Message ID | 7742335b-f05e-edcd-ad80-fd09fb295eab@codeaurora.org (mailing list archive) |
---|---|
State | Changes Requested |
Delegated to: | Johannes Berg |
Headers | show |
"Lazar, Alexei Avshalom" <ailizaro@codeaurora.org> writes: >>From b739abb6f29dc43a86b8b2b60e893b4441f8aa1f Mon Sep 17 00:00:00 2001 > From: Alexei Avshalom Lazar <ailizaro@codeaurora.org> > Date: Sun, 6 Nov 2016 16:21:20 +0200 > Subject: [PATCH] cfg80211: add set/get link loss profile > > Introduce NL80211_CMD_SET_LINK_LOSS_PROFILE and > NL80211_CMD_GET_LINK_LOSS_PROFILE as it required by the user space > to configure the link loss profile. > The link loss profile represents the priority of maintaining link up > in different link quality environments. > Three types of behavior for link loss defined: > LINK_LOSS_PROFILE_RELAXED: prefer maintaining link up even in poor > link quality environment. > LINK_LOSS_PROFILE_DEFAULT: The default behavior for maintaining link > up vs link quality. > LINK_LOSS_PROFILE_AGGRESSIVE: prefer losing link up in poor link > quality environment. > New cfg80211 API also been added, set/get_link_loss_profile. > > Signed-off-by: Alexei Avshalom Lazar <ailizaro@codeaurora.org> What does the term "link" mean in this context? A connection to the AP or something else? The documentation doesn't give me any detailed information nor any examples and lefts me guessing what this is supposed to do.
Hi Kalle, In this context the term "link" does mean connection to AP. There are scenarios that require to lose connection faster once reduced link quality detected in order maintain better user experience (from both sides - AP and STA). Example case is FST (Fast Session Transfer), where we have a fast 11ad connection but close-range and a backup 11ac connection which is "always on". in such cases if we detect poor link quality in the active band we prefer to disconnect and switch to the backup connection with the good link quality. This will be reached by setting "aggressive" profile that will cause more rapid disconnect in such cases. Alexei Lazar Qualcomm Israel, on behalf of Qualcomm Innovation Center, Inc. The Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum a Linux Foundation Collaborative Project On 11/10/2016 3:55 PM, Kalle Valo wrote: > "Lazar, Alexei Avshalom" <ailizaro@codeaurora.org> writes: > >> >From b739abb6f29dc43a86b8b2b60e893b4441f8aa1f Mon Sep 17 00:00:00 2001 >> From: Alexei Avshalom Lazar <ailizaro@codeaurora.org> >> Date: Sun, 6 Nov 2016 16:21:20 +0200 >> Subject: [PATCH] cfg80211: add set/get link loss profile >> >> Introduce NL80211_CMD_SET_LINK_LOSS_PROFILE and >> NL80211_CMD_GET_LINK_LOSS_PROFILE as it required by the user space >> to configure the link loss profile. >> The link loss profile represents the priority of maintaining link up >> in different link quality environments. >> Three types of behavior for link loss defined: >> LINK_LOSS_PROFILE_RELAXED: prefer maintaining link up even in poor >> link quality environment. >> LINK_LOSS_PROFILE_DEFAULT: The default behavior for maintaining link >> up vs link quality. >> LINK_LOSS_PROFILE_AGGRESSIVE: prefer losing link up in poor link >> quality environment. >> New cfg80211 API also been added, set/get_link_loss_profile. >> >> Signed-off-by: Alexei Avshalom Lazar <ailizaro@codeaurora.org> > What does the term "link" mean in this context? A connection to the AP > or something else? The documentation doesn't give me any detailed > information nor any examples and lefts me guessing what this is supposed > to do. >
Hi Alexei, "Lazar, Alexei Avshalom" <ailizaro@codeaurora.org> writes: > In this context the term "link" does mean connection to AP. There are > scenarios that require to lose connection faster once reduced link > quality detected in order maintain better user experience (from both > sides - AP and STA). Example case is FST (Fast Session Transfer), > where we have a fast 11ad connection but close-range and a backup 11ac > connection which is "always on". in such cases if we detect poor link > quality in the active band we prefer to disconnect and switch to the > backup connection with the good link quality. This will be reached by > setting "aggressive" profile that will cause more rapid disconnect in > such cases. The documentation should clearly document what this command is supposed, currently it only tells me "maintain/lose link" and does not specify in detail what this is supposed to do. Also there's no mention in what modes this can be used. (STA, P2P?) Also please refrain from top posting in public mailing lists, the preferred way is to edit quotes and bottom post. It's much more natural way to have long detailed discussions.
> Introduce NL80211_CMD_SET_LINK_LOSS_PROFILE and > NL80211_CMD_GET_LINK_LOSS_PROFILE as it required by the user space > to configure the link loss profile. I'm not convinced. > The link loss profile represents the priority of maintaining link up > in different link quality environments. > Three types of behavior for link loss defined: > LINK_LOSS_PROFILE_RELAXED: prefer maintaining link up even in poor > link quality environment. > LINK_LOSS_PROFILE_DEFAULT: The default behavior for maintaining link > up vs link quality. > LINK_LOSS_PROFILE_AGGRESSIVE: prefer losing link up in poor link > quality environment. What you're doing here is *extremely* vague. No definitions of what this means, no description of how it's intended to be implemented, no note even on whether or not this is for full-MAC chips or not, etc. At least for mac80211, we can define signals to userspace regarding "link quality" and make decisions in wpa_s then? > --- a/drivers/net/wireless/ath/wil6210/cfg80211.c Those changes obviously don't belong into this patch anyway. johannes
On 11/28/2016 4:52 PM, Johannes Berg wrote: > >> Introduce NL80211_CMD_SET_LINK_LOSS_PROFILE and >> NL80211_CMD_GET_LINK_LOSS_PROFILE as it required by the user space >> to configure the link loss profile. > > I'm not convinced. > >> The link loss profile represents the priority of maintaining link up >> in different link quality environments. >> Three types of behavior for link loss defined: >> LINK_LOSS_PROFILE_RELAXED: prefer maintaining link up even in poor >> link quality environment. >> LINK_LOSS_PROFILE_DEFAULT: The default behavior for maintaining link >> up vs link quality. >> LINK_LOSS_PROFILE_AGGRESSIVE: prefer losing link up in poor link >> quality environment. > > What you're doing here is *extremely* vague. No definitions of what > this means, no description of how it's intended to be implemented, no > note even on whether or not this is for full-MAC chips or not, etc. > The patch is intended for full-MAC chips and I will update the commit message to include more details. > At least for mac80211, we can define signals to userspace regarding > "link quality" and make decisions in wpa_s then? > The policy is offloaded today to the FW and the decision being made by the FW. We are only modifying the behavior and not making the decision. > >> --- a/drivers/net/wireless/ath/wil6210/cfg80211.c > > Those changes obviously don't belong into this patch anyway. > > johannes >
On 11/18/2016 1:05 PM, Kalle Valo wrote: > Hi Alexei, > > "Lazar, Alexei Avshalom" <ailizaro@codeaurora.org> writes: > >> In this context the term "link" does mean connection to AP. There are >> scenarios that require to lose connection faster once reduced link >> quality detected in order maintain better user experience (from both >> sides - AP and STA). Example case is FST (Fast Session Transfer), >> where we have a fast 11ad connection but close-range and a backup 11ac >> connection which is "always on". in such cases if we detect poor link >> quality in the active band we prefer to disconnect and switch to the >> backup connection with the good link quality. This will be reached by >> setting "aggressive" profile that will cause more rapid disconnect in >> such cases. > > The documentation should clearly document what this command is supposed, > currently it only tells me "maintain/lose link" and does not specify in > detail what this is supposed to do. Also there's no mention in what > modes this can be used. (STA, P2P?) > Ok I will send a new patch with updated description. > Also please refrain from top posting in public mailing lists, the > preferred way is to edit quotes and bottom post. It's much more natural > way to have long detailed discussions. > Sorry for that, I will refrain from top posting.
On Mon, 2016-12-19 at 12:58 +0200, Lazar, Alexei Avshalom wrote: > > What you're doing here is *extremely* vague. No definitions of what > > this means, no description of how it's intended to be implemented, > > no note even on whether or not this is for full-MAC chips or not, > > etc. > > > > The patch is intended for full-MAC chips and I will update the commit > message to include more details. > > > At least for mac80211, we can define signals to userspace regarding > > "link quality" and make decisions in wpa_s then? > > > > The policy is offloaded today to the FW and the decision being made > by the FW. > We are only modifying the behavior and not making the decision. Fair enough, but I stand by saying that it's extremely vague - how could userspace possibly know what "level" to pick, if you don't give it any information about what the levels really mean? johannes
On 1/2/2017 12:45 PM, Johannes Berg wrote: > On Mon, 2016-12-19 at 12:58 +0200, Lazar, Alexei Avshalom wrote: > >>> What you're doing here is *extremely* vague. No definitions of what >>> this means, no description of how it's intended to be implemented, >>> no note even on whether or not this is for full-MAC chips or not, >>> etc. >>> >> >> The patch is intended for full-MAC chips and I will update the commit >> message to include more details. >> >>> At least for mac80211, we can define signals to userspace regarding >>> "link quality" and make decisions in wpa_s then? >>> >> >> The policy is offloaded today to the FW and the decision being made >> by the FW. >> We are only modifying the behavior and not making the decision. > > Fair enough, but I stand by saying that it's extremely vague - how > could userspace possibly know what "level" to pick, if you don't give > it any information about what the levels really mean? > > johannes > I have updated the commit message for clarifying what the levels really mean.
diff --git a/drivers/net/wireless/ath/wil6210/cfg80211.c b/drivers/net/wireless/ath/wil6210/cfg80211.c index d117240..f8d576b 100644 --- a/drivers/net/wireless/ath/wil6210/cfg80211.c +++ b/drivers/net/wireless/ath/wil6210/cfg80211.c @@ -1424,6 +1424,22 @@ static void wil_cfg80211_stop_p2p_device(struct wiphy *wiphy, mutex_unlock(&wil->mutex); } +static int +wil_cfg80211_set_link_loss_profile(struct wiphy *wiphy, + struct wireless_dev *wdev, + enum nl80211_link_loss_profile profile, + const u8 *addr) +{ + return -ENOTSUPP; +} + +static enum nl80211_link_loss_profile +wil_cfg80211_get_link_loss_profile(struct wiphy *wiphy, + struct wireless_dev *wdev, const u8 *addr) +{ + return -ENOTSUPP; +} + static struct cfg80211_ops wil_cfg80211_ops = { .add_virtual_intf = wil_cfg80211_add_iface, .del_virtual_intf = wil_cfg80211_del_iface, @@ -1450,6 +1466,8 @@ static struct cfg80211_ops wil_cfg80211_ops = { /* P2P device */ .start_p2p_device = wil_cfg80211_start_p2p_device, .stop_p2p_device = wil_cfg80211_stop_p2p_device, + .set_link_loss_profile = wil_cfg80211_set_link_loss_profile, + .get_link_loss_profile = wil_cfg80211_get_link_loss_profile, }; static void wil_wiphy_init(struct wiphy *wiphy) diff --git a/include/net/cfg80211.h b/include/net/cfg80211.h index 41ae3f5..2d7a0af 100644 --- a/include/net/cfg80211.h +++ b/include/net/cfg80211.h @@ -2744,6 +2744,10 @@ struct cfg80211_nan_func { * All other parameters must be ignored. * * @set_multicast_to_unicast: configure multicast to unicast conversion for BSS + * + * @set_link_loss_profile: Set link loss profile for specific connection. + * @get_link_loss_profile: Get the current link loss profile of specific + * connection. */ struct cfg80211_ops { int (*suspend)(struct wiphy *wiphy, struct cfg80211_wowlan *wow); @@ -3024,6 +3028,15 @@ struct cfg80211_ops { int (*set_multicast_to_unicast)(struct wiphy *wiphy, struct net_device *dev, const bool enabled); + + int (*set_link_loss_profile)(struct wiphy *wiphy, + struct wireless_dev *wdev, + enum nl80211_link_loss_profile profile, + const u8 *addr); + enum nl80211_link_loss_profile (*get_link_loss_profile)( + struct wiphy *wiphy, + struct wireless_dev *wdev, + const u8 *addr); }; /* diff --git a/include/uapi/linux/nl80211.h b/include/uapi/linux/nl80211.h index e21d23d..b68fad2 100644 --- a/include/uapi/linux/nl80211.h +++ b/include/uapi/linux/nl80211.h @@ -888,6 +888,11 @@ * This will contain a %NL80211_ATTR_NAN_MATCH nested attribute and * %NL80211_ATTR_COOKIE. * + * @NL80211_CMD_SET_LINK_LOSS_PROFILE: Set link loss profile (behavior) for + * specific connection. + * @NL80211_CMD_GET_LINK_LOSS_PROFILE: Get current link loss profile of specific + * connection. + * * @NL80211_CMD_MAX: highest used command number * @__NL80211_CMD_AFTER_LAST: internal use */ @@ -1085,6 +1090,9 @@ enum nl80211_commands { NL80211_CMD_SET_MULTICAST_TO_UNICAST, + NL80211_CMD_SET_LINK_LOSS_PROFILE, + NL80211_CMD_GET_LINK_LOSS_PROFILE, + /* add new commands above here */ /* used to define NL80211_CMD_MAX below */ @@ -1969,6 +1977,9 @@ enum nl80211_commands { * @NL80211_ATTR_MULTICAST_TO_UNICAST_ENABLED: Indicates whether or not multicast * packets should be send out as unicast to all stations (flag attribute). * + * @NL80211_ATTR_LINK_LOSS_PROFILE: attribute that indicate the link loss + * behavior using &enum nl80211_link_loss_profile values. + * * @NUM_NL80211_ATTR: total number of nl80211_attrs available * @NL80211_ATTR_MAX: highest attribute number currently defined * @__NL80211_ATTR_AFTER_LAST: internal use @@ -2373,6 +2384,8 @@ enum nl80211_attrs { NL80211_ATTR_MULTICAST_TO_UNICAST_ENABLED, + NL80211_ATTR_LINK_LOSS_PROFILE, + /* add attributes here, update the policy in nl80211.c */ __NL80211_ATTR_AFTER_LAST, @@ -5174,4 +5187,23 @@ enum nl80211_nan_match_attributes { NL80211_NAN_MATCH_ATTR_MAX = NUM_NL80211_NAN_MATCH_ATTR - 1 }; +/** + * enum nl80211_link_loss_profile. + * + * Used by set_link_loss_profile() and get_link_loss_profile() + * to set/get link loss behavior. + * + * @NL80211_LINK_LOSS_PROFILE_RELAXED: prefer maintaining link + * up even in poor link quality environment + * @NL80211_LINK_LOSS_PROFILE_DEFAULT: The default behavior for + * maintaining link up vs link quality. + * @NL80211_LINK_LOSS_PROFILE_AGGRESSIVE: prefer losing link + * up in poor link quality environment + */ +enum nl80211_link_loss_profile { + NL80211_LINK_LOSS_PROFILE_RELAXED, + NL80211_LINK_LOSS_PROFILE_DEFAULT, + NL80211_LINK_LOSS_PROFILE_AGGRESSIVE +}; + #endif /* __LINUX_NL80211_H */ diff --git a/net/wireless/nl80211.c b/net/wireless/nl80211.c index 46b2e8c..c8dc7e2 100644 --- a/net/wireless/nl80211.c +++ b/net/wireless/nl80211.c @@ -418,6 +418,7 @@ static const struct nla_policy nl80211_policy[NUM_NL80211_ATTR] = { .len = FILS_MAX_KEK_LEN }, [NL80211_ATTR_FILS_NONCES] = { .len = 2 * FILS_NONCE_LEN }, [NL80211_ATTR_MULTICAST_TO_UNICAST_ENABLED] = { .type = NLA_FLAG, }, + [NL80211_ATTR_LINK_LOSS_PROFILE] = { .type = NLA_U8 }, }; /* policy for the key attributes */ @@ -11792,6 +11793,60 @@ static int nl80211_set_multicast_to_unicast(struct sk_buff *skb, return rdev_set_multicast_to_unicast(rdev, dev, enabled); } +static int nl80211_set_link_loss_profile(struct sk_buff *skb, + struct genl_info *info) +{ + struct cfg80211_registered_device *rdev = info->user_ptr[0]; + struct wireless_dev *wdev = info->user_ptr[1]; + enum nl80211_link_loss_profile profile; + const u8 *addr; + int ret; + + if (!info->attrs[NL80211_ATTR_LINK_LOSS_PROFILE] || + !info->attrs[NL80211_ATTR_MAC]) + return -EINVAL; + + profile = nla_get_u8(info->attrs[NL80211_ATTR_LINK_LOSS_PROFILE]); + addr = nla_data(info->attrs[NL80211_ATTR_MAC]); + + if (profile != NL80211_LINK_LOSS_PROFILE_RELAXED && + profile != NL80211_LINK_LOSS_PROFILE_DEFAULT && + profile != NL80211_LINK_LOSS_PROFILE_AGGRESSIVE) + return -EINVAL; + + if (!rdev->ops->set_link_loss_profile) + return -EOPNOTSUPP; + + wdev_lock(wdev); + ret = rdev_set_link_loss_profile(rdev, wdev, profile, addr); + wdev_unlock(wdev); + + return ret; +} + +static int nl80211_get_link_loss_profile(struct sk_buff *skb, + struct genl_info *info) +{ + struct cfg80211_registered_device *rdev = info->user_ptr[0]; + struct wireless_dev *wdev = info->user_ptr[1]; + const u8 *addr; + enum nl80211_link_loss_profile profile; + + if (!info->attrs[NL80211_ATTR_MAC]) + return -EINVAL; + + addr = nla_data(info->attrs[NL80211_ATTR_MAC]); + + if (!rdev->ops->get_link_loss_profile) + return -EOPNOTSUPP; + + wdev_lock(wdev); + profile = rdev_get_link_loss_profile(rdev, wdev, addr); + wdev_unlock(wdev); + + return profile; +} + #define NL80211_FLAG_NEED_WIPHY 0x01 #define NL80211_FLAG_NEED_NETDEV 0x02 #define NL80211_FLAG_NEED_RTNL 0x04 @@ -12659,6 +12714,21 @@ static const struct genl_ops nl80211_ops[] = { .internal_flags = NL80211_FLAG_NEED_NETDEV | NL80211_FLAG_NEED_RTNL, }, + { + .cmd = NL80211_CMD_SET_LINK_LOSS_PROFILE, + .doit = nl80211_set_link_loss_profile, + .policy = nl80211_policy, + .flags = GENL_ADMIN_PERM, + .internal_flags = NL80211_FLAG_NEED_NETDEV_UP | + NL80211_FLAG_NEED_RTNL, + }, + { + .cmd = NL80211_CMD_GET_LINK_LOSS_PROFILE, + .doit = nl80211_get_link_loss_profile, + .policy = nl80211_policy, + .internal_flags = NL80211_FLAG_NEED_NETDEV_UP | + NL80211_FLAG_NEED_RTNL, + }, }; /* notification functions */ diff --git a/net/wireless/rdev-ops.h b/net/wireless/rdev-ops.h index e9ecf23..9a377c8 100644 --- a/net/wireless/rdev-ops.h +++ b/net/wireless/rdev-ops.h @@ -1141,4 +1141,32 @@ rdev_set_coalesce(struct cfg80211_registered_device *rdev, trace_rdev_return_int(&rdev->wiphy, ret); return ret; } + +static inline int +rdev_set_link_loss_profile(struct cfg80211_registered_device *rdev, + struct wireless_dev *wdev, + enum nl80211_link_loss_profile profile, + const u8 *addr) +{ + int ret; + + trace_rdev_set_link_loss_profile(&rdev->wiphy, wdev, profile, addr); + ret = rdev->ops->set_link_loss_profile(&rdev->wiphy, wdev, profile, + addr); + trace_rdev_return_int(&rdev->wiphy, ret); + return ret; +} + +static inline enum nl80211_link_loss_profile +rdev_get_link_loss_profile(struct cfg80211_registered_device *rdev, + struct wireless_dev *wdev, + const u8 *addr) +{ + enum nl80211_link_loss_profile profile; + + trace_rdev_get_link_loss_profile(&rdev->wiphy, wdev, addr); + profile = rdev->ops->get_link_loss_profile(&rdev->wiphy, wdev, addr); + trace_rdev_return_int(&rdev->wiphy, profile); + return profile; +} #endif /* __CFG80211_RDEV_OPS */ diff --git a/net/wireless/trace.h b/net/wireless/trace.h index 25f8318..d93a4b1 100644 --- a/net/wireless/trace.h +++ b/net/wireless/trace.h @@ -2212,6 +2212,45 @@ TRACE_EVENT(rdev_tdls_cancel_channel_switch, WIPHY_PR_ARG, NETDEV_PR_ARG, MAC_PR_ARG(addr)) ); +TRACE_EVENT(rdev_set_link_loss_profile, + TP_PROTO(struct wiphy *wiphy, struct wireless_dev *wdev, + enum nl80211_link_loss_profile profile, const u8 *addr), + TP_ARGS(wiphy, wdev, profile, addr), + TP_STRUCT__entry( + WIPHY_ENTRY + WDEV_ENTRY + __field(enum nl80211_link_loss_profile, profile) + MAC_ENTRY(addr) + ), + TP_fast_assign( + WIPHY_ASSIGN; + WDEV_ASSIGN; + __entry->profile = profile; + MAC_ASSIGN(addr, addr); + ), + TP_printk(WIPHY_PR_FMT ", " WDEV_PR_FMT ", " MAC_PR_FMT ", PROFILE %d", + WIPHY_PR_ARG, WDEV_PR_ARG, MAC_PR_ARG(addr), + __entry->profile) +); + +TRACE_EVENT(rdev_get_link_loss_profile, + TP_PROTO(struct wiphy *wiphy, struct wireless_dev *wdev, + const u8 *addr), + TP_ARGS(wiphy, wdev, addr), + TP_STRUCT__entry( + WIPHY_ENTRY + WDEV_ENTRY + MAC_ENTRY(addr) + ), + TP_fast_assign( + WIPHY_ASSIGN; + WDEV_ASSIGN; + MAC_ASSIGN(addr, addr); + ), + TP_printk(WIPHY_PR_FMT ", " WDEV_PR_FMT ", " MAC_PR_FMT, WIPHY_PR_ARG, + WDEV_PR_ARG, MAC_PR_ARG(addr)) +); + /************************************************************* * cfg80211 exported functions traces * *************************************************************/