Re: performance tools on xen

Re: performance tools on xen

[Dante Cinco]

[-- Attachment #1.1: Type: text/plain, Size: 1351 bytes --]

Jeremy,

I searched Mark Mail for "xenoprof pvops" and found this email from dulloor:

http://markmail.org/search/?q=xenoprof+pvops#query:xenoprof%20pvops+page:1+mid:yyzbymymat3rtivo+state:results

Do you know if dulloor's patches for xenoprof for linux pvops got applied?

Dante

On Fri, Oct 30, 2009 at 7:30 AM, Jeremy Fitzhardinge <jeremy@goop.org>wrote:

> On 10/29/09 19:36, Santos, Jose Renato G wrote:
> > Dante,
> >
> > To run OProfile in Xen you need support in the kernel and in the
> > OProfile user level tools.
> > The user level patches were never merged on mainline OProfile. You
> > need to apply a patch to the OProfile sources before compiling them. I
> > did not have time to port the patch to OProfile 0.9.5 yet, so the
> > latest available patch is for Oprofile 0.9.3. You can download that
> > from http://xenoprof.sourceforge.net/.
> >
> > Regarding the kernel I am not sure if the OProfile extensions for Xen
> > were already merged in the pvops kernel. You will need to ask that to
> > Jeremy. I know that OProfile works with the 2.6.18-xen and the XCI
> > 2.6.27-xen versions. I have never tried running the pvops kernel yet.
>
> I have not looked at Xen+Oprofile, so there's no support in the pvops
> kernel as yet.  If someone who's familiar with that stuff could generate
> a patch, I'd be happy to include it.
>
>    J
>

[-- Attachment #1.2: Type: text/html, Size: 2010 bytes --]

[-- Attachment #2: Type: text/plain, Size: 138 bytes --]

_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel

Re: performance tools on xen

[Jeremy Fitzhardinge]

On 11/02/09 09:56, Dante Cinco wrote:
> Jeremy,
>
> I searched Mark Mail for "xenoprof pvops" and found this email from
> dulloor:
>
> http://markmail.org/search/?q=xenoprof+pvops#query:xenoprof%20pvops+page:1+mid:yyzbymymat3rtivo+state:results
>
> Do you know if dulloor's patches for xenoprof for linux pvops got applied?

No, they fell through the cracks; I think there was some problem that I
could never connect to his git server.

Dulloor, do you still have a current set of patched we can try?

Thanks,
    J

Re: performance tools on xen

[Dulloor]

Jeremy, I thought you were able to access my git repo (per your last
mail). I should be having the patches somewhere around. Will send them
over soon.

-dulloor

On Mon, Nov 2, 2009 at 1:47 PM, Jeremy Fitzhardinge <jeremy@goop.org> wrote:
> On 11/02/09 09:56, Dante Cinco wrote:
>> Jeremy,
>>
>> I searched Mark Mail for "xenoprof pvops" and found this email from
>> dulloor:
>>
>> http://markmail.org/search/?q=xenoprof+pvops#query:xenoprof%20pvops+page:1+mid:yyzbymymat3rtivo+state:results
>>
>> Do you know if dulloor's patches for xenoprof for linux pvops got applied?
>
> No, they fell through the cracks; I think there was some problem that I
> could never connect to his git server.
>
> Dulloor, do you still have a current set of patched we can try?
>
> Thanks,
>    J
>

Re: performance tools on xen

[Dulloor]

[-- Attachment #1: Type: text/plain, Size: 1229 bytes --]

I have attached the patch that will apply to the current pvops tree.
Tested both on 32-bit and 64-bit platforms.

I lost some of the cleanups that I had done to get rid of some
(CONFIG_XEN)ifdefs in a couple of files in oprofile directory. But, I
will work on them incrementally, if this is checked in.

-dulloor

On Wed, Nov 4, 2009 at 12:46 AM, Dulloor <dulloor@gmail.com> wrote:
> Jeremy, I thought you were able to access my git repo (per your last
> mail). I should be having the patches somewhere around. Will send them
> over soon.
>
> -dulloor
>
> On Mon, Nov 2, 2009 at 1:47 PM, Jeremy Fitzhardinge <jeremy@goop.org> wrote:
>> On 11/02/09 09:56, Dante Cinco wrote:
>>> Jeremy,
>>>
>>> I searched Mark Mail for "xenoprof pvops" and found this email from
>>> dulloor:
>>>
>>> http://markmail.org/search/?q=xenoprof+pvops#query:xenoprof%20pvops+page:1+mid:yyzbymymat3rtivo+state:results
>>>
>>> Do you know if dulloor's patches for xenoprof for linux pvops got applied?
>>
>> No, they fell through the cracks; I think there was some problem that I
>> could never connect to his git server.
>>
>> Dulloor, do you still have a current set of patched we can try?
>>
>> Thanks,
>>    J
>>
>

[-- Attachment #2: xenoprof.patch --]
[-- Type: text/x-diff, Size: 49569 bytes --]

diff --git a/arch/x86/include/asm/xen/hypercall.h b/arch/x86/include/asm/xen/hypercall.h
index 7de93f3..04d8f38 100644
--- a/arch/x86/include/asm/xen/hypercall.h
+++ b/arch/x86/include/asm/xen/hypercall.h
@@ -450,6 +450,12 @@ HYPERVISOR_nmi_op(unsigned long op, unsigned long arg)
 	return _hypercall2(int, nmi_op, op, arg);
 }
 
+static inline int
+HYPERVISOR_xenoprof_op(unsigned int op, void *arg)
+{
+	return _hypercall2(int, xenoprof_op, op, arg);
+}
+
 static inline void
 MULTI_fpu_taskswitch(struct multicall_entry *mcl, int set)
 {
diff --git a/arch/x86/oprofile/Makefile b/arch/x86/oprofile/Makefile
index 446902b..6a976e6 100644
--- a/arch/x86/oprofile/Makefile
+++ b/arch/x86/oprofile/Makefile
@@ -6,6 +6,12 @@ DRIVER_OBJS = $(addprefix ../../../drivers/oprofile/, \
 		oprofilefs.o oprofile_stats.o  \
 		timer_int.o )
 
+ifdef CONFIG_XEN
+XENOPROF_COMMON_OBJS = $(addprefix ../../../drivers/xen/xenoprof/, \
+			 xenoprofile.o)
+DRIVER_OBJS				:= $(DRIVER_OBJS) \
+					   $(XENOPROF_COMMON_OBJS) xenoprof.o
+endif 
 oprofile-y				:= $(DRIVER_OBJS) init.o backtrace.o
 oprofile-$(CONFIG_X86_LOCAL_APIC) 	+= nmi_int.o op_model_amd.o \
 					   op_model_ppro.o op_model_p4.o
diff --git a/arch/x86/oprofile/xenoprof.c b/arch/x86/oprofile/xenoprof.c
new file mode 100644
index 0000000..e86f1d0
--- /dev/null
+++ b/arch/x86/oprofile/xenoprof.c
@@ -0,0 +1,172 @@
+/**
+ * @file xenoprof.c
+ *
+ * @remark Copyright 2002 OProfile authors
+ * @remark Read the file COPYING
+ *
+ * @author John Levon <levon@movementarian.org>
+ *
+ * Modified by Aravind Menon and Jose Renato Santos for Xen
+ * These modifications are:
+ * Copyright (C) 2005 Hewlett-Packard Co.
+ *
+ * x86-specific part
+ * Copyright (c) 2006 Isaku Yamahata <yamahata at valinux co jp>
+ *                    VA Linux Systems Japan K.K.
+ */
+
+#include <linux/init.h>
+#include <linux/oprofile.h>
+#include <linux/sched.h>
+#include <linux/vmalloc.h>
+#include <asm/pgtable.h>
+
+#include <xen/interface/xen.h>
+#include <asm/xen/hypercall.h>
+#include <xen/xen-ops.h>
+#include <xen/interface/xenoprof.h>
+#include <xen/xenoprof.h>
+#include "op_counter.h"
+
+static unsigned int num_events = 0;
+struct op_counter_config xen_counter_config[OP_MAX_COUNTER];
+
+void __init xenoprof_arch_init_counter(struct xenoprof_init *init)
+{
+	num_events = init->num_events;
+	/* just in case - make sure we do not overflow event list 
+	   (i.e. xen_counter_config list) */
+	if (num_events > OP_MAX_COUNTER) {
+		num_events = OP_MAX_COUNTER;
+		init->num_events = num_events;
+	}
+}
+
+void xenoprof_arch_counter(void)
+{
+	int i;
+	struct xenoprof_counter counter;
+
+	for (i=0; i<num_events; i++) {
+		counter.ind       = i;
+		counter.count     = (uint64_t)xen_counter_config[i].count;
+		counter.enabled   = (uint32_t)xen_counter_config[i].enabled;
+		counter.event     = (uint32_t)xen_counter_config[i].event;
+		counter.kernel    = (uint32_t)xen_counter_config[i].kernel;
+		counter.user      = (uint32_t)xen_counter_config[i].user;
+		counter.unit_mask = (uint64_t)xen_counter_config[i].unit_mask;
+		WARN_ON(HYPERVISOR_xenoprof_op(XENOPROF_counter,
+					       &counter));
+	}
+}
+
+void xenoprof_arch_start(void) 
+{
+	/* nothing */
+}
+
+void xenoprof_arch_stop(void)
+{
+	/* nothing */
+}
+
+void xenoprof_arch_unmap_shared_buffer(struct xenoprof_shared_buffer * sbuf)
+{
+	if (sbuf->buffer) {
+		vunmap(sbuf->buffer);
+		sbuf->buffer = NULL;
+	}
+}
+
+int xenoprof_arch_map_shared_buffer(struct xenoprof_get_buffer * get_buffer,
+				    struct xenoprof_shared_buffer * sbuf)
+{
+	int npages, ret;
+	struct vm_struct *area;
+
+	sbuf->buffer = NULL;
+	if ( (ret = HYPERVISOR_xenoprof_op(XENOPROF_get_buffer, get_buffer)) )
+		return ret;
+
+	npages = (get_buffer->bufsize * get_buffer->nbuf - 1) / PAGE_SIZE + 1;
+
+	area = alloc_vm_area(npages * PAGE_SIZE);
+	if (area == NULL)
+		return -ENOMEM;
+
+	if ( (ret = xen_remap_domain_kernel_mfn_range(
+		      (unsigned long)area->addr,
+		      get_buffer->buf_gmaddr >> PAGE_SHIFT,
+		      npages, __pgprot(_KERNPG_TABLE),
+		      DOMID_SELF)) ) {
+		vunmap(area->addr);
+		return ret;
+	}
+
+	sbuf->buffer = area->addr;
+	return ret;
+}
+
+int xenoprof_arch_set_passive(struct xenoprof_passive * pdomain,
+			      struct xenoprof_shared_buffer * sbuf)
+{
+	int ret;
+	int npages;
+	struct vm_struct *area;
+	pgprot_t prot = __pgprot(_KERNPG_TABLE);
+
+	sbuf->buffer = NULL;
+
+	ret = HYPERVISOR_xenoprof_op(XENOPROF_set_passive, pdomain);
+	if (ret)
+		goto out;
+
+	npages = (pdomain->bufsize * pdomain->nbuf - 1) / PAGE_SIZE + 1;
+
+	area = alloc_vm_area(npages * PAGE_SIZE);
+	if (area == NULL) {
+		ret = -ENOMEM;
+		goto out;
+	}
+
+	ret = xen_remap_domain_kernel_mfn_range(
+		(unsigned long)area->addr & PAGE_MASK,
+		pdomain->buf_gmaddr >> PAGE_SHIFT,
+		npages, prot, DOMID_SELF);
+	if (ret) {
+		vunmap(area->addr);
+		goto out;
+	}
+	sbuf->buffer = area->addr;
+
+out:
+	return ret;
+}
+
+
+int xenoprof_create_files(struct super_block * sb, struct dentry * root)
+{
+	unsigned int i;
+
+	for (i = 0; i < num_events; ++i) {
+		struct dentry * dir;
+		char buf[2];
+ 
+		snprintf(buf, 2, "%d", i);
+		dir = oprofilefs_mkdir(sb, root, buf);
+		oprofilefs_create_ulong(sb, dir, "enabled",
+					&xen_counter_config[i].enabled);
+		oprofilefs_create_ulong(sb, dir, "event",
+					&xen_counter_config[i].event);
+		oprofilefs_create_ulong(sb, dir, "count",
+					&xen_counter_config[i].count);
+		oprofilefs_create_ulong(sb, dir, "unit_mask",
+					&xen_counter_config[i].unit_mask);
+		oprofilefs_create_ulong(sb, dir, "kernel",
+					&xen_counter_config[i].kernel);
+		oprofilefs_create_ulong(sb, dir, "user",
+					&xen_counter_config[i].user);
+	}
+
+	return 0;
+}
diff --git a/arch/x86/xen/mmu.c b/arch/x86/xen/mmu.c
index c5e31cb..7a222eb 100644
--- a/arch/x86/xen/mmu.c
+++ b/arch/x86/xen/mmu.c
@@ -2351,30 +2351,30 @@ static int remap_area_mfn_pte_fn(pte_t *ptep, pgtable_t token,
 				 unsigned long addr, void *data)
 {
 	struct remap_data *rmd = data;
-	pte_t pte = pte_mkspecial(pfn_pte(rmd->mfn++, rmd->prot));
+	pte_t pte = pte_mkspecial(pfn_pte(rmd->mfn, rmd->prot));
 
 	rmd->mmu_update->ptr = arbitrary_virt_to_machine(ptep).maddr;
 	rmd->mmu_update->val = pte_val_ma(pte);
+
+	rmd->mfn++;
 	rmd->mmu_update++;
 
 	return 0;
 }
 
-int xen_remap_domain_mfn_range(struct vm_area_struct *vma,
-			       unsigned long addr,
-			       unsigned long mfn, int nr,
-			       pgprot_t prot, unsigned domid)
+static int __xen_remap_domain_mfn_range(struct mm_struct *mm,
+				unsigned long addr,
+				unsigned long mfn, int nr,
+				pgprot_t prot, unsigned domid)
 {
 	struct remap_data rmd;
 	struct mmu_update mmu_update[REMAP_BATCH_SIZE];
 	int batch;
 	unsigned long range;
-	int err = 0;
+	int err;
 
 	prot = __pgprot(pgprot_val(prot) | _PAGE_IOMAP);
 
-	vma->vm_flags |= VM_IO | VM_RESERVED | VM_PFNMAP;
-
 	rmd.mfn = mfn;
 	rmd.prot = prot;
 
@@ -2383,14 +2383,16 @@ int xen_remap_domain_mfn_range(struct vm_area_struct *vma,
 		range = (unsigned long)batch << PAGE_SHIFT;
 
 		rmd.mmu_update = mmu_update;
-		err = apply_to_page_range(vma->vm_mm, addr, range,
+
+		err = apply_to_page_range(mm, addr, range,
 					  remap_area_mfn_pte_fn, &rmd);
 		if (err)
 			goto out;
 
-		err = -EFAULT;
-		if (HYPERVISOR_mmu_update(mmu_update, batch, NULL, domid) < 0)
+		if (HYPERVISOR_mmu_update(mmu_update, batch, NULL, domid) < 0) {
+			err = -EFAULT;
 			goto out;
+		}
 
 		nr -= batch;
 		addr += range;
@@ -2398,13 +2400,33 @@ int xen_remap_domain_mfn_range(struct vm_area_struct *vma,
 
 	err = 0;
 out:
-
 	flush_tlb_all();
-
 	return err;
 }
+
+int xen_remap_domain_mfn_range(struct vm_area_struct *vma,
+			       unsigned long addr,
+			       unsigned long mfn, int nr,
+			       pgprot_t prot, unsigned domid)
+{
+
+	vma->vm_flags |= VM_IO | VM_RESERVED | VM_PFNMAP;
+
+	return __xen_remap_domain_mfn_range(vma->vm_mm, addr, 
+					mfn, nr, prot, domid);
+}
 EXPORT_SYMBOL_GPL(xen_remap_domain_mfn_range);
 
+
+int xen_remap_domain_kernel_mfn_range(unsigned long addr,
+			       unsigned long mfn, int nr,
+			       pgprot_t prot, unsigned domid)
+{
+	return __xen_remap_domain_mfn_range(&init_mm, addr, 
+					mfn, nr, prot, domid);
+}
+EXPORT_SYMBOL_GPL(xen_remap_domain_kernel_mfn_range);
+
 #ifdef CONFIG_XEN_DEBUG_FS
 
 static struct dentry *d_mmu_debug;
diff --git a/drivers/oprofile/buffer_sync.c b/drivers/oprofile/buffer_sync.c
index 8574622..856139d 100644
--- a/drivers/oprofile/buffer_sync.c
+++ b/drivers/oprofile/buffer_sync.c
@@ -42,6 +42,10 @@ static cpumask_var_t marked_cpus;
 static DEFINE_SPINLOCK(task_mortuary);
 static void process_task_mortuary(void);
 
+#ifdef CONFIG_XEN
+static int cpu_current_xen_domain[NR_CPUS];
+#endif
+
 /* Take ownership of the task struct and place it on the
  * list for processing. Only after two full buffer syncs
  * does the task eventually get freed, because by then
@@ -154,10 +158,16 @@ int sync_start(void)
 {
 	int err;
 
+#ifdef CONFIG_XEN
+	int i;
+
+	for (i = 0; i < NR_CPUS; i++)
+		cpu_current_xen_domain[i] = XEN_COORDINATOR_DOMAIN;
+#endif
+
 	if (!alloc_cpumask_var(&marked_cpus, GFP_KERNEL))
 		return -ENOMEM;
 	cpumask_clear(marked_cpus);
-
 	start_cpu_work();
 
 	err = task_handoff_register(&task_free_nb);
@@ -285,14 +295,37 @@ static void add_cpu_switch(int i)
 	last_cookie = INVALID_COOKIE;
 }
 
-static void add_kernel_ctx_switch(unsigned int in_kernel)
+static void add_cpu_mode_switch(unsigned int cpu_mode)
+{
+	add_event_entry(ESCAPE_CODE);
+	switch(cpu_mode)
+	{
+		case CPU_MODE_USER:
+			add_event_entry(USER_ENTER_SWITCH_CODE);
+			break;
+		case CPU_MODE_KERNEL:
+			add_event_entry(KERNEL_ENTER_SWITCH_CODE);
+			break;
+#ifdef CONFIG_XEN
+		case CPU_MODE_XEN:
+			add_event_entry(XEN_ENTER_SWITCH_CODE);
+			break;
+#endif
+		default:
+			break;
+	}
+
+	return;
+}
+
+#ifdef CONFIG_XEN
+static void add_xen_domain_switch(unsigned long domain_id)
 {
 	add_event_entry(ESCAPE_CODE);
-	if (in_kernel)
-		add_event_entry(KERNEL_ENTER_SWITCH_CODE);
-	else
-		add_event_entry(KERNEL_EXIT_SWITCH_CODE);
+	add_event_entry(XEN_DOMAIN_SWITCH_CODE);
+	add_event_entry(domain_id);
 }
+#endif
 
 static void
 add_user_ctx_switch(struct task_struct const *task, unsigned long cookie)
@@ -372,12 +405,12 @@ static inline void add_sample_entry(unsigned long offset, unsigned long event)
  * for later lookup from userspace. Return 0 on failure.
  */
 static int
-add_sample(struct mm_struct *mm, struct op_sample *s, int in_kernel)
+add_sample(struct mm_struct *mm, struct op_sample *s, int cpu_mode)
 {
 	unsigned long cookie;
 	off_t offset;
 
-	if (in_kernel) {
+	if (cpu_mode >= CPU_MODE_KERNEL) {
 		add_sample_entry(s->eip, s->event);
 		return 1;
 	}
@@ -502,7 +535,7 @@ void sync_buffer(int cpu)
 	unsigned long val;
 	struct task_struct *new;
 	unsigned long cookie = 0;
-	int in_kernel = 1;
+	int cpu_mode = CPU_MODE_KERNEL;
 	sync_buffer_state state = sb_buffer_start;
 	unsigned int i;
 	unsigned long available;
@@ -514,6 +547,13 @@ void sync_buffer(int cpu)
 
 	add_cpu_switch(cpu);
 
+#ifdef CONFIG_XEN
+	/* We need to assign the first samples in this CPU buffer to the
+	 * same domain that we were processing at the last sync_buffer */
+	if(cpu_current_xen_domain[cpu] != XEN_COORDINATOR_DOMAIN)
+		add_xen_domain_switch(cpu_current_xen_domain[cpu]);
+#endif
+
 	op_cpu_buffer_reset(cpu);
 	available = op_cpu_buffer_entries(cpu);
 
@@ -530,10 +570,11 @@ void sync_buffer(int cpu)
 			}
 			if (flags & KERNEL_CTX_SWITCH) {
 				/* kernel/userspace switch */
-				in_kernel = flags & IS_KERNEL;
+				/* XXX: crap change this to use cpu_mode explicitly */
+				cpu_mode = flags & CPU_MODE_MASK;
 				if (state == sb_buffer_start)
 					state = sb_sample_start;
-				add_kernel_ctx_switch(flags & IS_KERNEL);
+				add_cpu_mode_switch(cpu_mode);
 			}
 			if (flags & USER_CTX_SWITCH
 			    && op_cpu_buffer_get_data(&entry, &val)) {
@@ -546,16 +587,32 @@ void sync_buffer(int cpu)
 					cookie = get_exec_dcookie(mm);
 				add_user_ctx_switch(new, cookie);
 			}
+#ifdef CONFIG_XEN
+			/* xen domain switch */
+			if (flags & XEN_DOMAIN_SWITCH
+				&& op_cpu_buffer_get_data(&entry, &val)) {
+				cpu_current_xen_domain[cpu] = val;
+				add_xen_domain_switch(val);
+			}
+#endif
 			if (op_cpu_buffer_get_size(&entry))
 				add_data(&entry, mm);
 			continue;
 		}
 
+#ifdef CONFIG_XEN
+		if(cpu_current_xen_domain[cpu] != XEN_COORDINATOR_DOMAIN)
+		{
+			add_sample_entry(sample->eip, sample->event);
+			continue;
+		}
+#endif
+			
 		if (state < sb_bt_start)
 			/* ignore sample */
 			continue;
 
-		if (add_sample(mm, sample, in_kernel))
+		if (add_sample(mm, sample, cpu_mode))
 			continue;
 
 		/* ignore backtraces if failed to add a sample */
diff --git a/drivers/oprofile/cpu_buffer.c b/drivers/oprofile/cpu_buffer.c
index 242257b..21959f1 100644
--- a/drivers/oprofile/cpu_buffer.c
+++ b/drivers/oprofile/cpu_buffer.c
@@ -55,6 +55,11 @@ static void wq_sync_buffer(struct work_struct *work);
 #define DEFAULT_TIMER_EXPIRE (HZ / 10)
 static int work_enabled;
 
+
+#ifdef CONFIG_XEN
+static int current_xen_domain = XEN_COORDINATOR_DOMAIN;
+#endif
+
 unsigned long oprofile_get_cpu_buffer_size(void)
 {
 	return oprofile_cpu_buffer_size;
@@ -99,7 +104,7 @@ int alloc_cpu_buffers(void)
 		struct oprofile_cpu_buffer *b = &per_cpu(cpu_buffer, i);
 
 		b->last_task = NULL;
-		b->last_is_kernel = -1;
+		b->last_cpu_mode = -1;
 		b->tracing = 0;
 		b->buffer_size = buffer_size;
 		b->sample_received = 0;
@@ -217,7 +222,7 @@ unsigned long op_cpu_buffer_entries(int cpu)
 
 static int
 op_add_code(struct oprofile_cpu_buffer *cpu_buf, unsigned long backtrace,
-	    int is_kernel, struct task_struct *task)
+	    int cpu_mode, struct task_struct *task)
 {
 	struct op_entry entry;
 	struct op_sample *sample;
@@ -229,17 +234,20 @@ op_add_code(struct oprofile_cpu_buffer *cpu_buf, unsigned long backtrace,
 	if (backtrace)
 		flags |= TRACE_BEGIN;
 
-	/* notice a switch from user->kernel or vice versa */
-	is_kernel = !!is_kernel;
-	if (cpu_buf->last_is_kernel != is_kernel) {
-		cpu_buf->last_is_kernel = is_kernel;
-		flags |= KERNEL_CTX_SWITCH;
-		if (is_kernel)
-			flags |= IS_KERNEL;
+	/* switch in cpu_mode */
+	if (cpu_buf->last_cpu_mode != cpu_mode) {
+		cpu_buf->last_cpu_mode = cpu_mode;
+		flags |= (KERNEL_CTX_SWITCH | cpu_mode);
 	}
 
 	/* notice a task switch */
+/* XXX: yuck ! do something about this too. */
+#ifndef CONFIG_XEN
 	if (cpu_buf->last_task != task) {
+#else
+	if ((cpu_buf->last_task != task)
+		&& (current_xen_domain == XEN_COORDINATOR_DOMAIN)) {
+#endif
 		cpu_buf->last_task = task;
 		flags |= USER_CTX_SWITCH;
 	}
@@ -288,14 +296,14 @@ op_add_sample(struct oprofile_cpu_buffer *cpu_buf,
 /*
  * This must be safe from any context.
  *
- * is_kernel is needed because on some architectures you cannot
- * tell if you are in kernel or user space simply by looking at
- * pc. We tag this in the buffer by generating kernel enter/exit
+ * cpu_mode is needed because on some architectures you cannot
+ * tell if you are in user/kernel(/xen) space simply by looking at
+ * pc. We tag this in the buffer by generating user/kernel(/xen) enter
  * events whenever is_kernel changes
  */
 static int
 log_sample(struct oprofile_cpu_buffer *cpu_buf, unsigned long pc,
-	   unsigned long backtrace, int is_kernel, unsigned long event)
+	   unsigned long backtrace, int cpu_mode, unsigned long event)
 {
 	cpu_buf->sample_received++;
 
@@ -304,7 +312,7 @@ log_sample(struct oprofile_cpu_buffer *cpu_buf, unsigned long pc,
 		return 0;
 	}
 
-	if (op_add_code(cpu_buf, backtrace, is_kernel, current))
+	if (op_add_code(cpu_buf, backtrace, cpu_mode, current))
 		goto fail;
 
 	if (op_add_sample(cpu_buf, pc, event))
@@ -414,12 +422,27 @@ int oprofile_write_commit(struct op_entry *entry)
 	return op_cpu_buffer_write_commit(entry);
 }
 
+/* XXX: yuck ! Needs clean-up */
 void oprofile_add_pc(unsigned long pc, int is_kernel, unsigned long event)
 {
 	struct oprofile_cpu_buffer *cpu_buf = &__get_cpu_var(cpu_buffer);
 	log_sample(cpu_buf, pc, 0, is_kernel, event);
 }
 
+/*
+ * Equivalent to log_sample(b, ESCAPE_CODE, 1, cpu_mode, CPU_TRACE_BEGIN),
+ * Previously accessible through oprofile_add_pc().
+ */
+void oprofile_add_mode(int cpu_mode)
+{
+	struct oprofile_cpu_buffer *cpu_buf = &__get_cpu_var(cpu_buffer);
+
+	if (op_add_code(cpu_buf, 1, cpu_mode, current))
+		cpu_buf->sample_lost_overflow++;
+
+	return;
+}
+
 void oprofile_add_trace(unsigned long pc)
 {
 	struct oprofile_cpu_buffer *cpu_buf = &__get_cpu_var(cpu_buffer);
@@ -444,6 +467,28 @@ fail:
 	return;
 }
 
+#ifdef CONFIG_XEN
+int oprofile_add_domain_switch(int32_t domain_id)
+{
+	struct op_entry entry;
+	struct op_sample *sample;
+
+	sample = op_cpu_buffer_write_reserve(&entry, 1);
+	if (!sample)
+        	return 0;
+
+	sample->eip = ESCAPE_CODE;
+	sample->event = XEN_DOMAIN_SWITCH;
+
+	op_cpu_buffer_add_data(&entry, domain_id);
+	op_cpu_buffer_write_commit(&entry);
+
+	current_xen_domain = domain_id;
+
+	return 1;
+}
+#endif
+
 /*
  * This serves to avoid cpu buffer overflow, and makes sure
  * the task mortuary progresses
diff --git a/drivers/oprofile/cpu_buffer.h b/drivers/oprofile/cpu_buffer.h
index 272995d..95be4c7 100644
--- a/drivers/oprofile/cpu_buffer.h
+++ b/drivers/oprofile/cpu_buffer.h
@@ -40,7 +40,7 @@ struct op_entry;
 struct oprofile_cpu_buffer {
 	unsigned long buffer_size;
 	struct task_struct *last_task;
-	int last_is_kernel;
+	int last_cpu_mode;
 	int tracing;
 	unsigned long sample_received;
 	unsigned long sample_lost_overflow;
@@ -62,7 +62,7 @@ static inline void op_cpu_buffer_reset(int cpu)
 {
 	struct oprofile_cpu_buffer *cpu_buf = &per_cpu(cpu_buffer, cpu);
 
-	cpu_buf->last_is_kernel = -1;
+	cpu_buf->last_cpu_mode = -1;
 	cpu_buf->last_task = NULL;
 }
 
@@ -111,10 +111,22 @@ int op_cpu_buffer_get_data(struct op_entry *entry, unsigned long *val)
 	return size;
 }
 
+/* data flags */
+/* cpu modes */
+/* */
+#define CPU_MODE_BEGIN		(0UL)
+#define CPU_MODE_USER		(CPU_MODE_BEGIN + 0x0)
+#define CPU_MODE_KERNEL		(CPU_MODE_BEGIN + 0x1)
+#ifdef CONFIG_XEN
+#define CPU_MODE_XEN		(CPU_MODE_BEGIN + 0x2)
+#endif
+#define CPU_MODE_END		(CPU_MODE_BEGIN + 0x3)
+#define CPU_MODE_MASK		0x3
+
 /* extra data flags */
-#define KERNEL_CTX_SWITCH	(1UL << 0)
-#define IS_KERNEL		(1UL << 1)
 #define TRACE_BEGIN		(1UL << 2)
 #define USER_CTX_SWITCH		(1UL << 3)
+#define KERNEL_CTX_SWITCH	(1UL << 4)
+#define XEN_DOMAIN_SWITCH	(1UL << 5)
 
 #endif /* OPROFILE_CPU_BUFFER_H */
diff --git a/drivers/oprofile/event_buffer.h b/drivers/oprofile/event_buffer.h
index 4e70749..9f19d42 100644
--- a/drivers/oprofile/event_buffer.h
+++ b/drivers/oprofile/event_buffer.h
@@ -30,6 +30,11 @@ void wake_up_buffer_waiter(void);
 #define INVALID_COOKIE ~0UL
 #define NO_COOKIE 0UL
 
+#ifdef CONFIG_XEN
+#define XEN_COORDINATOR_DOMAIN -1
+#endif
+
+
 extern const struct file_operations event_buffer_fops;
 
 /* mutex between sync_cpu_buffers() and the
diff --git a/drivers/oprofile/oprof.c b/drivers/oprofile/oprof.c
index 3cffce9..c55cc02 100644
--- a/drivers/oprofile/oprof.c
+++ b/drivers/oprofile/oprof.c
@@ -20,6 +20,11 @@
 #include "buffer_sync.h"
 #include "oprofile_stats.h"
 
+#ifdef CONFIG_XEN
+#include <xen/xenoprof.h>
+#include <asm/xen/hypervisor.h>
+#endif
+
 struct oprofile_operations oprofile_ops;
 
 unsigned long oprofile_started;
@@ -33,6 +38,35 @@ static DEFINE_MUTEX(start_mutex);
  */
 static int timer = 0;
 
+#ifdef CONFIG_XEN
+int oprofile_xen_set_active(int active_domains[], unsigned int adomains)
+{
+	int err;
+
+	if (!oprofile_ops.xen_set_active)
+		return -EINVAL;
+
+	mutex_lock(&start_mutex);
+	err = oprofile_ops.xen_set_active(active_domains, adomains);
+	mutex_unlock(&start_mutex);
+	return err;
+}
+
+int oprofile_xen_set_passive(int passive_domains[], unsigned int pdomains)
+{
+	int err;
+
+	if (!oprofile_ops.xen_set_passive)
+		return -EINVAL;
+
+	mutex_lock(&start_mutex);
+	err = oprofile_ops.xen_set_passive(passive_domains, pdomains);
+	mutex_unlock(&start_mutex);
+	return err;
+}
+#endif
+
+
 int oprofile_setup(void)
 {
 	int err;
@@ -182,8 +216,21 @@ out:
 static int __init oprofile_init(void)
 {
 	int err;
-
-	err = oprofile_arch_init(&oprofile_ops);
+    int (*oprofile_arch_init_func)(struct oprofile_operations * ops);
+    void (*oprofile_arch_exit_func)(void);
+
+    if (xen_initial_domain())
+    {
+        oprofile_arch_init_func = xenoprofile_init;
+        oprofile_arch_exit_func = xenoprofile_exit;
+    }
+    else
+    {
+        oprofile_arch_init_func = oprofile_arch_init;
+        oprofile_arch_exit_func = oprofile_arch_exit;
+    }
+
+	err = oprofile_arch_init_func(&oprofile_ops);
 
 	if (err < 0 || timer) {
 		printk(KERN_INFO "oprofile: using timer interrupt.\n");
@@ -192,7 +239,7 @@ static int __init oprofile_init(void)
 
 	err = oprofilefs_register();
 	if (err)
-		oprofile_arch_exit();
+		oprofile_arch_exit_func();
 
 	return err;
 }
diff --git a/drivers/oprofile/oprof.h b/drivers/oprofile/oprof.h
index c288d3c..e659728 100644
--- a/drivers/oprofile/oprof.h
+++ b/drivers/oprofile/oprof.h
@@ -36,4 +36,9 @@ void oprofile_timer_init(struct oprofile_operations *ops);
 
 int oprofile_set_backtrace(unsigned long depth);
 
+#ifdef CONFIG_XEN
+int oprofile_xen_set_active(int active_domains[], unsigned int adomains);
+int oprofile_xen_set_passive(int passive_domains[], unsigned int pdomains);
+#endif
+
 #endif /* OPROF_H */
diff --git a/drivers/oprofile/oprofile_files.c b/drivers/oprofile/oprofile_files.c
index 5d36ffc..c1318c9 100644
--- a/drivers/oprofile/oprofile_files.c
+++ b/drivers/oprofile/oprofile_files.c
@@ -9,6 +9,11 @@
 
 #include <linux/fs.h>
 #include <linux/oprofile.h>
+#include <asm/uaccess.h>
+
+#include <linux/slab.h>
+#include <linux/ctype.h>
+#include <linux/gfp.h>
 
 #include "event_buffer.h"
 #include "oprofile_stats.h"
@@ -123,6 +128,207 @@ static const struct file_operations dump_fops = {
 	.write		= dump_write,
 };
 
+#ifdef CONFIG_XEN
+
+#define TMPBUFSIZE 512
+
+static unsigned int adomains = 0;
+static int active_domains[MAX_OPROF_DOMAINS + 1];
+static DEFINE_MUTEX(adom_mutex);
+
+static ssize_t adomain_write(struct file * file, char const __user * buf,
+			     size_t count, loff_t * offset)
+{
+	char *tmpbuf;
+	char *startp, *endp;
+	int i;
+	unsigned long val;
+	ssize_t retval = count;
+
+	if (*offset)
+		return -EINVAL;
+	if (count > TMPBUFSIZE - 1)
+		return -EINVAL;
+
+	if (!(tmpbuf = kmalloc(TMPBUFSIZE, GFP_KERNEL)))
+		return -ENOMEM;
+
+	if (copy_from_user(tmpbuf, buf, count)) {
+		kfree(tmpbuf);
+		return -EFAULT;
+	}
+	tmpbuf[count] = 0;
+
+	mutex_lock(&adom_mutex);
+
+	startp = tmpbuf;
+	/* Parse one more than MAX_OPROF_DOMAINS, for easy error checking */
+	for (i = 0; i <= MAX_OPROF_DOMAINS; i++) {
+		val = simple_strtoul(startp, &endp, 0);
+		if (endp == startp)
+			break;
+		while (ispunct(*endp) || isspace(*endp))
+			endp++;
+		active_domains[i] = val;
+		if (active_domains[i] != val)
+			/* Overflow, force error below */
+			i = MAX_OPROF_DOMAINS + 1;
+		startp = endp;
+	}
+	/* Force error on trailing junk */
+	adomains = *startp ? MAX_OPROF_DOMAINS + 1 : i;
+
+	kfree(tmpbuf);
+
+	if (adomains > MAX_OPROF_DOMAINS
+	    || oprofile_xen_set_active(active_domains, adomains)) {
+		adomains = 0;
+		retval = -EINVAL;
+	}
+
+	mutex_unlock(&adom_mutex);
+	return retval;
+}
+
+static ssize_t adomain_read(struct file * file, char __user * buf,
+			    size_t count, loff_t * offset)
+{
+	char * tmpbuf;
+	size_t len;
+	int i;
+	ssize_t retval;
+
+	if (!(tmpbuf = kmalloc(TMPBUFSIZE, GFP_KERNEL)))
+		return -ENOMEM;
+
+	mutex_lock(&adom_mutex);
+
+	len = 0;
+	for (i = 0; i < adomains; i++)
+		len += snprintf(tmpbuf + len,
+				len < TMPBUFSIZE ? TMPBUFSIZE - len : 0,
+				"%u ", active_domains[i]);
+	WARN_ON(len > TMPBUFSIZE);
+	if (len != 0 && len <= TMPBUFSIZE)
+		tmpbuf[len-1] = '\n';
+
+	mutex_unlock(&adom_mutex);
+
+	retval = simple_read_from_buffer(buf, count, offset, tmpbuf, len);
+
+	kfree(tmpbuf);
+	return retval;
+}
+
+
+static const struct file_operations active_domain_ops = {
+	.read		= adomain_read,
+	.write		= adomain_write,
+};
+
+static unsigned int pdomains = 0;
+static int passive_domains[MAX_OPROF_DOMAINS];
+static DEFINE_MUTEX(pdom_mutex);
+
+static ssize_t pdomain_write(struct file * file, char const __user * buf,
+			     size_t count, loff_t * offset)
+{
+	char *tmpbuf;
+	char *startp, *endp;
+	int i;
+	unsigned long val;
+	ssize_t retval = count;
+
+	if (*offset)
+		return -EINVAL;
+	if (count > TMPBUFSIZE - 1)
+		return -EINVAL;
+
+	if (!(tmpbuf = kmalloc(TMPBUFSIZE, GFP_KERNEL)))
+		return -ENOMEM;
+
+	if (copy_from_user(tmpbuf, buf, count)) {
+		kfree(tmpbuf);
+		return -EFAULT;
+	}
+	tmpbuf[count] = 0;
+
+	mutex_lock(&pdom_mutex);
+
+	startp = tmpbuf;
+	/* Parse one more than MAX_OPROF_DOMAINS, for easy error checking */
+	for (i = 0; i <= MAX_OPROF_DOMAINS; i++) {
+		val = simple_strtoul(startp, &endp, 0);
+		if (endp == startp)
+			break;
+		while (ispunct(*endp) || isspace(*endp))
+			endp++;
+		passive_domains[i] = val;
+		if (passive_domains[i] != val)
+			/* Overflow, force error below */
+			i = MAX_OPROF_DOMAINS + 1;
+		startp = endp;
+	}
+	/* Force error on trailing junk */
+	pdomains = *startp ? MAX_OPROF_DOMAINS + 1 : i;
+
+	kfree(tmpbuf);
+
+	if (pdomains > MAX_OPROF_DOMAINS)
+	{
+		pdomains = 0;
+		retval = -EINVAL;
+		goto out;
+	}
+	
+	if(oprofile_xen_set_passive(passive_domains, pdomains))
+	{
+		pdomains = 0;
+		retval = -EINVAL;
+		goto out;
+	}
+
+out:
+	mutex_unlock(&pdom_mutex);
+	return retval;
+}
+
+static ssize_t pdomain_read(struct file * file, char __user * buf,
+			    size_t count, loff_t * offset)
+{
+	char * tmpbuf;
+	size_t len;
+	int i;
+	ssize_t retval;
+
+	if (!(tmpbuf = kmalloc(TMPBUFSIZE, GFP_KERNEL)))
+		return -ENOMEM;
+
+	mutex_lock(&pdom_mutex);
+
+	len = 0;
+	for (i = 0; i < pdomains; i++)
+		len += snprintf(tmpbuf + len,
+				len < TMPBUFSIZE ? TMPBUFSIZE - len : 0,
+				"%u ", passive_domains[i]);
+	WARN_ON(len > TMPBUFSIZE);
+	if (len != 0 && len <= TMPBUFSIZE)
+		tmpbuf[len-1] = '\n';
+
+	mutex_unlock(&pdom_mutex);
+
+	retval = simple_read_from_buffer(buf, count, offset, tmpbuf, len);
+
+	kfree(tmpbuf);
+	return retval;
+}
+
+static const struct file_operations passive_domain_ops = {
+	.read		= pdomain_read,
+	.write		= pdomain_write,
+};
+
+#endif /* CONFIG_XEN */
 void oprofile_create_files(struct super_block *sb, struct dentry *root)
 {
 	/* reinitialize default values */
@@ -132,6 +338,10 @@ void oprofile_create_files(struct super_block *sb, struct dentry *root)
 
 	oprofilefs_create_file(sb, root, "enable", &enable_fops);
 	oprofilefs_create_file_perm(sb, root, "dump", &dump_fops, 0666);
+#ifdef CONFIG_XEN
+	oprofilefs_create_file(sb, root, "active_domains", &active_domain_ops);
+	oprofilefs_create_file(sb, root, "passive_domains", &passive_domain_ops);
+#endif
 	oprofilefs_create_file(sb, root, "buffer", &event_buffer_fops);
 	oprofilefs_create_ulong(sb, root, "buffer_size", &oprofile_buffer_size);
 	oprofilefs_create_ulong(sb, root, "buffer_watershed", &oprofile_buffer_watershed);
diff --git a/drivers/xen/xenoprof/xenoprofile.c b/drivers/xen/xenoprof/xenoprofile.c
new file mode 100644
index 0000000..116b617
--- /dev/null
+++ b/drivers/xen/xenoprof/xenoprofile.c
@@ -0,0 +1,547 @@
+/**
+ * @file xenoprofile.c
+ *
+ * @remark Copyright 2002 OProfile authors
+ * @remark Read the file COPYING
+ *
+ * @author John Levon <levon@movementarian.org>
+ *
+ * Modified by Aravind Menon and Jose Renato Santos for Xen
+ * These modifications are:
+ * Copyright (C) 2005 Hewlett-Packard Co.
+ *
+ * Separated out arch-generic part
+ * Copyright (c) 2006 Isaku Yamahata <yamahata at valinux co jp>
+ *                    VA Linux Systems Japan K.K.
+ */
+
+#include <linux/init.h>
+#include <linux/notifier.h>
+#include <linux/smp.h>
+#include <linux/oprofile.h>
+#include <linux/sysdev.h>
+#include <linux/slab.h>
+#include <linux/interrupt.h>
+#include <linux/vmalloc.h>
+#include <asm/pgtable.h>
+#include <xen/evtchn.h>
+#include <xen/events.h>
+#include <xen/xenoprof.h>
+#include <xen/interface/xen.h>
+#include <xen/interface/xenoprof.h>
+#include "../../../drivers/oprofile/cpu_buffer.h"
+#include "../../../drivers/oprofile/event_buffer.h"
+
+#define MAX_XENOPROF_SAMPLES 16
+
+/* sample buffers shared with Xen */
+static xenoprof_buf_t *xenoprof_buf[MAX_VIRT_CPUS];
+/* Shared buffer area */
+static struct xenoprof_shared_buffer shared_buffer;
+
+/* Passive sample buffers shared with Xen */
+static xenoprof_buf_t *p_xenoprof_buf[MAX_OPROF_DOMAINS][MAX_VIRT_CPUS];
+/* Passive shared buffer area */
+static struct xenoprof_shared_buffer p_shared_buffer[MAX_OPROF_DOMAINS];
+
+static int xenoprof_start(void);
+static void xenoprof_stop(void);
+
+static int xenoprof_enabled = 0;
+static int xenoprof_is_primary = 0;
+static int active_defined;
+
+extern unsigned long oprofile_backtrace_depth;
+
+/* Number of buffers in shared area (one per VCPU) */
+static int nbuf;
+/* Mappings of VIRQ_XENOPROF to irq number (per cpu) */
+static int ovf_irq[NR_CPUS];
+/* cpu model type string - copied from Xen on XENOPROF_init command */
+static char cpu_type[XENOPROF_CPU_TYPE_SIZE];
+
+#ifdef CONFIG_PM
+
+static int xenoprof_suspend(struct sys_device * dev, pm_message_t state)
+{
+	if (xenoprof_enabled == 1)
+		xenoprof_stop();
+	return 0;
+}
+
+
+static int xenoprof_resume(struct sys_device * dev)
+{
+	if (xenoprof_enabled == 1)
+		xenoprof_start();
+	return 0;
+}
+
+
+static struct sysdev_class oprofile_sysclass = {
+	.name 		= "oprofile",
+	.resume		= xenoprof_resume,
+	.suspend	= xenoprof_suspend
+};
+
+
+static struct sys_device device_oprofile = {
+	.id	= 0,
+	.cls	= &oprofile_sysclass,
+};
+
+
+static int __init init_driverfs(void)
+{
+	int error;
+	if (!(error = sysdev_class_register(&oprofile_sysclass)))
+		error = sysdev_register(&device_oprofile);
+	return error;
+}
+
+
+static void exit_driverfs(void)
+{
+	sysdev_unregister(&device_oprofile);
+	sysdev_class_unregister(&oprofile_sysclass);
+}
+
+#else
+#define init_driverfs() do { } while (0)
+#define exit_driverfs() do { } while (0)
+#endif /* CONFIG_PM */
+
+static unsigned long long oprofile_samples;
+static unsigned long long p_oprofile_samples;
+
+static unsigned int pdomains;
+static struct xenoprof_passive passive_domains[MAX_OPROF_DOMAINS];
+
+/* Check whether the given entry is an escape code */
+static int xenoprof_is_escape(xenoprof_buf_t * buf, int tail)
+{
+	return (buf->event_log[tail].eip == XENOPROF_ESCAPE_CODE);
+}
+
+/* Get the event at the given entry  */
+static uint8_t xenoprof_get_event(xenoprof_buf_t * buf, int tail)
+{
+	return (buf->event_log[tail].event);
+}
+
+static void xenoprof_add_pc(xenoprof_buf_t *buf, int is_passive)
+{
+	int head, tail, size;
+	int tracing = 0;
+
+	head = buf->event_head;
+	tail = buf->event_tail;
+	size = buf->event_size;
+
+	while (tail != head) {
+		if (xenoprof_is_escape(buf, tail) &&
+		    xenoprof_get_event(buf, tail) == XENOPROF_TRACE_BEGIN) {
+			tracing=1;
+			oprofile_add_pc(ESCAPE_CODE, buf->event_log[tail].mode, 
+					TRACE_BEGIN); 
+			if (!is_passive)
+				oprofile_samples++;
+			else
+				p_oprofile_samples++;
+			
+		} else {
+			oprofile_add_pc(buf->event_log[tail].eip,
+					buf->event_log[tail].mode,
+					buf->event_log[tail].event);
+			if (!tracing) {
+				if (!is_passive)
+					oprofile_samples++;
+				else
+					p_oprofile_samples++;
+			}
+       
+		}
+		tail++;
+		if(tail==size)
+		    tail=0;
+	}
+	buf->event_tail = tail;
+}
+
+static void xenoprof_handle_passive(void)
+{
+	int i, j;
+	int flag_domain, flag_switch = 0;
+	
+	for (i = 0; i < pdomains; i++) {
+		flag_domain = 0;
+		for (j = 0; j < passive_domains[i].nbuf; j++) {
+			xenoprof_buf_t *buf = p_xenoprof_buf[i][j];
+			if (buf->event_head == buf->event_tail)
+				continue;
+			if (!flag_domain) {
+				if (!oprofile_add_domain_switch(
+					passive_domains[i].domain_id))
+					goto done;
+				flag_domain = 1;
+			}
+			xenoprof_add_pc(buf, 1);
+			flag_switch = 1;
+		}
+	}
+done:
+	if (flag_switch)
+		oprofile_add_domain_switch(XEN_COORDINATOR_DOMAIN);
+}
+
+static irqreturn_t 
+xenoprof_ovf_interrupt(int irq, void * dev_id)
+{
+	struct xenoprof_buf * buf;
+	static unsigned long flag;
+
+	buf = xenoprof_buf[smp_processor_id()];
+
+	xenoprof_add_pc(buf, 0);
+
+	if (xenoprof_is_primary && !test_and_set_bit(0, &flag)) {
+		xenoprof_handle_passive();
+		smp_mb__before_clear_bit();
+		clear_bit(0, &flag);
+	}
+
+	return IRQ_HANDLED;
+}
+
+
+static void unbind_virq(void)
+{
+	unsigned int i;
+
+	for_each_online_cpu(i) {
+		if (ovf_irq[i] >= 0) {
+			unbind_from_irqhandler(ovf_irq[i], NULL);
+			ovf_irq[i] = -1;
+		}
+	}
+}
+
+
+static int bind_virq(void)
+{
+	unsigned int i;
+	int result;
+
+	for_each_online_cpu(i) {
+		result = bind_virq_to_irqhandler(VIRQ_XENOPROF,
+						 i,
+						 xenoprof_ovf_interrupt,
+						 IRQF_DISABLED|IRQF_NOBALANCING,
+						 "xenoprof",
+						 NULL);
+
+		if (result < 0) {
+			unbind_virq();
+			return result;
+		}
+
+		ovf_irq[i] = result;
+	}
+		
+	return 0;
+}
+
+
+static void unmap_passive_list(void)
+{
+	int i;
+	for (i = 0; i < pdomains; i++)
+		xenoprof_arch_unmap_shared_buffer(&p_shared_buffer[i]);
+	pdomains = 0;
+}
+
+
+static int map_xenoprof_buffer(int max_samples)
+{
+	struct xenoprof_get_buffer get_buffer;
+	struct xenoprof_buf *buf;
+	int ret, i;
+
+	if ( shared_buffer.buffer )
+		return 0;
+
+	get_buffer.max_samples = max_samples;
+	ret = xenoprof_arch_map_shared_buffer(&get_buffer, &shared_buffer);
+	if (ret)
+		return ret;
+	nbuf = get_buffer.nbuf;
+
+	for (i=0; i< nbuf; i++) {
+		buf = (struct xenoprof_buf*) 
+			&shared_buffer.buffer[i * get_buffer.bufsize];
+		BUG_ON(buf->vcpu_id >= MAX_VIRT_CPUS);
+		xenoprof_buf[buf->vcpu_id] = buf;
+	}
+
+	return 0;
+}
+
+
+static int xenoprof_setup(void)
+{
+	int ret;
+
+	if ( (ret = map_xenoprof_buffer(MAX_XENOPROF_SAMPLES)) )
+		return ret;
+
+	if ( (ret = bind_virq()) )
+		return ret;
+
+	if (xenoprof_is_primary) {
+		/* Define dom0 as an active domain if not done yet */
+		if (!active_defined) {
+			domid_t domid;
+			ret = HYPERVISOR_xenoprof_op(
+				XENOPROF_reset_active_list, NULL);
+			if (ret)
+				goto err;
+			domid = 0;
+			ret = HYPERVISOR_xenoprof_op(
+				XENOPROF_set_active, &domid);
+			if (ret)
+				goto err;
+			active_defined = 1;
+		}
+
+		if (oprofile_backtrace_depth > 0) {
+			ret = HYPERVISOR_xenoprof_op(XENOPROF_set_backtrace, 
+						     &oprofile_backtrace_depth);
+			if (ret)
+				oprofile_backtrace_depth = 0;
+		}
+
+		ret = HYPERVISOR_xenoprof_op(XENOPROF_reserve_counters, NULL);
+		if (ret)
+			goto err;
+		
+		xenoprof_arch_counter();
+		ret = HYPERVISOR_xenoprof_op(XENOPROF_setup_events, NULL);
+		if (ret)
+			goto err;
+	}
+
+	ret = HYPERVISOR_xenoprof_op(XENOPROF_enable_virq, NULL);
+	if (ret)
+		goto err;
+
+	xenoprof_enabled = 1;
+	return 0;
+ err:
+	unbind_virq();
+	return ret;
+}
+
+
+static void xenoprof_shutdown(void)
+{
+	xenoprof_enabled = 0;
+
+	WARN_ON(HYPERVISOR_xenoprof_op(XENOPROF_disable_virq, NULL));
+
+	if (xenoprof_is_primary) {
+		WARN_ON(HYPERVISOR_xenoprof_op(XENOPROF_release_counters,
+					       NULL));
+		active_defined = 0;
+	}
+
+	unbind_virq();
+
+	xenoprof_arch_unmap_shared_buffer(&shared_buffer);
+	if (xenoprof_is_primary)
+		unmap_passive_list();
+}
+
+
+static int xenoprof_start(void)
+{
+	int ret = 0;
+
+	if (xenoprof_is_primary)
+		ret = HYPERVISOR_xenoprof_op(XENOPROF_start, NULL);
+	if (!ret)
+		xenoprof_arch_start();
+	return ret;
+}
+
+
+static void xenoprof_stop(void)
+{
+	if (xenoprof_is_primary)
+		WARN_ON(HYPERVISOR_xenoprof_op(XENOPROF_stop, NULL));
+	xenoprof_arch_stop();
+}
+
+
+static int xenoprof_set_active(int * active_domains,
+			       unsigned int adomains)
+{
+	int ret = 0;
+	int i;
+	int set_dom0 = 0;
+	domid_t domid;
+
+	if (!xenoprof_is_primary)
+		return 0;
+
+	if (adomains > MAX_OPROF_DOMAINS)
+		return -E2BIG;
+
+	ret = HYPERVISOR_xenoprof_op(XENOPROF_reset_active_list, NULL);
+	if (ret)
+		return ret;
+
+	for (i=0; i<adomains; i++) {
+		domid = active_domains[i];
+		if (domid != active_domains[i]) {
+			ret = -EINVAL;
+			goto out;
+		}
+		ret = HYPERVISOR_xenoprof_op(XENOPROF_set_active, &domid);
+		if (ret)
+			goto out;
+		if (active_domains[i] == 0)
+			set_dom0 = 1;
+	}
+	/* dom0 must always be active but may not be in the list */ 
+	if (!set_dom0) {
+		domid = 0;
+		ret = HYPERVISOR_xenoprof_op(XENOPROF_set_active, &domid);
+	}
+
+out:
+	if (ret)
+		WARN_ON(HYPERVISOR_xenoprof_op(XENOPROF_reset_active_list,
+					       NULL));
+	active_defined = !ret;
+	return ret;
+}
+
+static int xenoprof_set_passive(int * p_domains,
+                                unsigned int pdoms)
+{
+	int ret;
+	unsigned int i, j;
+	struct xenoprof_buf *buf;
+
+	if (!xenoprof_is_primary)
+        	return 0;
+
+	if (pdoms > MAX_OPROF_DOMAINS)
+		return -E2BIG;
+
+	ret = HYPERVISOR_xenoprof_op(XENOPROF_reset_passive_list, NULL);
+	if (ret)
+		return ret;
+	unmap_passive_list();
+
+	for (i = 0; i < pdoms; i++) {
+		passive_domains[i].domain_id = p_domains[i];
+		passive_domains[i].max_samples = 2048;
+		ret = xenoprof_arch_set_passive(&passive_domains[i],
+						&p_shared_buffer[i]);
+		if (ret)
+		{
+			goto out;
+		}
+		for (j = 0; j < passive_domains[i].nbuf; j++) {
+			buf = (struct xenoprof_buf *)
+				&p_shared_buffer[i].buffer[
+				j * passive_domains[i].bufsize];
+			BUG_ON(buf->vcpu_id >= MAX_VIRT_CPUS);
+			p_xenoprof_buf[i][buf->vcpu_id] = buf;
+		}
+	}
+
+	pdomains = pdoms;
+	return 0;
+
+out:
+	for (j = 0; j < i; j++)
+		xenoprof_arch_unmap_shared_buffer(&p_shared_buffer[i]);
+
+	return ret;
+}
+
+
+/* The dummy backtrace function to keep oprofile happy
+ * The real backtrace is done in xen
+ */
+static void xenoprof_dummy_backtrace(struct pt_regs * const regs, 
+				     unsigned int depth)
+{
+	/* this should never be called */
+	BUG();
+	return;
+}
+
+
+static struct oprofile_operations xenoprof_ops = {
+#ifdef HAVE_XENOPROF_CREATE_FILES
+	.create_files 	= xenoprof_create_files,
+#endif
+	.xen_set_active	= xenoprof_set_active,
+	.xen_set_passive    = xenoprof_set_passive,
+	.setup 		= xenoprof_setup,
+	.shutdown	= xenoprof_shutdown,
+	.start		= xenoprof_start,
+	.stop		= xenoprof_stop,
+	.backtrace	= xenoprof_dummy_backtrace
+};
+
+
+/* in order to get driverfs right */
+static int using_xenoprof;
+
+int __init xenoprofile_init(struct oprofile_operations * ops)
+{
+	struct xenoprof_init init;
+	unsigned int i;
+	int ret;
+
+	ret = HYPERVISOR_xenoprof_op(XENOPROF_init, &init);
+	if (!ret) {
+		xenoprof_arch_init_counter(&init);
+		xenoprof_is_primary = init.is_primary;
+
+		/*  cpu_type is detected by Xen */
+		cpu_type[XENOPROF_CPU_TYPE_SIZE-1] = 0;
+		strncpy(cpu_type, init.cpu_type, XENOPROF_CPU_TYPE_SIZE - 1);
+		xenoprof_ops.cpu_type = cpu_type;
+
+		init_driverfs();
+		using_xenoprof = 1;
+		*ops = xenoprof_ops;
+
+		for (i=0; i<NR_CPUS; i++)
+			ovf_irq[i] = -1;
+
+		active_defined = 0;
+	}
+
+	printk(KERN_INFO "%s: ret %d, events %d, xenoprof_is_primary %d\n",
+	       __func__, ret, init.num_events, xenoprof_is_primary);
+	return ret;
+}
+
+
+void xenoprofile_exit(void)
+{
+	if (using_xenoprof)
+		exit_driverfs();
+
+	xenoprof_arch_unmap_shared_buffer(&shared_buffer);
+	if (xenoprof_is_primary) {
+		unmap_passive_list();
+		WARN_ON(HYPERVISOR_xenoprof_op(XENOPROF_shutdown, NULL));
+        }
+}
diff --git a/include/linux/oprofile.h b/include/linux/oprofile.h
index 1d9518b..bd55065 100644
--- a/include/linux/oprofile.h
+++ b/include/linux/oprofile.h
@@ -16,6 +16,9 @@
 #include <linux/types.h>
 #include <linux/spinlock.h>
 #include <asm/atomic.h>
+#ifdef CONFIG_XEN
+#include <xen/interface/xenoprof.h>
+#endif
  
 /* Each escaped entry is prefixed by ESCAPE_CODE
  * then one of the following codes, then the
@@ -28,14 +31,18 @@
 #define CPU_SWITCH_CODE			2
 #define COOKIE_SWITCH_CODE		3
 #define KERNEL_ENTER_SWITCH_CODE	4
-#define KERNEL_EXIT_SWITCH_CODE		5
+#define USER_ENTER_SWITCH_CODE		5
 #define MODULE_LOADED_CODE		6
 #define CTX_TGID_CODE			7
 #define TRACE_BEGIN_CODE		8
 #define TRACE_END_CODE			9
 #define XEN_ENTER_SWITCH_CODE		10
+#ifndef CONFIG_XEN
 #define SPU_PROFILING_CODE		11
 #define SPU_CTX_SWITCH_CODE		12
+#else
+#define XEN_DOMAIN_SWITCH_CODE		11
+#endif
 #define IBS_FETCH_CODE			13
 #define IBS_OP_CODE			14
 
@@ -49,6 +56,12 @@ struct oprofile_operations {
 	/* create any necessary configuration files in the oprofile fs.
 	 * Optional. */
 	int (*create_files)(struct super_block * sb, struct dentry * root);
+#ifdef CONFIG_XEN
+	/* setup active domains with Xen */
+	int (*xen_set_active)(int *active_domains, unsigned int adomains);
+	/* setup passive domains with Xen */
+	int (*xen_set_passive)(int *passive_domains, unsigned int pdomains);
+#endif
 	/* Do any necessary interrupt setup. Optional. */
 	int (*setup)(void);
 	/* Do any necessary interrupt shutdown. Optional. */
@@ -104,9 +117,16 @@ void oprofile_add_ext_sample(unsigned long pc, struct pt_regs * const regs,
  * backtrace. */
 void oprofile_add_pc(unsigned long pc, int is_kernel, unsigned long event);
 
+/* Record when the cpu mode switches between user/kernel/xen(hypervisor) */
+void oprofile_add_mode(int cpu_mode);
+
 /* add a backtrace entry, to be called from the ->backtrace callback */
 void oprofile_add_trace(unsigned long eip);
 
+#ifdef CONFIG_XEN
+/* add a xen domain switch entry */
+int oprofile_add_domain_switch(int32_t domain_id);
+#endif
 
 /**
  * Create a file of the given name as a child of the given root, with
diff --git a/include/xen/interface/xen.h b/include/xen/interface/xen.h
index 812ffd5..0054a3f 100644
--- a/include/xen/interface/xen.h
+++ b/include/xen/interface/xen.h
@@ -79,6 +79,8 @@
 #define VIRQ_CONSOLE    2  /* (DOM0) Bytes received on emergency console. */
 #define VIRQ_DOM_EXC    3  /* (DOM0) Exceptional event for some domain.   */
 #define VIRQ_DEBUGGER   6  /* (DOM0) A domain has paused for debugging.   */
+#define VIRQ_XENOPROF   7  /* V. XenOprofile interrupt: new sample available */
+
 
 /* Architecture-specific VIRQ definitions. */
 #define VIRQ_ARCH_0    16
diff --git a/include/xen/interface/xenoprof.h b/include/xen/interface/xenoprof.h
new file mode 100644
index 0000000..8ff3a56
--- /dev/null
+++ b/include/xen/interface/xenoprof.h
@@ -0,0 +1,140 @@
+/******************************************************************************
+ * xenoprof.h
+ * 
+ * Interface for enabling system wide profiling based on hardware performance
+ * counters
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a copy
+ * of this software and associated documentation files (the "Software"), to
+ * deal in the Software without restriction, including without limitation the
+ * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
+ * sell copies of the Software, and to permit persons to whom the Software is
+ * furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+ * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
+ * DEALINGS IN THE SOFTWARE.
+ *
+ * Copyright (C) 2005 Hewlett-Packard Co.
+ * Written by Aravind Menon & Jose Renato Santos
+ */
+
+#ifndef __XEN_PUBLIC_XENOPROF_H__
+#define __XEN_PUBLIC_XENOPROF_H__
+
+#include "xen.h"
+
+/*
+ * Commands to HYPERVISOR_xenoprof_op().
+ */
+#define XENOPROF_init                0
+#define XENOPROF_reset_active_list   1
+#define XENOPROF_reset_passive_list  2
+#define XENOPROF_set_active          3
+#define XENOPROF_set_passive         4
+#define XENOPROF_reserve_counters    5
+#define XENOPROF_counter             6
+#define XENOPROF_setup_events        7
+#define XENOPROF_enable_virq         8
+#define XENOPROF_start               9
+#define XENOPROF_stop               10
+#define XENOPROF_disable_virq       11
+#define XENOPROF_release_counters   12
+#define XENOPROF_shutdown           13
+#define XENOPROF_get_buffer         14
+#define XENOPROF_set_backtrace      15
+#define XENOPROF_last_op            15
+
+#define MAX_OPROF_EVENTS    32
+#define MAX_OPROF_DOMAINS   25
+#define XENOPROF_CPU_TYPE_SIZE 64
+
+#define DEFINE_XEN_GUEST_HANDLE(x)
+
+/* Xenoprof performance events (not Xen events) */
+struct event_log {
+    uint64_t eip;
+    uint8_t mode;
+    uint8_t event;
+};
+
+/* PC value that indicates a special code */
+#define XENOPROF_ESCAPE_CODE ~0UL
+/* Transient events for the xenoprof->oprofile cpu buf */
+#define XENOPROF_TRACE_BEGIN 1
+
+/* Xenoprof buffer shared between Xen and domain - 1 per VCPU */
+struct xenoprof_buf {
+    uint32_t event_head;
+    uint32_t event_tail;
+    uint32_t event_size;
+    uint32_t vcpu_id;
+    uint64_t xen_samples;
+    uint64_t kernel_samples;
+    uint64_t user_samples;
+    uint64_t lost_samples;
+    struct event_log event_log[1];
+};
+#ifndef __XEN__
+typedef struct xenoprof_buf xenoprof_buf_t;
+DEFINE_XEN_GUEST_HANDLE(xenoprof_buf_t);
+#endif
+
+struct xenoprof_init {
+    int32_t  num_events;
+    int32_t  is_primary;
+    char cpu_type[XENOPROF_CPU_TYPE_SIZE];
+};
+typedef struct xenoprof_init xenoprof_init_t;
+DEFINE_XEN_GUEST_HANDLE(xenoprof_init_t);
+
+struct xenoprof_get_buffer {
+    int32_t  max_samples;
+    int32_t  nbuf;
+    int32_t  bufsize;
+    uint64_t buf_gmaddr;
+};
+typedef struct xenoprof_get_buffer xenoprof_get_buffer_t;
+DEFINE_XEN_GUEST_HANDLE(xenoprof_get_buffer_t);
+
+struct xenoprof_counter {
+    uint32_t ind;
+    uint64_t count;
+    uint32_t enabled;
+    uint32_t event;
+    uint32_t hypervisor;
+    uint32_t kernel;
+    uint32_t user;
+    uint64_t unit_mask;
+};
+typedef struct xenoprof_counter xenoprof_counter_t;
+DEFINE_XEN_GUEST_HANDLE(xenoprof_counter_t);
+
+typedef struct xenoprof_passive {
+    uint16_t domain_id;
+    int32_t  max_samples;
+    int32_t  nbuf;
+    int32_t  bufsize;
+    uint64_t buf_gmaddr;
+} xenoprof_passive_t;
+DEFINE_XEN_GUEST_HANDLE(xenoprof_passive_t);
+
+
+#endif /* __XEN_PUBLIC_XENOPROF_H__ */
+
+/*
+ * Local variables:
+ * mode: C
+ * c-set-style: "BSD"
+ * c-basic-offset: 4
+ * tab-width: 4
+ * indent-tabs-mode: nil
+ * End:
+ */
diff --git a/include/xen/xen-ops.h b/include/xen/xen-ops.h
index 9769738..1922ef3 100644
--- a/include/xen/xen-ops.h
+++ b/include/xen/xen-ops.h
@@ -24,5 +24,8 @@ int xen_remap_domain_mfn_range(struct vm_area_struct *vma,
 			       unsigned long mfn, int nr,
 			       pgprot_t prot, unsigned domid);
 
+int xen_remap_domain_kernel_mfn_range(unsigned long addr,
+			       unsigned long mfn, int nr,
+			       pgprot_t prot, unsigned domid);
 
 #endif /* INCLUDE_XEN_OPS_H */
diff --git a/include/xen/xenoprof.h b/include/xen/xenoprof.h
new file mode 100644
index 0000000..2a9a119
--- /dev/null
+++ b/include/xen/xenoprof.h
@@ -0,0 +1,69 @@
+/******************************************************************************
+ * xen/xenoprof.h
+ *
+ * Copyright (c) 2006 Isaku Yamahata <yamahata at valinux co jp>
+ *                    VA Linux Systems Japan K.K.
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; either version 2 of the License, or
+ * (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
+ *
+ */
+
+#ifndef __XEN_XENOPROF_H__
+#define __XEN_XENOPROF_H__
+#ifdef CONFIG_XEN
+
+#if 0
+#include <asm/xenoprof.h>
+#endif
+
+#if defined(CONFIG_X86) || defined(CONFIG_X86_64)
+/* xenoprof x86 specific */
+struct super_block;
+struct dentry;
+int xenoprof_create_files(struct super_block * sb, struct dentry * root);
+#define HAVE_XENOPROF_CREATE_FILES
+
+struct xenoprof_init;
+void xenoprof_arch_init_counter(struct xenoprof_init *init);
+void xenoprof_arch_counter(void);
+void xenoprof_arch_start(void);
+void xenoprof_arch_stop(void);
+
+struct xenoprof_arch_shared_buffer {
+	/* nothing */
+};
+struct xenoprof_shared_buffer;
+void xenoprof_arch_unmap_shared_buffer(struct xenoprof_shared_buffer* sbuf);
+struct xenoprof_get_buffer;
+int xenoprof_arch_map_shared_buffer(struct xenoprof_get_buffer* get_buffer, struct xenoprof_shared_buffer* sbuf);
+struct xenoprof_passive;
+int xenoprof_arch_set_passive(struct xenoprof_passive* pdomain, struct xenoprof_shared_buffer* sbuf);
+#endif
+
+/* xenoprof common */
+struct oprofile_operations;
+int xenoprofile_init(struct oprofile_operations * ops);
+void xenoprofile_exit(void);
+
+struct xenoprof_shared_buffer {
+	char					*buffer;
+	struct xenoprof_arch_shared_buffer	arch;
+};
+#else
+#define xenoprofile_init(ops)	(-ENOSYS)
+#define xenoprofile_exit()	do { } while (0)
+
+#endif /* CONFIG_XEN */
+#endif /* __XEN_XENOPROF_H__ */

[-- Attachment #3: Type: text/plain, Size: 138 bytes --]

_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel

Re: performance tools on xen

[Jeremy Fitzhardinge]

On 11/04/09 04:18, Dulloor wrote:
> I have attached the patch that will apply to the current pvops tree.
> Tested both on 32-bit and 64-bit platforms.
>
> I lost some of the cleanups that I had done to get rid of some
> (CONFIG_XEN)ifdefs in a couple of files in oprofile directory. But, I
> will work on them incrementally, if this is checked in.
>   

Thanks, I've committed it to the xen/oprofile branch.  I haven't merged
it into xen/master yet.

    J

Re: performance tools on xen

[Fasiha Ashraf]

[-- Attachment #1.1: Type: text/plain, Size: 2078 bytes --]

Please help me starting profiling my xen3.4.1 with pv kernel 2.6.31.4 on fedora11 platform. I have confusions in installing and using latest version of Oprofile with xenoprof.

Regards,
Fasiha Ashraf

--- On Wed, 4/11/09, Dulloor <dulloor@gmail.com> wrote:

From: Dulloor <dulloor@gmail.com>
Subject: Re: [Xen-devel] performance tools on xen
To: "Jeremy Fitzhardinge" <jeremy@goop.org>
Cc: "Xen-devel" <xen-devel@lists.xensource.com>, "Santos, Jose Renato G" <joserenato.santos@hp.com>, "Dante Cinco" <dantecinco@gmail.com>
Date: Wednesday, 4 November, 2009, 5:18 PM

I have attached the patch that will apply to the current pvops tree.
Tested both on 32-bit and 64-bit platforms.

I lost some of the cleanups that I had done to get rid of some
(CONFIG_XEN)ifdefs in a couple of files in oprofile directory. But, I
will work on them incrementally, if this is checked in.

-dulloor

On Wed, Nov 4, 2009 at 12:46 AM, Dulloor <dulloor@gmail.com> wrote:
> Jeremy, I thought you were able to access my git repo (per your last
> mail). I should be having the patches somewhere around. Will send them
> over soon.
>
> -dulloor
>
> On Mon, Nov 2, 2009 at 1:47 PM, Jeremy Fitzhardinge <jeremy@goop.org> wrote:
>> On 11/02/09 09:56, Dante Cinco wrote:
>>> Jeremy,
>>>
>>> I searched Mark Mail for "xenoprof pvops" and found this email from
>>> dulloor:
>>>
>>> http://markmail.org/search/?q=xenoprof+pvops#query:xenoprof%20pvops+page:1+mid:yyzbymymat3rtivo+state:results
>>>
>>> Do you know if dulloor's patches for xenoprof for linux pvops got applied?
>>
>> No, they fell through the cracks; I think there was some problem that I
>> could never connect to his git server.
>>
>> Dulloor, do you still have a current set of patched we can try?
>>
>> Thanks,
>>    J
>>
>

-----Inline Attachment Follows-----

_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel



      Yahoo! India has a new look. Take a sneak peek http://in.yahoo.com/trynew

[-- Attachment #1.2: Type: text/html, Size: 3275 bytes --]

[-- Attachment #2: Type: text/plain, Size: 138 bytes --]

_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel

Re: performance tools on xen

[Dulloor]

[-- Attachment #1.1: Type: text/plain, Size: 2789 bytes --]

Even the latest version of oprofile doesn't have the passive-domain
profiling option, which I personally prefer. You could get the patches from
http://xenoprof.sourceforge.net
A useful guide for xenoprof is
http://xen.org/files/summit_3/xenoprof_tutorial.pdf

Maybe it's time those changes are checked-in too (into oprofile). But, Jose
knows better.

-dulloor

On Thu, Nov 5, 2009 at 4:12 AM, Fasiha Ashraf <feehapk@yahoo.co.in> wrote:

> Please help me starting profiling my xen3.4.1 with pv kernel 2.6.31.4 on
> fedora11 platform. I have confusions in installing and using latest version
> of Oprofile with xenoprof.
>
> Regards,
> Fasiha Ashraf
>
> --- On *Wed, 4/11/09, Dulloor <dulloor@gmail.com>* wrote:
>
>
> From: Dulloor <dulloor@gmail.com>
> Subject: Re: [Xen-devel] performance tools on xen
> To: "Jeremy Fitzhardinge" <jeremy@goop.org>
> Cc: "Xen-devel" <xen-devel@lists.xensource.com>, "Santos, Jose Renato G" <
> joserenato.santos@hp.com>, "Dante Cinco" <dantecinco@gmail.com>
> Date: Wednesday, 4 November, 2009, 5:18 PM
>
>
> I have attached the patch that will apply to the current pvops tree.
> Tested both on 32-bit and 64-bit platforms.
>
> I lost some of the cleanups that I had done to get rid of some
> (CONFIG_XEN)ifdefs in a couple of files in oprofile directory. But, I
> will work on them incrementally, if this is checked in.
>
> -dulloor
>
> On Wed, Nov 4, 2009 at 12:46 AM, Dulloor <dulloor@gmail.com<http://mc/compose?to=dulloor@gmail.com>>
> wrote:
> > Jeremy, I thought you were able to access my git repo (per your last
> > mail). I should be having the patches somewhere around. Will send them
> > over soon.
> >
> > -dulloor
> >
> > On Mon, Nov 2, 2009 at 1:47 PM, Jeremy Fitzhardinge <jeremy@goop.org<http://mc/compose?to=jeremy@goop.org>>
> wrote:
> >> On 11/02/09 09:56, Dante Cinco wrote:
> >>> Jeremy,
> >>>
> >>> I searched Mark Mail for "xenoprof pvops" and found this email from
> >>> dulloor:
> >>>
> >>>
> http://markmail.org/search/?q=xenoprof+pvops#query:xenoprof%20pvops+page:1+mid:yyzbymymat3rtivo+state:results
> >>>
> >>> Do you know if dulloor's patches for xenoprof for linux pvops got
> applied?
> >>
> >> No, they fell through the cracks; I think there was some problem that I
> >> could never connect to his git server.
> >>
> >> Dulloor, do you still have a current set of patched we can try?
> >>
> >> Thanks,
> >>    J
> >>
> >
>
> -----Inline Attachment Follows-----
>
>
> _______________________________________________
> Xen-devel mailing list
> Xen-devel@lists.xensource.com<http://mc/compose?to=Xen-devel@lists.xensource.com>
> http://lists.xensource.com/xen-devel
>
>
> ------------------------------
> Try the new Yahoo! India Homepage. Click here<http://in.rd.yahoo.com/tagline_metro_1/*http://in.yahoo.com/trynew>
> .
>

[-- Attachment #1.2: Type: text/html, Size: 4695 bytes --]

[-- Attachment #2: Type: text/plain, Size: 138 bytes --]

_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel

RE: performance tools on xen

[Santos, Jose Renato G]

[-- Attachment #1.1: Type: text/plain, Size: 3936 bytes --]

Having the Xen passive domain support not integrated into mainline OProfile user level tools is a pain.
The problem is that OProfile mantainers are not not happy with the way passive domain samples are stored today (as fake modules named domainX, domainY, etc...) which I agree is ugly . We need to do this properly and separate passive domains samples using OProfile sample directory hierarchy in the same way different CPUs samples are stored. This requires some reasonable amount of work and I don't have the time to work on this myself.
I have been looking for some volunteer to work on that for a while but nobody had taken that task yet.
I would be happy to help anyone who is willing to work on that. Any volunteers?:)

Renato

________________________________
From: Dulloor [mailto:dulloor@gmail.com]
Sent: Thursday, November 05, 2009 1:21 AM
To: Fasiha Ashraf
Cc: Jeremy Fitzhardinge; Xen-devel; Santos, Jose Renato G; Dante Cinco
Subject: Re: [Xen-devel] performance tools on xen

Even the latest version of oprofile doesn't have the passive-domain profiling option, which I personally prefer. You could get the patches from http://xenoprof.sourceforge.net
A useful guide for xenoprof is http://xen.org/files/summit_3/xenoprof_tutorial.pdf

Maybe it's time those changes are checked-in too (into oprofile). But, Jose knows better.

-dulloor

On Thu, Nov 5, 2009 at 4:12 AM, Fasiha Ashraf <feehapk@yahoo.co.in<mailto:feehapk@yahoo.co.in>> wrote:
Please help me starting profiling my xen3.4.1 with pv kernel 2.6.31.4 on fedora11 platform. I have confusions in installing and using latest version of Oprofile with xenoprof.

Regards,
Fasiha Ashraf

--- On Wed, 4/11/09, Dulloor <dulloor@gmail.com<mailto:dulloor@gmail.com>> wrote:

From: Dulloor <dulloor@gmail.com<mailto:dulloor@gmail.com>>
Subject: Re: [Xen-devel] performance tools on xen
To: "Jeremy Fitzhardinge" <jeremy@goop.org<mailto:jeremy@goop.org>>
Cc: "Xen-devel" <xen-devel@lists.xensource.com<mailto:xen-devel@lists.xensource.com>>, "Santos, Jose Renato G" <joserenato.santos@hp.com<mailto:joserenato.santos@hp.com>>, "Dante Cinco" <dantecinco@gmail.com<mailto:dantecinco@gmail.com>>
Date: Wednesday, 4 November, 2009, 5:18 PM


I have attached the patch that will apply to the current pvops tree.
Tested both on 32-bit and 64-bit platforms.

I lost some of the cleanups that I had done to get rid of some
(CONFIG_XEN)ifdefs in a couple of files in oprofile directory. But, I
will work on them incrementally, if this is checked in.

-dulloor

On Wed, Nov 4, 2009 at 12:46 AM, Dulloor <dulloor@gmail.com<http://mc/compose?to=dulloor@gmail.com>> wrote:
> Jeremy, I thought you were able to access my git repo (per your last
> mail). I should be having the patches somewhere around. Will send them
> over soon.
>
> -dulloor
>
> On Mon, Nov 2, 2009 at 1:47 PM, Jeremy Fitzhardinge <jeremy@goop.org<http://mc/compose?to=jeremy@goop.org>> wrote:
>> On 11/02/09 09:56, Dante Cinco wrote:
>>> Jeremy,
>>>
>>> I searched Mark Mail for "xenoprof pvops" and found this email from
>>> dulloor:
>>>
>>> http://markmail.org/search/?q=xenoprof+pvops#query:xenoprof%20pvops+page:1+mid:yyzbymymat3rtivo+state:results
>>>
>>> Do you know if dulloor's patches for xenoprof for linux pvops got applied?
>>
>> No, they fell through the cracks; I think there was some problem that I
>> could never connect to his git server.
>>
>> Dulloor, do you still have a current set of patched we can try?
>>
>> Thanks,
>>    J
>>
>

-----Inline Attachment Follows-----


_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com<http://mc/compose?to=Xen-devel@lists.xensource.com>
http://lists.xensource.com/xen-devel


________________________________
Try the new Yahoo! India Homepage. Click here<http://in.rd.yahoo.com/tagline_metro_1/*http://in.yahoo.com/trynew>.


[-- Attachment #1.2: Type: text/html, Size: 7612 bytes --]

[-- Attachment #2: Type: text/plain, Size: 138 bytes --]

_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel

Re: performance tools on xen

[Mukesh Rathor]

On Thu, 29 Oct 2009 14:10:01 +0000
George Dunlap <dunlapg@gmail.com> wrote:

> On Thu, Oct 29, 2009 at 7:29 AM, Keir Fraser
> <keir.fraser@eu.citrix.com> wrote:
> > On 29/10/2009 06:04, "Juergen Gross" <juergen.gross@ts.fujitsu.com>
> > wrote:
> >
> >>> So far I've found:
> >>>   - Xenoprof : is this working currently for 64bit VMs on 32bit
> >>> dom0?
> >>>   - xentop based data
> >>>   - xenmon: does this support SMP?
> >>
> >> xenlockprof    lock profiling of the hypervisor
> >> xenperf        software counters of the hypervisor
> >>
> >> Both need to be enabled at compile time of the hypervisor.
> >
> > Also xentrace, for trace-based analysis.
> 
> And xenalyze, to analyze the resulting traces:
> 
> http://xenbits.xensource.com/ext/xenalyze.hg
> 
>  -George


Thanks all. After I get familiar with them, I'll post a summary on a 
blog, or create a 'xen performance tools' page somewhere.

mukesh

RE: performance tools on xen

[Santos, Jose Renato G]

 

> -----Original Message-----
> From: xen-devel-bounces@lists.xensource.com 
> [mailto:xen-devel-bounces@lists.xensource.com] On Behalf Of 
> Mukesh Rathor
> Sent: Wednesday, October 28, 2009 8:07 PM
> To: Xen-devel@lists.xensource.com
> Subject: [Xen-devel] performance tools on xen
> 
> Hi guys,
> 
> I'm trying to find available tools on xen to do performance
> analysis: hypervisor + dom0 + domU. I'm thinking CPU usage, 
> lock contentions, page faults, etc type of info.
> 
> So far I've found:
>   - Xenoprof : is this working currently for 64bit VMs on 32bit dom0?

  I am afraid not. If you are interested in working on this I would be glad to help. :)

Renato

>   - xentop based data
>   - xenmon: does this support SMP? 
> 
> 
> thanks,
> Mukesh
> 
> 
> 
> _______________________________________________
> Xen-devel mailing list
> Xen-devel@lists.xensource.com
> http://lists.xensource.com/xen-devel
> 

Re: performance tools on xen

[George Dunlap]

On Thu, Oct 29, 2009 at 7:29 AM, Keir Fraser <keir.fraser@eu.citrix.com> wrote:
> On 29/10/2009 06:04, "Juergen Gross" <juergen.gross@ts.fujitsu.com> wrote:
>
>>> So far I've found:
>>>   - Xenoprof : is this working currently for 64bit VMs on 32bit dom0?
>>>   - xentop based data
>>>   - xenmon: does this support SMP?
>>
>> xenlockprof    lock profiling of the hypervisor
>> xenperf        software counters of the hypervisor
>>
>> Both need to be enabled at compile time of the hypervisor.
>
> Also xentrace, for trace-based analysis.

And xenalyze, to analyze the resulting traces:

http://xenbits.xensource.com/ext/xenalyze.hg

 -George

Re: performance tools on xen

[Keir Fraser]

On 29/10/2009 06:04, "Juergen Gross" <juergen.gross@ts.fujitsu.com> wrote:

>> So far I've found:
>>   - Xenoprof : is this working currently for 64bit VMs on 32bit dom0?
>>   - xentop based data
>>   - xenmon: does this support SMP?
> 
> xenlockprof    lock profiling of the hypervisor
> xenperf        software counters of the hypervisor
> 
> Both need to be enabled at compile time of the hypervisor.

Also xentrace, for trace-based analysis.

 -- Keir

Re: performance tools on xen

[Juergen Gross]

Mukesh Rathor wrote:
> Hi guys,
> 
> I'm trying to find available tools on xen to do performance
> analysis: hypervisor + dom0 + domU. I'm thinking CPU usage,
> lock contentions, page faults, etc type of info.
> 
> So far I've found:
>   - Xenoprof : is this working currently for 64bit VMs on 32bit dom0?
>   - xentop based data
>   - xenmon: does this support SMP? 

xenlockprof    lock profiling of the hypervisor
xenperf        software counters of the hypervisor

Both need to be enabled at compile time of the hypervisor.


Juergen

-- 
Juergen Gross                 Principal Developer Operating Systems
TSP ES&S SWE OS6                       Telephone: +49 (0) 89 636 47950
Fujitsu Technolgy Solutions               e-mail: juergen.gross@ts.fujitsu.com
Otto-Hahn-Ring 6                        Internet: ts.fujitsu.com
D-81739 Muenchen                 Company details: ts.fujitsu.com/imprint.html

performance tools on xen

[Mukesh Rathor]

Hi guys,

I'm trying to find available tools on xen to do performance
analysis: hypervisor + dom0 + domU. I'm thinking CPU usage,
lock contentions, page faults, etc type of info.

So far I've found:
  - Xenoprof : is this working currently for 64bit VMs on 32bit dom0?
  - xentop based data
  - xenmon: does this support SMP? 


thanks,
Mukesh