PCI: Rework config space locking, add INTx masking services

index fdaa42a..0c4c717 100644
@@ -127,20 +127,20 @@ EXPORT_SYMBOL(pci_write_vpd);
  * We have a bit per device to indicate it's blocked and a global wait queue
  * for callers to sleep on until devices are unblocked.
  */
-static DECLARE_WAIT_QUEUE_HEAD(pci_ucfg_wait);
+static DECLARE_WAIT_QUEUE_HEAD(pci_cfg_wait);
 
-static noinline void pci_wait_ucfg(struct pci_dev *dev)
+static noinline void pci_wait_cfg(struct pci_dev *dev)
 {
 	DECLARE_WAITQUEUE(wait, current);
 
-	__add_wait_queue(&pci_ucfg_wait, &wait);
+	__add_wait_queue(&pci_cfg_wait, &wait);
 	do {
 		set_current_state(TASK_UNINTERRUPTIBLE);
 		raw_spin_unlock_irq(&pci_lock);
 		schedule();
 		raw_spin_lock_irq(&pci_lock);
-	} while (dev->block_ucfg_access);
-	__remove_wait_queue(&pci_ucfg_wait, &wait);
+	} while (dev->block_cfg_access);
+	__remove_wait_queue(&pci_cfg_wait, &wait);
 }
 
 /* Returns 0 on success, negative values indicate error. */

@@ -153,7 +153,8 @@ int pci_user_read_config_##size						\
 	if (PCI_##size##_BAD)						\
 		return -EINVAL;						\
 	raw_spin_lock_irq(&pci_lock);				\
-	if (unlikely(dev->block_ucfg_access)) pci_wait_ucfg(dev);	\
+	if (unlikely(dev->block_cfg_access))				\
+		pci_wait_cfg(dev);					\
 	ret = dev->bus->ops->read(dev->bus, dev->devfn,			\
 					pos, sizeof(type), &data);	\
 	raw_spin_unlock_irq(&pci_lock);				\

@@ -172,7 +173,8 @@ int pci_user_write_config_##size					\
 	if (PCI_##size##_BAD)						\
 		return -EINVAL;						\
 	raw_spin_lock_irq(&pci_lock);				\
-	if (unlikely(dev->block_ucfg_access)) pci_wait_ucfg(dev);	\
+	if (unlikely(dev->block_cfg_access))				\
+		pci_wait_cfg(dev);					\
 	ret = dev->bus->ops->write(dev->bus, dev->devfn,		\
 					pos, sizeof(type), val);	\
 	raw_spin_unlock_irq(&pci_lock);				\

@@ -401,36 +403,56 @@ int pci_vpd_truncate(struct pci_dev *dev, size_t size)
 EXPORT_SYMBOL(pci_vpd_truncate);
 
 /**
- * pci_block_user_cfg_access - Block userspace PCI config reads/writes
+ * pci_cfg_access_lock - Lock PCI config reads/writes
  * @dev:	pci device struct
  *
- * When user access is blocked, any reads or writes to config space will
- * sleep until access is unblocked again.  We don't allow nesting of
- * block/unblock calls.
+ * When access is locked, any userspace reads or writes to config
+ * space and concurrent lock requests will sleep until access is
+ * allowed via pci_cfg_access_unlocked again.
  */
-void pci_block_user_cfg_access(struct pci_dev *dev)
+void pci_cfg_access_lock(struct pci_dev *dev)
+{
+	might_sleep();
+
+	raw_spin_lock_irq(&pci_lock);
+	if (dev->block_cfg_access)
+		pci_wait_cfg(dev);
+	dev->block_cfg_access = 1;
+	raw_spin_unlock_irq(&pci_lock);
+}
+EXPORT_SYMBOL_GPL(pci_cfg_access_lock);
+
+/**
+ * pci_cfg_access_trylock - try to lock PCI config reads/writes
+ * @dev:	pci device struct
+ *
+ * Same as pci_cfg_access_lock, but will return 0 if access is
+ * already locked, 1 otherwise. This function can be used from
+ * atomic contexts.
+ */
+bool pci_cfg_access_trylock(struct pci_dev *dev)
 {
 	unsigned long flags;
-	int was_blocked;
+	bool locked = true;
 
 	raw_spin_lock_irqsave(&pci_lock, flags);
-	was_blocked = dev->block_ucfg_access;
-	dev->block_ucfg_access = 1;
+	if (dev->block_cfg_access)
+		locked = false;
+	else
+		dev->block_cfg_access = 1;
 	raw_spin_unlock_irqrestore(&pci_lock, flags);
 
-	/* If we BUG() inside the pci_lock, we're guaranteed to hose
-	 * the machine */
-	BUG_ON(was_blocked);
+	return locked;
 }
-EXPORT_SYMBOL_GPL(pci_block_user_cfg_access);
+EXPORT_SYMBOL_GPL(pci_cfg_access_trylock);
 
 /**
- * pci_unblock_user_cfg_access - Unblock userspace PCI config reads/writes
+ * pci_cfg_access_unlock - Unlock PCI config reads/writes
  * @dev:	pci device struct
  *
- * This function allows userspace PCI config accesses to resume.
+ * This function allows PCI config accesses to resume.
  */
-void pci_unblock_user_cfg_access(struct pci_dev *dev)
+void pci_cfg_access_unlock(struct pci_dev *dev)
 {
 	unsigned long flags;
 

@@ -438,10 +460,10 @@ void pci_unblock_user_cfg_access(struct pci_dev *dev)
 
 	/* This indicates a problem in the caller, but we don't need
 	 * to kill them, unlike a double-block above. */
-	WARN_ON(!dev->block_ucfg_access);
+	WARN_ON(!dev->block_cfg_access);
 
-	dev->block_ucfg_access = 0;
-	wake_up_all(&pci_ucfg_wait);
+	dev->block_cfg_access = 0;
+	wake_up_all(&pci_cfg_wait);
 	raw_spin_unlock_irqrestore(&pci_lock, flags);
 }
-EXPORT_SYMBOL_GPL(pci_unblock_user_cfg_access);
+EXPORT_SYMBOL_GPL(pci_cfg_access_unlock);

index b82c155..c8f9aee 100644
@@ -341,10 +341,10 @@ static int sriov_enable(struct pci_dev *dev, int nr_virtfn)
 	}
 
 	iov->ctrl |= PCI_SRIOV_CTRL_VFE | PCI_SRIOV_CTRL_MSE;
-	pci_block_user_cfg_access(dev);
+	pci_cfg_access_lock(dev);
 	pci_write_config_word(dev, iov->pos + PCI_SRIOV_CTRL, iov->ctrl);
 	msleep(100);
-	pci_unblock_user_cfg_access(dev);
+	pci_cfg_access_unlock(dev);
 
 	iov->initial = initial;
 	if (nr_virtfn < initial)

@@ -372,10 +372,10 @@ failed:
 		virtfn_remove(dev, j, 0);
 
 	iov->ctrl &= ~(PCI_SRIOV_CTRL_VFE | PCI_SRIOV_CTRL_MSE);
-	pci_block_user_cfg_access(dev);
+	pci_cfg_access_lock(dev);
 	pci_write_config_word(dev, iov->pos + PCI_SRIOV_CTRL, iov->ctrl);
 	ssleep(1);
-	pci_unblock_user_cfg_access(dev);
+	pci_cfg_access_unlock(dev);
 
 	if (iov->link != dev->devfn)
 		sysfs_remove_link(&dev->dev.kobj, "dep_link");

@@ -398,10 +398,10 @@ static void sriov_disable(struct pci_dev *dev)
 		virtfn_remove(dev, i, 0);
 
 	iov->ctrl &= ~(PCI_SRIOV_CTRL_VFE | PCI_SRIOV_CTRL_MSE);
-	pci_block_user_cfg_access(dev);
+	pci_cfg_access_lock(dev);
 	pci_write_config_word(dev, iov->pos + PCI_SRIOV_CTRL, iov->ctrl);
 	ssleep(1);
-	pci_unblock_user_cfg_access(dev);
+	pci_cfg_access_unlock(dev);
 
 	if (iov->link != dev->devfn)
 		sysfs_remove_link(&dev->dev.kobj, "dep_link");

index 6f45a73..e40fdaa 100644
@@ -2958,7 +2958,7 @@ static int pci_dev_reset(struct pci_dev *dev, int probe)
 	might_sleep();
 
 	if (!probe) {
-		pci_block_user_cfg_access(dev);
+		pci_cfg_access_lock(dev);
 		/* block PM suspend, driver probe, etc. */
 		device_lock(&dev->dev);
 	}

@@ -2983,7 +2983,7 @@ static int pci_dev_reset(struct pci_dev *dev, int probe)
 done:
 	if (!probe) {
 		device_unlock(&dev->dev);
-		pci_unblock_user_cfg_access(dev);
+		pci_cfg_access_unlock(dev);
 	}
 
 	return rc;

110 lines not on this page · raw message

index ac84736..b13f9cc 100644
@@ -1387,6 +1387,7 @@ struct ipr_ioa_cfg {
 	u8 msi_received:1;
 	u8 sis64:1;
 	u8 dump_timeout:1;
+	u8 cfg_locked:1;
 
 	u8 revid;
 

index 02bd47b..56d00c62 100644
@@ -55,7 +55,8 @@ static irqreturn_t irqhandler(int irq, struct uio_info *info)
 	BUILD_BUG_ON(PCI_COMMAND % 4);
 	BUILD_BUG_ON(PCI_COMMAND + 2 != PCI_STATUS);
 
-	pci_block_user_cfg_access(pdev);
+	if (!pci_cfg_access_trylock(pdev))
+		goto error;
 
 	/* Read both command and status registers in a single 32-bit operation.
 	 * Note: we could cache the value for command and move the status read

@@ -79,7 +80,7 @@ static irqreturn_t irqhandler(int irq, struct uio_info *info)
 	ret = IRQ_HANDLED;
 done:
 
-	pci_unblock_user_cfg_access(pdev);
+	pci_cfg_access_lock(pdev);
 	return ret;
 }
 

@@ -91,7 +92,7 @@ static int __devinit verify_pci_2_3(struct pci_dev *pdev)
 	u16 orig, new;
 	int err = 0;
 
-	pci_block_user_cfg_access(pdev);
+	pci_cfg_access_lock(pdev);
 	pci_read_config_word(pdev, PCI_COMMAND, &orig);
 	pci_write_config_word(pdev, PCI_COMMAND,
 			      orig ^ PCI_COMMAND_INTX_DISABLE);

@@ -114,7 +115,7 @@ static int __devinit verify_pci_2_3(struct pci_dev *pdev)
 	/* Now restore the original value. */
 	pci_write_config_word(pdev, PCI_COMMAND, orig);
 err:
-	pci_unblock_user_cfg_access(pdev);
+	pci_cfg_access_unlock(pdev);
 	return err;
 }
 

index 7cda65b..033fe76 100644
@@ -308,7 +308,7 @@ struct pci_dev {
 	unsigned int	is_added:1;
 	unsigned int	is_busmaster:1; /* device is busmaster */
 	unsigned int	no_msi:1;	/* device may not use msi */
-	unsigned int	block_ucfg_access:1;	/* userspace config space access is blocked */
+	unsigned int	block_cfg_access:1;	/* config space access is blocked */
 	unsigned int	broken_parity_status:1;	/* Device generates false positive parity */
 	unsigned int	irq_reroute_variant:2;	/* device needs IRQ rerouting variant */
 	unsigned int 	msi_enabled:1;

@@ -1084,8 +1084,9 @@ int  ht_create_irq(struct pci_dev *dev, int idx);
 void ht_destroy_irq(unsigned int irq);
 #endif /* CONFIG_HT_IRQ */
 
-extern void pci_block_user_cfg_access(struct pci_dev *dev);
-extern void pci_unblock_user_cfg_access(struct pci_dev *dev);
+extern void pci_cfg_access_lock(struct pci_dev *dev);
+extern bool pci_cfg_access_trylock(struct pci_dev *dev);
+extern void pci_cfg_access_unlock(struct pci_dev *dev);
 
 /*
  * PCI domain support.  Sometimes called PCI segment (eg by ACPI),

@@ -1282,10 +1283,13 @@ static inline void pci_release_regions(struct pci_dev *dev)
 
 #define pci_dma_burst_advice(pdev, strat, strategy_parameter) do { } while (0)
 
-static inline void pci_block_user_cfg_access(struct pci_dev *dev)
+static inline void pci_block_cfg_access(struct pci_dev *dev)
 { }
 
-static inline void pci_unblock_user_cfg_access(struct pci_dev *dev)
+static inline int pci_block_cfg_access_in_atomic(struct pci_dev *dev)
+{ return 0; }
+
+static inline void pci_unblock_cfg_access(struct pci_dev *dev)
 { }
 
 static inline struct pci_bus *pci_find_next_bus(const struct pci_bus *from)

index 0c4c717..2a58164 100644
@@ -13,7 +13,7 @@
  * configuration space.
  */
 
-static DEFINE_RAW_SPINLOCK(pci_lock);
+DEFINE_RAW_SPINLOCK(pci_lock);
 
 /*
  *  Wrappers for all PCI configuration access functions.  They just check

116 lines not on this page · raw message

index b74084e..3b6e4ed 100644
@@ -136,6 +136,8 @@ static inline void pci_remove_legacy_files(struct pci_bus *bus) { return; }
 /* Lock for read/write access to pci device and bus lists */
 extern struct rw_semaphore pci_bus_sem;
 
+extern raw_spinlock_t pci_lock;
+
 extern unsigned int pci_pm_d3_delay;
 
 #ifdef CONFIG_PCI_MSI

index 033fe76..1cc3ac1 100644
@@ -803,6 +803,9 @@ int __must_check pci_set_mwi(struct pci_dev *dev);
 int pci_try_set_mwi(struct pci_dev *dev);
 void pci_clear_mwi(struct pci_dev *dev);
 void pci_intx(struct pci_dev *dev, int enable);
+bool pci_intx_mask_supported(struct pci_dev *dev);
+bool pci_check_and_mask_intx(struct pci_dev *dev);
+bool pci_check_and_unmask_intx(struct pci_dev *dev);
 void pci_msi_off(struct pci_dev *dev);
 int pci_set_dma_max_seg_size(struct pci_dev *dev, unsigned int size);
 int pci_set_dma_seg_boundary(struct pci_dev *dev, unsigned long mask);

index 56d00c62..0bd08ef 100644
@@ -45,78 +45,12 @@ to_uio_pci_generic_dev(struct uio_info *info)
 static irqreturn_t irqhandler(int irq, struct uio_info *info)
 {
 	struct uio_pci_generic_dev *gdev = to_uio_pci_generic_dev(info);
-	struct pci_dev *pdev = gdev->pdev;
-	irqreturn_t ret = IRQ_NONE;
-	u32 cmd_status_dword;
-	u16 origcmd, newcmd, status;
-
-	/* We do a single dword read to retrieve both command and status.
-	 * Document assumptions that make this possible. */
-	BUILD_BUG_ON(PCI_COMMAND % 4);
-	BUILD_BUG_ON(PCI_COMMAND + 2 != PCI_STATUS);
-
-	if (!pci_cfg_access_trylock(pdev))
-		goto error;
-
-	/* Read both command and status registers in a single 32-bit operation.
-	 * Note: we could cache the value for command and move the status read
-	 * out of the lock if there was a way to get notified of user changes
-	 * to command register through sysfs. Should be good for shared irqs. */
-	pci_read_config_dword(pdev, PCI_COMMAND, &cmd_status_dword);
-	origcmd = cmd_status_dword;
-	status = cmd_status_dword >> 16;
-
-	/* Check interrupt status register to see whether our device
-	 * triggered the interrupt. */
-	if (!(status & PCI_STATUS_INTERRUPT))
-		goto done;
-
-	/* We triggered the interrupt, disable it. */
-	newcmd = origcmd | PCI_COMMAND_INTX_DISABLE;
-	if (newcmd != origcmd)
-		pci_write_config_word(pdev, PCI_COMMAND, newcmd);
 
-	/* UIO core will signal the user process. */
-	ret = IRQ_HANDLED;
-done:
-
-	pci_cfg_access_lock(pdev);
-	return ret;
-}
+	if (!pci_check_and_mask_intx(gdev->pdev))
+		return IRQ_NONE;
 
-/* Verify that the device supports Interrupt Disable bit in command register,
- * per PCI 2.3, by flipping this bit and reading it back: this bit was readonly
- * in PCI 2.2. */
-static int __devinit verify_pci_2_3(struct pci_dev *pdev)
-{
-	u16 orig, new;
-	int err = 0;
-
-	pci_cfg_access_lock(pdev);
-	pci_read_config_word(pdev, PCI_COMMAND, &orig);
-	pci_write_config_word(pdev, PCI_COMMAND,
-			      orig ^ PCI_COMMAND_INTX_DISABLE);
-	pci_read_config_word(pdev, PCI_COMMAND, &new);
-	/* There's no way to protect against
-	 * hardware bugs or detect them reliably, but as long as we know
-	 * what the value should be, let's go ahead and check it. */
-	if ((new ^ orig) & ~PCI_COMMAND_INTX_DISABLE) {
-		err = -EBUSY;
-		dev_err(&pdev->dev, "Command changed from 0x%x to 0x%x: "
-			"driver or HW bug?\n", orig, new);
-		goto err;
-	}
-	if (!((new ^ orig) & PCI_COMMAND_INTX_DISABLE)) {
-		dev_warn(&pdev->dev, "Device does not support "
-			 "disabling interrupts: unable to bind.\n");
-		err = -ENODEV;
-		goto err;
-	}
-	/* Now restore the original value. */
-	pci_write_config_word(pdev, PCI_COMMAND, orig);
-err:
-	pci_cfg_access_unlock(pdev);
-	return err;
+	/* UIO core will signal the user process. */
+	return IRQ_HANDLED;
 }
 
 static int __devinit probe(struct pci_dev *pdev,

@@ -139,9 +73,10 @@ static int __devinit probe(struct pci_dev *pdev,
 		return -ENODEV;
 	}
 
-	err = verify_pci_2_3(pdev);
-	if (err)
+	if (!pci_intx_mask_supported(pdev)) {
+		err = -ENODEV;
 		goto err_verify;
+	}
 
 	gdev = kzalloc(sizeof(struct uio_pci_generic_dev), GFP_KERNEL);
 	if (!gdev) {