adding inode dirty tracing....

2 files changed, 407 insertions, 0 deletions

diff --git a/patches/linux-2.6.37-rc1-ahci-alpm-accounting.patch b/patches/linux-2.6.37-rc1-ahci-alpm-accounting.patch
new file mode 100644
index 0000000..bd451e2
--- /dev/null
+++ b/patches/linux-2.6.37-rc1-ahci-alpm-accounting.patch
@@ -0,0 +1,302 @@
+From: Arjan van de Ven <arjan@linux.intel.com>
+Subject: [PATCH] libata: Add ALPM power state accounting to the AHCI driver
+
+PowerTOP wants to be able to show the user how effective the ALPM link
+power management is for the user. ALPM is worth around 0.5W on a quiet
+link; PowerTOP wants to be able to find cases where the "quiet link" isn't
+actually quiet.
+
+This patch adds state accounting functionality to the AHCI driver for
+PowerTOP to use.
+The parts of the patch are
+1) the sysfs logic of exposing the stats for each state in sysfs
+2) the basic accounting logic that gets update on link change interrupts
+   (or when the user accesses the info from sysfs)
+3) a "accounting enable" flag; in order to get the accounting to work,
+   the driver needs to get phyrdy interrupts on link status changes.
+   Normally and currently this is disabled by the driver when ALPM is
+   on (to reduce overhead); when PowerTOP is running this will need
+   to be on to get usable statistics... hence the sysfs tunable.
+
+The PowerTOP output currently looks like this:
+
+Recent SATA AHCI link activity statistics
+Active	Partial	Slumber	Device name
+  0.5%	 99.5%	  0.0%	host0
+
+(work to resolve "host0" to a more human readable name is in progress)
+
+Signed-off-by: Arjan van de Ven <arjan@linux.intel.com>
+
+---
+ drivers/ata/ahci.h    |   15 ++++
+ drivers/ata/libahci.c |  187 +++++++++++++++++++++++++++++++++++++++++++++++++-
+ 2 files changed, 200 insertions(+), 2 deletions(-)
+
+Index: linux.trees.git/drivers/ata/ahci.h
+===================================================================
+--- linux.trees.git.orig/drivers/ata/ahci.h
++++ linux.trees.git/drivers/ata/ahci.h
+@@ -262,6 +262,13 @@ struct ahci_em_priv {
+ 	unsigned long led_state;
+ };
+ 
++enum ahci_port_states {
++	AHCI_PORT_NOLINK = 0,
++	AHCI_PORT_ACTIVE = 1,
++	AHCI_PORT_PARTIAL = 2,
++	AHCI_PORT_SLUMBER = 3
++};
++
+ struct ahci_port_priv {
+ 	struct ata_link		*active_link;
+ 	struct ahci_cmd_hdr	*cmd_slot;
+@@ -280,6 +287,14 @@ struct ahci_port_priv {
+ 	int			fbs_last_dev;	/* save FBS.DEV of last FIS */
+ 	/* enclosure management info per PM slot */
+ 	struct ahci_em_priv	em_priv[EM_MAX_SLOTS];
++
++	/* ALPM accounting state and stats */
++	unsigned int 		accounting_active:1;
++	u64			active_jiffies;
++	u64			partial_jiffies;
++	u64			slumber_jiffies;
++	int			previous_state;
++	int			previous_jiffies;
+ };
+ 
+ struct ahci_host_priv {
+Index: linux.trees.git/drivers/ata/libahci.c
+===================================================================
+--- linux.trees.git.orig/drivers/ata/libahci.c
++++ linux.trees.git/drivers/ata/libahci.c
+@@ -58,6 +58,17 @@ MODULE_PARM_DESC(ignore_sss, "Ignore sta
+ 
+ static int ahci_set_lpm(struct ata_link *link, enum ata_lpm_policy policy,
+ 			unsigned hints);
++static ssize_t ahci_alpm_show_active(struct device *dev,
++				  struct device_attribute *attr, char *buf);
++static ssize_t ahci_alpm_show_slumber(struct device *dev,
++				  struct device_attribute *attr, char *buf);
++static ssize_t ahci_alpm_show_partial(struct device *dev,
++				  struct device_attribute *attr, char *buf);
++static ssize_t ahci_alpm_show_accounting(struct device *dev,
++				  struct device_attribute *attr, char *buf);
++static ssize_t ahci_alpm_set_accounting(struct device *dev,
++				  struct device_attribute *attr,
++				  const char *buf, size_t count);
+ static ssize_t ahci_led_show(struct ata_port *ap, char *buf);
+ static ssize_t ahci_led_store(struct ata_port *ap, const char *buf,
+ 			      size_t size);
+@@ -117,6 +128,12 @@ static DEVICE_ATTR(ahci_host_caps, S_IRU
+ static DEVICE_ATTR(ahci_host_cap2, S_IRUGO, ahci_show_host_cap2, NULL);
+ static DEVICE_ATTR(ahci_host_version, S_IRUGO, ahci_show_host_version, NULL);
+ static DEVICE_ATTR(ahci_port_cmd, S_IRUGO, ahci_show_port_cmd, NULL);
++static DEVICE_ATTR(ahci_alpm_active, S_IRUGO, ahci_alpm_show_active, NULL);
++static DEVICE_ATTR(ahci_alpm_partial, S_IRUGO, ahci_alpm_show_partial, NULL);
++static DEVICE_ATTR(ahci_alpm_slumber, S_IRUGO, ahci_alpm_show_slumber, NULL);
++static DEVICE_ATTR(ahci_alpm_accounting, S_IRUGO | S_IWUSR,
++		ahci_alpm_show_accounting, ahci_alpm_set_accounting);
++
+ static DEVICE_ATTR(em_buffer, S_IWUSR | S_IRUGO,
+ 		   ahci_read_em_buffer, ahci_store_em_buffer);
+ 
+@@ -128,6 +145,10 @@ struct device_attribute *ahci_shost_attr
+ 	&dev_attr_ahci_host_cap2,
+ 	&dev_attr_ahci_host_version,
+ 	&dev_attr_ahci_port_cmd,
++	&dev_attr_ahci_alpm_active,
++	&dev_attr_ahci_alpm_partial,
++	&dev_attr_ahci_alpm_slumber,
++	&dev_attr_ahci_alpm_accounting,
+ 	&dev_attr_em_buffer,
+ 	NULL
+ };
+@@ -653,9 +674,14 @@ static int ahci_set_lpm(struct ata_link 
+ 		 * Disable interrupts on Phy Ready. This keeps us from
+ 		 * getting woken up due to spurious phy ready
+ 		 * interrupts.
++		 *
++		 * However, when accounting_active is set, we do want
++		 * the interrupts for accounting purposes.
+ 		 */
+-		pp->intr_mask &= ~PORT_IRQ_PHYRDY;
+-		writel(pp->intr_mask, port_mmio + PORT_IRQ_MASK);
++		if (!pp->accounting_active) {
++			pp->intr_mask &= ~PORT_IRQ_PHYRDY;
++			writel(pp->intr_mask, port_mmio + PORT_IRQ_MASK);
++		}
+ 
+ 		sata_link_scr_lpm(link, policy, false);
+ 	}
+@@ -1570,6 +1596,162 @@ static void ahci_error_intr(struct ata_p
+ 		ata_port_abort(ap);
+ }
+ 
++static int get_current_alpm_state(struct ata_port *ap)
++{
++	u32 status = 0;
++
++	ahci_scr_read(&ap->link, SCR_STATUS, &status);
++
++	/* link status is in bits 11-8 */
++	status = status >> 8;
++	status = status & 0x7;
++
++	if (status == 6)
++		return AHCI_PORT_SLUMBER;
++	if (status == 2)
++		return AHCI_PORT_PARTIAL;
++	if (status == 1)
++		return AHCI_PORT_ACTIVE;
++	return AHCI_PORT_NOLINK;
++}
++
++static void account_alpm_stats(struct ata_port *ap)
++{
++	struct ahci_port_priv *pp;
++
++	int new_state;
++	u64 new_jiffies, jiffies_delta;
++
++	if (ap == NULL)
++		return;
++	pp = ap->private_data;
++
++	if (!pp) return;
++
++	new_state = get_current_alpm_state(ap);
++	new_jiffies = jiffies;
++
++	jiffies_delta = new_jiffies - pp->previous_jiffies;
++
++	switch (pp->previous_state) {
++	case AHCI_PORT_NOLINK:
++		pp->active_jiffies = 0;
++		pp->partial_jiffies = 0;
++		pp->slumber_jiffies = 0;
++		break;
++	case AHCI_PORT_ACTIVE:
++		pp->active_jiffies += jiffies_delta;
++		break;
++	case AHCI_PORT_PARTIAL:
++		pp->partial_jiffies += jiffies_delta;
++		break;
++	case AHCI_PORT_SLUMBER:
++		pp->slumber_jiffies += jiffies_delta;
++		break;
++	default:
++		break;
++	}
++	pp->previous_state = new_state;
++	pp->previous_jiffies = new_jiffies;
++}
++
++static ssize_t ahci_alpm_show_active(struct device *dev,
++				   struct device_attribute *attr, char *buf)
++{
++	struct Scsi_Host *shost = class_to_shost(dev);
++	struct ata_port *ap = ata_shost_to_port(shost);
++	struct ahci_port_priv *pp;
++
++	if (!ap || ata_port_is_dummy(ap))
++		return -EINVAL;
++	pp = ap->private_data;
++	account_alpm_stats(ap);
++
++	return sprintf(buf, "%u\n", jiffies_to_msecs(pp->active_jiffies));
++}
++
++static ssize_t ahci_alpm_show_partial(struct device *dev,
++				   struct device_attribute *attr, char *buf)
++{
++	struct Scsi_Host *shost = class_to_shost(dev);
++	struct ata_port *ap = ata_shost_to_port(shost);
++	struct ahci_port_priv *pp;
++
++	if (!ap || ata_port_is_dummy(ap))
++		return -EINVAL;
++
++	pp = ap->private_data;
++	account_alpm_stats(ap);
++
++	return sprintf(buf, "%u\n", jiffies_to_msecs(pp->partial_jiffies));
++}
++
++static ssize_t ahci_alpm_show_slumber(struct device *dev,
++				   struct device_attribute *attr, char *buf)
++{
++	struct Scsi_Host *shost = class_to_shost(dev);
++	struct ata_port *ap = ata_shost_to_port(shost);
++	struct ahci_port_priv *pp;
++
++	if (!ap || ata_port_is_dummy(ap))
++		return -EINVAL;
++
++	pp = ap->private_data;
++
++	account_alpm_stats(ap);
++
++	return sprintf(buf, "%u\n", jiffies_to_msecs(pp->slumber_jiffies));
++}
++
++static ssize_t ahci_alpm_show_accounting(struct device *dev,
++				  struct device_attribute *attr, char *buf)
++{
++	struct Scsi_Host *shost = class_to_shost(dev);
++	struct ata_port *ap = ata_shost_to_port(shost);
++	struct ahci_port_priv *pp;
++
++	if (!ap || ata_port_is_dummy(ap))
++		return -EINVAL;
++
++	pp = ap->private_data;
++
++	return sprintf(buf, "%u\n", pp->accounting_active);
++}
++
++static ssize_t ahci_alpm_set_accounting(struct device *dev,
++				  struct device_attribute *attr,
++				  const char *buf, size_t count)
++{
++	unsigned long flags;
++	struct Scsi_Host *shost = class_to_shost(dev);
++	struct ata_port *ap = ata_shost_to_port(shost);
++	struct ahci_port_priv *pp;
++	void __iomem *port_mmio;
++
++	if (!ap || ata_port_is_dummy(ap))
++		return 1;
++
++	pp = ap->private_data;
++	port_mmio = ahci_port_base(ap);
++
++	if (!pp)
++		return 1;
++	if (buf[0] == '0')
++		pp->accounting_active = 0;
++	if (buf[0] == '1')
++		pp->accounting_active = 1;
++
++	/* we need to enable the PHYRDY interrupt when we want accounting */
++	if (pp->accounting_active) {
++		spin_lock_irqsave(ap->lock, flags);
++		pp->intr_mask |= PORT_IRQ_PHYRDY;
++		writel(pp->intr_mask, port_mmio + PORT_IRQ_MASK);
++		spin_unlock_irqrestore(ap->lock, flags);
++	}
++	return count;
++}
++
++
+ static void ahci_port_intr(struct ata_port *ap)
+ {
+ 	void __iomem *port_mmio = ahci_port_base(ap);
+@@ -1590,6 +1772,7 @@ static void ahci_port_intr(struct ata_po
+ 	/* if LPM is enabled, PHYRDY doesn't mean anything */
+ 	if (ap->link.lpm_policy > ATA_LPM_MAX_POWER) {
+ 		status &= ~PORT_IRQ_PHYRDY;
++		account_alpm_stats(ap);
+ 		ahci_scr_write(&ap->link, SCR_ERROR, SERR_PHYRDY_CHG);
+ 	}
+ 
diff --git a/patches/linux-2.6.37-rc3-vfs-dirty-inode.patch b/patches/linux-2.6.37-rc3-vfs-dirty-inode.patch
new file mode 100644
index 0000000..8365f94
--- /dev/null
+++ b/patches/linux-2.6.37-rc3-vfs-dirty-inode.patch
@@ -0,0 +1,105 @@
+From 880a17e27f0b90d32c0c0b9cf4cd9b4e5d779c73 Mon Sep 17 00:00:00 2001
+From: Arjan van de Ven <arjan@linux.intel.com>
+Date: Fri, 26 Nov 2010 12:18:03 -0800
+Subject: [PATCH] vfs: Add a trace point in the mark_inode_dirty function
+
+PowerTOP would like to be able to show who is keeping the disk
+busy by dirtying data. The most logical spot for this is in the vfs
+in the mark_inode_dirty() function, doing this on the block level
+is not possible because by the time the IO hits the block layer the
+guilty party can no longer be found ("kjournald" and "pdflush" are not
+useful answers to "who caused this file to be dirty).
+
+The trace point follows the same logic/style as the block_dump code
+and pretty much dumps the same data, just not to dmesg (and thus to
+/var/log/messages) but via the trace events streams.
+
+Eventually we should be able to phase out the block dump code, but that's
+for later on after a transition time.
+
+Signed-of-by: Arjan van de Ven <arjan@linux.intel.com>
+---
+ fs/fs-writeback.c          |    4 ++++
+ fs/inode.c                 |    4 ++++
+ include/trace/events/vfs.h |   35 +++++++++++++++++++++++++++++++++++
+ 3 files changed, 43 insertions(+), 0 deletions(-)
+ create mode 100644 include/trace/events/vfs.h
+
+diff --git a/fs/fs-writeback.c b/fs/fs-writeback.c
+index 3d06ccc..6c93517 100644
+--- a/fs/fs-writeback.c
++++ b/fs/fs-writeback.c
+@@ -27,6 +27,7 @@
+ #include <linux/backing-dev.h>
+ #include <linux/buffer_head.h>
+ #include <linux/tracepoint.h>
++#include <trace/events/vfs.h>
+ #include "internal.h"
+ 
+ /*
+@@ -942,6 +943,9 @@ void __mark_inode_dirty(struct inode *inode, int flags)
+ 			sb->s_op->dirty_inode(inode);
+ 	}
+ 
++	if (flags & (I_DIRTY_SYNC | I_DIRTY_DATASYNC | I_DIRTY_PAGES))
++		trace_inode_dirty(inode);
++
+ 	/*
+ 	 * make sure that changes are seen by all cpus before we test i_state
+ 	 * -- mikulas
+diff --git a/fs/inode.c b/fs/inode.c
+index ae2727a..e8b1d42 100644
+--- a/fs/inode.c
++++ b/fs/inode.c
+@@ -1708,3 +1708,7 @@ void inode_init_owner(struct inode *inode, const struct inode *dir,
+ 	inode->i_mode = mode;
+ }
+ EXPORT_SYMBOL(inode_init_owner);
++
++#define CREATE_TRACE_POINTS
++#include <trace/events/vfs.h>
++
+diff --git a/include/trace/events/vfs.h b/include/trace/events/vfs.h
+new file mode 100644
+index 0000000..7e5c135
+--- /dev/null
++++ b/include/trace/events/vfs.h
+@@ -0,0 +1,35 @@
++#undef TRACE_SYSTEM
++#define TRACE_SYSTEM vfs
++
++#if !defined(_TRACE_VFS_H) || defined(TRACE_HEADER_MULTI_READ)
++#define _TRACE_VFS_H
++
++/*
++ * Tracepoint for dirtying an inode:
++ */
++TRACE_EVENT(inode_dirty,
++
++	TP_PROTO(struct inode *inode),
++
++	TP_ARGS(inode),
++
++	TP_STRUCT__entry(
++		__field(	int,	dev_major		)
++		__field(	int,	dev_minor		)
++		__field(	ino_t,	ino			)
++	),
++
++	TP_fast_assign(
++		__entry->dev_major = MAJOR(inode->i_sb->s_dev);
++		__entry->dev_minor = MINOR(inode->i_sb->s_dev);
++		__entry->ino	   = inode->i_ino;
++	),
++
++	TP_printk("dev %d,%d ino %lu ", __entry->dev_major, __entry->dev_minor,
++		  (unsigned long) __entry->ino)
++);
++
++#endif /* _TRACE_VFS_H */
++
++/* This part must be outside protection */
++#include <trace/define_trace.h>
+-- 
+1.7.2.3
+