Skip to contentSkip to conversation

net: Assorted fixes

PATCH New Anton Vorontsov 2 participants 6 messages Dec 30, 2009 – Dec 31, 2009

Subsystems (3)
lore
Share & tools
lore
#0 lore raw
Quoted 7 lines · David Miller

Actually I'm reverting both patches:

drivers/net/phy/mdio_bus.c:268: warning: ‘mdio_bus_phy_may_suspend’ defined but not used

Please resubmit these two patches with this warning fixed.

CONFIG_PM is not always enabled :-)

Ugh, sorry about that, and thanks for catching!

Here are the amended patches + one more fix.

Thanks,

Signature · 7 lines
Anton Vorontsov
email: [email protected]
irc://irc.freenode.net/bd2
_______________________________________________
Linuxppc-dev mailing list
[email protected]
https://lists.ozlabs.org/listinfo/linuxppc-dev
#1 lore raw

PATCH 1/4 phylib: Fix deadlock on resume

Sometimes kernel hangs on resume with the following trace:

ucc_geth e0102000.ucc: resume

 INFO: task bash:1764 blocked for more than 120 seconds.
 "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message.
 bash          D 0fecf43c     0  1764   1763 0x00000000
 Call Trace:
 [cf9a7c10] [c0012868] ret_from_except+0x0/0x14 (unreliable)
 --- Exception: cf9a7ce0 at __switch_to+0x4c/0x6c
     LR = 0xcf9a7cc0
 [cf9a7cd0] [c0008c14] __switch_to+0x4c/0x6c (unreliable)
 [cf9a7ce0] [c028bcfc] schedule+0x158/0x260
 [cf9a7d10] [c028c720] __mutex_lock_slowpath+0x80/0xd8
 [cf9a7d40] [c01cf388] phy_stop+0x20/0x70
 [cf9a7d50] [c01d514c] ugeth_resume+0x6c/0x13c
 [...]

Here is why.

On suspend:

- PM core starts suspending devices, ucc_geth_suspend gets called;

- ucc_geth calls phy_stop() on suspend. Note that phy_stop() is mostly asynchronous so it doesn't block ucc_geth's suspend routine, it just sets PHY_HALTED state and disables PHY's interrupts;

- Suddenly the state machine gets scheduled, it grabs the phydev->lock mutex and tries to process the PHY_HALTED state, so it calls phydev->adjust_link(phydev->attached_dev). In ucc_geth case adjust_link() calls msleep(), which reschedules the code flow back to PM core, which now finishes suspend and so we end up sleeping with phydev->lock mutex held.

On resume:

- PM core starts resuming devices (notice that nobody rescheduled the state machine yet, so the mutex is still held), the core calls ucc_geth's resume routine;

- ucc_geth_resume restarts the PHY with phy_stop()/phy_start() sequence, and the phy_*() calls are trying to grab the phydev->lock mutex. Here comes the deadlock.

This patch fixes the issue by stopping the state machine on suspend and starting it again on resume.

---
 drivers/net/phy/mdio_bus.c |   24 ++++++++++++++++++++++--
 1 files changed, 22 insertions(+), 2 deletions(-)
drivers/net/phy/mdio_bus.c +22 -2
index bd4e8d7..49252d3 100644
@@ -303,8 +303,18 @@ static int mdio_bus_suspend(struct device * dev, pm_message_t state)
 	struct phy_driver *phydrv = to_phy_driver(dev->driver);
 	struct phy_device *phydev = to_phy_device(dev);
 
+	/*
+	 * We must stop the state machine manually, otherwise it stops out of
+	 * control, possibly with the phydev->lock held. Upon resume, netdev
+	 * may call phy routines that try to grab the same lock, and that may
+	 * lead to a deadlock.
+	 */
+	if (phydev->attached_dev)
+		phy_stop_machine(phydev);
+
 	if (!mdio_bus_phy_may_suspend(phydev))
 		return 0;
+
 	return phydrv->suspend(phydev);
 }
 
@@ -312,10 +322,20 @@ static int mdio_bus_resume(struct device * dev)
 {
 	struct phy_driver *phydrv = to_phy_driver(dev->driver);
 	struct phy_device *phydev = to_phy_device(dev);
+	int ret;
 
 	if (!mdio_bus_phy_may_suspend(phydev))
-		return 0;
-	return phydrv->resume(phydev);
+		goto no_resume;
+
+	ret = phydrv->resume(phydev);
+	if (ret < 0)
+		return ret;
+
+no_resume:
+	if (phydev->attached_dev)
+		phy_start_machine(phydev, NULL);
+
+	return 0;
 }
 
 struct bus_type mdio_bus_type = {
Signature · 1 line
1.6.5.7
#2 lore raw

PATCH 2/4 phylib: Properly reinitialize PHYs after hibernation

In reply to Anton Vorontsov

Since hibernation assumes power loss, we should fully reinitialize PHYs (including platform fixups), as if PHYs were just attached.

This patch factors phy_init_hw() out of phy_attach_direct(), then converts mdio_bus to dev_pm_ops and adds an appropriate restore() callback.

---
 drivers/net/phy/mdio_bus.c   |   50 ++++++++++++++++++++++++++++++++++++------
 drivers/net/phy/phy_device.c |   30 ++++++++++++------------
 include/linux/phy.h          |    1 +
 3 files changed, 59 insertions(+), 22 deletions(-)
drivers/net/phy/mdio_bus.c +43 -7
index 49252d3..e17b702 100644
@@ -264,6 +264,8 @@ static int mdio_bus_match(struct device *dev, struct device_driver *drv)
 		(phydev->phy_id & phydrv->phy_id_mask));
 }
 
+#ifdef CONFIG_PM
+
 static bool mdio_bus_phy_may_suspend(struct phy_device *phydev)
 {
 	struct device_driver *drv = phydev->dev.driver;
@@ -295,10 +297,7 @@ static bool mdio_bus_phy_may_suspend(struct phy_device *phydev)
 	return true;
 }
 
-/* Suspend and resume.  Copied from platform_suspend and
- * platform_resume
- */
-static int mdio_bus_suspend(struct device * dev, pm_message_t state)
+static int mdio_bus_suspend(struct device *dev)
 {
 	struct phy_driver *phydrv = to_phy_driver(dev->driver);
 	struct phy_device *phydev = to_phy_device(dev);
@@ -318,7 +317,7 @@ static int mdio_bus_suspend(struct device * dev, pm_message_t state)
 	return phydrv->suspend(phydev);
 }
 
-static int mdio_bus_resume(struct device * dev)
+static int mdio_bus_resume(struct device *dev)
 {
 	struct phy_driver *phydrv = to_phy_driver(dev->driver);
 	struct phy_device *phydev = to_phy_device(dev);
@@ -338,11 +337,48 @@ no_resume:
 	return 0;
 }
 
+static int mdio_bus_restore(struct device *dev)
+{
+	struct phy_device *phydev = to_phy_device(dev);
+	struct net_device *netdev = phydev->attached_dev;
+	int ret;
+
+	if (!netdev)
+		return 0;
+
+	ret = phy_init_hw(phydev);
+	if (ret < 0)
+		return ret;
+
+	/* The PHY needs to renegotiate. */
+	phydev->link = 0;
+	phydev->state = PHY_UP;
+
+	phy_start_machine(phydev, NULL);
+
+	return 0;
+}
+
+static struct dev_pm_ops mdio_bus_pm_ops = {
+	.suspend = mdio_bus_suspend,
+	.resume = mdio_bus_resume,
+	.freeze = mdio_bus_suspend,
+	.thaw = mdio_bus_resume,
+	.restore = mdio_bus_restore,
+};
+
+#define MDIO_BUS_PM_OPS (&mdio_bus_pm_ops)
+
+#else
+
+#define MDIO_BUS_PM_OPS NULL
+
+#endif /* CONFIG_PM */
+
 struct bus_type mdio_bus_type = {
 	.name		= "mdio_bus",
 	.match		= mdio_bus_match,
-	.suspend	= mdio_bus_suspend,
-	.resume		= mdio_bus_resume,
+	.pm		= MDIO_BUS_PM_OPS,
 };
 EXPORT_SYMBOL(mdio_bus_type);
 
drivers/net/phy/phy_device.c +15 -15
index b10fedd..8212b2b 100644
@@ -378,6 +378,20 @@ void phy_disconnect(struct phy_device *phydev)
 }
 EXPORT_SYMBOL(phy_disconnect);
 
+int phy_init_hw(struct phy_device *phydev)
+{
+	int ret;
+
+	if (!phydev->drv || !phydev->drv->config_init)
+		return 0;
+
+	ret = phy_scan_fixups(phydev);
+	if (ret < 0)
+		return ret;
+
+	return phydev->drv->config_init(phydev);
+}
+
 /**
  * phy_attach_direct - attach a network device to a given PHY device pointer
  * @dev: network device to attach
@@ -425,21 +439,7 @@ int phy_attach_direct(struct net_device *dev, struct phy_device *phydev,
 	/* Do initial configuration here, now that
 	 * we have certain key parameters
 	 * (dev_flags and interface) */
-	if (phydev->drv->config_init) {
-		int err;
-
-		err = phy_scan_fixups(phydev);
-
-		if (err < 0)
-			return err;
-
-		err = phydev->drv->config_init(phydev);
-
-		if (err < 0)
-			return err;
-	}
-
-	return 0;
+	return phy_init_hw(phydev);
 }
 EXPORT_SYMBOL(phy_attach_direct);
 
include/linux/phy.h +1 -0
index b1368b8..7968def 100644
@@ -447,6 +447,7 @@ struct phy_device* get_phy_device(struct mii_bus *bus, int addr);
 int phy_device_register(struct phy_device *phy);
 int phy_clear_interrupt(struct phy_device *phydev);
 int phy_config_interrupt(struct phy_device *phydev, u32 interrupts);
+int phy_init_hw(struct phy_device *phydev);
 int phy_attach_direct(struct net_device *dev, struct phy_device *phydev,
 		u32 flags, phy_interface_t interface);
 struct phy_device * phy_attach(struct net_device *dev,
Signature · 1 line
1.6.5.7
#3 lore raw

PATCH 3/4 ucc_geth: Fix netdev watchdog triggering on suspend

In reply to Anton Vorontsov

Sometimes ucc_geth fails to suspend with the following trace:

ucc_geth e0103000.ucc: suspend
ucc_geth e0102000.ucc: suspend
NETDEV WATCHDOG: eth0 (ucc_geth): transmit queue 0 timed out

 ------------[ cut here ]------------
 Badness at net/sched/sch_generic.c:255
 NIP: c021cb5c LR: c021cb5c CTR: c01ab4b4
 [...]
 NIP [c021cb5c] dev_watchdog+0x298/0x2a8
 LR [c021cb5c] dev_watchdog+0x298/0x2a8
 Call Trace:
 [c0389da0] [c021cb5c] dev_watchdog+0x298/0x2a8 (unreliable)
 [c0389e00] [c0031ed8] run_timer_softirq+0x16c/0x1dc
 [c0389e50] [c002c638] __do_softirq+0xa4/0x11c
 [...]

This patch fixes the issue by properly detaching the device on suspend, and attaching it back on resume.

---
 drivers/net/ucc_geth.c |    3 ++-
 1 files changed, 2 insertions(+), 1 deletions(-)
drivers/net/ucc_geth.c +2 -1
index 41ad2f3..96bdc0b 100644
@@ -3607,6 +3607,7 @@ static int ucc_geth_suspend(struct of_device *ofdev, pm_message_t state)
 	if (!netif_running(ndev))
 		return 0;
 
+	netif_device_detach(ndev);
 	napi_disable(&ugeth->napi);
 
 	/*
@@ -3665,7 +3666,7 @@ static int ucc_geth_resume(struct of_device *ofdev)
 	phy_start(ugeth->phydev);
 
 	napi_enable(&ugeth->napi);
-	netif_start_queue(ndev);
+	netif_device_attach(ndev);
 
 	return 0;
 }
Signature · 1 line
1.6.5.7
#4 lore raw

PATCH 4/4 fsl_pq_mdio: Fix iomem unmapping for non-eTSEC2.0 controllers

In reply to Anton Vorontsov

We use a rather complicated logic to support eTSEC and eTSEC2.0 registers maps in a single driver. Currently, the code tries to unmap 'regs', but for non-eTSEC2.0 controllers 'regs' doesn't point to a mapping start, and this might cause badness on probe failure or module removal:

Freescale PowerQUICC MII Bus: probed
Trying to vfree() nonexistent vm area (e107f000)

 ------------[ cut here ]------------
 Badness at c00a7754 [verbose debug info unavailable]
 NIP: c00a7754 LR: c00a7754 CTR: c02231ec
 [...]
 NIP [c00a7754] __vunmap+0xec/0xf4
 LR [c00a7754] __vunmap+0xec/0xf4
 Call Trace:
 [df827e50] [c00a7754] __vunmap+0xec/0xf4 (unreliable)
 [df827e70] [c001519c] iounmap+0x44/0x54
 [df827e80] [c028b924] fsl_pq_mdio_probe+0x1cc/0x2fc
 [df827eb0] [c02fb9b4] of_platform_device_probe+0x5c/0x84
 [df827ed0] [c0229928] really_probe+0x78/0x1a8
 [df827ef0] [c0229b20] __driver_attach+0xa4/0xa8

Fix this by introducing a proper priv structure (finally!), which now holds 'regs' and 'map' fields separately.

---
 drivers/net/fsl_pq_mdio.c |   30 +++++++++++++++++++++++-------
 1 files changed, 23 insertions(+), 7 deletions(-)
drivers/net/fsl_pq_mdio.c +23 -7
index 25fabb3..d5160ed 100644
@@ -46,6 +46,11 @@
 #include "gianfar.h"
 #include "fsl_pq_mdio.h"
 
+struct fsl_pq_mdio_priv {
+	void __iomem *map;
+	struct fsl_pq_mdio __iomem *regs;
+};
+
 /*
  * Write value to the PHY at mii_id at register regnum,
  * on the bus attached to the local interface, which may be different from the
@@ -105,7 +110,9 @@ int fsl_pq_local_mdio_read(struct fsl_pq_mdio __iomem *regs,
 
 static struct fsl_pq_mdio __iomem *fsl_pq_mdio_get_regs(struct mii_bus *bus)
 {
-	return (void __iomem __force *)bus->priv;
+	struct fsl_pq_mdio_priv *priv = bus->priv;
+
+	return priv->regs;
 }
 
 /*
@@ -266,6 +273,7 @@ static int fsl_pq_mdio_probe(struct of_device *ofdev,
 {
 	struct device_node *np = ofdev->node;
 	struct device_node *tbi;
+	struct fsl_pq_mdio_priv *priv;
 	struct fsl_pq_mdio __iomem *regs = NULL;
 	void __iomem *map;
 	u32 __iomem *tbipa;
@@ -274,14 +282,19 @@ static int fsl_pq_mdio_probe(struct of_device *ofdev,
 	u64 addr = 0, size = 0;
 	int err = 0;
 
+	priv = kzalloc(sizeof(*priv), GFP_KERNEL);
+	if (!priv)
+		return -ENOMEM;
+
 	new_bus = mdiobus_alloc();
 	if (NULL == new_bus)
-		return -ENOMEM;
+		goto err_free_priv;
 
 	new_bus->name = "Freescale PowerQUICC MII Bus",
 	new_bus->read = &fsl_pq_mdio_read,
 	new_bus->write = &fsl_pq_mdio_write,
 	new_bus->reset = &fsl_pq_mdio_reset,
+	new_bus->priv = priv;
 	fsl_pq_mdio_bus_name(new_bus->id, np);
 
 	/* Set the PHY base address */
@@ -291,6 +304,7 @@ static int fsl_pq_mdio_probe(struct of_device *ofdev,
 		err = -ENOMEM;
 		goto err_free_bus;
 	}
+	priv->map = map;
 
 	if (of_device_is_compatible(np, "fsl,gianfar-mdio") ||
 			of_device_is_compatible(np, "fsl,gianfar-tbi") ||
@@ -298,8 +312,7 @@ static int fsl_pq_mdio_probe(struct of_device *ofdev,
 			of_device_is_compatible(np, "ucc_geth_phy"))
 		map -= offsetof(struct fsl_pq_mdio, miimcfg);
 	regs = map;
-
-	new_bus->priv = (void __force *)regs;
+	priv->regs = regs;
 
 	new_bus->irq = kcalloc(PHY_MAX_ADDR, sizeof(int), GFP_KERNEL);
 
@@ -392,10 +405,11 @@ static int fsl_pq_mdio_probe(struct of_device *ofdev,
 err_free_irqs:
 	kfree(new_bus->irq);
 err_unmap_regs:
-	iounmap(regs);
+	iounmap(priv->map);
 err_free_bus:
 	kfree(new_bus);
-
+err_free_priv:
+	kfree(priv);
 	return err;
 }
 
@@ -404,14 +418,16 @@ static int fsl_pq_mdio_remove(struct of_device *ofdev)
 {
 	struct device *device = &ofdev->dev;
 	struct mii_bus *bus = dev_get_drvdata(device);
+	struct fsl_pq_mdio_priv *priv = bus->priv;
 
 	mdiobus_unregister(bus);
 
 	dev_set_drvdata(device, NULL);
 
-	iounmap(fsl_pq_mdio_get_regs(bus));
+	iounmap(priv->map);
 	bus->priv = NULL;
 	mdiobus_free(bus);
+	kfree(priv);
 
 	return 0;
 }
Signature · 1 line
1.6.5.7
maintainer
#5 lore raw
In reply to Anton Vorontsov

From: Anton Vorontsov <[email protected]>
Date: Wed, 30 Dec 2009 21:21:46 +0300

Quoted 5 lines · David Miller
Quoted 4 lines

Actually I'm reverting both patches:

drivers/net/phy/mdio_bus.c:268: warning: ‘mdio_bus_phy_may_suspend’ defined but not used

Quoted 7 lines
Quoted 3 lines

Please resubmit these two patches with this warning fixed.

CONFIG_PM is not always enabled :-)

Ugh, sorry about that, and thanks for catching!

Here are the amended patches + one more fix.

All applied, thanks.