* [PATCH v2] mac80211: mesh: Add support for HW RC implementation
@ 2016-06-30 15:30 Maxim Altshul
2016-07-06 12:35 ` Johannes Berg
0 siblings, 1 reply; 4+ messages in thread
From: Maxim Altshul @ 2016-06-30 15:30 UTC (permalink / raw)
To: linux-kernel
Cc: Maxim Altshul, Johannes Berg, David S. Miller, linux-wireless, netdev
Mesh HWMP module will be able to rely on the HW
RC algorithm if it exists, for path metric calculations.
This allows the metric calculation mechanism to calculate
a correct metric, based on PER and last TX rate both via
HW RC algorithm if it exists or via parameters collected
by the SW.
Signed-off-by: Maxim Altshul <maxim.altshul@ti.com>
---
Changed the function to return u32, I agree that this
is much clearer.
As for the rate, two things:
1. I had to divide the returned value by 100, since
drv_get_expected_throughput returns values in units of Kbps.
On the contrary, the function cfg80211_calculate_bitrate
returns in units of 100Kbps, so a correction is needed.
2. Why return the value into rate?
As I understand, rate here is actually bitrate,
and so, we have two possible outcomes:
- A SW/HW RC algo does exist, and an estimated throughput is
returned. err is set to 0 (as it is already included in the RC algo)
and the airtime is calculated using the estimated throughput.
- A SW/HW RC algo does not exist, and thus the regular calculation
takes place, in which an estimated throughput is calculated
using the bitrate and the err parameter.
>From this calculation the airtime is calculated.
net/mac80211/mesh_hwmp.c | 25 +++++++++++++++++--------
net/mac80211/sta_info.c | 23 +++++++++++++++++++----
net/mac80211/sta_info.h | 2 ++
3 files changed, 38 insertions(+), 12 deletions(-)
diff --git a/net/mac80211/mesh_hwmp.c b/net/mac80211/mesh_hwmp.c
index c6be0b4..ad67f46 100644
--- a/net/mac80211/mesh_hwmp.c
+++ b/net/mac80211/mesh_hwmp.c
@@ -322,19 +322,28 @@ static u32 airtime_link_metric_get(struct ieee80211_local *local,
int device_constant = 1 << ARITH_SHIFT;
int test_frame_len = TEST_FRAME_LEN << ARITH_SHIFT;
int s_unit = 1 << ARITH_SHIFT;
- int rate, err;
+ int rate = 0, err = 0;
u32 tx_time, estimated_retx;
u64 result;
- if (sta->mesh->fail_avg >= 100)
- return MAX_METRIC;
+ /* Try to get rate based on HW/SW RC algorithm.
+ * Rate is returned in units of Kbps, correct this
+ * to comply with airtime calculation units
+ */
+ rate = sta_get_expected_throughput(sta) / 100;
+
+ /* if we did not get a rate */
+ if (!rate) {
+ if (sta->mesh->fail_avg >= 100)
+ return MAX_METRIC;
- sta_set_rate_info_tx(sta, &sta->tx_stats.last_rate, &rinfo);
- rate = cfg80211_calculate_bitrate(&rinfo);
- if (WARN_ON(!rate))
- return MAX_METRIC;
+ sta_set_rate_info_tx(sta, &sta->tx_stats.last_rate, &rinfo);
+ rate = cfg80211_calculate_bitrate(&rinfo);
+ if (WARN_ON(!rate))
+ return MAX_METRIC;
- err = (sta->mesh->fail_avg << ARITH_SHIFT) / 100;
+ err = (sta->mesh->fail_avg << ARITH_SHIFT) / 100;
+ }
/* bitrate is in units of 100 Kbps, while we need rate in units of
* 1Mbps. This will be corrected on tx_time computation.
diff --git a/net/mac80211/sta_info.c b/net/mac80211/sta_info.c
index 63ea6cb..fa2c507 100644
--- a/net/mac80211/sta_info.c
+++ b/net/mac80211/sta_info.c
@@ -2069,14 +2069,29 @@ void sta_set_sinfo(struct sta_info *sta, struct station_info *sinfo)
if (test_sta_flag(sta, WLAN_STA_TDLS_PEER))
sinfo->sta_flags.set |= BIT(NL80211_STA_FLAG_TDLS_PEER);
+ thr = sta_get_expected_throughput(sta);
+
+ if (thr != 0) {
+ sinfo->filled |= BIT(NL80211_STA_INFO_EXPECTED_THROUGHPUT);
+ sinfo->expected_throughput = thr;
+ }
+}
+
+u32 sta_get_expected_throughput(struct sta_info *sta)
+{
+ struct ieee80211_sub_if_data *sdata = sta->sdata;
+ struct ieee80211_local *local = sdata->local;
+ struct rate_control_ref *ref = NULL;
+ u32 thr = 0;
+
+ if (test_sta_flag(sta, WLAN_STA_RATE_CONTROL))
+ ref = local->rate_ctrl;
+
/* check if the driver has a SW RC implementation */
if (ref && ref->ops->get_expected_throughput)
thr = ref->ops->get_expected_throughput(sta->rate_ctrl_priv);
else
thr = drv_get_expected_throughput(local, &sta->sta);
- if (thr != 0) {
- sinfo->filled |= BIT(NL80211_STA_INFO_EXPECTED_THROUGHPUT);
- sinfo->expected_throughput = thr;
- }
+ return thr;
}
diff --git a/net/mac80211/sta_info.h b/net/mac80211/sta_info.h
index 2cafb21..5c9c8be 100644
--- a/net/mac80211/sta_info.h
+++ b/net/mac80211/sta_info.h
@@ -661,6 +661,8 @@ void sta_set_rate_info_tx(struct sta_info *sta,
struct rate_info *rinfo);
void sta_set_sinfo(struct sta_info *sta, struct station_info *sinfo);
+u32 sta_get_expected_throughput(struct sta_info *sta);
+
void ieee80211_sta_expire(struct ieee80211_sub_if_data *sdata,
unsigned long exp_time);
u8 sta_info_tx_streams(struct sta_info *sta);
--
2.9.0
^ permalink raw reply related [flat|nested] 4+ messages in thread* Re: [PATCH v2] mac80211: mesh: Add support for HW RC implementation
2016-06-30 15:30 [PATCH v2] mac80211: mesh: Add support for HW RC implementation Maxim Altshul
@ 2016-07-06 12:35 ` Johannes Berg
0 siblings, 0 replies; 4+ messages in thread
From: Johannes Berg @ 2016-07-06 12:35 UTC (permalink / raw)
To: Maxim Altshul, linux-kernel; +Cc: David S. Miller, linux-wireless, netdev
On Thu, 2016-06-30 at 18:30 +0300, Maxim Altshul wrote:
> Mesh HWMP module will be able to rely on the HW
> RC algorithm if it exists, for path metric calculations.
>
> This allows the metric calculation mechanism to calculate
> a correct metric, based on PER and last TX rate both via
> HW RC algorithm if it exists or via parameters collected
> by the SW.
>
> Signed-off-by: Maxim Altshul <maxim.altshul@ti.com>
> ---
> Changed the function to return u32, I agree that this
> is much clearer.
> As for the rate, two things:
> 1. I had to divide the returned value by 100, since
> drv_get_expected_throughput returns values in units of Kbps.
> On the contrary, the function cfg80211_calculate_bitrate
> returns in units of 100Kbps, so a correction is needed.
> 2. Why return the value into rate?
> As I understand, rate here is actually bitrate,
> and so, we have two possible outcomes:
> - A SW/HW RC algo does exist, and an estimated throughput is
> returned. err is set to 0 (as it is already included in the RC algo)
> and the airtime is calculated using the estimated throughput.
> - A SW/HW RC algo does not exist, and thus the regular calculation
> takes place, in which an estimated throughput is calculated
> using the bitrate and the err parameter.
> From this calculation the airtime is calculated.
>
> net/mac80211/mesh_hwmp.c | 25 +++++++++++++++++--------
> net/mac80211/sta_info.c | 23 +++++++++++++++++++----
> net/mac80211/sta_info.h | 2 ++
> 3 files changed, 38 insertions(+), 12 deletions(-)
>
> diff --git a/net/mac80211/mesh_hwmp.c b/net/mac80211/mesh_hwmp.c
> index c6be0b4..ad67f46 100644
> --- a/net/mac80211/mesh_hwmp.c
> +++ b/net/mac80211/mesh_hwmp.c
> @@ -322,19 +322,28 @@ static u32 airtime_link_metric_get(struct
> ieee80211_local *local,
> int device_constant = 1 << ARITH_SHIFT;
> int test_frame_len = TEST_FRAME_LEN << ARITH_SHIFT;
> int s_unit = 1 << ARITH_SHIFT;
> - int rate, err;
> + int rate = 0, err = 0;
The rate init is wrong - you overwrite it immediately.
The err init is questionable - I think it might be better to write
if (rate) {
err = 0;
} else {
...
}
below?
> u32 tx_time, estimated_retx;
> u64 result;
>
> - if (sta->mesh->fail_avg >= 100)
> - return MAX_METRIC;
> + /* Try to get rate based on HW/SW RC algorithm.
> + * Rate is returned in units of Kbps, correct this
> + * to comply with airtime calculation units
> + */
> + rate = sta_get_expected_throughput(sta) / 100;
> +
> + /* if we did not get a rate */
> + if (!rate) {
Maybe you want DIV_ROUND_UP to account for getting a very lot estimated
rate (< 100Kbps) returned, which isn't "no rate"?
Ohterwise looks fine.
johannes
^ permalink raw reply [flat|nested] 4+ messages in thread
* [PATCH v2] mac80211: mesh: Add support for HW RC implementation
@ 2016-06-23 13:53 Maxim Altshul
2016-06-23 19:26 ` Johannes Berg
0 siblings, 1 reply; 4+ messages in thread
From: Maxim Altshul @ 2016-06-23 13:53 UTC (permalink / raw)
To: linux-kernel; +Cc: Maxim Altshul, Johannes Berg, linux-wireless
Mesh HWMP module will be able to rely on the HW
RC algorithm if it exists, for path metric calculations.
This allows the metric calculation mechanism to calculate
a correct metric, based on PER and last TX rate both via
HW RC algorithm if it exists or via parameters collected
by the SW.
Signed-off-by: Maxim Altshul <maxim.altshul@ti.com>
---
Changed previous implementation according to Johannes's
comment. Combined the RC test and throughput assignment
to a sta_get_expected_throughput function.
Using the function both from airtime_link_metric and
from sta_set_sinfo.
net/mac80211/mesh_hwmp.c | 22 ++++++++++++++--------
net/mac80211/sta_info.c | 22 +++++++++++++++++-----
net/mac80211/sta_info.h | 2 ++
3 files changed, 33 insertions(+), 13 deletions(-)
diff --git a/net/mac80211/mesh_hwmp.c b/net/mac80211/mesh_hwmp.c
index c6be0b4..39f9541 100644
--- a/net/mac80211/mesh_hwmp.c
+++ b/net/mac80211/mesh_hwmp.c
@@ -322,19 +322,25 @@ static u32 airtime_link_metric_get(struct ieee80211_local *local,
int device_constant = 1 << ARITH_SHIFT;
int test_frame_len = TEST_FRAME_LEN << ARITH_SHIFT;
int s_unit = 1 << ARITH_SHIFT;
- int rate, err;
+ int rate = 0, err = 0;
u32 tx_time, estimated_retx;
u64 result;
- if (sta->mesh->fail_avg >= 100)
- return MAX_METRIC;
+ /* try to get rate based on HW/SW RC algorithm */
+ sta_get_expected_throughput(sta, &rate);
- sta_set_rate_info_tx(sta, &sta->tx_stats.last_rate, &rinfo);
- rate = cfg80211_calculate_bitrate(&rinfo);
- if (WARN_ON(!rate))
- return MAX_METRIC;
+ /* if we did not get a rate */
+ if (!rate) {
+ if (sta->mesh->fail_avg >= 100)
+ return MAX_METRIC;
- err = (sta->mesh->fail_avg << ARITH_SHIFT) / 100;
+ sta_set_rate_info_tx(sta, &sta->tx_stats.last_rate, &rinfo);
+ rate = cfg80211_calculate_bitrate(&rinfo);
+ if (WARN_ON(!rate))
+ return MAX_METRIC;
+
+ err = (sta->mesh->fail_avg << ARITH_SHIFT) / 100;
+ }
/* bitrate is in units of 100 Kbps, while we need rate in units of
* 1Mbps. This will be corrected on tx_time computation.
diff --git a/net/mac80211/sta_info.c b/net/mac80211/sta_info.c
index 63ea6cb..9580946 100644
--- a/net/mac80211/sta_info.c
+++ b/net/mac80211/sta_info.c
@@ -2069,14 +2069,26 @@ void sta_set_sinfo(struct sta_info *sta, struct station_info *sinfo)
if (test_sta_flag(sta, WLAN_STA_TDLS_PEER))
sinfo->sta_flags.set |= BIT(NL80211_STA_FLAG_TDLS_PEER);
- /* check if the driver has a SW RC implementation */
- if (ref && ref->ops->get_expected_throughput)
- thr = ref->ops->get_expected_throughput(sta->rate_ctrl_priv);
- else
- thr = drv_get_expected_throughput(local, &sta->sta);
+ sta_get_expected_throughput(sta, &thr);
if (thr != 0) {
sinfo->filled |= BIT(NL80211_STA_INFO_EXPECTED_THROUGHPUT);
sinfo->expected_throughput = thr;
}
}
+
+void sta_get_expected_throughput(struct sta_info *sta, u32 *thr)
+{
+ struct ieee80211_sub_if_data *sdata = sta->sdata;
+ struct ieee80211_local *local = sdata->local;
+ struct rate_control_ref *ref = NULL;
+
+ if (test_sta_flag(sta, WLAN_STA_RATE_CONTROL))
+ ref = local->rate_ctrl;
+
+ /* check if the driver has a SW RC implementation */
+ if (ref && ref->ops->get_expected_throughput)
+ *thr = ref->ops->get_expected_throughput(sta->rate_ctrl_priv);
+ else
+ *thr = drv_get_expected_throughput(local, &sta->sta);
+}
diff --git a/net/mac80211/sta_info.h b/net/mac80211/sta_info.h
index 2cafb21..8535443 100644
--- a/net/mac80211/sta_info.h
+++ b/net/mac80211/sta_info.h
@@ -661,6 +661,8 @@ void sta_set_rate_info_tx(struct sta_info *sta,
struct rate_info *rinfo);
void sta_set_sinfo(struct sta_info *sta, struct station_info *sinfo);
+void sta_get_expected_throughput(struct sta_info *sta, u32 *thr);
+
void ieee80211_sta_expire(struct ieee80211_sub_if_data *sdata,
unsigned long exp_time);
u8 sta_info_tx_streams(struct sta_info *sta);
--
2.9.0
^ permalink raw reply related [flat|nested] 4+ messages in thread* Re: [PATCH v2] mac80211: mesh: Add support for HW RC implementation
2016-06-23 13:53 Maxim Altshul
@ 2016-06-23 19:26 ` Johannes Berg
0 siblings, 0 replies; 4+ messages in thread
From: Johannes Berg @ 2016-06-23 19:26 UTC (permalink / raw)
To: Maxim Altshul, linux-kernel; +Cc: linux-wireless
> +void sta_get_expected_throughput(struct sta_info *sta, u32 *thr)
> +{
> + struct ieee80211_sub_if_data *sdata = sta->sdata;
> + struct ieee80211_local *local = sdata->local;
> + struct rate_control_ref *ref = NULL;
> +
> + if (test_sta_flag(sta, WLAN_STA_RATE_CONTROL))
> + ref = local->rate_ctrl;
> +
> + /* check if the driver has a SW RC implementation */
> + if (ref && ref->ops->get_expected_throughput)
> + *thr = ref->ops->get_expected_throughput(sta-
> >rate_ctrl_priv);
> + else
> + *thr = drv_get_expected_throughput(local, &sta-
> >sta);
> +}
I don't really see a reason for the function to take a u32 *thr, rather
than return u32? Would be clearer, IMHO, to just return the value.
Then in the user, it seems that you shouldn't use "rate" as the
variable that gets this result?
johannes
^ permalink raw reply [flat|nested] 4+ messages in thread
end of thread, other threads:[~2016-07-06 12:35 UTC | newest]
Thread overview: 4+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2016-06-30 15:30 [PATCH v2] mac80211: mesh: Add support for HW RC implementation Maxim Altshul
2016-07-06 12:35 ` Johannes Berg
-- strict thread matches above, loose matches on Subject: below --
2016-06-23 13:53 Maxim Altshul
2016-06-23 19:26 ` Johannes Berg
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).