| Message ID | 20240319204008.792040-1-Chaitanya.Tata@nordicsemi.no |
|---|---|
| State | Changes Requested |
| Headers | show |
| Series | wpa_supplicant: Add an option to disable SNR capping | expand |
On 3/19/24 13:40, Chaitanya Tata wrote: > When selecting a network the default behaviour of WPA supplicant is to > prefer higher throughput, it does this by capping the SNR. > > But in certain environments, reliability is important over throughput > and choosing a lower SNR (thought it is greater than "Great SNR") might > be sub-optimal. > > Introduce a configuration option to choose the two options (throughput > or reliability). Could you make this a run-time configurable instead of compile time option? That seems a lot more flexible to me... Thanks, Ben > > Signed-off-by: Chaitanya Tata <Chaitanya.Tata@nordicsemi.no> > --- > wpa_supplicant/Makefile | 4 ++++ > wpa_supplicant/defconfig | 11 +++++++++++ > wpa_supplicant/scan.h | 10 ++++++++++ > 3 files changed, 25 insertions(+) > > diff --git a/wpa_supplicant/Makefile b/wpa_supplicant/Makefile > index ce1c8b2e3..dc3dbf0ca 100644 > --- a/wpa_supplicant/Makefile > +++ b/wpa_supplicant/Makefile > @@ -1877,6 +1877,10 @@ ifdef CONFIG_NO_TKIP > CFLAGS += -DCONFIG_NO_TKIP > endif > > +ifdef CONFIG_NW_SEL_RELIABILITY > +CFLAGS += -DCONFIG_NW_SEL_RELIABILITY > +endif
On Wed, Mar 20, 2024 at 2:19 AM Ben Greear <greearb@candelatech.com> wrote: > > On 3/19/24 13:40, Chaitanya Tata wrote: > > When selecting a network the default behaviour of WPA supplicant is to > > prefer higher throughput, it does this by capping the SNR. > > > > But in certain environments, reliability is important over throughput > > and choosing a lower SNR (thought it is greater than "Great SNR") might > > be sub-optimal. > > > > Introduce a configuration option to choose the two options (throughput > > or reliability). > > Could you make this a run-time configurable instead of compile time option? > > That seems a lot more flexible to me... Runtime is a bit trickier as the WPA supplicant maintains an aggregated scan list (maintained using ageing), so, if we change the criterion in between, we might have to flush the existing scan results and start afresh, hence went with compile time. But I will think more about this (this was a fresh bug) while I await any other feedback.
On Wed, Mar 20, 2024 at 02:48:39AM +0530, Krishna Chaitanya wrote: > Runtime is a bit trickier as the WPA supplicant maintains an > aggregated scan list (maintained > using ageing), so, if we change the criterion in between, we might > have to flush the existing > scan results and start afresh, hence went with compile time. > > But I will think more about this (this was a fresh bug) while I await > any other feedback. I do not really like adding build time configuration parameters for something like this, so I would be much more likely to apply a version that allows configuration to be done from the runtime configuration file. I don't see this really being toggled that frequently, but it makes no sense for distros to be forced to distribute two different binaries if they want to support both options.
On Sun, Mar 24, 2024 at 1:19 AM Jouni Malinen <j@w1.fi> wrote: > > On Wed, Mar 20, 2024 at 02:48:39AM +0530, Krishna Chaitanya wrote: > > Runtime is a bit trickier as the WPA supplicant maintains an > > aggregated scan list (maintained > > using ageing), so, if we change the criterion in between, we might > > have to flush the existing > > scan results and start afresh, hence went with compile time. > > > > But I will think more about this (this was a fresh bug) while I await > > any other feedback. > > I do not really like adding build time configuration parameters for > something like this, so I would be much more likely to apply a version > that allows configuration to be done from the runtime configuration > file. I don't see this really being toggled that frequently, but it > makes no sense for distros to be forced to distribute two different > binaries if they want to support both options. I have chosen compile time as this option is seldom changed, but I understand your point about maintaining separate binaries. I will add a runtime configuration option for the scan, thanks.
diff --git a/wpa_supplicant/Makefile b/wpa_supplicant/Makefile index ce1c8b2e3..dc3dbf0ca 100644 --- a/wpa_supplicant/Makefile +++ b/wpa_supplicant/Makefile @@ -1877,6 +1877,10 @@ ifdef CONFIG_NO_TKIP CFLAGS += -DCONFIG_NO_TKIP endif +ifdef CONFIG_NW_SEL_RELIABILITY +CFLAGS += -DCONFIG_NW_SEL_RELIABILITY +endif + dynamic_eap_methods: $(EAPDYN) _OBJS_VAR := OBJS_priv diff --git a/wpa_supplicant/defconfig b/wpa_supplicant/defconfig index 708a82385..0b87fb5df 100644 --- a/wpa_supplicant/defconfig +++ b/wpa_supplicant/defconfig @@ -628,3 +628,14 @@ CONFIG_DPP=y # design is still subject to change. As such, this should not yet be enabled in # production use. #CONFIG_PASN=y + +# When selecting a network, prefer networks with a better signal strength +# without any cap a.k.a "Great SNR". +# +# If this is enabled then higher SNR is preferred without any cap over higher band +# (higher throughput) to achieve better reliability. +# +# If this is disabled then for networks with capped SNR, the preference is given +# to the network with higher band (5GHz or 6GHz) and then to the network with +# higher SNR. +#CONFIG_NW_SEL_RELIABILITY=y diff --git a/wpa_supplicant/scan.h b/wpa_supplicant/scan.h index d1780eb09..175dc37ef 100644 --- a/wpa_supplicant/scan.h +++ b/wpa_supplicant/scan.h @@ -18,6 +18,15 @@ #define DEFAULT_NOISE_FLOOR_5GHZ (-92) #define DEFAULT_NOISE_FLOOR_6GHZ (-92) +#ifdef CONFIG_NW_SEL_RELIABILITY +/* + * This basically disables capping the SNR when choosing the best BSS. This + * means that the BSS with the highest SNR will be chosen, independent of the + * band. This is useful for scenarios where reliability is more important than + * throughput. + */ +#define GREAT_SNR 999 +#else /* * Channels with a great SNR can operate at full rate. What is a great SNR? * This doc https://supportforums.cisco.com/docs/DOC-12954 says, "the general @@ -29,6 +38,7 @@ * somewhat conservative value here. */ #define GREAT_SNR 25 +#endif /* CONFIG_NW_SEL_RELIABILITY */ #define IS_2P4GHZ(n) (n >= 2412 && n <= 2484) #define IS_5GHZ(n) (n > 4000 && n < 5895)
When selecting a network the default behaviour of WPA supplicant is to prefer higher throughput, it does this by capping the SNR. But in certain environments, reliability is important over throughput and choosing a lower SNR (thought it is greater than "Great SNR") might be sub-optimal. Introduce a configuration option to choose the two options (throughput or reliability). Signed-off-by: Chaitanya Tata <Chaitanya.Tata@nordicsemi.no> --- wpa_supplicant/Makefile | 4 ++++ wpa_supplicant/defconfig | 11 +++++++++++ wpa_supplicant/scan.h | 10 ++++++++++ 3 files changed, 25 insertions(+)