Donate to e Foundation | Murena handsets with /e/OS | Own a part of Murena! Learn more

Commit a076aa4f authored by Robert Richter's avatar Robert Richter
Browse files

Merge branch 'oprofile/cell' into oprofile/oprofile-for-tip

parents d2852b93 25006644

arch/powerpc/include/asm/cell-pmu.h

+2 −0
Original line number Diff line number Diff line
@@ -37,9 +37,11 @@ 
#define CBE_PM_STOP_AT_MAX                 0x40000000

#define CBE_PM_TRACE_MODE_GET(pm_control)  (((pm_control) >> 28) & 0x3)

#define CBE_PM_TRACE_MODE_SET(mode)        (((mode)  & 0x3) << 28)

#define CBE_PM_TRACE_BUF_OVFLW(bit)        (((bit) & 0x1) << 17)

#define CBE_PM_COUNT_MODE_SET(count)       (((count) & 0x3) << 18)

#define CBE_PM_FREEZE_ALL_CTRS             0x00100000

#define CBE_PM_ENABLE_EXT_TRACE            0x00008000

#define CBE_PM_SPU_ADDR_TRACE_SET(msk)     (((msk) & 0x3) << 9)



/* Macros for the trace_address register. */

#define CBE_PM_TRACE_BUF_FULL              0x00000800

arch/powerpc/include/asm/oprofile_impl.h

+6 −0
Original line number Diff line number Diff line
@@ -32,6 +32,12 @@ struct op_system_config { 
	unsigned long mmcr0;

	unsigned long mmcr1;

	unsigned long mmcra;

#ifdef CONFIG_OPROFILE_CELL

	/* Register for oprofile user tool to check cell kernel profiling

	 * suport.

	 */

	unsigned long cell_support;

#endif

#endif

	unsigned long enable_kernel;

	unsigned long enable_user;

arch/powerpc/oprofile/cell/pr_util.h

+8 −3
Original line number Diff line number Diff line
@@ -30,6 +30,10 @@ 
extern struct delayed_work spu_work;

extern int spu_prof_running;



#define TRACE_ARRAY_SIZE 1024



extern spinlock_t oprof_spu_smpl_arry_lck;



struct spu_overlay_info {	/* map of sections within an SPU overlay */

	unsigned int vma;	/* SPU virtual memory address from elf */

	unsigned int size;	/* size of section from elf */
@@ -89,10 +93,11 @@ void vma_map_free(struct vma_to_fileoffset_map *map); 
 * Entry point for SPU profiling.

 * cycles_reset is the SPU_CYCLES count value specified by the user.

 */

int start_spu_profiling(unsigned int cycles_reset);



void stop_spu_profiling(void);

int start_spu_profiling_cycles(unsigned int cycles_reset);

void start_spu_profiling_events(void);



void stop_spu_profiling_cycles(void);

void stop_spu_profiling_events(void);



/* add the necessary profiling hooks */

int spu_sync_start(void);

arch/powerpc/oprofile/cell/spu_profiler.c

+43 −13
Original line number Diff line number Diff line
@@ -18,11 +18,21 @@ 
#include <asm/cell-pmu.h>

#include "pr_util.h"



#define TRACE_ARRAY_SIZE 1024

#define SCALE_SHIFT 14



static u32 *samples;



/* spu_prof_running is a flag used to indicate if spu profiling is enabled

 * or not.  It is set by the routines start_spu_profiling_cycles() and

 * start_spu_profiling_events().  The flag is cleared by the routines

 * stop_spu_profiling_cycles() and stop_spu_profiling_events().  These

 * routines are called via global_start() and global_stop() which are called in

 * op_powerpc_start() and op_powerpc_stop().  These routines are called once

 * per system as a result of the user starting/stopping oprofile.  Hence, only

 * one CPU per user at a time will be changing  the value of spu_prof_running.

 * In general, OProfile does not protect against multiple users trying to run

 * OProfile at a time.

 */

int spu_prof_running;

static unsigned int profiling_interval;
@@ -31,8 +41,8 @@ static unsigned int profiling_interval; 


#define SPU_PC_MASK	     0xFFFF



static DEFINE_SPINLOCK(sample_array_lock);

unsigned long sample_array_lock_flags;

DEFINE_SPINLOCK(oprof_spu_smpl_arry_lck);

unsigned long oprof_spu_smpl_arry_lck_flags;



void set_spu_profiling_frequency(unsigned int freq_khz, unsigned int cycles_reset)

{
@@ -145,13 +155,13 @@ static enum hrtimer_restart profile_spus(struct hrtimer *timer) 
		 * sample array must be loaded and then processed for a given

		 * cpu.	 The sample array is not per cpu.

		 */

		spin_lock_irqsave(&sample_array_lock,

				  sample_array_lock_flags);

		spin_lock_irqsave(&oprof_spu_smpl_arry_lck,

				  oprof_spu_smpl_arry_lck_flags);

		num_samples = cell_spu_pc_collection(cpu);



		if (num_samples == 0) {

			spin_unlock_irqrestore(&sample_array_lock,

					       sample_array_lock_flags);

			spin_unlock_irqrestore(&oprof_spu_smpl_arry_lck,

					       oprof_spu_smpl_arry_lck_flags);

			continue;

		}
@@ -162,8 +172,8 @@ static enum hrtimer_restart profile_spus(struct hrtimer *timer) 
					num_samples);

		}



		spin_unlock_irqrestore(&sample_array_lock,

				       sample_array_lock_flags);

		spin_unlock_irqrestore(&oprof_spu_smpl_arry_lck,

				       oprof_spu_smpl_arry_lck_flags);



	}

	smp_wmb();	/* insure spu event buffer updates are written */
@@ -182,13 +192,13 @@ static enum hrtimer_restart profile_spus(struct hrtimer *timer) 


static struct hrtimer timer;

/*

 * Entry point for SPU profiling.

 * Entry point for SPU cycle profiling.

 * NOTE:  SPU profiling is done system-wide, not per-CPU.

 *

 * cycles_reset is the count value specified by the user when

 * setting up OProfile to count SPU_CYCLES.

 */

int start_spu_profiling(unsigned int cycles_reset)

int start_spu_profiling_cycles(unsigned int cycles_reset)

{

	ktime_t kt;
@@ -212,10 +222,30 @@ int start_spu_profiling(unsigned int cycles_reset) 
	return 0;

}



void stop_spu_profiling(void)

/*

 * Entry point for SPU event profiling.

 * NOTE:  SPU profiling is done system-wide, not per-CPU.

 *

 * cycles_reset is the count value specified by the user when

 * setting up OProfile to count SPU_CYCLES.

 */

void start_spu_profiling_events(void)

{

	spu_prof_running = 1;

	schedule_delayed_work(&spu_work, DEFAULT_TIMER_EXPIRE);



	return;

}



void stop_spu_profiling_cycles(void)

{

	spu_prof_running = 0;

	hrtimer_cancel(&timer);

	kfree(samples);

	pr_debug("SPU_PROF: stop_spu_profiling issued\n");

	pr_debug("SPU_PROF: stop_spu_profiling_cycles issued\n");

}



void stop_spu_profiling_events(void)

{

	spu_prof_running = 0;

}

arch/powerpc/oprofile/common.c

+22 −0
Original line number Diff line number Diff line
@@ -132,6 +132,28 @@ static int op_powerpc_create_files(struct super_block *sb, struct dentry *root) 
	oprofilefs_create_ulong(sb, root, "mmcr0", &sys.mmcr0);

	oprofilefs_create_ulong(sb, root, "mmcr1", &sys.mmcr1);

	oprofilefs_create_ulong(sb, root, "mmcra", &sys.mmcra);

#ifdef CONFIG_OPROFILE_CELL

	/* create a file the user tool can check to see what level of profiling

	 * support exits with this kernel. Initialize bit mask to indicate

	 * what support the kernel has:

	 * bit 0      -  Supports SPU event profiling in addition to PPU

	 *               event and cycles; and SPU cycle profiling

	 * bits 1-31  -  Currently unused.

	 *

	 * If the file does not exist, then the kernel only supports SPU

	 * cycle profiling, PPU event and cycle profiling.

	 */

	oprofilefs_create_ulong(sb, root, "cell_support", &sys.cell_support);

	sys.cell_support = 0x1; /* Note, the user OProfile tool must check

				 * that this bit is set before attempting to

				 * user SPU event profiling.  Older kernels

				 * will not have this file, hence the user

				 * tool is not allowed to do SPU event

				 * profiling on older kernels.  Older kernels

				 * will accept SPU events but collected data

				 * is garbage.

				 */

#endif

#endif



	for (i = 0; i < model->num_counters; ++i) {