[POWERPC] spufs: Add marker-based tracing facility

This adds markers two important points in the spufs code and a new
module (sputrace.ko) that allows reading these out through a proc file.

Long-term I'd rather see something like lttng extended to use the spufs
instrumentation, but for now I think this is a good enough quick
solution.  We'll probably want to add various addition event in addition
to that ones I have already.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Jeremy Kerr <jk@ozlabs.org>
Signed-off-by: Paul Mackerras <paulus@samba.org>

arch/powerpc/platforms/cell/Kconfig

@@ -54,6 +54,13 @@ config SPU_FS_64K_LS
 	  uses 4K pages. This can improve performances of applications
 	  using multiple SPEs by lowering the TLB pressure on them.
 
+config SPU_TRACE
+	tristate "SPU event tracing support"
+	depends on SPU_FS && MARKERS
+	help
+	  This option allows reading a trace of spu-related events through
+	  the sputrace file in procfs.
+
 config SPU_BASE
 	bool
 	default n

arch/powerpc/platforms/cell/spufs/Makefile

@@ -4,6 +4,8 @@ spufs-y += inode.o file.o context.o syscalls.o coredump.o
 spufs-y += sched.o backing_ops.o hw_ops.o run.o gang.o
 spufs-y += switch.o fault.o lscsa_alloc.o
 
+obj-$(CONFIG_SPU_TRACE)	+= sputrace.o
+
 # Rules to build switch.o with the help of SPU tool chain
 SPU_CROSS	:= spu-
 SPU_CC		:= $(SPU_CROSS)gcc

arch/powerpc/platforms/cell/spufs/file.c

@@ -29,6 +29,7 @@
 #include <linux/poll.h>
 #include <linux/ptrace.h>
 #include <linux/seq_file.h>
+#include <linux/marker.h>
 
 #include <asm/io.h>
 #include <asm/semaphore.h>
@@ -358,6 +359,8 @@ static unsigned long spufs_ps_nopfn(struct vm_area_struct *vma,
 	struct spu_context *ctx = vma->vm_file->private_data;
 	unsigned long area, offset = address - vma->vm_start;
 
+	spu_context_nospu_trace(spufs_ps_nopfn__enter, ctx);
+
 	offset += vma->vm_pgoff << PAGE_SHIFT;
 	if (offset >= ps_size)
 		return NOPFN_SIGBUS;
@@ -375,11 +378,14 @@ static unsigned long spufs_ps_nopfn(struct vm_area_struct *vma,
 
 	if (ctx->state == SPU_STATE_SAVED) {
 		up_read(&current->mm->mmap_sem);
+		spu_context_nospu_trace(spufs_ps_nopfn__sleep, ctx);
 		spufs_wait(ctx->run_wq, ctx->state == SPU_STATE_RUNNABLE);
+		spu_context_trace(spufs_ps_nopfn__wake, ctx, ctx->spu);
 		down_read(&current->mm->mmap_sem);
 	} else {
 		area = ctx->spu->problem_phys + ps_offs;
 		vm_insert_pfn(vma, address, (area + offset) >> PAGE_SHIFT);
+		spu_context_trace(spufs_ps_nopfn__insert, ctx, ctx->spu);
 	}
 
 	spu_release(ctx);

arch/powerpc/platforms/cell/spufs/sched.c

@@ -39,6 +39,7 @@
 #include <linux/pid_namespace.h>
 #include <linux/proc_fs.h>
 #include <linux/seq_file.h>
+#include <linux/marker.h>
 
 #include <asm/io.h>
 #include <asm/mmu_context.h>
@@ -216,8 +217,8 @@ void do_notify_spus_active(void)
  */
 static void spu_bind_context(struct spu *spu, struct spu_context *ctx)
 {
-	pr_debug("%s: pid=%d SPU=%d NODE=%d\n", __FUNCTION__, current->pid,
-		 spu->number, spu->node);
+	spu_context_trace(spu_bind_context__enter, ctx, spu);
+
 	spuctx_switch_state(ctx, SPU_UTIL_SYSTEM);
 
 	if (ctx->flags & SPU_CREATE_NOSCHED)
@@ -399,8 +400,8 @@ static int has_affinity(struct spu_context *ctx)
  */
 static void spu_unbind_context(struct spu *spu, struct spu_context *ctx)
 {
-	pr_debug("%s: unbind pid=%d SPU=%d NODE=%d\n", __FUNCTION__,
-		 spu->pid, spu->number, spu->node);
+	spu_context_trace(spu_unbind_context__enter, ctx, spu);
+
 	spuctx_switch_state(ctx, SPU_UTIL_SYSTEM);
 
  	if (spu->ctx->flags & SPU_CREATE_NOSCHED)
@@ -528,6 +529,8 @@ static struct spu *spu_get_idle(struct spu_context *ctx)
 	struct spu *spu, *aff_ref_spu;
 	int node, n;
 
+	spu_context_nospu_trace(spu_get_idle__enter, ctx);
+
 	if (ctx->gang) {
 		mutex_lock(&ctx->gang->aff_mutex);
 		if (has_affinity(ctx)) {
@@ -546,8 +549,7 @@ static struct spu *spu_get_idle(struct spu_context *ctx)
 			if (atomic_dec_and_test(&ctx->gang->aff_sched_count))
 				ctx->gang->aff_ref_spu = NULL;
 			mutex_unlock(&ctx->gang->aff_mutex);
-
-			return NULL;
+			goto not_found;
 		}
 		mutex_unlock(&ctx->gang->aff_mutex);
 	}
@@ -565,12 +567,14 @@ static struct spu *spu_get_idle(struct spu_context *ctx)
 		mutex_unlock(&cbe_spu_info[node].list_mutex);
 	}
 
+ not_found:
+	spu_context_nospu_trace(spu_get_idle__not_found, ctx);
 	return NULL;
 
  found:
 	spu->alloc_state = SPU_USED;
 	mutex_unlock(&cbe_spu_info[node].list_mutex);
-	pr_debug("Got SPU %d %d\n", spu->number, spu->node);
+	spu_context_trace(spu_get_idle__found, ctx, spu);
 	spu_init_channels(spu);
 	return spu;
 }
@@ -587,6 +591,8 @@ static struct spu *find_victim(struct spu_context *ctx)
 	struct spu *spu;
 	int node, n;
 
+	spu_context_nospu_trace(spu_find_vitim__enter, ctx);
+
 	/*
 	 * Look for a possible preemption candidate on the local node first.
 	 * If there is no candidate look at the other nodes.  This isn't
@@ -640,6 +646,8 @@ static struct spu *find_victim(struct spu_context *ctx)
 				goto restart;
 			}
 
+			spu_context_trace(__spu_deactivate__unload, ctx, spu);
+
 			mutex_lock(&cbe_spu_info[node].list_mutex);
 			cbe_spu_info[node].nr_active--;
 			spu_unbind_context(spu, victim);
@@ -822,6 +830,7 @@ static int __spu_deactivate(struct spu_context *ctx, int force, int max_prio)
  */
 void spu_deactivate(struct spu_context *ctx)
 {
+	spu_context_nospu_trace(spu_deactivate__enter, ctx);
 	__spu_deactivate(ctx, 1, MAX_PRIO);
 }
 
@@ -835,6 +844,7 @@ void spu_deactivate(struct spu_context *ctx)
  */
 void spu_yield(struct spu_context *ctx)
 {
+	spu_context_nospu_trace(spu_yield__enter, ctx);
 	if (!(ctx->flags & SPU_CREATE_NOSCHED)) {
 		mutex_lock(&ctx->state_mutex);
 		__spu_deactivate(ctx, 0, MAX_PRIO);
@@ -864,11 +874,15 @@ static noinline void spusched_tick(struct spu_context *ctx)
 		goto out;
 
 	spu = ctx->spu;
+
+	spu_context_trace(spusched_tick__preempt, ctx, spu);
+
 	new = grab_runnable_context(ctx->prio + 1, spu->node);
 	if (new) {
 		spu_unschedule(spu, ctx);
 		spu_add_to_rq(ctx);
 	} else {
+		spu_context_nospu_trace(spusched_tick__newslice, ctx);
 		ctx->time_slice++;
 	}
 out:

arch/powerpc/platforms/cell/spufs/spufs.h

@@ -325,4 +325,9 @@ extern void spu_free_lscsa(struct spu_state *csa);
 extern void spuctx_switch_state(struct spu_context *ctx,
 		enum spu_utilization_state new_state);
 
+#define spu_context_trace(name, ctx, spu) \
+	trace_mark(name, "%p %p", ctx, spu);
+#define spu_context_nospu_trace(name, ctx) \
+	trace_mark(name, "%p", ctx);
+
 #endif