Re: [RFC] ath9k: Fix ath9k prevents CPU to enter C3 states

On Wed, Feb 09, 2011 at 06:33:27AM -0800, Mohammed Shajakhan wrote:

From: Mohammed Shafi Shajakhan <redacted>

The DMA latency issue is observed only in Intel pinetrail platforms but
in the driver we had a default PM-QOS value of 55. This caused
unnecessary power consumption and battery drain in other platforms.
Address this issue by disabling PM-QOS by default by setting it's value
as '0' and making code changes appropriately.This addresses the bug:
https://bugzilla.kernel.org/show_bug.cgi?id=27532
	If the user sees some DMA latency issue he can still use the pmqos as a
module parameter to trade power for throughput as below:
sudo modprobe ath9k pmqos=55

How many times are we going to write some work around for some
platform unpatched BIOS bug ?

mcgrof@tux ~/wireless-testing (git::master)$ git grep pm_qos_update_request drivers/
drivers/net/e1000e/netdev.c:                    pm_qos_update_request(&adapter->netdev->pm_qos_req, 55);
drivers/net/e1000e/netdev.c:                    pm_qos_update_request(&adapter->netdev->pm_qos_req,
drivers/net/wireless/ath/ath9k/main.c:  pm_qos_update_request(&sc->pm_qos_req, ath9k_pm_qos_value);
drivers/net/wireless/ath/ath9k/main.c:  pm_qos_update_request(&sc->pm_qos_req, PM_QOS_DEFAULT_VALUE);
drivers/net/wireless/ipw2x00/ipw2100.c: pm_qos_update_request(&ipw2100_pm_qos_req, 175);
drivers/net/wireless/ipw2x00/ipw2100.c: pm_qos_update_request(&ipw2100_pm_qos_req, PM_QOS_DEFAULT_VALUE);

Can't we punt this crap to userspace?

http://www.kernel.org/pub/linux/kernel/people/mcgrof/scripts/cpudmalatency.c

Can distributions just autodetect this issue and call something like this?

  Luis

On Thursday 10 February 2011 01:28 AM, Luis Rodriguez wrote:

On Wed, Feb 09, 2011 at 06:33:27AM -0800, Mohammed Shajakhan wrote:
   

quoted

From: Mohammed Shafi Shajakhan<redacted>

The DMA latency issue is observed only in Intel pinetrail platforms but
in the driver we had a default PM-QOS value of 55. This caused
unnecessary power consumption and battery drain in other platforms.
Address this issue by disabling PM-QOS by default by setting it's value
as '0' and making code changes appropriately.This addresses the bug:
https://bugzilla.kernel.org/show_bug.cgi?id=27532
	If the user sees some DMA latency issue he can still use the pmqos as a
module parameter to trade power for throughput as below:
sudo modprobe ath9k pmqos=55
     

How many times are we going to write some work around for some
platform unpatched BIOS bug ?

mcgrof@tux ~/wireless-testing (git::master)$ git grep pm_qos_update_request drivers/
drivers/net/e1000e/netdev.c:                    pm_qos_update_request(&adapter->netdev->pm_qos_req, 55);
drivers/net/e1000e/netdev.c:                    pm_qos_update_request(&adapter->netdev->pm_qos_req,
drivers/net/wireless/ath/ath9k/main.c:  pm_qos_update_request(&sc->pm_qos_req, ath9k_pm_qos_value);
drivers/net/wireless/ath/ath9k/main.c:  pm_qos_update_request(&sc->pm_qos_req, PM_QOS_DEFAULT_VALUE);
drivers/net/wireless/ipw2x00/ipw2100.c: pm_qos_update_request(&ipw2100_pm_qos_req, 175);
drivers/net/wireless/ipw2x00/ipw2100.c: pm_qos_update_request(&ipw2100_pm_qos_req, PM_QOS_DEFAULT_VALUE);

Can't we punt this crap to userspace?
   

Luis thanks for the script and sorry I missed it in the bugzilla. Luis I 
agree these workarounds are better placed to be in user place and the 
only thing is those have issues must be aware of the script.
I am ccing Senthil to have his views.

thanks,
shafi

http://www.kernel.org/pub/linux/kernel/people/mcgrof/scripts/cpudmalatency.c

Can distributions just autodetect this issue and call something like this?

   Luis

   

quoted

Signed-off-by: Mohammed Shafi Shajakhan<redacted>
---
  drivers/net/wireless/ath/ath9k/ath9k.h |    4 +++-
  drivers/net/wireless/ath/ath9k/init.c  |    9 +++++----
  drivers/net/wireless/ath/ath9k/main.c  |    9 +++++----
  3 files changed, 13 insertions(+), 9 deletions(-)

diff --git a/drivers/net/wireless/ath/ath9k/ath9k.h b/drivers/net/wireless/ath/ath9k/ath9k.h
index 9272278..55fcd01 100644
--- a/drivers/net/wireless/ath/ath9k/ath9k.h
+++ b/drivers/net/wireless/ath/ath9k/ath9k.h

@@ -57,7 +57,9 @@ struct ath_node;

  #define A_MAX(a, b) ((a)>  (b) ? (a) : (b))

-#define ATH9K_PM_QOS_DEFAULT_VALUE	55
+/* By default PM QOS is disabled */
+#define ATH9K_PM_QOS_DEFAULT_VALUE	0
+

  #define TSF_TO_TU(_h,_l) \
  	((((u32)(_h))<<  22) | (((u32)(_l))>>  10))

diff --git a/drivers/net/wireless/ath/ath9k/init.c b/drivers/net/wireless/ath/ath9k/init.c
index e5c1eea..658ea02 100644
--- a/drivers/net/wireless/ath/ath9k/init.c
+++ b/drivers/net/wireless/ath/ath9k/init.c

@@ -43,7 +43,7 @@ MODULE_PARM_DESC(btcoex_enable, "Enable wifi-BT coexistence");

  int ath9k_pm_qos_value = ATH9K_PM_QOS_DEFAULT_VALUE;
  module_param_named(pmqos, ath9k_pm_qos_value, int, S_IRUSR | S_IRGRP | S_IROTH);
-MODULE_PARM_DESC(pmqos, "User specified PM-QOS value");
+MODULE_PARM_DESC(pmqos, "User specified PM-QOS value, by default disabled");

  bool is_ath9k_unloaded;
  /* We use the hw_value as an index into our private channel structure */

@@ -759,8 +759,8 @@ int ath9k_init_device(u16 devid, struct ath_softc *sc, u16 subsysid,

  	ath_init_leds(sc);
  	ath_start_rfkill_poll(sc);
-
-	pm_qos_add_request(&sc->pm_qos_req, PM_QOS_CPU_DMA_LATENCY,
+	if (ath9k_pm_qos_value)
+		pm_qos_add_request(&sc->pm_qos_req, PM_QOS_CPU_DMA_LATENCY,
  			   PM_QOS_DEFAULT_VALUE);

  	return 0;

@@ -819,7 +819,8 @@ void ath9k_deinit_device(struct ath_softc *sc)
  	ath9k_ps_restore(sc);

  	ieee80211_unregister_hw(hw);
-	pm_qos_remove_request(&sc->pm_qos_req);
+	if (ath9k_pm_qos_value)
+		pm_qos_remove_request(&sc->pm_qos_req);
  	ath_rx_cleanup(sc);
  	ath_tx_cleanup(sc);
  	ath9k_deinit_softc(sc);

diff --git a/drivers/net/wireless/ath/ath9k/main.c b/drivers/net/wireless/ath/ath9k/main.c
index 4ed43b2..3855092 100644
--- a/drivers/net/wireless/ath/ath9k/main.c
+++ b/drivers/net/wireless/ath/ath9k/main.c

@@ -1119,9 +1119,10 @@ static int ath9k_start(struct ieee80211_hw *hw)

  	/* User has the option to provide pm-qos value as a module
  	 * parameter rather than using the default value of
-	 * 'ATH9K_PM_QOS_DEFAULT_VALUE'.
+	 * 'ATH9K_PM_QOS_DEFAULT_VALUE' which disables PM QOS.
  	 */
-	pm_qos_update_request(&sc->pm_qos_req, ath9k_pm_qos_value);
+	if (ath9k_pm_qos_value)
+		pm_qos_update_request(&sc->pm_qos_req, ath9k_pm_qos_value);

  	if (ah->caps.pcie_lcr_extsync_en&&  common->bus_ops->extn_synch_en)
  		common->bus_ops->extn_synch_en(common);

@@ -1266,8 +1267,8 @@ static void ath9k_stop(struct ieee80211_hw *hw)
  	ath_radio_disable(sc, hw);

  	sc->sc_flags |= SC_OP_INVALID;
-
-	pm_qos_update_request(&sc->pm_qos_req, PM_QOS_DEFAULT_VALUE);
+	if (ath9k_pm_qos_value)
+		pm_qos_update_request(&sc->pm_qos_req, PM_QOS_DEFAULT_VALUE);

  	mutex_unlock(&sc->mutex);

-- 
1.7.0.4

     

.