[Cbe-oss-dev] [PATCH] spufs: marker-based tracing facility
Christoph Hellwig
hch at lst.de
Thu Dec 6 19:22:24 EST 2007
On Tue, Dec 04, 2007 at 10:23:33PM +1100, Jeremy Kerr wrote:
> Hi Christoph,
>
> > > 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.
>
> Sounds good! However:
>
> > --- linux-2.6.orig/arch/powerpc/platforms/cell/spufs/sputrace.c
> > +++ linux-2.6/arch/powerpc/platforms/cell/spufs/sputrace.c
>
> I don't have any original sputrace.c to patch against.
Oops, looks like I still had the file around due to some quilt messup.
The same messup also made the patch create a .rej file.
Here's a version without those messups fixed but no other changes:
Index: linux-2.6/arch/powerpc/platforms/cell/spufs/Makefile
===================================================================
--- linux-2.6.orig/arch/powerpc/platforms/cell/spufs/Makefile 2007-12-06 08:13:42.000000000 +0100
+++ linux-2.6/arch/powerpc/platforms/cell/spufs/Makefile 2007-12-06 08:14:21.000000000 +0100
@@ -4,6 +4,8 @@ spufs-y += inode.o file.o context.o sysc
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
Index: linux-2.6/arch/powerpc/platforms/cell/spufs/sched.c
===================================================================
--- linux-2.6.orig/arch/powerpc/platforms/cell/spufs/sched.c 2007-12-06 08:13:47.000000000 +0100
+++ linux-2.6/arch/powerpc/platforms/cell/spufs/sched.c 2007-12-06 08:14:21.000000000 +0100
@@ -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>
@@ -239,8 +240,8 @@ EXPORT_SYMBOL_GPL(spu_switch_event_unreg
*/
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)
@@ -422,8 +423,8 @@ static int has_affinity(struct spu_conte
*/
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)
@@ -551,6 +552,8 @@ static struct spu *spu_get_idle(struct s
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)) {
@@ -569,8 +572,7 @@ static struct spu *spu_get_idle(struct s
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);
}
@@ -588,12 +590,14 @@ static struct spu *spu_get_idle(struct s
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;
}
@@ -610,6 +614,8 @@ static struct spu *find_victim(struct sp
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
@@ -663,6 +669,8 @@ static struct spu *find_victim(struct sp
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);
@@ -845,6 +853,7 @@ static int __spu_deactivate(struct spu_c
*/
void spu_deactivate(struct spu_context *ctx)
{
+ spu_context_nospu_trace(spu_deactivate__enter, ctx);
__spu_deactivate(ctx, 1, MAX_PRIO);
}
@@ -858,6 +867,7 @@ void spu_deactivate(struct spu_context *
*/
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);
@@ -887,11 +897,15 @@ static noinline void spusched_tick(struc
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:
Index: linux-2.6/arch/powerpc/platforms/cell/spufs/file.c
===================================================================
--- linux-2.6.orig/arch/powerpc/platforms/cell/spufs/file.c 2007-12-06 08:13:47.000000000 +0100
+++ linux-2.6/arch/powerpc/platforms/cell/spufs/file.c 2007-12-06 08:14:21.000000000 +0100
@@ -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(stru
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(stru
if (ctx->state == SPU_STATE_SAVED) {
up_read(¤t->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(¤t->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);
Index: linux-2.6/arch/powerpc/platforms/cell/spufs/spufs.h
===================================================================
--- linux-2.6.orig/arch/powerpc/platforms/cell/spufs/spufs.h 2007-12-06 08:13:47.000000000 +0100
+++ linux-2.6/arch/powerpc/platforms/cell/spufs/spufs.h 2007-12-06 08:14:21.000000000 +0100
@@ -325,4 +325,9 @@ extern void spu_free_lscsa(struct spu_st
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
Index: linux-2.6/arch/powerpc/platforms/cell/Kconfig
===================================================================
--- linux-2.6.orig/arch/powerpc/platforms/cell/Kconfig 2007-12-06 08:12:34.000000000 +0100
+++ linux-2.6/arch/powerpc/platforms/cell/Kconfig 2007-12-06 08:14:21.000000000 +0100
@@ -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
Index: linux-2.6/arch/powerpc/platforms/cell/spufs/sputrace.c
===================================================================
--- /dev/null 1970-01-01 00:00:00.000000000 +0000
+++ linux-2.6/arch/powerpc/platforms/cell/spufs/sputrace.c 2007-12-06 08:14:44.000000000 +0100
@@ -0,0 +1,250 @@
+/*
+ * Copyright (C) 2007 IBM Deutschland Entwicklung GmbH
+ * Released under GPL v2.
+ *
+ * Partially based on net/ipv4/tcp_probe.c.
+ *
+ * Simple tracing facility for spu contexts.
+ */
+#include <linux/sched.h>
+#include <linux/kernel.h>
+#include <linux/module.h>
+#include <linux/marker.h>
+#include <linux/proc_fs.h>
+#include <linux/wait.h>
+#include <asm/atomic.h>
+#include <asm/uaccess.h>
+#include "spufs.h"
+
+struct spu_probe {
+ const char *name;
+ const char *format;
+ marker_probe_func *probe_func;
+};
+
+struct sputrace {
+ ktime_t tstamp;
+ int owner_tid; /* owner */
+ int curr_tid;
+ const char *name;
+ int number;
+};
+
+static int bufsize __read_mostly = 16384;
+MODULE_PARM_DESC(bufsize, "Log buffer size (number of records)");
+module_param(bufsize, int, 0);
+
+
+static DEFINE_SPINLOCK(sputrace_lock);
+static DECLARE_WAIT_QUEUE_HEAD(sputrace_wait);
+static ktime_t sputrace_start;
+static unsigned long sputrace_head, sputrace_tail;
+static struct sputrace *sputrace_log;
+
+static int sputrace_used(void)
+{
+ return (sputrace_head - sputrace_tail) % bufsize;
+}
+
+static inline int sputrace_avail(void)
+{
+ return bufsize - sputrace_used();
+}
+
+static int sputrace_sprint(char *tbuf, int n)
+{
+ const struct sputrace *t = sputrace_log + sputrace_tail % bufsize;
+ struct timespec tv =
+ ktime_to_timespec(ktime_sub(t->tstamp, sputrace_start));
+
+ return snprintf(tbuf, n,
+ "[%lu.%09lu] %d: %s (thread = %d, spu = %d)\n",
+ (unsigned long) tv.tv_sec,
+ (unsigned long) tv.tv_nsec,
+ t->owner_tid,
+ t->name,
+ t->curr_tid,
+ t->number);
+}
+
+static ssize_t sputrace_read(struct file *file, char __user *buf,
+ size_t len, loff_t *ppos)
+{
+ int error = 0, cnt = 0;
+
+ if (!buf || len < 0)
+ return -EINVAL;
+
+ while (cnt < len) {
+ char tbuf[128];
+ int width;
+
+ error = wait_event_interruptible(sputrace_wait,
+ sputrace_used() > 0);
+ if (error)
+ break;
+
+ spin_lock(&sputrace_lock);
+ if (sputrace_head == sputrace_tail) {
+ spin_unlock(&sputrace_lock);
+ continue;
+ }
+
+ width = sputrace_sprint(tbuf, sizeof(tbuf));
+ if (width < len)
+ sputrace_tail = (sputrace_tail + 1) % bufsize;
+ spin_unlock(&sputrace_lock);
+
+ if (width >= len)
+ break;
+
+ error = copy_to_user(buf + cnt, tbuf, width);
+ if (error)
+ break;
+ cnt += width;
+ }
+
+ return cnt == 0 ? error : cnt;
+}
+
+static int tcpprobe_open(struct inode * inode, struct file * file)
+{
+ spin_lock(&sputrace_lock);
+ sputrace_head = sputrace_tail = 0;
+ sputrace_start = ktime_get();
+ spin_unlock(&sputrace_lock);
+
+ return 0;
+}
+
+static const struct file_operations sputrace_fops = {
+ .owner = THIS_MODULE,
+ .open = tcpprobe_open,
+ .read = sputrace_read,
+};
+
+static void sputrace_log_item(const char *name, struct spu_context *ctx,
+ struct spu *spu)
+{
+ spin_lock(&sputrace_lock);
+ if (sputrace_avail() > 1) {
+ struct sputrace *t = sputrace_log + sputrace_head;
+
+ t->tstamp = ktime_get();
+ t->owner_tid = ctx->tid;
+ t->name = name;
+ t->curr_tid = current->pid;
+ t->number = spu ? spu->number : -1;
+
+ sputrace_head = (sputrace_head + 1) % bufsize;
+ } else {
+ printk(KERN_WARNING
+ "sputrace: lost samples due to full buffer.\n");
+ }
+ spin_unlock(&sputrace_lock);
+
+ wake_up(&sputrace_wait);
+}
+
+static void spu_context_event(const struct marker *mdata,
+ void *private, const char *format, ...)
+{
+ struct spu_probe *p = mdata->private;
+ va_list ap;
+ struct spu_context *ctx;
+ struct spu *spu;
+
+ va_start(ap, format);
+ ctx = va_arg(ap, struct spu_context *);
+ spu = va_arg(ap, struct spu *);
+
+ sputrace_log_item(p->name, ctx, spu);
+ va_end(ap);
+}
+
+static void spu_context_nospu_event(const struct marker *mdata,
+ void *private, const char *format, ...)
+{
+ struct spu_probe *p = mdata->private;
+ va_list ap;
+ struct spu_context *ctx;
+
+ va_start(ap, format);
+ ctx = va_arg(ap, struct spu_context *);
+
+ sputrace_log_item(p->name, ctx, NULL);
+ va_end(ap);
+}
+
+struct spu_probe spu_probes[] = {
+ { "spu_bind_context__enter", "%p %p", spu_context_event },
+ { "spu_unbind_context__enter", "%p %p", spu_context_event },
+ { "spu_get_idle__enter", "%p", spu_context_nospu_event },
+ { "spu_get_idle__found", "%p %p", spu_context_event },
+ { "spu_get_idle__not_found", "%p", spu_context_nospu_event },
+ { "spu_find_victim__enter", "%p", spu_context_nospu_event },
+ { "spusched_tick__preempt", "%p %p", spu_context_event },
+ { "spusched_tick__newslice", "%p", spu_context_nospu_event },
+ { "spu_yield__enter", "%p", spu_context_nospu_event },
+ { "spu_deactivate__enter", "%p", spu_context_nospu_event },
+ { "__spu_deactivate__unload", "%p %p", spu_context_event },
+ { "spufs_ps_nopfn__enter", "%p", spu_context_nospu_event },
+ { "spufs_ps_nopfn__sleep", "%p", spu_context_nospu_event },
+ { "spufs_ps_nopfn__wake", "%p %p", spu_context_event },
+ { "spufs_ps_nopfn__insert", "%p %p", spu_context_event },
+ { "spu_acquire_saved__enter", "%p", spu_context_nospu_event },
+ { "destroy_spu_context__enter", "%p", spu_context_nospu_event },
+};
+
+static int __init sputrace_init(void)
+{
+ struct proc_dir_entry *entry;
+ int error = -ENOMEM, i;
+
+ sputrace_log = kcalloc(sizeof(struct sputrace),
+ bufsize, GFP_KERNEL);
+ if (!sputrace_log)
+ goto out;
+
+ entry = create_proc_entry("sputrace", S_IRUSR, NULL);
+ if (!entry)
+ goto out_free_log;
+ entry->proc_fops = &sputrace_fops;
+
+ for (i = 0; i < ARRAY_SIZE(spu_probes); i++) {
+ struct spu_probe *p = &spu_probes[i];
+
+ error = marker_probe_register(p->name, p->format,
+ p->probe_func, p);
+ if (error)
+ printk(KERN_INFO "Unable to register probe %s\n",
+ p->name);
+
+ error = marker_arm(p->name);
+ if (error)
+ printk(KERN_INFO "Unable to arm probe %s\n", p->name);
+ }
+
+ return 0;
+
+ out_free_log:
+ kfree(sputrace_log);
+ out:
+ return -ENOMEM;
+}
+
+static void __exit sputrace_exit(void)
+{
+ int i;
+
+ for (i = 0; i < ARRAY_SIZE(spu_probes); i++)
+ marker_probe_unregister(spu_probes[i].name);
+
+ remove_proc_entry("sputrace", NULL);
+ kfree(sputrace_log);
+}
+
+module_init(sputrace_init);
+module_exit(sputrace_exit);
+
+MODULE_LICENSE("GPL");
More information about the cbe-oss-dev
mailing list