Failed to toggle loopback state\n" ); +if (if_running) +ice_open (netdev); +} +/** * ice_set_features - set the netdev feature flags * @netdev: ptr to the netdev being adjusted
@@ -5910,6 +5923,7 @@ ice_set_vlan_features(struct net_device *netdev, netdev_features_t features)static int ice_set_features (struct net_device *netdev,netdev_features_t features) { +netdev_features_t changed = netdev->features ^ features;struct ice_netdev_priv *np =netdev_priv (netdev);struct ice_vsi *vsi = np->vsi;struct ice_pf *pf = vsi->back;
@@ -5917,37 +5931,33 @@ ice_set_features(struct net_device *netdev, netdev_features_t features)/* Don't set any netdev advanced features with device in Safe Mode */ if (ice_is_safe_mode (vsi->back)) { -dev_err (ice_pf_to_dev (vsi->back),"Device is in Safe Mode - not enabling advanced netdev features\n" ); +dev_err (ice_pf_to_dev (vsi->back), +"Device is in Safe Mode - not enabling advanced netdev features\n" );return ret; }/* Do not change setting during reset */ if (ice_is_reset_in_progress (pf->state)) { -dev_err (ice_pf_to_dev (vsi->back),"Device is resetting, changing advanced netdev features temporarily unavailable.\n" ); +dev_err (ice_pf_to_dev (vsi->back), +"Device is resetting, changing advanced netdev features temporarily unavailable.\n" );return -EBUSY ; }/* Multiple features can be changed in one call so keep features in * separate if/else statements to guarantee each feature is checked */ -if (features &NETIF_F_RXHASH && !(netdev->features &NETIF_F_RXHASH )) -ice_vsi_manage_rss_lut (vsi, true); -else if (!(features &NETIF_F_RXHASH ) && - netdev->features &NETIF_F_RXHASH ) -ice_vsi_manage_rss_lut (vsi, false); +if (changed &NETIF_F_RXHASH ) +ice_vsi_manage_rss_lut (vsi, !!(features &NETIF_F_RXHASH )); ret =ice_set_vlan_features (netdev, features);if (ret)return ret; -if ((features &NETIF_F_NTUPLE ) && - !(netdev->features &NETIF_F_NTUPLE )) { -ice_vsi_manage_fdir (vsi, true); -ice_init_arfs (vsi); - }else if (!(features &NETIF_F_NTUPLE ) && - (netdev->features &NETIF_F_NTUPLE )) { -ice_vsi_manage_fdir (vsi, false); -ice_clear_arfs (vsi); +if (changed &NETIF_F_NTUPLE ) { +bool ena = !!(features &NETIF_F_NTUPLE ); + +ice_vsi_manage_fdir (vsi, ena); + ena ?ice_init_arfs (vsi) :ice_clear_arfs (vsi); }/* don't turn off hw_tc_offload when ADQ is already enabled */
@@ -5956,11 +5966,15 @@ ice_set_features(struct net_device *netdev, netdev_features_t features)return -EACCES ; } -if ((features &NETIF_F_HW_TC ) && - !(netdev->features &NETIF_F_HW_TC )) -set_bit (ICE_FLAG_CLS_FLOWER , pf->flags); -else -clear_bit (ICE_FLAG_CLS_FLOWER , pf->flags); +if (changed &NETIF_F_HW_TC ) { +bool ena = !!(features &NETIF_F_HW_TC ); + + ena ?set_bit (ICE_FLAG_CLS_FLOWER , pf->flags) : +clear_bit (ICE_FLAG_CLS_FLOWER , pf->flags); + } + +if (changed &NETIF_F_LOOPBACK ) +ice_set_loopback (pf, netdev, !!(features &NETIF_F_LOOPBACK ));return 0 ; }
Signature · 1 line
2.27.0
PATCH v2 1/10 ice: allow toggling loopback mode via ndo_set_features callback
From: Maciej Fijalkowski <[email protected]>
Date: Tue, 14 Jun 2022 19:47:40 +0200
Quoted 22 lines
Add support for
NETIF_F_LOOPBACK. Also, simplify checks throughout wholeice_set_features().CC Alexandr Lobakin <[email protected]>Signed-off-by Maciej Fijalkowski <[email protected]>--- drivers/net/ethernet/intel/ice/ice_main.c | 54 ++++++++++++++--------- 1 file changed, 34 insertions(+), 20 deletions(-)drivers/net/ethernet/intel/ice/ice_main.c +1
-0@@ -3358,6 +3358,7 @@ static void ice_set_netdev_features(struct net_device *netdev) netdev->features |= netdev->hw_features; netdev->hw_features |=NETIF_F_HW_TC ; + netdev->hw_features |=NETIF_F_LOOPBACK ;/* encap and VLAN devices inherit default, csumo and tso features */ netdev->hw_enc_features |= dflt_features | csumo_features |
drivers/net/ethernet/intel/ice/ice_main.c ice_set_vlan_features(struct net_device *netdev, netdev_features_t features) · line 5906
Quoted 4 lines
drivers/net/ethernet/intel/ice/ice_main.c +1
-0@@ -5902,6 +5903,18 @@ ice_set_vlan_features(struct net_device *netdev, netdev_features_t features)return 0 ; } +static void ice_set_loopback (struct ice_pf *pf,struct net_device *netdev,bool ena)
I feel like the first argument is redundant in here since we can do
const struct ice_netdev_priv *np = netdev_priv(netdev); struct ice_pf *pf = np->vsi->back;
But at the same time one function argument can be cheaper than two jumps, so up to you.
drivers/net/ethernet/intel/ice/ice_main.c ice_set_vlan_features(struct net_device *netdev, netdev_features_t features) · line 5913
Quoted 7 lines
drivers/net/ethernet/intel/ice/ice_main.c +7
-0+{ +bool if_running =netif_running (netdev); + +if (if_running) +ice_stop (netdev); +if (ice_aq_set_mac_loopback (&pf->hw, ena,NULL )) +dev_err (ice_pf_to_dev (pf),"Failed to toggle loopback state\n" );
netdev_err() instead probably? I guess dev_err() is used for consistency with the rest of ice_set_features(), but I'd recommend using the former anyways.
Quoted 16 lines
drivers/net/ethernet/intel/ice/ice_main.c +5
-0+if (if_running) +ice_open (netdev); +} +/** * ice_set_features - set the netdev feature flags * @netdev: ptr to the netdev being adjusted @@ -5910,6 +5923,7 @@ ice_set_vlan_features(struct net_device *netdev, netdev_features_t features)static int ice_set_features (struct net_device *netdev,netdev_features_t features) { +netdev_features_t changed = netdev->features ^ features;struct ice_netdev_priv *np =netdev_priv (netdev);struct ice_vsi *vsi = np->vsi;struct ice_pf *pf = vsi->back;
drivers/net/ethernet/intel/ice/ice_main.c ice_set_features(struct net_device *netdev, netdev_features_t features) · line 5935
Quoted 6 lines
drivers/net/ethernet/intel/ice/ice_main.c +2
-1@@ -5917,37 +5931,33 @@ ice_set_features(struct net_device *netdev, netdev_features_t features)/* Don't set any netdev advanced features with device in Safe Mode */ if (ice_is_safe_mode (vsi->back)) { -dev_err (ice_pf_to_dev (vsi->back),"Device is in Safe Mode - not enabling advanced netdev features\n" ); +dev_err (ice_pf_to_dev (vsi->back), +"Device is in Safe Mode - not enabling advanced netdev features\n" );
This (I mean the whole ice_set_features() cleanup) deserves to be in a separate patch to me. In can be a past of this series for sure.
Quoted 65 lines
drivers/net/ethernet/intel/ice/ice_main.c +18
-19return ret; }/* Do not change setting during reset */ if (ice_is_reset_in_progress (pf->state)) { -dev_err (ice_pf_to_dev (vsi->back),"Device is resetting, changing advanced netdev features temporarily unavailable.\n" ); +dev_err (ice_pf_to_dev (vsi->back), +"Device is resetting, changing advanced netdev features temporarily unavailable.\n" );return -EBUSY ; }/* Multiple features can be changed in one call so keep features in * separate if/else statements to guarantee each feature is checked */ -if (features &NETIF_F_RXHASH && !(netdev->features &NETIF_F_RXHASH )) -ice_vsi_manage_rss_lut (vsi, true); -else if (!(features &NETIF_F_RXHASH ) && - netdev->features &NETIF_F_RXHASH ) -ice_vsi_manage_rss_lut (vsi, false); +if (changed &NETIF_F_RXHASH ) +ice_vsi_manage_rss_lut (vsi, !!(features &NETIF_F_RXHASH )); ret =ice_set_vlan_features (netdev, features);if (ret)return ret; -if ((features &NETIF_F_NTUPLE ) && - !(netdev->features &NETIF_F_NTUPLE )) { -ice_vsi_manage_fdir (vsi, true); -ice_init_arfs (vsi); - }else if (!(features &NETIF_F_NTUPLE ) && - (netdev->features &NETIF_F_NTUPLE )) { -ice_vsi_manage_fdir (vsi, false); -ice_clear_arfs (vsi); +if (changed &NETIF_F_NTUPLE ) { +bool ena = !!(features &NETIF_F_NTUPLE ); + +ice_vsi_manage_fdir (vsi, ena); + ena ?ice_init_arfs (vsi) :ice_clear_arfs (vsi); }/* don't turn off hw_tc_offload when ADQ is already enabled */ @@ -5956,11 +5966,15 @@ ice_set_features(struct net_device *netdev, netdev_features_t features)return -EACCES ; } -if ((features &NETIF_F_HW_TC ) && - !(netdev->features &NETIF_F_HW_TC )) -set_bit (ICE_FLAG_CLS_FLOWER , pf->flags); -else -clear_bit (ICE_FLAG_CLS_FLOWER , pf->flags); +if (changed &NETIF_F_HW_TC ) { +bool ena = !!(features &NETIF_F_HW_TC ); + + ena ?set_bit (ICE_FLAG_CLS_FLOWER , pf->flags) : +clear_bit (ICE_FLAG_CLS_FLOWER , pf->flags); + } + +if (changed &NETIF_F_LOOPBACK ) +ice_set_loopback (pf, netdev, !!(features &NETIF_F_LOOPBACK ));return 0 ; }Signature · 1 line
2.27.0
The rest is good, I like wiring up standard interfaces with the existing hardware functionality :) Loopback mode is useful for testing stuff.
Thanks!
Olek
Maciej Fijalkowski
maintainerRe: [PATCH v2 bpf-next 01/10] ice: allow toggling loopback mode via ndo_set_features callback
Quoted 38 lines · Alexander Lobakin
From: Maciej Fijalkowski <[email protected]>
Date: Tue, 14 Jun 2022 19:47:40 +0200Quoted 22 lines
Add support for
NETIF_F_LOOPBACK. Also, simplify checks throughout wholeice_set_features().CC Alexandr Lobakin <[email protected]>Signed-off-by Maciej Fijalkowski <[email protected]>--- drivers/net/ethernet/intel/ice/ice_main.c | 54 ++++++++++++++--------- 1 file changed, 34 insertions(+), 20 deletions(-)drivers/net/ethernet/intel/ice/ice_main.c +1
-0@@ -3358,6 +3358,7 @@ static void ice_set_netdev_features(struct net_device *netdev) netdev->features |= netdev->hw_features; netdev->hw_features |=NETIF_F_HW_TC ; + netdev->hw_features |=NETIF_F_LOOPBACK ;/* encap and VLAN devices inherit default, csumo and tso features */ netdev->hw_enc_features |= dflt_features | csumo_features |drivers/net/ethernet/intel/ice/ice_main.c ice_set_vlan_features(struct net_device *netdev, netdev_features_t features) · line 5906
Quoted 4 lines
drivers/net/ethernet/intel/ice/ice_main.c +1
-0@@ -5902,6 +5903,18 @@ ice_set_vlan_features(struct net_device *netdev, netdev_features_t features)return 0 ; } +static void ice_set_loopback (struct ice_pf *pf,struct net_device *netdev,bool ena)I feel like the first argument is redundant in here since we can do
const struct ice_netdev_priv *np = netdev_priv(netdev); struct ice_pf *pf = np->vsi->back;But at the same time one function argument can be cheaper than two jumps, so up to you.
Quoted 11 lines
drivers/net/ethernet/intel/ice/ice_main.c ice_set_vlan_features(struct net_device *netdev, netdev_features_t features) · line 5913
Quoted 7 lines
drivers/net/ethernet/intel/ice/ice_main.c +7
-0+{ +bool if_running =netif_running (netdev); + +if (if_running) +ice_stop (netdev); +if (ice_aq_set_mac_loopback (&pf->hw, ena,NULL )) +dev_err (ice_pf_to_dev (pf),"Failed to toggle loopback state\n" );
netdev_err()instead probably? I guessdev_err()is used for consistency with the rest ofice_set_features(), but I'd recommend using the former anyways.
So let's use netdev_err and drop the pf from arguments that are passed
Quoted 18 lines
Quoted 18 lines
drivers/net/ethernet/intel/ice/ice_main.c +5
-0+if (if_running) +ice_open (netdev); +} +/** * ice_set_features - set the netdev feature flags * @netdev: ptr to the netdev being adjusted @@ -5910,6 +5923,7 @@ ice_set_vlan_features(struct net_device *netdev, netdev_features_t features)static int ice_set_features (struct net_device *netdev,netdev_features_t features) { +netdev_features_t changed = netdev->features ^ features;struct ice_netdev_priv *np =netdev_priv (netdev);struct ice_vsi *vsi = np->vsi;struct ice_pf *pf = vsi->back;@@ -5917,37 +5931,33 @@ ice_set_features(struct net_device *netdev, netdev_features_t features)
Quoted 8 lines
drivers/net/ethernet/intel/ice/ice_main.c ice_set_features(struct net_device *netdev, netdev_features_t features) · line 5935
Quoted 5 lines
drivers/net/ethernet/intel/ice/ice_main.c +2
-1/* Don't set any netdev advanced features with device in Safe Mode */ if (ice_is_safe_mode (vsi->back)) { -dev_err (ice_pf_to_dev (vsi->back),"Device is in Safe Mode - not enabling advanced netdev features\n" ); +dev_err (ice_pf_to_dev (vsi->back), +"Device is in Safe Mode - not enabling advanced netdev features\n" );This (I mean the whole
ice_set_features()cleanup) deserves to be in a separate patch to me. In can be a past of this series for sure.
ack, will pull out this to a standalone patch followed by loopback enablement
Quoted 62 lines
Quoted 62 lines
drivers/net/ethernet/intel/ice/ice_main.c +18
-19return ret; }/* Do not change setting during reset */ if (ice_is_reset_in_progress (pf->state)) { -dev_err (ice_pf_to_dev (vsi->back),"Device is resetting, changing advanced netdev features temporarily unavailable.\n" ); +dev_err (ice_pf_to_dev (vsi->back), +"Device is resetting, changing advanced netdev features temporarily unavailable.\n" );return -EBUSY ; }/* Multiple features can be changed in one call so keep features in * separate if/else statements to guarantee each feature is checked */ -if (features &NETIF_F_RXHASH && !(netdev->features &NETIF_F_RXHASH )) -ice_vsi_manage_rss_lut (vsi, true); -else if (!(features &NETIF_F_RXHASH ) && - netdev->features &NETIF_F_RXHASH ) -ice_vsi_manage_rss_lut (vsi, false); +if (changed &NETIF_F_RXHASH ) +ice_vsi_manage_rss_lut (vsi, !!(features &NETIF_F_RXHASH )); ret =ice_set_vlan_features (netdev, features);if (ret)return ret; -if ((features &NETIF_F_NTUPLE ) && - !(netdev->features &NETIF_F_NTUPLE )) { -ice_vsi_manage_fdir (vsi, true); -ice_init_arfs (vsi); - }else if (!(features &NETIF_F_NTUPLE ) && - (netdev->features &NETIF_F_NTUPLE )) { -ice_vsi_manage_fdir (vsi, false); -ice_clear_arfs (vsi); +if (changed &NETIF_F_NTUPLE ) { +bool ena = !!(features &NETIF_F_NTUPLE ); + +ice_vsi_manage_fdir (vsi, ena); + ena ?ice_init_arfs (vsi) :ice_clear_arfs (vsi); }/* don't turn off hw_tc_offload when ADQ is already enabled */ @@ -5956,11 +5966,15 @@ ice_set_features(struct net_device *netdev, netdev_features_t features)return -EACCES ; } -if ((features &NETIF_F_HW_TC ) && - !(netdev->features &NETIF_F_HW_TC )) -set_bit (ICE_FLAG_CLS_FLOWER , pf->flags); -else -clear_bit (ICE_FLAG_CLS_FLOWER , pf->flags); +if (changed &NETIF_F_HW_TC ) { +bool ena = !!(features &NETIF_F_HW_TC ); + + ena ?set_bit (ICE_FLAG_CLS_FLOWER , pf->flags) : +clear_bit (ICE_FLAG_CLS_FLOWER , pf->flags); + } + +if (changed &NETIF_F_LOOPBACK ) +ice_set_loopback (pf, netdev, !!(features &NETIF_F_LOOPBACK ));
Quoted 8 lines
Quoted 4 lines
drivers/net/ethernet/intel/ice/ice_main.c +0
-0return 0 ; }Signature · 1 line
2.27.0The rest is good, I like wiring up standard interfaces with the existing hardware functionality :) Loopback mode is useful for testing stuff.
thanks for review!
Quoted 3 lines
Thanks!
Olek
Maciej Fijalkowski
maintainerRe: [PATCH v2 bpf-next 01/10] ice: allow toggling loopback mode via ndo_set_features callback
Quoted 51 lines · Maciej Fijalkowski
Quoted 38 lines · Alexander Lobakin
From: Maciej Fijalkowski <[email protected]>
Date: Tue, 14 Jun 2022 19:47:40 +0200Quoted 22 lines
Add support for
NETIF_F_LOOPBACK. Also, simplify checks throughout wholeice_set_features().CC Alexandr Lobakin <[email protected]>Signed-off-by Maciej Fijalkowski <[email protected]>--- drivers/net/ethernet/intel/ice/ice_main.c | 54 ++++++++++++++--------- 1 file changed, 34 insertions(+), 20 deletions(-)drivers/net/ethernet/intel/ice/ice_main.c +1
-0@@ -3358,6 +3358,7 @@ static void ice_set_netdev_features(struct net_device *netdev) netdev->features |= netdev->hw_features; netdev->hw_features |=NETIF_F_HW_TC ; + netdev->hw_features |=NETIF_F_LOOPBACK ;/* encap and VLAN devices inherit default, csumo and tso features */ netdev->hw_enc_features |= dflt_features | csumo_features |drivers/net/ethernet/intel/ice/ice_main.c ice_set_vlan_features(struct net_device *netdev, netdev_features_t features) · line 5906
Quoted 4 lines
drivers/net/ethernet/intel/ice/ice_main.c +1
-0@@ -5902,6 +5903,18 @@ ice_set_vlan_features(struct net_device *netdev, netdev_features_t features)return 0 ; } +static void ice_set_loopback (struct ice_pf *pf,struct net_device *netdev,bool ena)I feel like the first argument is redundant in here since we can do
const struct ice_netdev_priv *np = netdev_priv(netdev); struct ice_pf *pf = np->vsi->back;But at the same time one function argument can be cheaper than two jumps, so up to you.
Quoted 11 lines
drivers/net/ethernet/intel/ice/ice_main.c ice_set_vlan_features(struct net_device *netdev, netdev_features_t features) · line 5913
Quoted 7 lines
drivers/net/ethernet/intel/ice/ice_main.c +7
-0+{ +bool if_running =netif_running (netdev); + +if (if_running) +ice_stop (netdev); +if (ice_aq_set_mac_loopback (&pf->hw, ena,NULL )) +dev_err (ice_pf_to_dev (pf),"Failed to toggle loopback state\n" );
netdev_err()instead probably? I guessdev_err()is used for consistency with the rest ofice_set_features(), but I'd recommend using the former anyways.
Quoted 1 line
So let's use
netdev_errand drop the pf from arguments that are passed
brain fart, we need ice_hw ptr for ice_aq_set_mac_loopback(), so i'll pass this then.
Quoted 6 lines
Quoted 6 lines
Quoted 6 lines
drivers/net/ethernet/intel/ice/ice_main.c +4
-0+if (if_running) +ice_open (netdev); +} +
(...)
Jakub Kicinski
maintainerRe: [PATCH v2 bpf-next 01/10] ice: allow toggling loopback mode via ndo_set_features callback
In reply to Maciej Fijalkowskidrivers/net/ethernet/intel/ice/ice_main.c ice_set_vlan_features(struct net_device *netdev, netdev_features_t features) · line 5915
Quoted 6 lines · On Tue, 14 Jun 2022 19:47:40 +0200 Maciej Fijalkowski wrote:
drivers/net/ethernet/intel/ice/ice_main.c +6
-0+if (if_running) +ice_stop (netdev); +if (ice_aq_set_mac_loopback (&pf->hw, ena,NULL )) +dev_err (ice_pf_to_dev (pf),"Failed to toggle loopback state\n" ); +if (if_running) +ice_open (netdev);
Loopback or not, I don't think we should be accepting the shutdown -> set config -> pray approach in modern drivers. ice_open() seems to be allocating all the Rx memory, and can fail.
Maciej Fijalkowski
maintainerRe: [PATCH v2 bpf-next 01/10] ice: allow toggling loopback mode via ndo_set_features callback
Quoted 11 lines · Jakub Kicinski
drivers/net/ethernet/intel/ice/ice_main.c ice_set_vlan_features(struct net_device *netdev, netdev_features_t features) · line 5915
Quoted 6 lines · On Tue, 14 Jun 2022 19:47:40 +0200 Maciej Fijalkowski wrote:
drivers/net/ethernet/intel/ice/ice_main.c +6
-0+if (if_running) +ice_stop (netdev); +if (ice_aq_set_mac_loopback (&pf->hw, ena,NULL )) +dev_err (ice_pf_to_dev (pf),"Failed to toggle loopback state\n" ); +if (if_running) +ice_open (netdev);Loopback or not, I don't think we should be accepting the shutdown -> set config -> pray approach in modern drivers.
ice_open()seems to be allocating all the Rx memory, and can fail.
They say that those who sing pray twice, so why don't we sing? :)
But seriously, I'll degrade this to ice_down/up and check retvals. I think I just mimicked flow from ice_self_test(), which should be fixed as well...
I'll send v4.
Jakub Kicinski
maintainerRe: [PATCH v2 bpf-next 01/10] ice: allow toggling loopback mode via ndo_set_features callback
Quoted 11 lines · On Thu, 16 Jun 2022 18:00:40 +0200 Maciej Fijalkowski wrote:
Quoted 3 lines
Loopback or not, I don't think we should be accepting the shutdown -> set config -> pray approach in modern drivers.
ice_open()seems to be allocating all the Rx memory, and can fail.They say that those who sing pray twice, so why don't we sing? :)
But seriously, I'll degrade this to ice_down/up and check retvals. I think I just mimicked flow from
ice_self_test(), which should be fixed as well...I'll send v4.
checking retval is not enough, does ice not have the ability to allocate resources upfront? I think iavf was already restructured to follow the "resilient" paradigm, time for ice to follow suit?
This is something DaveM complained about in the first Ethernet driver I upstreamed, which must have been a decade ago by now. It's time we all get on board :)
Maciej Fijalkowski
maintainerRe: [PATCH v2 bpf-next 01/10] ice: allow toggling loopback mode via ndo_set_features callback
Quoted 11 lines · Jakub Kicinski
Quoted 10 lines · On Thu, 16 Jun 2022 18:00:40 +0200 Maciej Fijalkowski wrote:
Quoted 3 lines
Loopback or not, I don't think we should be accepting the shutdown -> set config -> pray approach in modern drivers.
ice_open()seems to be allocating all the Rx memory, and can fail.They say that those who sing pray twice, so why don't we sing? :)
But seriously, I'll degrade this to ice_down/up and check retvals. I think I just mimicked flow from
ice_self_test(), which should be fixed as well...
Quoted 5 lines
Quoted 1 line
I'll send v4.
checking retval is not enough, does ice not have the ability to allocate resources upfront? I think iavf was already restructured to follow the "resilient" paradigm, time for ice to follow suit?
I'm not aware of such restructure TBH. FWIW ice_down() won't free irqs/rings. I said I'll switch to it plus check its retval whereas I feel like you took it like I would want to keep the ice_stop() and check its retval.
Quoted 4 lines
This is something DaveM complained about in the first Ethernet driver I upstreamed, which must have been a decade ago by now. It's time we all get on board :)
Maciej Fijalkowski
maintainerPATCH v2 2/10 ice: check DD bit on Rx descriptor rather than (EOP | RS)
In reply to Maciej FijalkowskiTx side sets EOP and RS bits on descriptors to indicate that a particular descriptor is the last one and needs to generate an irq when it was sent. These bits should not be checked on completion path regardless whether it's the Tx or the Rx. DD bit serves this purpose and it indicates that a particular descriptor is either for Rx or was successfully Txed.
Look at DD bit being set in ice_lbtest_receive_frames() instead of EOP and RS pair.
--- drivers/net/ethernet/intel/ice/ice_ethtool.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/net/ethernet/intel/ice/ice_ethtool.c +1 -1
@@ -658,7 +658,7 @@ static int ice_lbtest_receive_frames(struct ice_rx_ring *rx_ring) rx_desc =ICE_RX_DESC (rx_ring, i);if (!(rx_desc->wb.status_error0 & -cpu_to_le16 (ICE_TX_DESC_CMD_EOP |ICE_TX_DESC_CMD_RS ))) +cpu_to_le16 (BIT (ICE_RX_FLEX_DESC_STATUS0_DD_S ))))continue ; rx_buf = &rx_ring->rx_buf[i];
Signature · 1 line
2.27.0
Maciej Fijalkowski
maintainerPATCH v2 3/10 ice: do not setup vlan for loopback VSI
In reply to Maciej FijalkowskiCurrently loopback test is failiing due to the error returned from ice_vsi_vlan_setup(). Skip calling it when preparing loopback VSI.
--- drivers/net/ethernet/intel/ice/ice_main.c | 8 +++++--- 1 file changed, 5 insertions(+), 3 deletions(-)
drivers/net/ethernet/intel/ice/ice_main.c +5 -3
@@ -6011,10 +6011,12 @@ int ice_vsi_cfg(struct ice_vsi *vsi)if (vsi->netdev) {ice_set_rx_mode (vsi->netdev); - err =ice_vsi_vlan_setup (vsi); +if (vsi->type !=ICE_VSI_LB ) { + err =ice_vsi_vlan_setup (vsi); -if (err) -return err; +if (err) +return err; + } }ice_vsi_cfg_dcb_rings (vsi);
Signature · 1 line
2.27.0
Maciej Fijalkowski
maintainerPATCH v2 4/10 selftests: xsk: query for native XDP support
In reply to Maciej FijalkowskiCurrently, xdpxceiver assumes that underlying device supports XDP in native mode - it is fine by now since tests can run only on a veth pair. Future commit is going to allow running test suite against physical devices, so let us query the device if it is capable of running XDP programs in native mode. This way xdpxceiver will not try to run TEST_MODE_DRV if device being tested is not supporting it.
--- tools/testing/selftests/bpf/xdpxceiver.c | 36 ++++++++++++++++++++++-- 1 file changed, 34 insertions(+), 2 deletions(-)
tools/testing/selftests/bpf/xdpxceiver.c +34 -2
@@ -98,6 +98,8 @@#include <unistd.h> #include <stdatomic.h> #include <bpf/xsk.h> +#include <bpf/bpf.h> +#include <linux/filter.h> #include "xdpxceiver.h" #include "../kselftest.h"
@@ -1605,10 +1607,37 @@ static void ifobject_delete(struct ifobject *ifobj)free (ifobj); } +static bool is_xdp_supported (struct ifobject *ifobject) +{ +int flags =XDP_FLAGS_DRV_MODE ; + + LIBBPF_OPTS(bpf_link_create_opts, opts, .flags = flags); +struct bpf_insn insns[2 ] = { +BPF_MOV64_IMM (BPF_REG_0 ,XDP_PASS ), +BPF_EXIT_INSN () + }; +int ifindex =if_nametoindex (ifobject->ifname); +int prog_fd, insn_cnt =ARRAY_SIZE (insns); +int err; + + prog_fd =bpf_prog_load (BPF_PROG_TYPE_XDP ,NULL ,"GPL" , insns, insn_cnt,NULL ); +if (prog_fd <0 ) +return false; + + err =bpf_xdp_attach (ifindex, prog_fd, flags,NULL ); +if (err) +return false; + +bpf_xdp_detach (ifindex, flags,NULL ); + +return true; +} +int main (int argc,char **argv) {struct pkt_stream *pkt_stream_default;struct ifobject *ifobj_tx, *ifobj_rx; +int modes =TEST_MODE_SKB +1 ;u32 i, j, failed_tests =0 ;struct test_spec test;
@@ -1636,15 +1665,18 @@ int main(int argc, char **argv)init_iface (ifobj_rx,MAC2 ,MAC1 ,IP2 ,IP1 ,UDP_PORT2 ,UDP_PORT1 , worker_testapp_validate_rx); +if (is_xdp_supported (ifobj_tx)) + modes++; +test_spec_init (&test, ifobj_tx, ifobj_rx,0 ); pkt_stream_default =pkt_stream_generate (ifobj_tx->umem,DEFAULT_PKT_CNT ,PKT_SIZE );if (!pkt_stream_default)exit_with_error (ENOMEM ); test.pkt_stream_default = pkt_stream_default; -ksft_set_plan (TEST_MODE_MAX *TEST_TYPE_MAX ); +ksft_set_plan (modes *TEST_TYPE_MAX ); -for (i =0 ; i <TEST_MODE_MAX ; i++) +for (i =0 ; i < modes; i++)for (j =0 ; j <TEST_TYPE_MAX ; j++) {test_spec_init (&test, ifobj_tx, ifobj_rx, i);run_pkt_test (&test, i, j);
Signature · 1 line
2.27.0
Re: [PATCH v2 bpf-next 04/10] selftests: xsk: query for native XDP support
Quoted 7 lines · Maciej Fijalkowski
Currently, xdpxceiver assumes that underlying device supports XDP in native mode - it is fine by now since tests can run only on a veth pair. Future commit is going to allow running test suite against physical devices, so let us query the device if it is capable of running XDP programs in native mode. This way xdpxceiver will not try to run
TEST_MODE_DRVif device being tested is not supporting it.
Quoted 80 lines
Signed-off-by Maciej Fijalkowski <[email protected]>--- tools/testing/selftests/bpf/xdpxceiver.c | 36 ++++++++++++++++++++++-- 1 file changed, 34 insertions(+), 2 deletions(-)tools/testing/selftests/bpf/xdpxceiver.c +34
-3@@ -98,6 +98,8 @@#include <unistd.h> #include <stdatomic.h> #include <bpf/xsk.h> +#include <bpf/bpf.h> +#include <linux/filter.h> #include "xdpxceiver.h" #include "../kselftest.h" @@ -1605,10 +1607,37 @@ static void ifobject_delete(struct ifobject *ifobj)free (ifobj); } +static bool is_xdp_supported (struct ifobject *ifobject) +{ +int flags =XDP_FLAGS_DRV_MODE ; + + LIBBPF_OPTS(bpf_link_create_opts, opts, .flags = flags); +struct bpf_insn insns[2 ] = { +BPF_MOV64_IMM (BPF_REG_0 ,XDP_PASS ), +BPF_EXIT_INSN () + }; +int ifindex =if_nametoindex (ifobject->ifname); +int prog_fd, insn_cnt =ARRAY_SIZE (insns); +int err; + + prog_fd =bpf_prog_load (BPF_PROG_TYPE_XDP ,NULL ,"GPL" , insns, insn_cnt,NULL ); +if (prog_fd <0 ) +return false; + + err =bpf_xdp_attach (ifindex, prog_fd, flags,NULL ); +if (err) +return false; + +bpf_xdp_detach (ifindex, flags,NULL ); + +return true; +} +int main (int argc,char **argv) {struct pkt_stream *pkt_stream_default;struct ifobject *ifobj_tx, *ifobj_rx; +int modes =TEST_MODE_SKB +1 ;u32 i, j, failed_tests =0 ;struct test_spec test;@@ -1636,15 +1665,18 @@ int main(int argc, char **argv)init_iface (ifobj_rx,MAC2 ,MAC1 ,IP2 ,IP1 ,UDP_PORT2 ,UDP_PORT1 , worker_testapp_validate_rx); +if (is_xdp_supported (ifobj_tx)) + modes++; +test_spec_init (&test, ifobj_tx, ifobj_rx,0 ); pkt_stream_default =pkt_stream_generate (ifobj_tx->umem,DEFAULT_PKT_CNT ,PKT_SIZE );if (!pkt_stream_default)exit_with_error (ENOMEM ); test.pkt_stream_default = pkt_stream_default; -ksft_set_plan (TEST_MODE_MAX *TEST_TYPE_MAX ); +ksft_set_plan (modes *TEST_TYPE_MAX ); -for (i =0 ; i <TEST_MODE_MAX ; i++) +for (i =0 ; i < modes; i++)for (j =0 ; j <TEST_TYPE_MAX ; j++) {test_spec_init (&test, ifobj_tx, ifobj_rx, i);run_pkt_test (&test, i, j); --2.27.0
Maciej Fijalkowski
maintainerPATCH v2 5/10 selftests: xsk: add missing close() on netns fd
In reply to Maciej FijalkowskiCommit 1034b03e54ac ("selftests: xsk: Simplify cleanup of ifobjects") removed close on netns fd, which is not correct, so let us restore it.
--- tools/testing/selftests/bpf/xdpxceiver.c | 4 ++++ 1 file changed, 4 insertions(+)
tools/testing/selftests/bpf/xdpxceiver.c +4 -0
@@ -1591,6 +1591,8 @@ static struct ifobject *ifobject_create(void)if (!ifobj->umem) goto out_umem; + ifobj->ns_fd =-1 ; +return ifobj; out_umem:
@@ -1602,6 +1604,8 @@ static struct ifobject *ifobject_create(void)static void ifobject_delete (struct ifobject *ifobj) { +if (ifobj->ns_fd !=-1 ) +close (ifobj->ns_fd);free (ifobj->umem);free (ifobj->xsk_arr);free (ifobj);
Signature · 1 line
2.27.0
Maciej Fijalkowski
maintainerPATCH v2 6/10 selftests: xsk: introduce default Rx pkt stream
In reply to Maciej FijalkowskiIn order to prepare xdpxceiver for physical device testing, let us introduce default Rx pkt stream. Reason for doing it is that physical device testing will use a UMEM with a doubled size where half of it will be used by Tx and other half by Rx. This means that pkt addresses will differ for Tx and Rx streams. Rx thread will initialize the xsk_umem_info::base_addr that is added here so that pkt_set(), when working on Rx UMEM will add this offset and second half of UMEM space will be used. Note that currently base_addr is 0 on both sides. Future commit will do the mentioned initialization.
Previously, veth based testing worked on separate UMEMs, so single default stream was fine.
--- tools/testing/selftests/bpf/xdpxceiver.c | 74 +++++++++++++++--------- tools/testing/selftests/bpf/xdpxceiver.h | 4 +- 2 files changed, 51 insertions(+), 27 deletions(-)
tools/testing/selftests/bpf/xdpxceiver.c +48 -26
@@ -428,15 +428,16 @@ static void __test_spec_init(struct test_spec *test, struct ifobject *ifobj_tx, ifobj->use_poll = false; ifobj->use_fill_ring = true; ifobj->release_rx = true; - ifobj->pkt_stream = test->pkt_stream_default; ifobj->validation_func =NULL ;if (i ==0 ) { ifobj->rx_on = false; ifobj->tx_on = true; + ifobj->pkt_stream = test->tx_pkt_stream_default; }else { ifobj->rx_on = true; ifobj->tx_on = false; + ifobj->pkt_stream = test->rx_pkt_stream_default; }memset (ifobj->umem,0 ,sizeof (*ifobj->umem));
@@ -460,12 +461,15 @@ static void __test_spec_init(struct test_spec *test, struct ifobject *ifobj_tx,static void test_spec_init (struct test_spec *test,struct ifobject *ifobj_tx,struct ifobject *ifobj_rx,enum test_mode mode) { -struct pkt_stream *pkt_stream; +struct pkt_stream *tx_pkt_stream; +struct pkt_stream *rx_pkt_stream;u32 i; - pkt_stream = test->pkt_stream_default; + tx_pkt_stream = test->tx_pkt_stream_default; + rx_pkt_stream = test->rx_pkt_stream_default;memset (test,0 ,sizeof (*test)); - test->pkt_stream_default = pkt_stream; + test->tx_pkt_stream_default = tx_pkt_stream; + test->rx_pkt_stream_default = rx_pkt_stream;for (i =0 ; i <MAX_INTERFACES ; i++) {struct ifobject *ifobj = i ? ifobj_rx : ifobj_tx;
@@ -526,16 +530,17 @@ static void pkt_stream_delete(struct pkt_stream *pkt_stream)static void pkt_stream_restore_default (struct test_spec *test) {struct pkt_stream *tx_pkt_stream = test->ifobj_tx->pkt_stream; +struct pkt_stream *rx_pkt_stream = test->ifobj_rx->pkt_stream; -if (tx_pkt_stream != test->pkt_stream_default) { +if (tx_pkt_stream != test->tx_pkt_stream_default) {pkt_stream_delete (test->ifobj_tx->pkt_stream); - test->ifobj_tx->pkt_stream = test->pkt_stream_default; + test->ifobj_tx->pkt_stream = test->tx_pkt_stream_default; } -if (test->ifobj_rx->pkt_stream != test->pkt_stream_default && - test->ifobj_rx->pkt_stream != tx_pkt_stream) +if (rx_pkt_stream != test->rx_pkt_stream_default) {pkt_stream_delete (test->ifobj_rx->pkt_stream); - test->ifobj_rx->pkt_stream = test->pkt_stream_default; + test->ifobj_rx->pkt_stream = test->rx_pkt_stream_default; + } }static struct pkt_stream *__pkt_stream_alloc (u32 nb_pkts)
@@ -558,7 +563,7 @@ static struct pkt_stream *__pkt_stream_alloc(u32 nb_pkts)static void pkt_set (struct xsk_umem_info *umem,struct pkt *pkt,u64 addr,u32 len) { - pkt->addr = addr; + pkt->addr = addr + umem->base_addr; pkt->len = len;if (len > umem->frame_size -XDP_PACKET_HEADROOM -MIN_PKT_SIZE *2 - umem->frame_headroom) pkt->valid = false;
@@ -597,22 +602,29 @@ static void pkt_stream_replace(struct test_spec *test, u32 nb_pkts, u32 pkt_len) pkt_stream =pkt_stream_generate (test->ifobj_tx->umem, nb_pkts, pkt_len); test->ifobj_tx->pkt_stream = pkt_stream; + pkt_stream =pkt_stream_generate (test->ifobj_rx->umem, nb_pkts, pkt_len); test->ifobj_rx->pkt_stream = pkt_stream; } -static void pkt_stream_replace_half (struct test_spec *test,u32 pkt_len,int offset) +static void __pkt_stream_replace_half (struct ifobject *ifobj,u32 pkt_len, +int offset) { -struct xsk_umem_info *umem = test->ifobj_tx->umem; +struct xsk_umem_info *umem = ifobj->umem;struct pkt_stream *pkt_stream;u32 i; - pkt_stream =pkt_stream_clone (umem, test->pkt_stream_default); -for (i =1 ; i < test->pkt_stream_default->nb_pkts; i +=2 ) + pkt_stream =pkt_stream_clone (umem, ifobj->pkt_stream); +for (i =1 ; i < ifobj->pkt_stream->nb_pkts; i +=2 )pkt_set (umem, &pkt_stream->pkts[i], (i % umem->num_frames) * umem->frame_size + offset, pkt_len); - test->ifobj_tx->pkt_stream = pkt_stream; - test->ifobj_rx->pkt_stream = pkt_stream; + ifobj->pkt_stream = pkt_stream; +} + +static void pkt_stream_replace_half (struct test_spec *test,u32 pkt_len,int offset) +{ +__pkt_stream_replace_half (test->ifobj_tx, pkt_len, offset); +__pkt_stream_replace_half (test->ifobj_rx, pkt_len, offset); }static void pkt_stream_receive_half (struct test_spec *test)
@@ -654,7 +666,8 @@ static struct pkt *pkt_generate(struct ifobject *ifobject, u32 pkt_nb)return pkt; } -static void pkt_stream_generate_custom (struct test_spec *test,struct pkt *pkts,u32 nb_pkts) +static void __pkt_stream_generate_custom (struct ifobject *ifobj, +struct pkt *pkts,u32 nb_pkts) {struct pkt_stream *pkt_stream;u32 i;
@@ -663,15 +676,20 @@ static void pkt_stream_generate_custom(struct test_spec *test, struct pkt *pkts,if (!pkt_stream)exit_with_error (ENOMEM ); - test->ifobj_tx->pkt_stream = pkt_stream; - test->ifobj_rx->pkt_stream = pkt_stream; -for (i =0 ; i < nb_pkts; i++) { - pkt_stream->pkts[i].addr = pkts[i].addr; + pkt_stream->pkts[i].addr = pkts[i].addr + ifobj->umem->base_addr; pkt_stream->pkts[i].len = pkts[i].len; pkt_stream->pkts[i].payload = i; pkt_stream->pkts[i].valid = pkts[i].valid; } + + ifobj->pkt_stream = pkt_stream; +} + +static void pkt_stream_generate_custom (struct test_spec *test,struct pkt *pkts,u32 nb_pkts) +{ +__pkt_stream_generate_custom (test->ifobj_tx, pkts, nb_pkts); +__pkt_stream_generate_custom (test->ifobj_rx, pkts, nb_pkts); }static void pkt_dump (void *pkt,u32 len)
@@ -1639,7 +1657,8 @@ static bool is_xdp_supported(struct ifobject *ifobject)int main (int argc,char **argv) { -struct pkt_stream *pkt_stream_default; +struct pkt_stream *rx_pkt_stream_default; +struct pkt_stream *tx_pkt_stream_default;struct ifobject *ifobj_tx, *ifobj_rx;int modes =TEST_MODE_SKB +1 ;u32 i, j, failed_tests =0 ;
@@ -1673,10 +1692,12 @@ int main(int argc, char **argv) modes++;test_spec_init (&test, ifobj_tx, ifobj_rx,0 ); - pkt_stream_default =pkt_stream_generate (ifobj_tx->umem,DEFAULT_PKT_CNT ,PKT_SIZE ); -if (!pkt_stream_default) + tx_pkt_stream_default =pkt_stream_generate (ifobj_tx->umem,DEFAULT_PKT_CNT ,PKT_SIZE ); + rx_pkt_stream_default =pkt_stream_generate (ifobj_rx->umem,DEFAULT_PKT_CNT ,PKT_SIZE ); +if (!tx_pkt_stream_default || !rx_pkt_stream_default)exit_with_error (ENOMEM ); - test.pkt_stream_default = pkt_stream_default; + test.tx_pkt_stream_default = tx_pkt_stream_default; + test.rx_pkt_stream_default = rx_pkt_stream_default;ksft_set_plan (modes *TEST_TYPE_MAX );
@@ -1690,7 +1711,8 @@ int main(int argc, char **argv) failed_tests++; } -pkt_stream_delete (pkt_stream_default); +pkt_stream_delete (tx_pkt_stream_default); +pkt_stream_delete (rx_pkt_stream_default);ifobject_delete (ifobj_tx);ifobject_delete (ifobj_rx);
tools/testing/selftests/bpf/xdpxceiver.h +3 -1
@@ -95,6 +95,7 @@ struct xsk_umem_info {u32 frame_headroom;void *buffer;u32 frame_size; +u32 base_addr;bool unaligned_mode; };
@@ -155,7 +156,8 @@ struct ifobject {struct test_spec {struct ifobject *ifobj_tx;struct ifobject *ifobj_rx; -struct pkt_stream *pkt_stream_default; +struct pkt_stream *tx_pkt_stream_default; +struct pkt_stream *rx_pkt_stream_default;u16 total_steps;u16 current_step;u16 nb_sockets;
Signature · 1 line
2.27.0
Maciej Fijalkowski
maintainerPATCH v2 7/10 selftests: xsk: add support for executing tests on physical device
In reply to Maciej FijalkowskiCurrently, architecture of xdpxceiver is designed strictly for conducting veth based tests. Veth pair is created together with a network namespace and one of the veth interfaces is moved to the mentioned netns. Then, separate threads for Tx and Rx are spawned which will utilize described setup.
Infrastructure described in the paragraph above can not be used for testing AF_XDP support on physical devices. That testing will be conducted on a single network interface and same queue. Xdpxceiver needs to be extended to distinguish between veth tests and physical interface tests.
Since same iface/queue id pair will be used by both Tx/Rx threads for physical device testing, Tx thread, which happen to run after the Rx thread, is going to create XSK socket with shared umem flag. In order to track this setting throughout the lifetime of spawned threads, introduce 'shared_umem' boolean variable to struct ifobject and set it to true when xdpxceiver is run against physical device. In such case, UMEM size needs to be doubled, so half of it will be used by Rx thread and other half by Tx thread. For two step based test types, value of XSKMAP element under key 0 has to be updated as there is now another socket for the second step. Also, to avoid race conditions when destroying XSK resources, move this activity to the main thread after spawned Rx and Tx threads have finished its job. This way it is possible to gracefully remove shared umem without introducing synchronization mechanisms.
To run xsk selftests suite on physical device, append "-i $IFACE" when invoking test_xsk.sh. For veth based tests, simply skip it. When "-i $IFACE" is in place, under the hood test_xsk.sh will use $IFACE for both interfaces supplied to xdpxceiver, which in turn will interpret that this execution of test suite is for a physical device.
Note that currently this makes it possible only to test SKB and DRV mode (in case underlying device has native XDP support). ZC testing support is added in a later patch.
--- tools/testing/selftests/bpf/test_xsk.sh | 52 +++++-- tools/testing/selftests/bpf/xdpxceiver.c | 189 ++++++++++++++--------- tools/testing/selftests/bpf/xdpxceiver.h | 1 + 3 files changed, 156 insertions(+), 86 deletions(-)
tools/testing/selftests/bpf/test_xsk.sh +36 -16
@@ -73,14 +73,20 @@ # # Run and dump packet contents: # sudo ./test_xsk.sh -D +# +# Run test suite for physical device in loopback mode +# sudo ./test_xsk.sh -i IFACE . xsk_prereqs.sh -while getopts "vD" flag +ETH="" + +while getopts "vDi:" flag do case "${flag}" in v) verbose=1;; D) dump_pkts=1;; + i) ETH=${OPTARG};; esac done
@@ -132,18 +138,25 @@ setup_vethPairs() { ip link set ${VETH0} up } -validate_root_exec -validate_veth_support ${VETH0} -validate_ip_utility -setup_vethPairs - -retval=$? -if [ $retval -ne 0 ]; then - test_status $retval "${TEST_NAME}" - cleanup_exit ${VETH0} ${VETH1} ${NS1} - exit $retval +if [ ! -z $ETH ]; then + VETH0=${ETH} + VETH1=${ETH} + NS1="" +else + validate_root_exec + validate_veth_support ${VETH0} + validate_ip_utility + setup_vethPairs + + retval=$? + if [ $retval -ne 0 ]; then + test_status $retval "${TEST_NAME}" + cleanup_exit ${VETH0} ${VETH1} ${NS1} + exit $retval + fi fi + if [[ $verbose -eq 1 ]]; then ARGS+="-v " fi
@@ -152,26 +165,33 @@ if [[ $dump_pkts -eq 1 ]]; then ARGS="-D " fi +retval=$? test_status $retval "${TEST_NAME}" ## START TESTS statusList=() -TEST_NAME="XSK_SELFTESTS_SOFTIRQ" +TEST_NAME="XSK_SELFTESTS_${VETH0}_SOFTIRQ" execxdpxceiver -cleanup_exit ${VETH0} ${VETH1} ${NS1} -TEST_NAME="XSK_SELFTESTS_BUSY_POLL" +if [ -z $ETH ]; then + cleanup_exit ${VETH0} ${VETH1} ${NS1} +fi +TEST_NAME="XSK_SELFTESTS_${VETH0}_BUSY_POLL" busy_poll=1 -setup_vethPairs +if [ -z $ETH ]; then + setup_vethPairs +fi execxdpxceiver ## END TESTS -cleanup_exit ${VETH0} ${VETH1} ${NS1} +if [ -z $ETH ]; then + cleanup_exit ${VETH0} ${VETH1} ${NS1} +fi failures=0 echo -e "\nSummary:"
tools/testing/selftests/bpf/xdpxceiver.c +119 -70
308 lines not on this page · raw message
tools/testing/selftests/bpf/xdpxceiver.h +1 -0
@@ -149,6 +149,7 @@ struct ifobject {bool busy_poll;bool use_fill_ring;bool release_rx; +bool shared_umem;u8 dst_mac[ETH_ALEN ];u8 src_mac[ETH_ALEN ]; };
Signature · 1 line
2.27.0
Maciej Fijalkowski
maintainerPATCH v2 8/10 selftests: xsk: rely on pkts_in_flight in wait_for_tx_completion()
In reply to Maciej FijalkowskiSome of the drivers that implement support for AF_XDP Zero Copy (like ice) can have lazy approach for cleaning Tx descriptors. For ZC, when descriptor is cleaned, it is placed onto AF_XDP completion queue. This means that current implementation of wait_for_tx_completion() in xdpxceiver can get onto infinite loop, as some of the descriptors can never reach CQ.
This function can be changed to rely on pkts_in_flight instead.
--- tools/testing/selftests/bpf/xdpxceiver.c | 3 ++- tools/testing/selftests/bpf/xdpxceiver.h | 2 +- 2 files changed, 3 insertions(+), 2 deletions(-)
tools/testing/selftests/bpf/xdpxceiver.c +2 -1
@@ -965,7 +965,7 @@ static int __send_pkts(struct ifobject *ifobject, u32 *pkt_nb)static void wait_for_tx_completion (struct xsk_socket_info *xsk) { -while (xsk->outstanding_tx) +while (pkts_in_flight)complete_pkts (xsk,BATCH_SIZE ); }
@@ -1269,6 +1269,7 @@ static void *worker_testapp_validate_rx(void *arg)pthread_mutex_unlock (&pacing_mutex); } + pkts_in_flight =0 ;pthread_exit (NULL ); }
tools/testing/selftests/bpf/xdpxceiver.h +1 -1
@@ -170,6 +170,6 @@ pthread_barrier_t barr;pthread_mutex_t pacing_mutex =PTHREAD_MUTEX_INITIALIZER ;pthread_cond_t pacing_cond =PTHREAD_COND_INITIALIZER ; -int pkts_in_flight; +volatile int pkts_in_flight;#endif /* XDPXCEIVER_H */
Signature · 1 line
2.27.0
Maciej Fijalkowski
maintainerPATCH v2 9/10 selftests: xsk: remove struct xsk_socket_info::outstanding_tx
In reply to Maciej FijalkowskiPrevious change makes xsk->outstanding_tx a dead code, so let's remove it.
--- tools/testing/selftests/bpf/xdpxceiver.c | 20 +++----------------- tools/testing/selftests/bpf/xdpxceiver.h | 1 - 2 files changed, 3 insertions(+), 18 deletions(-)
tools/testing/selftests/bpf/xdpxceiver.c +3 -17
@@ -815,7 +815,7 @@ static void kick_rx(struct xsk_socket_info *xsk)exit_with_error (errno); } -static int complete_pkts (struct xsk_socket_info *xsk,int batch_size) +static void complete_pkts (struct xsk_socket_info *xsk,int batch_size) {unsigned int rcvd;u32 idx;
@@ -824,20 +824,8 @@ static int complete_pkts(struct xsk_socket_info *xsk, int batch_size)kick_tx (xsk); rcvd =xsk_ring_cons__peek (&xsk->umem->cq, batch_size, &idx); -if (rcvd) { -if (rcvd > xsk->outstanding_tx) { -u64 addr = *xsk_ring_cons__comp_addr (&xsk->umem->cq, idx + rcvd -1 ); - -ksft_print_msg ("[%s] Too many packets completed\n" , __func__); -ksft_print_msg ("Last completion address: %llx\n" , addr); -return TEST_FAILURE ; - } - +if (rcvd)xsk_ring_cons__release (&xsk->umem->cq, rcvd); - xsk->outstanding_tx -= rcvd; - } - -return TEST_PASS ; }static int receive_pkts (struct ifobject *ifobj,struct pollfd *fds)
@@ -955,9 +943,7 @@ static int __send_pkts(struct ifobject *ifobject, u32 *pkt_nb)pthread_mutex_unlock (&pacing_mutex);xsk_ring_prod__submit (&xsk->tx, i); - xsk->outstanding_tx += valid_pkts; -if (complete_pkts (xsk, i)) -return TEST_FAILURE ; +complete_pkts (xsk, i);usleep (10 );return TEST_PASS ;
tools/testing/selftests/bpf/xdpxceiver.h +0 -1
@@ -104,7 +104,6 @@ struct xsk_socket_info {struct xsk_ring_prod tx;struct xsk_umem_info *umem;struct xsk_socket *xsk; -u32 outstanding_tx;u32 rxqsize; };
Signature · 1 line
2.27.0
Maciej Fijalkowski
maintainerPATCH v2 10/10 selftests: xsk: add support for zero copy testing
In reply to Maciej FijalkowskiIntroduce new mode to xdpxceiver responsible for testing AF_XDP zero copy support of driver that serves underlying physical device. When setting up test suite, determine whether driver has ZC support or not by trying to bind XSK ZC socket to the interface. If it succeeded, interpret it as ZC support being in place and do softirq and busy poll tests for zero copy mode.
Note that Rx dropped tests are skipped since ZC path is not touching rx_dropped stat at all.
--- tools/testing/selftests/bpf/xdpxceiver.c | 76 ++++++++++++++++++++++-- tools/testing/selftests/bpf/xdpxceiver.h | 2 + 2 files changed, 74 insertions(+), 4 deletions(-)
tools/testing/selftests/bpf/xdpxceiver.c +72 -4
@@ -124,9 +124,20 @@ static void __exit_with_error(int error, const char *file, const char *func, int }#define exit_with_error (error) __exit_with_error(error, __FILE__, __func__, __LINE__) - -#define mode_string (test) (test)->ifobj_tx->xdp_flags & XDP_FLAGS_SKB_MODE ? "SKB" : "DRV"#define busy_poll_string (test) (test)->ifobj_tx->busy_poll ? "BUSY-POLL " : "" +static char *mode_string (struct test_spec *test) +{ +switch (test->mode) { +case TEST_MODE_SKB : +return "SKB" ; +case TEST_MODE_DRV : +return "DRV" ; +case TEST_MODE_ZC : +return "ZC" ; +default : +return "BOGUS" ; + } +}static void report_failure (struct test_spec *test) {
@@ -317,6 +328,51 @@ static int __xsk_configure_socket(struct xsk_socket_info *xsk, struct xsk_umem_ireturn xsk_socket__create (&xsk->xsk, ifobject->ifname,0 , umem->umem, rxr, txr, &cfg); } +static bool ifobj_zc_avail (struct ifobject *ifobject) +{ +size_t umem_sz =DEFAULT_UMEM_BUFFERS *XSK_UMEM__DEFAULT_FRAME_SIZE ; +int mmap_flags =MAP_PRIVATE |MAP_ANONYMOUS |MAP_NORESERVE ; +struct xsk_socket_info *xsk; +struct xsk_umem_info *umem; +bool zc_avail = false; +void *bufs; +int ret; + + bufs =mmap (NULL , umem_sz,PROT_READ |PROT_WRITE , mmap_flags,-1 ,0 ); +if (bufs ==MAP_FAILED ) +exit_with_error (errno); + + umem =calloc (1 ,sizeof (struct xsk_umem_info )); +if (!umem) { +munmap (bufs, umem_sz); +exit_with_error (-ENOMEM ); + } + umem->frame_size =XSK_UMEM__DEFAULT_FRAME_SIZE ; + ret =xsk_configure_umem (umem, bufs, umem_sz); +if (ret) +exit_with_error (-ret); + + xsk =calloc (1 ,sizeof (struct xsk_socket_info )); +if (!xsk) + goto out; + ifobject->xdp_flags =XDP_FLAGS_UPDATE_IF_NOEXIST ; + ifobject->xdp_flags |=XDP_FLAGS_DRV_MODE ; + ifobject->bind_flags =XDP_USE_NEED_WAKEUP |XDP_ZEROCOPY ; + ifobject->rx_on = true; + xsk->rxqsize =XSK_RING_CONS__DEFAULT_NUM_DESCS ; + ret =__xsk_configure_socket (xsk, umem, ifobject, false); +if (!ret) + zc_avail = true; + +xsk_socket__delete (xsk->xsk); +free (xsk); +out: +munmap (umem->buffer, umem_sz); +xsk_umem__delete (umem->umem); +free (umem); +return zc_avail; +} +static struct option long_options[] = { {"interface" , required_argument,0 ,'i' }, {"busy-poll" , no_argument,0 ,'b' },
@@ -483,9 +539,14 @@ static void test_spec_init(struct test_spec *test, struct ifobject *ifobj_tx,else ifobj->xdp_flags |=XDP_FLAGS_DRV_MODE ; - ifobj->bind_flags =XDP_USE_NEED_WAKEUP |XDP_COPY ; + ifobj->bind_flags =XDP_USE_NEED_WAKEUP ; +if (mode ==TEST_MODE_ZC ) + ifobj->bind_flags |=XDP_ZEROCOPY ; +else + ifobj->bind_flags |=XDP_COPY ; } + test->mode = mode;__test_spec_init (test, ifobj_tx, ifobj_rx); }
@@ -1543,6 +1604,10 @@ static void run_pkt_test(struct test_spec *test, enum test_mode mode, enum test_ {switch (type) {case TEST_TYPE_STATS_RX_DROPPED : +if (mode ==TEST_MODE_ZC ) { +ksft_test_result_skip ("Can not run RX_DROPPED test for ZC mode\n" ); +return ; + }testapp_stats_rx_dropped (test);break ;case TEST_TYPE_STATS_TX_INVALID_DESCS :
@@ -1721,8 +1786,11 @@ int main(int argc, char **argv)init_iface (ifobj_rx,MAC2 ,MAC1 ,IP2 ,IP1 ,UDP_PORT2 ,UDP_PORT1 , worker_testapp_validate_rx); -if (is_xdp_supported (ifobj_tx)) +if (is_xdp_supported (ifobj_tx)) { modes++; +if (ifobj_zc_avail (ifobj_tx)) + modes++; + }test_spec_init (&test, ifobj_tx, ifobj_rx,0 ); tx_pkt_stream_default =pkt_stream_generate (ifobj_tx->umem,DEFAULT_PKT_CNT ,PKT_SIZE );
tools/testing/selftests/bpf/xdpxceiver.h +2 -0
@@ -61,6 +61,7 @@enum test_mode {TEST_MODE_SKB ,TEST_MODE_DRV , +TEST_MODE_ZC ,TEST_MODE_MAX };
@@ -162,6 +163,7 @@ struct test_spec {u16 current_step;u16 nb_sockets;bool fail; +enum test_mode mode;char name[MAX_TEST_NAME_SIZE ]; };
Signature · 1 line
2.27.0
Re: [PATCH v2 bpf-next 10/10] selftests: xsk: add support for zero copy testing
Quoted 10 lines · Maciej Fijalkowski
Introduce new mode to xdpxceiver responsible for testing
AF_XDPzero copy support of driver that serves underlying physical device. When setting up test suite, determine whether driver has ZC support or not by trying to bind XSK ZC socket to the interface. If it succeeded, interpret it as ZC support being in place and do softirq and busy poll tests for zero copy mode.Note that Rx dropped tests are skipped since ZC path is not touching
rx_droppedstat at all.
And again, thanks for working on this. This feature will be very useful to increase test coverage.
Quoted 148 lines
Signed-off-by Maciej Fijalkowski <[email protected]>--- tools/testing/selftests/bpf/xdpxceiver.c | 76 ++++++++++++++++++++++-- tools/testing/selftests/bpf/xdpxceiver.h | 2 + 2 files changed, 74 insertions(+), 4 deletions(-)tools/testing/selftests/bpf/xdpxceiver.c +72
-4@@ -124,9 +124,20 @@ static void __exit_with_error(int error, const char *file, const char *func, int }#define exit_with_error (error) __exit_with_error(error, __FILE__, __func__, __LINE__) - -#define mode_string (test) (test)->ifobj_tx->xdp_flags & XDP_FLAGS_SKB_MODE ? "SKB" : "DRV"#define busy_poll_string (test) (test)->ifobj_tx->busy_poll ? "BUSY-POLL " : "" +static char *mode_string (struct test_spec *test) +{ +switch (test->mode) { +case TEST_MODE_SKB : +return "SKB" ; +case TEST_MODE_DRV : +return "DRV" ; +case TEST_MODE_ZC : +return "ZC" ; +default : +return "BOGUS" ; + } +}static void report_failure (struct test_spec *test) {@@ -317,6 +328,51 @@ static int __xsk_configure_socket(struct xsk_socket_info *xsk, struct xsk_umem_ireturn xsk_socket__create (&xsk->xsk, ifobject->ifname,0 , umem->umem, rxr, txr, &cfg); } +static bool ifobj_zc_avail (struct ifobject *ifobject) +{ +size_t umem_sz =DEFAULT_UMEM_BUFFERS *XSK_UMEM__DEFAULT_FRAME_SIZE ; +int mmap_flags =MAP_PRIVATE |MAP_ANONYMOUS |MAP_NORESERVE ; +struct xsk_socket_info *xsk; +struct xsk_umem_info *umem; +bool zc_avail = false; +void *bufs; +int ret; + + bufs =mmap (NULL , umem_sz,PROT_READ |PROT_WRITE , mmap_flags,-1 ,0 ); +if (bufs ==MAP_FAILED ) +exit_with_error (errno); + + umem =calloc (1 ,sizeof (struct xsk_umem_info )); +if (!umem) { +munmap (bufs, umem_sz); +exit_with_error (-ENOMEM ); + } + umem->frame_size =XSK_UMEM__DEFAULT_FRAME_SIZE ; + ret =xsk_configure_umem (umem, bufs, umem_sz); +if (ret) +exit_with_error (-ret); + + xsk =calloc (1 ,sizeof (struct xsk_socket_info )); +if (!xsk) + goto out; + ifobject->xdp_flags =XDP_FLAGS_UPDATE_IF_NOEXIST ; + ifobject->xdp_flags |=XDP_FLAGS_DRV_MODE ; + ifobject->bind_flags =XDP_USE_NEED_WAKEUP |XDP_ZEROCOPY ; + ifobject->rx_on = true; + xsk->rxqsize =XSK_RING_CONS__DEFAULT_NUM_DESCS ; + ret =__xsk_configure_socket (xsk, umem, ifobject, false); +if (!ret) + zc_avail = true; + +xsk_socket__delete (xsk->xsk); +free (xsk); +out: +munmap (umem->buffer, umem_sz); +xsk_umem__delete (umem->umem); +free (umem); +return zc_avail; +} +static struct option long_options[] = { {"interface" , required_argument,0 ,'i' }, {"busy-poll" , no_argument,0 ,'b' },@@ -483,9 +539,14 @@ static void test_spec_init(struct test_spec *test, struct ifobject *ifobj_tx,else ifobj->xdp_flags |=XDP_FLAGS_DRV_MODE ; - ifobj->bind_flags =XDP_USE_NEED_WAKEUP |XDP_COPY ; + ifobj->bind_flags =XDP_USE_NEED_WAKEUP ; +if (mode ==TEST_MODE_ZC ) + ifobj->bind_flags |=XDP_ZEROCOPY ; +else + ifobj->bind_flags |=XDP_COPY ; } + test->mode = mode;__test_spec_init (test, ifobj_tx, ifobj_rx); }@@ -1543,6 +1604,10 @@ static void run_pkt_test(struct test_spec *test, enum test_mode mode, enum test_ {switch (type) {case TEST_TYPE_STATS_RX_DROPPED : +if (mode ==TEST_MODE_ZC ) { +ksft_test_result_skip ("Can not run RX_DROPPED test for ZC mode\n" ); +return ; + }testapp_stats_rx_dropped (test);break ;case TEST_TYPE_STATS_TX_INVALID_DESCS :@@ -1721,8 +1786,11 @@ int main(int argc, char **argv)init_iface (ifobj_rx,MAC2 ,MAC1 ,IP2 ,IP1 ,UDP_PORT2 ,UDP_PORT1 , worker_testapp_validate_rx); -if (is_xdp_supported (ifobj_tx)) +if (is_xdp_supported (ifobj_tx)) { modes++; +if (ifobj_zc_avail (ifobj_tx)) + modes++; + }test_spec_init (&test, ifobj_tx, ifobj_rx,0 ); tx_pkt_stream_default =pkt_stream_generate (ifobj_tx->umem,DEFAULT_PKT_CNT ,PKT_SIZE );tools/testing/selftests/bpf/xdpxceiver.h +2
-1@@ -61,6 +61,7 @@enum test_mode {TEST_MODE_SKB ,TEST_MODE_DRV , +TEST_MODE_ZC ,TEST_MODE_MAX };@@ -162,6 +163,7 @@ struct test_spec {u16 current_step;u16 nb_sockets;bool fail; +enum test_mode mode;char name[MAX_TEST_NAME_SIZE ]; }; --2.27.0