[ath9k-devel] [RFC v3 0/4] SMP fixes: serialization and Disablig D1 and D2

[ath9k-devel] [RFC v3 0/4] SMP fixes: serialization and Disablig D1 and D2

[Luis R. Rodriguez]

From: Luis R. Rodriguez <lrodriguez@atheros.com>

Ok, here's a respin of my SMP fixes series (previously just serialization).
I got an SMP box at work now and did manage to get a PCI AR5416 card and
did reproduce a hang.

The good news is the hang is gone but I'm not sure if this will fix the hangs
reported here with AR5416 and SMP boxen so please give it a spin and let me
know.

This also now deregisters the CPU notifier which we were not doing before.

This applies on top of today's wireless-testing. You can get the full
set of 4 patches in one file here:

http://www.kernel.org/pub/linux/kernel/people/mcgrof/patches/ath9k/2009-02-24/serialization-v3.patch

I'd like a few reports on this, please let me know.

Luis R. Rodriguez (4):
  ath9k: implement IO serialization
  ath9k: AR9280 PCI devices must serialize IO as well
  ath9k: add cpu notifier to enhance device configuration
  ath9k: disable D1 and D2 on PCI multicore systems

 drivers/net/wireless/ath9k/ath9k.h |    7 ++++
 drivers/net/wireless/ath9k/hw.c    |   71 +++++++++++++++++++++++++++++++++++-
 drivers/net/wireless/ath9k/hw.h    |   11 +++++-
 drivers/net/wireless/ath9k/main.c  |   30 +++++++++++++++
 drivers/net/wireless/ath9k/pci.c   |   46 +++++++++++++++++++++++
 5 files changed, 162 insertions(+), 3 deletions(-)

[ath9k-devel] [RFC v3 1/4] ath9k: implement IO serialization

[Luis R. Rodriguez]

From: Luis R. Rodriguez <lrodriguez@atheros.com>

All 802.11n PCI devices (Cardbus, PCI, mini-PCI) require
serialization of IO when on non-uniprocessor systems. PCI
express devices not not require this.

Signed-off-by: Luis R. Rodriguez <lrodriguez@atheros.com>
---
 drivers/net/wireless/ath9k/ath9k.h |    1 +
 drivers/net/wireless/ath9k/hw.c    |   49 ++++++++++++++++++++++++++++++++++++
 drivers/net/wireless/ath9k/hw.h    |    8 ++++-
 drivers/net/wireless/ath9k/main.c  |    1 +
 4 files changed, 57 insertions(+), 2 deletions(-)

diff --git a/drivers/net/wireless/ath9k/ath9k.h b/drivers/net/wireless/ath9k/ath9k.h
index 6481ea4..7245797 100644
--- a/drivers/net/wireless/ath9k/ath9k.h
+++ b/drivers/net/wireless/ath9k/ath9k.h
@@ -575,6 +575,7 @@ struct ath_softc {
 	void __iomem *mem;
 	int irq;
 	spinlock_t sc_resetlock;
+	spinlock_t sc_serial_rw;
 	struct mutex mutex;
 
 	u8 curbssid[ETH_ALEN];
diff --git a/drivers/net/wireless/ath9k/hw.c b/drivers/net/wireless/ath9k/hw.c
index 2acbb84..005dd51 100644
--- a/drivers/net/wireless/ath9k/hw.c
+++ b/drivers/net/wireless/ath9k/hw.c
@@ -37,6 +37,36 @@ static u32 ath9k_hw_ini_fixup(struct ath_hw *ah,
 static void ath9k_hw_9280_spur_mitigate(struct ath_hw *ah, struct ath9k_channel *chan);
 static void ath9k_hw_spur_mitigate(struct ath_hw *ah, struct ath9k_channel *chan);
 
+/*
+ * Read and write, they both share the same lock. The exception is when
+ * in an the ISR -- in that case we don't need serialization as it should
+ * be "interrupting" everything and IRS are not reentrant.
+ */
+
+void ath9k_iowrite32(struct ath_hw *ah, u32 reg_offset, u32 val)
+{
+	if (ah->config.serialize_regmode == SER_REG_MODE_ON && !in_irq()) {
+		unsigned long flags;
+		spin_lock_irqsave(&ah->ah_sc->sc_serial_rw, flags);
+		iowrite32(val, ah->ah_sc->mem + reg_offset);
+		spin_unlock_irqrestore(&ah->ah_sc->sc_serial_rw, flags);
+	} else
+		iowrite32(val, ah->ah_sc->mem + reg_offset);
+}
+
+unsigned int ath9k_ioread32(struct ath_hw *ah, u32 reg_offset)
+{
+	u32 val;
+	if (ah->config.serialize_regmode == SER_REG_MODE_ON && !in_irq()) {
+		unsigned long flags;
+		spin_lock_irqsave(&ah->ah_sc->sc_serial_rw, flags);
+		val = ioread32(ah->ah_sc->mem + reg_offset);
+		spin_unlock_irqrestore(&ah->ah_sc->sc_serial_rw, flags);
+	} else
+		val = ioread32(ah->ah_sc->mem + reg_offset);
+	return val;
+}
+
 /********************/
 /* Helper Functions */
 /********************/
@@ -391,6 +421,25 @@ static void ath9k_hw_set_defaults(struct ath_hw *ah)
 	}
 
 	ah->config.intr_mitigation = 1;
+
+	/*
+	 * We need this for PCI devices only (Cardbus, PCI, miniPCI)
+	 * _and_ if on non-uniprocessor systems (Multiprocessor/HT).
+	 * This means we use it for all AR5416 devices, and the few
+	 * minor PCI AR9280 devices out there.
+	 *
+	 * Serialization is required because these devices do not handle
+	 * well the case of two concurrent reads/writes due to the latency
+	 * involved. During one read/write another read/write can be issued
+	 * on another CPU while the previous read/write may still be working
+	 * on our hardware, if we hit this case the hardware poops in a loop.
+	 * We prevent this by serializing reads and writes.
+	 *
+	 * This issue is not present on PCI-Express devices or pre-AR5416
+	 * devices (legacy, 802.11abg).
+	 */
+	if (num_possible_cpus() > 1)
+		ah->config.serialize_regmode = SER_REG_MODE_AUTO;
 }
 
 static struct ath_hw *ath9k_hw_newstate(u16 devid, struct ath_softc *sc,
diff --git a/drivers/net/wireless/ath9k/hw.h b/drivers/net/wireless/ath9k/hw.h
index 5ec416b..778137f 100644
--- a/drivers/net/wireless/ath9k/hw.h
+++ b/drivers/net/wireless/ath9k/hw.h
@@ -42,8 +42,8 @@
 #define AR5416_MAGIC		0x19641014
 
 /* Register read/write primitives */
-#define REG_WRITE(_ah, _reg, _val) iowrite32(_val, _ah->ah_sc->mem + _reg)
-#define REG_READ(_ah, _reg) ioread32(_ah->ah_sc->mem + _reg)
+#define REG_WRITE(_ah, _reg, _val) ath9k_iowrite32(_ah, _reg, _val)
+#define REG_READ(_ah, _reg) ath9k_ioread32(_ah, _reg)
 
 #define SM(_v, _f)  (((_v) << _f##_S) & _f)
 #define MS(_v, _f)  (((_v) & _f) >> _f##_S)
@@ -566,6 +566,10 @@ struct ath_hw {
 	struct ar5416IniArray iniModesTxGain;
 };
 
+/* Read and write */
+void ath9k_iowrite32(struct ath_hw *ah, u32 reg_offset, u32 val);
+unsigned int ath9k_ioread32(struct ath_hw *ah, u32 reg_offset);
+
 /* Attach, Detach, Reset */
 const char *ath9k_hw_probe(u16 vendorid, u16 devid);
 void ath9k_hw_detach(struct ath_hw *ah);
diff --git a/drivers/net/wireless/ath9k/main.c b/drivers/net/wireless/ath9k/main.c
index 1e38242..d25907d 100644
--- a/drivers/net/wireless/ath9k/main.c
+++ b/drivers/net/wireless/ath9k/main.c
@@ -1358,6 +1358,7 @@ static int ath_init(u16 devid, struct ath_softc *sc)
 		printk(KERN_ERR "Unable to create debugfs files\n");
 
 	spin_lock_init(&sc->sc_resetlock);
+	spin_lock_init(&sc->sc_serial_rw);
 	mutex_init(&sc->mutex);
 	tasklet_init(&sc->intr_tq, ath9k_tasklet, (unsigned long)sc);
 	tasklet_init(&sc->bcon_tasklet, ath9k_beacon_tasklet,
-- 
1.6.0.3

[ath9k-devel] [RFC v3 2/4] ath9k: AR9280 PCI devices must serialize IO as well

[Luis R. Rodriguez]

From: Luis R. Rodriguez <lrodriguez@atheros.com>

Signed-off-by: Luis R. Rodriguez <lrodriguez@atheros.com>
---
 drivers/net/wireless/ath9k/hw.c |    9 ++++++++-
 1 files changed, 8 insertions(+), 1 deletions(-)

diff --git a/drivers/net/wireless/ath9k/hw.c b/drivers/net/wireless/ath9k/hw.c
index 005dd51..f46aa3f 100644
--- a/drivers/net/wireless/ath9k/hw.c
+++ b/drivers/net/wireless/ath9k/hw.c
@@ -658,8 +658,15 @@ static struct ath_hw *ath9k_hw_do_attach(u16 devid, struct ath_softc *sc,
 		goto bad;
 	}
 
+	/*
+	 * All PCI devices should be put here.
+	 * XXX: remove ah->is_pciexpress and use pdev->is_pcie, then
+	 * we can just check for !pdev->is_pcie here, but
+	 * consideration must be taken for handling AHB as well.
+	 */
 	if (ah->config.serialize_regmode == SER_REG_MODE_AUTO) {
-		if (ah->hw_version.macVersion == AR_SREV_VERSION_5416_PCI) {
+		if (ah->hw_version.macVersion == AR_SREV_VERSION_5416_PCI ||
+		    (AR_SREV_9280(ah) && !ah->is_pciexpress)) {
 			ah->config.serialize_regmode =
 				SER_REG_MODE_ON;
 		} else {
-- 
1.6.0.3

[ath9k-devel] [RFC v3 3/4] ath9k: add cpu notifier to enhance device configuration

[Luis R. Rodriguez]

From: Luis R. Rodriguez <lrodriguez@atheros.com>

Upon CPU Hotplug the number of CPUs can change and we can tweak
our driver configuration as such. Keep in mind CPU hotplug also
occurs during suspend/resume.

Since we have a notifier now we can rely on num_present_cpus()
rather than num_possible_cpus() to enable/disable serialization.

Signed-off-by: Luis R. Rodriguez <lrodriguez@atheros.com>
---
 drivers/net/wireless/ath9k/ath9k.h |    6 ++++
 drivers/net/wireless/ath9k/hw.c    |   49 ++++++++++++++++++++++-------------
 drivers/net/wireless/ath9k/hw.h    |    3 ++
 drivers/net/wireless/ath9k/main.c  |   29 +++++++++++++++++++++
 drivers/net/wireless/ath9k/pci.c   |    4 +++
 5 files changed, 73 insertions(+), 18 deletions(-)

diff --git a/drivers/net/wireless/ath9k/ath9k.h b/drivers/net/wireless/ath9k/ath9k.h
index 7245797..5eaa3bc 100644
--- a/drivers/net/wireless/ath9k/ath9k.h
+++ b/drivers/net/wireless/ath9k/ath9k.h
@@ -22,6 +22,7 @@
 #include <net/mac80211.h>
 #include <linux/leds.h>
 #include <linux/rfkill.h>
+#include <linux/cpu.h>
 
 #include "hw.h"
 #include "rc.h"
@@ -566,9 +567,14 @@ struct ath_bus_ops {
 	bool		(*eeprom_read)(struct ath_hw *ah, u32 off, u16 *data);
 };
 
+int ath9k_cpu_callback(struct notifier_block *nfb,
+		       unsigned long action, void *hcpu);
+
 struct ath_softc {
 	struct ieee80211_hw *hw;
 	struct device *dev;
+	/* This should or _needs_ to be __cpuinit (?) */
+	struct notifier_block cpu_notifer;
 	struct tasklet_struct intr_tq;
 	struct tasklet_struct bcon_tasklet;
 	struct ath_hw *sc_ah;
diff --git a/drivers/net/wireless/ath9k/hw.c b/drivers/net/wireless/ath9k/hw.c
index f46aa3f..abcbe23 100644
--- a/drivers/net/wireless/ath9k/hw.c
+++ b/drivers/net/wireless/ath9k/hw.c
@@ -16,6 +16,7 @@
 
 #include <linux/io.h>
 #include <asm/unaligned.h>
+#include <linux/cpu.h>
 
 #include "ath9k.h"
 #include "initvals.h"
@@ -383,6 +384,35 @@ static const char *ath9k_hw_devname(u16 devid)
 	return NULL;
 }
 
+/*
+ * We can slap on to here things we need to tune depending
+ * on the number of CPUs. This will happen on CPU hotplug
+ * which is also used for suspend/resume.
+ */
+void ath9k_hw_config_for_cpus(struct ath_hw *ah)
+{
+	/*
+	 * We need this for PCI devices only (Cardbus, PCI, miniPCI)
+	 * _and_ if on non-uniprocessor systems (Multiprocessor/HT).
+	 * This means we use it for all AR5416 devices, and the few
+	 * minor PCI AR9280 devices out there.
+	 *
+	 * Serialization is required because these devices do not handle
+	 * well the case of two concurrent reads/writes due to the latency
+	 * involved. During one read/write another read/write can be issued
+	 * on another CPU while the previous read/write may still be working
+	 * on our hardware, if we hit this case the hardware poops in a loop.
+	 * We prevent this by serializing reads and writes.
+	 *
+	 * This issue is not present on PCI-Express devices or pre-AR5416
+	 * devices (legacy, 802.11abg).
+	 */
+	if (num_present_cpus() > 1)
+		ah->config.serialize_regmode = SER_REG_MODE_AUTO;
+	else
+		ah->config.serialize_regmode = SER_REG_MODE_OFF;
+}
+
 static void ath9k_hw_set_defaults(struct ath_hw *ah)
 {
 	int i;
@@ -422,24 +452,7 @@ static void ath9k_hw_set_defaults(struct ath_hw *ah)
 
 	ah->config.intr_mitigation = 1;
 
-	/*
-	 * We need this for PCI devices only (Cardbus, PCI, miniPCI)
-	 * _and_ if on non-uniprocessor systems (Multiprocessor/HT).
-	 * This means we use it for all AR5416 devices, and the few
-	 * minor PCI AR9280 devices out there.
-	 *
-	 * Serialization is required because these devices do not handle
-	 * well the case of two concurrent reads/writes due to the latency
-	 * involved. During one read/write another read/write can be issued
-	 * on another CPU while the previous read/write may still be working
-	 * on our hardware, if we hit this case the hardware poops in a loop.
-	 * We prevent this by serializing reads and writes.
-	 *
-	 * This issue is not present on PCI-Express devices or pre-AR5416
-	 * devices (legacy, 802.11abg).
-	 */
-	if (num_possible_cpus() > 1)
-		ah->config.serialize_regmode = SER_REG_MODE_AUTO;
+	ath9k_hw_config_for_cpus(ah);
 }
 
 static struct ath_hw *ath9k_hw_newstate(u16 devid, struct ath_softc *sc,
diff --git a/drivers/net/wireless/ath9k/hw.h b/drivers/net/wireless/ath9k/hw.h
index 778137f..778fb3c 100644
--- a/drivers/net/wireless/ath9k/hw.h
+++ b/drivers/net/wireless/ath9k/hw.h
@@ -570,6 +570,9 @@ struct ath_hw {
 void ath9k_iowrite32(struct ath_hw *ah, u32 reg_offset, u32 val);
 unsigned int ath9k_ioread32(struct ath_hw *ah, u32 reg_offset);
 
+/* Lets us tweak the device per CPU changes */
+void ath9k_hw_config_for_cpus(struct ath_hw *ah);
+
 /* Attach, Detach, Reset */
 const char *ath9k_hw_probe(u16 vendorid, u16 devid);
 void ath9k_hw_detach(struct ath_hw *ah);
diff --git a/drivers/net/wireless/ath9k/main.c b/drivers/net/wireless/ath9k/main.c
index d25907d..7823d37 100644
--- a/drivers/net/wireless/ath9k/main.c
+++ b/drivers/net/wireless/ath9k/main.c
@@ -2626,6 +2626,35 @@ struct ieee80211_ops ath9k_ops = {
 	.ampdu_action       = ath9k_ampdu_action,
 };
 
+int ath9k_cpu_callback(struct notifier_block *nfb,
+		       unsigned long action, void *hcpu)
+{
+	struct ath_softc *sc = container_of(nfb, struct ath_softc, cpu_notifer);
+	int old_serial_mode;
+
+	old_serial_mode = sc->sc_ah->config.serialize_regmode;
+
+	switch (action) {
+	case CPU_ONLINE:
+	case CPU_ONLINE_FROZEN:
+		ath9k_hw_config_for_cpus(sc->sc_ah);
+		break;
+	case CPU_DEAD:
+	case CPU_DEAD_FROZEN:
+		ath9k_hw_config_for_cpus(sc->sc_ah);
+		break;
+	}
+
+	if (unlikely(old_serial_mode != sc->sc_ah->config.serialize_regmode)) {
+		DPRINTF(sc, ATH_DBG_RESET,
+			"serialize_regmode has changed due to CPU "
+			"count to %d\n",
+			sc->sc_ah->config.serialize_regmode);
+	}
+
+	return NOTIFY_OK;
+}
+
 static struct {
 	u32 version;
 	const char * name;
diff --git a/drivers/net/wireless/ath9k/pci.c b/drivers/net/wireless/ath9k/pci.c
index eea9d3a..8711bc4 100644
--- a/drivers/net/wireless/ath9k/pci.c
+++ b/drivers/net/wireless/ath9k/pci.c
@@ -175,6 +175,9 @@ static int ath_pci_probe(struct pci_dev *pdev, const struct pci_device_id *id)
 		goto bad3;
 	}
 
+	sc->cpu_notifer.notifier_call = ath9k_cpu_callback;
+	register_hotcpu_notifier(&sc->cpu_notifer);
+
 	/* setup interrupt service routine */
 
 	if (request_irq(pdev->irq, ath_isr, IRQF_SHARED, "ath", sc)) {
@@ -216,6 +219,7 @@ static void ath_pci_remove(struct pci_dev *pdev)
 	struct ieee80211_hw *hw = pci_get_drvdata(pdev);
 	struct ath_softc *sc = hw->priv;
 
+	unregister_hotcpu_notifier(&sc->cpu_notifer);
 	ath_cleanup(sc);
 }
 
-- 
1.6.0.3

[ath9k-devel] [RFC v3 4/4] ath9k: disable D1 and D2 on PCI multicore systems

[Luis R. Rodriguez]

From: Luis R. Rodriguez <lrodriguez@atheros.com>

This fixes a hang seen on PCI devices present with AR5416.

Signed-off-by: Luis R. Rodriguez <lrodriguez@atheros.com>
---
 drivers/net/wireless/ath9k/pci.c |   42 ++++++++++++++++++++++++++++++++++++++
 1 files changed, 42 insertions(+), 0 deletions(-)

diff --git a/drivers/net/wireless/ath9k/pci.c b/drivers/net/wireless/ath9k/pci.c
index 8711bc4..c8bf28a 100644
--- a/drivers/net/wireless/ath9k/pci.c
+++ b/drivers/net/wireless/ath9k/pci.c
@@ -80,6 +80,44 @@ static struct ath_bus_ops ath_pci_bus_ops = {
 	.eeprom_read = ath_pci_eeprom_read,
 };
 
+/**
+ * ath9k_pci_disable_d1_d2 - disable D1 and D2 from PCI config space
+ * @pdev: PCI atheros device
+ *
+ * 0x41 is in vendor PCI config space, for Atheros 11n this consists
+ * of the "Power Management Capabilities and Next Item Pointer":
+ *
+ * 15:8		Next item pointer
+ * 19		PMECLK - if 0 indicates PCI clock is not required
+ * 		for PME, otherwise its set to 1.
+ * 24:22	Vaux current requirements (as per PCI spec)
+ * 25		D1 state support - 0 is not supported, 1 is supported
+ * 26		D2 state support - 0 is not supported, 1 is supported
+ * 30:27	PME power support. Indicates states from which PME
+ * 		can be generated.
+ * 		Bit 0 - PME state D0
+ * 		Bit 1 - PME state D1
+ * 		Bit 2 - PME state D2
+ * 		Bit 3 - PME state D3
+ *
+ * It seems on some systems D1 (light sleep) and D2 (deep sleep),
+ * which are optional as per the PCI spec, can create issues on SMP
+ * systems with PCI AR5416 so we disable it so later pci_pm_init()
+ * will detect them as not supported. Although this is only currently
+ * seen on AR5416 the AR9280 PCI devices are rare so we just enable
+ * this for any PCI and multiple CPU box.
+ */
+static void ath9k_pci_disable_d1_d2(struct pci_dev *pdev)
+{
+	u32 val;
+	if (pdev->is_pcie || num_possible_cpus() == 1)
+		return;
+	pci_read_config_dword(pdev, 0x41, &val);
+	val &= ~( 1 << 25 | 1 << 26);
+	pci_write_config_byte(pdev, 0x41, val);
+	printk(KERN_DEBUG "ath9k: Disabling D1 and D2\n");
+}
+
 static int ath_pci_probe(struct pci_dev *pdev, const struct pci_device_id *id)
 {
 	void __iomem *mem;
@@ -141,6 +179,8 @@ static int ath_pci_probe(struct pci_dev *pdev, const struct pci_device_id *id)
 	if ((val & 0x0000ff00) != 0)
 		pci_write_config_dword(pdev, 0x40, val & 0xffff00ff);
 
+	ath9k_pci_disable_d1_d2(pdev);
+
 	ret = pci_request_region(pdev, 0, "ath9k");
 	if (ret) {
 		dev_err(&pdev->dev, "PCI memory region reserve error\n");
@@ -264,6 +304,8 @@ static int ath_pci_resume(struct pci_dev *pdev)
 	if ((val & 0x0000ff00) != 0)
 		pci_write_config_dword(pdev, 0x40, val & 0xffff00ff);
 
+	ath9k_pci_disable_d1_d2(pdev);
+
 	/* Enable LED */
 	ath9k_hw_cfg_output(sc->sc_ah, ATH_LED_PIN,
 			    AR_GPIO_OUTPUT_MUX_AS_OUTPUT);
-- 
1.6.0.3

[ath9k-devel] [RFC v3 0/4] SMP fixes: serialization and Disablig D1 and D2

[W. van den Akker]

On Tuesday 24 February 2009 07:33:57 Luis R. Rodriguez wrote:
> From: Luis R. Rodriguez <lrodriguez@atheros.com>
>
> Ok, here's a respin of my SMP fixes series (previously just serialization).
> I got an SMP box at work now and did manage to get a PCI AR5416 card and
> did reproduce a hang.
>
> The good news is the hang is gone but I'm not sure if this will fix the
> hangs reported here with AR5416 and SMP boxen so please give it a spin and
> let me know.
>
> This also now deregisters the CPU notifier which we were not doing before.
>
> This applies on top of today's wireless-testing. You can get the full
> set of 4 patches in one file here:
>
> http://www.kernel.org/pub/linux/kernel/people/mcgrof/patches/ath9k/2009-02-
>24/serialization-v3.patch
>
> I'd like a few reports on this, please let me know.
>
> Luis R. Rodriguez (4):
>   ath9k: implement IO serialization
>   ath9k: AR9280 PCI devices must serialize IO as well
>   ath9k: add cpu notifier to enhance device configuration
>   ath9k: disable D1 and D2 on PCI multicore systems
>
>  drivers/net/wireless/ath9k/ath9k.h |    7 ++++
>  drivers/net/wireless/ath9k/hw.c    |   71
> +++++++++++++++++++++++++++++++++++- drivers/net/wireless/ath9k/hw.h    |  
> 11 +++++-
>  drivers/net/wireless/ath9k/main.c  |   30 +++++++++++++++
>  drivers/net/wireless/ath9k/pci.c   |   46 +++++++++++++++++++++++
>  5 files changed, 162 insertions(+), 3 deletions(-)
>
> _______________________________________________
> ath9k-devel mailing list
> ath9k-devel at lists.ath9k.org
> https://lists.ath9k.org/mailman/listinfo/ath9k-devel

I have applied them on the wireless-testing tree from 25-2.
Its running now.... 

To be continued....

gr,
Willem

-- 
This message has been scanned for viruses and
dangerous content by MailScanner, and is
believed to be clean.

[ath9k-devel] [RFC v3 0/4] SMP fixes: serialization and Disablig D1 and D2

[Luis R. Rodriguez]

On Thu, Feb 26, 2009 at 08:18:33AM -0800, W. van den Akker wrote:
> On Tuesday 24 February 2009 07:33:57 Luis R. Rodriguez wrote:
> > From: Luis R. Rodriguez <lrodriguez@atheros.com>
> >
> > Ok, here's a respin of my SMP fixes series (previously just serialization).
> > I got an SMP box at work now and did manage to get a PCI AR5416 card and
> > did reproduce a hang.
> >
> > The good news is the hang is gone but I'm not sure if this will fix the
> > hangs reported here with AR5416 and SMP boxen so please give it a spin and
> > let me know.
> >
> > This also now deregisters the CPU notifier which we were not doing before.
> >
> > This applies on top of today's wireless-testing. You can get the full
> > set of 4 patches in one file here:
> >
> > http://www.kernel.org/pub/linux/kernel/people/mcgrof/patches/ath9k/2009-02-
> >24/serialization-v3.patch
> >
> > I'd like a few reports on this, please let me know.
> >
> > Luis R. Rodriguez (4):
> >   ath9k: implement IO serialization
> >   ath9k: AR9280 PCI devices must serialize IO as well
> >   ath9k: add cpu notifier to enhance device configuration
> >   ath9k: disable D1 and D2 on PCI multicore systems
> >
> >  drivers/net/wireless/ath9k/ath9k.h |    7 ++++
> >  drivers/net/wireless/ath9k/hw.c    |   71
> > +++++++++++++++++++++++++++++++++++- drivers/net/wireless/ath9k/hw.h    |
> > 11 +++++-
> >  drivers/net/wireless/ath9k/main.c  |   30 +++++++++++++++
> >  drivers/net/wireless/ath9k/pci.c   |   46 +++++++++++++++++++++++
> >  5 files changed, 162 insertions(+), 3 deletions(-)
> >
> > _______________________________________________
> > ath9k-devel mailing list
> > ath9k-devel at lists.ath9k.org
> > https://lists.ath9k.org/mailman/listinfo/ath9k-devel
> 
> I have applied them on the wireless-testing tree from 25-2.
> Its running now....
> 
> To be continued....

How's it coming along with the patches.

  Luis

[ath9k-devel] [RFC v3 0/4] SMP fixes: serialization and Disablig D1 and D2

[W. van den Akker]

On Friday 27 February 2009 20:00:40 Luis R. Rodriguez wrote:
> On Thu, Feb 26, 2009 at 08:18:33AM -0800, W. van den Akker wrote:
> > On Tuesday 24 February 2009 07:33:57 Luis R. Rodriguez wrote:
> > > From: Luis R. Rodriguez <lrodriguez@atheros.com>
> > >
> > > Ok, here's a respin of my SMP fixes series (previously just
> > > serialization). I got an SMP box at work now and did manage to get a
> > > PCI AR5416 card and did reproduce a hang.
> > >
> > > The good news is the hang is gone but I'm not sure if this will fix the
> > > hangs reported here with AR5416 and SMP boxen so please give it a spin
> > > and let me know.
> > >
> > > This also now deregisters the CPU notifier which we were not doing
> > > before.
> > >
> > > This applies on top of today's wireless-testing. You can get the full
> > > set of 4 patches in one file here:
> > >
> > > http://www.kernel.org/pub/linux/kernel/people/mcgrof/patches/ath9k/2009
> > >-02- 24/serialization-v3.patch
> > >
> > > I'd like a few reports on this, please let me know.
> > >
> > > Luis R. Rodriguez (4):
> > >   ath9k: implement IO serialization
> > >   ath9k: AR9280 PCI devices must serialize IO as well
> > >   ath9k: add cpu notifier to enhance device configuration
> > >   ath9k: disable D1 and D2 on PCI multicore systems
> > >
> > >  drivers/net/wireless/ath9k/ath9k.h |    7 ++++
> > >  drivers/net/wireless/ath9k/hw.c    |   71
> > > +++++++++++++++++++++++++++++++++++- drivers/net/wireless/ath9k/hw.h   
> > > | 11 +++++-
> > >  drivers/net/wireless/ath9k/main.c  |   30 +++++++++++++++
> > >  drivers/net/wireless/ath9k/pci.c   |   46 +++++++++++++++++++++++
> > >  5 files changed, 162 insertions(+), 3 deletions(-)
> > >
> > > _______________________________________________
> > > ath9k-devel mailing list
> > > ath9k-devel at lists.ath9k.org
> > > https://lists.ath9k.org/mailman/listinfo/ath9k-devel
> >
> > I have applied them on the wireless-testing tree from 25-2.
> > Its running now....
> >
> > To be continued....
>
> How's it coming along with the patches.
>

Running since 26-2 17:00 (now 27 hours) without any problems.
Looking good :)




-- 
This message has been scanned for viruses and
dangerous content by MailScanner, and is
believed to be clean.

[ath9k-devel] [RFC v3 0/4] SMP fixes: serialization and Disablig D1 and D2

[Luis R. Rodriguez]

On Fri, Feb 27, 2009 at 11:18 AM, W. van den Akker <listsrv@wilsoft.nl> wrote:
> On Friday 27 February 2009 20:00:40 Luis R. Rodriguez wrote:
>> On Thu, Feb 26, 2009 at 08:18:33AM -0800, W. van den Akker wrote:
>> > On Tuesday 24 February 2009 07:33:57 Luis R. Rodriguez wrote:
>> > > From: Luis R. Rodriguez <lrodriguez@atheros.com>
>> > >
>> > > Ok, here's a respin of my SMP fixes series (previously just
>> > > serialization). I got an SMP box at work now and did manage to get a
>> > > PCI AR5416 card and did reproduce a hang.
>> > >
>> > > The good news is the hang is gone but I'm not sure if this will fix the
>> > > hangs reported here with AR5416 and SMP boxen so please give it a spin
>> > > and let me know.
>> > >
>> > > This also now deregisters the CPU notifier which we were not doing
>> > > before.
>> > >
>> > > This applies on top of today's wireless-testing. You can get the full
>> > > set of 4 patches in one file here:
>> > >
>> > > http://www.kernel.org/pub/linux/kernel/people/mcgrof/patches/ath9k/2009
>> > >-02- 24/serialization-v3.patch
>> > >
>> > > I'd like a few reports on this, please let me know.
>> > >
>> > > Luis R. Rodriguez (4):
>> > > ? ath9k: implement IO serialization
>> > > ? ath9k: AR9280 PCI devices must serialize IO as well
>> > > ? ath9k: add cpu notifier to enhance device configuration
>> > > ? ath9k: disable D1 and D2 on PCI multicore systems
>> > >
>> > > ?drivers/net/wireless/ath9k/ath9k.h | ? ?7 ++++
>> > > ?drivers/net/wireless/ath9k/hw.c ? ?| ? 71
>> > > +++++++++++++++++++++++++++++++++++- drivers/net/wireless/ath9k/hw.h
>> > > | 11 +++++-
>> > > ?drivers/net/wireless/ath9k/main.c ?| ? 30 +++++++++++++++
>> > > ?drivers/net/wireless/ath9k/pci.c ? | ? 46 +++++++++++++++++++++++
>> > > ?5 files changed, 162 insertions(+), 3 deletions(-)
>> > >
>> > > _______________________________________________
>> > > ath9k-devel mailing list
>> > > ath9k-devel at lists.ath9k.org
>> > > https://lists.ath9k.org/mailman/listinfo/ath9k-devel
>> >
>> > I have applied them on the wireless-testing tree from 25-2.
>> > Its running now....
>> >
>> > To be continued....
>>
>> How's it coming along with the patches.
>>
>
> Running since 26-2 17:00 (now 27 hours) without any problems.
> Looking good :)

Alright -- I just ran into a snag -- if I unload ath9k and the reload
it and actually use it I get a hang. But don't want to jinx you --
keep it running for a bit longer and if you're satisfied then try
unloading and loading to see if you also see the hang I'm seeing.

Unfortunately with my luck so far on this my SMP box doesn't have the
NMI watchdog working so I cannot troubleshoot the reload hang. Will
figure something out..

  Luis

[ath9k-devel] [RFC v3 0/4] SMP fixes: serialization and Disablig D1 and D2

[W. van den Akker]

On Friday 27 February 2009 20:18:31 W. van den Akker wrote:
> On Friday 27 February 2009 20:00:40 Luis R. Rodriguez wrote:
> > On Thu, Feb 26, 2009 at 08:18:33AM -0800, W. van den Akker wrote:
> > > On Tuesday 24 February 2009 07:33:57 Luis R. Rodriguez wrote:
> > > > From: Luis R. Rodriguez <lrodriguez@atheros.com>
> > > >
> > > > Ok, here's a respin of my SMP fixes series (previously just
> > > > serialization). I got an SMP box at work now and did manage to get a
> > > > PCI AR5416 card and did reproduce a hang.
> > > >
> > > > The good news is the hang is gone but I'm not sure if this will fix
> > > > the hangs reported here with AR5416 and SMP boxen so please give it a
> > > > spin and let me know.
> > > >
> > > > This also now deregisters the CPU notifier which we were not doing
> > > > before.
> > > >
> > > > This applies on top of today's wireless-testing. You can get the full
> > > > set of 4 patches in one file here:
> > > >
> > > > http://www.kernel.org/pub/linux/kernel/people/mcgrof/patches/ath9k/20
> > > >09 -02- 24/serialization-v3.patch
> > > >
> > > > I'd like a few reports on this, please let me know.
> > > >
> > > > Luis R. Rodriguez (4):
> > > >   ath9k: implement IO serialization
> > > >   ath9k: AR9280 PCI devices must serialize IO as well
> > > >   ath9k: add cpu notifier to enhance device configuration
> > > >   ath9k: disable D1 and D2 on PCI multicore systems
> > > >
> > > >  drivers/net/wireless/ath9k/ath9k.h |    7 ++++
> > > >  drivers/net/wireless/ath9k/hw.c    |   71
> > > > +++++++++++++++++++++++++++++++++++- drivers/net/wireless/ath9k/hw.h
> > > >
> > > > | 11 +++++-
> > > >
> > > >  drivers/net/wireless/ath9k/main.c  |   30 +++++++++++++++
> > > >  drivers/net/wireless/ath9k/pci.c   |   46 +++++++++++++++++++++++
> > > >  5 files changed, 162 insertions(+), 3 deletions(-)
> > > >
> > > > _______________________________________________
> > > > ath9k-devel mailing list
> > > > ath9k-devel at lists.ath9k.org
> > > > https://lists.ath9k.org/mailman/listinfo/ath9k-devel
> > >
> > > I have applied them on the wireless-testing tree from 25-2.
> > > Its running now....
> > >
> > > To be continued....
> >
> > How's it coming along with the patches.
>
> Running since 26-2 17:00 (now 27 hours) without any problems.
> Looking good :)

Hmmm, Houston we have a problem.

When I updated my notebook (which is connected through a wifi connection) I 
got again a hang.
I updated it with 'aptitude update' to get a updated list with packages. There 
I got the first hang. After a restart of the server an 'aptitude full-update' 
caused a second hang.
So it looks like if there is a high load on the wifi-connection it causes a 
hang.

I will test it with 'iperf' to see if I can also reproduce it.

Will report back.

Willem

-- 
This message has been scanned for viruses and
dangerous content by MailScanner, and is
believed to be clean.

[ath9k-devel] [RFC v3 0/4] SMP fixes: serialization and Disablig D1 and D2

[W. van den Akker]

On Friday 27 February 2009, W. van den Akker wrote:
> On Friday 27 February 2009 20:18:31 W. van den Akker wrote:
> > On Friday 27 February 2009 20:00:40 Luis R. Rodriguez wrote:
> > > On Thu, Feb 26, 2009 at 08:18:33AM -0800, W. van den Akker wrote:
> > > > On Tuesday 24 February 2009 07:33:57 Luis R. Rodriguez wrote:
> > > > > From: Luis R. Rodriguez <lrodriguez@atheros.com>
> > > > >
> > > > > Ok, here's a respin of my SMP fixes series (previously just
> > > > > serialization). I got an SMP box at work now and did manage to get
> > > > > a PCI AR5416 card and did reproduce a hang.
> > > > >
> > > > > The good news is the hang is gone but I'm not sure if this will fix
> > > > > the hangs reported here with AR5416 and SMP boxen so please give it
> > > > > a spin and let me know.
> > > > >
> > > > > This also now deregisters the CPU notifier which we were not doing
> > > > > before.
> > > > >
> > > > > This applies on top of today's wireless-testing. You can get the
> > > > > full set of 4 patches in one file here:
> > > > >
> > > > > http://www.kernel.org/pub/linux/kernel/people/mcgrof/patches/ath9k/
> > > > >20 09 -02- 24/serialization-v3.patch
> > > > >
> > > > > I'd like a few reports on this, please let me know.
> > > > >
> > > > > Luis R. Rodriguez (4):
> > > > >   ath9k: implement IO serialization
> > > > >   ath9k: AR9280 PCI devices must serialize IO as well
> > > > >   ath9k: add cpu notifier to enhance device configuration
> > > > >   ath9k: disable D1 and D2 on PCI multicore systems
> > > > >
> > > > >  drivers/net/wireless/ath9k/ath9k.h |    7 ++++
> > > > >  drivers/net/wireless/ath9k/hw.c    |   71
> > > > > +++++++++++++++++++++++++++++++++++-
> > > > > drivers/net/wireless/ath9k/hw.h
> > > > >
> > > > > | 11 +++++-
> > > > >
> > > > >  drivers/net/wireless/ath9k/main.c  |   30 +++++++++++++++
> > > > >  drivers/net/wireless/ath9k/pci.c   |   46 +++++++++++++++++++++++
> > > > >  5 files changed, 162 insertions(+), 3 deletions(-)
> > > > >
> > > > > _______________________________________________
> > > > > ath9k-devel mailing list
> > > > > ath9k-devel at lists.ath9k.org
> > > > > https://lists.ath9k.org/mailman/listinfo/ath9k-devel
> > > >
> > > > I have applied them on the wireless-testing tree from 25-2.
> > > > Its running now....
> > > >
> > > > To be continued....
> > >
> > > How's it coming along with the patches.
> >
> > Running since 26-2 17:00 (now 27 hours) without any problems.
> > Looking good :)
>
> Hmmm, Houston we have a problem.
>
> When I updated my notebook (which is connected through a wifi connection) I
> got again a hang.
> I updated it with 'aptitude update' to get a updated list with packages.
> There I got the first hang. After a restart of the server an 'aptitude
> full-update' caused a second hang.
> So it looks like if there is a high load on the wifi-connection it causes a
> hang.
>
> I will test it with 'iperf' to see if I can also reproduce it.
>

Did some testing with 'iperf'

Did about 20 tests whith 'iperf -c xxx'. Got 1 hang. Throughput about 12,8 
Mbps
Did about 10 tests whith 'iperf -c xx -P 2' Got 8 hangs. Thoughpunt about 6 
Mbps per session.

However not conclusive, but it looks like if there are more then 1 parallel 
sessions the server got a hangup. 

Any thoughts about this...?

gr, 
Willem

-- 
This message has been scanned for viruses and
dangerous content by MailScanner, and is
believed to be clean.

[ath9k-devel] [RFC v3 0/4] SMP fixes: serialization and Disablig D1 and D2

[Luis R. Rodriguez]

Ok, here's a respin of my SMP fixes series (previously just serialization).
I got an SMP box at work now and did manage to get a PCI AR5416 card and
did reproduce a hang.

The good news is the hang is gone but I'm not sure if this will fix the hangs
reported here with AR5416 and SMP boxen so please give it a spin and let me
know.

This also now deregisters the CPU notifier which we were not doing before.

This applies on top of today's wireless-testing. You can get the full
set of 4 patches in one file here:

http://www.kernel.org/pub/linux/kernel/people/mcgrof/patches/ath9k/2009-02-24/serialization-v3.patch

I'd like a few reports on this, please let me know.

Luis R. Rodriguez (4):
  ath9k: implement IO serialization
  ath9k: AR9280 PCI devices must serialize IO as well
  ath9k: add cpu notifier to enhance device configuration
  ath9k: disable D1 and D2 on PCI multicore systems

 drivers/net/wireless/ath9k/ath9k.h |    7 ++++
 drivers/net/wireless/ath9k/hw.c    |   71 +++++++++++++++++++++++++++++++++++-
 drivers/net/wireless/ath9k/hw.h    |   11 +++++-
 drivers/net/wireless/ath9k/main.c  |   30 +++++++++++++++
 drivers/net/wireless/ath9k/pci.c   |   46 +++++++++++++++++++++++
 5 files changed, 162 insertions(+), 3 deletions(-)