fix a STA PS bug and add PS support to mac80211_hwsim
Subsystems (2)
v2 cover fix a STA PS bug and add PS support to mac80211_hwsim
This patchset tries to make mac80211 power save testable with the hostap hwsim tests, and fixes a bug in STA power save.
Basic tests for AP and STA power save will be submitted to hostap separately.
Thomas Pedersen (3):
mac80211_hwsim: add power save support
mac80211: expose HW conf flags through debugfs
mac80211: consider QoS Null frames for STA_NULLFUNC_ACKED
drivers/net/wireless/mac80211_hwsim.c | 8 ++++++++ net/mac80211/debugfs.c | 3 +++ net/mac80211/status.c | 3 ++- 3 files changed, 13 insertions(+), 1 deletion(-)
Signature · 5 lines
v2: - got Johannes' email wrong - weird duplicate emails 2.20.1
PATCH v2 1/3 mac80211_hwsim: add power save support
Advertise the correct flags to mac80211 to indicate PS trigger frames and frame buffering should be handled by mac80211. This means mac80211_hwsim will now also have to release buffered multicast frames after a (DTIM) beacon.
--- drivers/net/wireless/mac80211_hwsim.c | 8 ++++++++ 1 file changed, 8 insertions(+)
drivers/net/wireless/mac80211_hwsim.c +8 -0
@@ -1595,6 +1595,11 @@ static void mac80211_hwsim_beacon_tx(void *arg, u8 *mac,mac80211_hwsim_tx_frame (hw, skb,rcu_dereference (vif->chanctx_conf)->def.chan); +while ((skb =ieee80211_get_buffered_bc (hw, vif)) !=NULL ) { +mac80211_hwsim_tx_frame (hw, skb, +rcu_dereference (vif->chanctx_conf)->def.chan); + } +if (vif->csa_active &&ieee80211_csa_is_complete (vif))ieee80211_csa_finish (vif); }
@@ -2925,6 +2930,9 @@ static int mac80211_hwsim_new_radio(struct genl_info *info,ieee80211_hw_set (hw,MFP_CAPABLE );ieee80211_hw_set (hw,SIGNAL_DBM );ieee80211_hw_set (hw,SUPPORTS_PS ); +ieee80211_hw_set (hw,REPORTS_TX_ACK_STATUS ); +ieee80211_hw_set (hw,HOST_BROADCAST_PS_BUFFERING ); +ieee80211_hw_set (hw,PS_NULLFUNC_STACK );ieee80211_hw_set (hw,TDLS_WIDER_BW );if (rctbl)ieee80211_hw_set (hw,SUPPORTS_RC_TABLE );
Signature · 1 line
2.20.1
PATCH v2 2/3 mac80211: expose HW conf flags through debugfs
In reply to Thomas PedersenThis is useful during testing to eg. check the currently configured HW power save state.
--- net/mac80211/debugfs.c | 3 +++ 1 file changed, 3 insertions(+)
net/mac80211/debugfs.c +3 -0
@@ -59,6 +59,8 @@ static const struct file_operations name## _ops = { \ debugfs_create_file(#name, mode, phyd, local, &name## _ops); +DEBUGFS_READONLY_FILE (hw_conf,"%x" , + local->hw.conf.flags);DEBUGFS_READONLY_FILE (user_power,"%d" , local->user_power_level); DEBUGFS_READONLY_FILE(power, "%d",
@@ -433,6 +435,7 @@ void debugfs_hw_add(struct ieee80211_local *local)DEBUGFS_ADD (hwflags);DEBUGFS_ADD (user_power);DEBUGFS_ADD (power); +DEBUGFS_ADD (hw_conf);DEBUGFS_ADD_MODE (force_tx_status,0600 );if (local->ops->wake_tx_queue)
Signature · 1 line
2.20.1
PATCH v2 3/3 mac80211: consider QoS Null frames for STA_NULLFUNC_ACKED
In reply to Thomas PedersenCommit 7b6ddeaf27ec ("mac80211: use QoS NDP for AP probing") let STAs send QoS Null frames as PS triggers if the AP was a QoS STA. However, the mac80211 PS stack relies on an interface flag IEEE80211_STA_NULLFUNC_ACKED for determining trigger frame ACK, which was not being set for acked non-QoS Null frames. The effect is an inability to trigger hardware sleep via IEEE80211_CONF_PS since the QoS Null frame was seemingly never acked.
This bug only applies to drivers which set bothIEEE80211_HW_REPORTS_TX_ACK_STATUS andIEEE80211_HW_PS_NULLFUNC_STACK.
Detect the acked QoS Null frame to restore STA power save.
--- net/mac80211/status.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-)
net/mac80211/status.c +2 -1
@@ -1030,7 +1030,8 @@ static void __ieee80211_tx_status(struct ieee80211_hw *hw,I802_DEBUG_INC (local->dot11FailedCount); } -if (ieee80211_is_nullfunc (fc) &&ieee80211_has_pm (fc) && +if ((ieee80211_is_nullfunc (fc) ||ieee80211_is_qos_nullfunc (fc)) && +ieee80211_has_pm (fc) &&ieee80211_hw_check (&local->hw,REPORTS_TX_ACK_STATUS ) && !(info->flags &IEEE80211_TX_CTL_INJECTED ) && local->ps_sdata && !(local->scanning)) {
Signature · 1 line
2.20.1
Johannes Berg
maintainerHi,
Quoted 5 lines · Thomas Pedersen
This patchset tries to make mac80211 power save testable with the hostap hwsim tests, and fixes a bug in STA power save.
Basic tests for AP and STA power save will be submitted to hostap separately.
Cool :)
I was going to apply all of these, but then it turned out that the first one causes a bunch of hwsim tests to fail?
It seems quite possible that this is just a mac80211 bug somewhere, but I'd want to look at this more closely, so I've dropped it back to "Under Review" in patchwork again.
Any thoughts?
johannes
Quoted 6 lines · Johannes Berg
Hi,
Quoted 3 lines · Thomas Pedersen
This patchset tries to make mac80211 power save testable with the hostap hwsim tests, and fixes a bug in STA power save.
Quoted 7 lines
Quoted 2 lines
Basic tests for AP and STA power save will be submitted to hostap separately.
Cool :)
I was going to apply all of these, but then it turned out that the first one causes a bunch of hwsim tests to fail?
OK. If you point me toward which tests started to fail I can take a look.
Quoted 8 lines
It seems quite possible that this is just a mac80211 bug somewhere, but I'd want to look at this more closely, so I've dropped it back to "Under Review" in patchwork again.
Any thoughts?
johannes
Signature · 1 line
-- thomas
Johannes Berg
maintainerQuoted 10 lines · Thomas Pedersen
Quoted 9 lines · Johannes Berg
Hi,
Quoted 5 lines · Thomas Pedersen
This patchset tries to make mac80211 power save testable with the hostap hwsim tests, and fixes a bug in STA power save.
Basic tests for AP and STA power save will be submitted to hostap separately.
Quoted 7 lines
Quoted 4 lines
Cool :)
I was going to apply all of these, but then it turned out that the first one causes a bunch of hwsim tests to fail?
OK. If you point me toward which tests started to fail I can take a look.
Various. Jouni just said mostly concurrency ones, e.g.
p2p_cli_inviteap_cipher_tkip_countermeasures_ap_mixed_modediscovery_while_go_p2p_devradius_macacl_unreachable
but I saw quite a few more failing as well, though not always entirely reproducibly.
johannes
Johannes Berg
maintainerQuoted 7 lines · Johannes Berg
Various. Jouni just said mostly concurrency ones, e.g.
p2p_cli_inviteap_cipher_tkip_countermeasures_ap_mixed_modediscovery_while_go_p2p_devradius_macacl_unreachable
Ah, still had my list open:
discovery_while_go radius_macacl_unreachable discovery_while_go_p2p_dev bgscan_learn_beacon_loss wifi_display_parsing ap_vlan_iface_cleanup_multibss ap_vlan_iface_cleanup_multibss_per_sta_vif autogo_2clibgscan_simple_beacon_loss rrm_lci_req_timeout rrm_ftm_range_req_timeoutp2p_cli_invite
johannes
Hi Johannes,
Quoted 11 lines · Johannes Berg
Quoted 7 lines · Johannes Berg
Various. Jouni just said mostly concurrency ones, e.g.
p2p_cli_inviteap_cipher_tkip_countermeasures_ap_mixed_modediscovery_while_go_p2p_devradius_macacl_unreachableAh, still had my list open:
Quoted 9 lines
discovery_while_goradius_macacl_unreachablediscovery_while_go_p2p_devbgscan_learn_beacon_losswifi_display_parsingap_vlan_iface_cleanup_multibssap_vlan_iface_cleanup_multibss_per_sta_vifautogo_2clibgscan_simple_beacon_lossrrm_lci_req_timeoutrrm_ftm_range_req_timeoutp2p_cli_invite
Is there a list of known passing hwsim tests somewhere?
Maybe a subset of suites / tests you like to see passing as a smoke test?
Or do you run through the 3300+ tests and expect them all to pass?
Thanks,
Signature · 1 line
-- thomas
Johannes Berg
maintainerHi Thomas,
Quoted 3 lines
Is there a list of known passing hwsim tests somewhere?
Maybe a subset of suites / tests you like to see passing as a smoke test?
Not really.
Quoted 1 line
Or do you run through the 3300+ tests and expect them all to pass?
Pretty much, yes. It takes <10 minutes on my quad-core desktop to run using the UML time-travel mode against recent kernels.
I don't always get 100% pass on the first round, but the handful that fails sometimes I expect to pass when running just those manually with vm-run.sh.
johannes
Quoted 8 lines · Johannes Berg
Hi Thomas,
Quoted 3 lines
Is there a list of known passing hwsim tests somewhere?
Maybe a subset of suites / tests you like to see passing as a smoke test?
Not really.
Quoted 8 lines
Quoted 1 line
Or do you run through the 3300+ tests and expect them all to pass?
Pretty much, yes. It takes <10 minutes on my quad-core desktop to run using the UML time-travel mode against recent kernels.
I don't always get 100% pass on the first round, but the handful that fails sometimes I expect to pass when running just those manually with
vm-run.sh.
Thanks, I'll give UML time-travel mode a try.
Signature · 1 line
-- thomas
Johannes Berg
maintainerQuoted 11 lines · Thomas Pedersen
Quoted 10 lines · Johannes Berg
Hi Thomas,
Quoted 3 lines
Is there a list of known passing hwsim tests somewhere?
Maybe a subset of suites / tests you like to see passing as a smoke test?
Not really.
Quoted 1 line
Or do you run through the 3300+ tests and expect them all to pass?
Quoted 8 lines
Quoted 6 lines
Pretty much, yes. It takes <10 minutes on my quad-core desktop to run using the UML time-travel mode against recent kernels.
I don't always get 100% pass on the first round, but the handful that fails sometimes I expect to pass when running just those manually with
vm-run.sh.Thanks, I'll give UML time-travel mode a try.
FWIW, I applied your patches yesterday. There was one issue remaining uncovered by them, but it wasn't actually related to your patches, just related to the test sending too many frames. I sent a workaround to the hostap list and also the "mac80211: use more bits for ack_frame_id" to the kernel to fix that.
johannes