Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]

Groups > linux.kernel > #1275193 > unrolled thread

Re: [PATCH tip/locking/core v10 5/7] locking/pvqspinlock: Collect slowpath lock statistics

Back to article view | Back to linux.kernel

This discussion starts older than the indexed window; earlier articles aren't shown. The article labeled Started by below is the oldest one visible, not the original post.

Contents

  Re: [PATCH tip/locking/core v10 5/7] locking/pvqspinlock: Collect  slowpath lock statistics Peter Zijlstra <peterz@infradead.org> - 2015-11-23 11:00 +0100
    Re: [PATCH tip/locking/core v10 5/7] locking/pvqspinlock: Collect  slowpath lock statistics Waiman Long <waiman.long@hpe.com> - 2015-11-25 20:10 +0100

#1275193 — Re: [PATCH tip/locking/core v10 5/7] locking/pvqspinlock: Collect slowpath lock statistics

On Mon, Nov 09, 2015 at 07:09:25PM -0500, Waiman Long wrote:
> +static ssize_t qstat_read(struct file *file, char __user *user_buf,
> +			  size_t count, loff_t *ppos)
> +{
> +	char buf[64];
> +	int cpu, counter, len;
> +	u64 stat = 0, kicks = 0;
> +
> +	/*
> +	 * Get the counter ID stored in file->f_inode->i_private
> +	 */
> +	if (!file->f_inode) {
> +		WARN_ON_ONCE(1);
> +		return -EBADF;
> +	}
> +	counter = (long)(file->f_inode->i_private);
> +
> +	if (counter >= qstat_num)
> +		return -EBADF;
> +
> +	for_each_possible_cpu(cpu) {
> +		stat += per_cpu(qstats[counter], cpu);
> +		/*
> +		 * Need to sum additional counter for some of them
> +		 */
> +		switch (counter) {
> +
> +		case qstat_pv_latency_kick:
> +		case qstat_pv_hash_hops:
> +			kicks += per_cpu(qstats[qstat_pv_kick_unlock], cpu);
> +			break;
> +
> +		case qstat_pv_latency_wake:
> +			kicks += per_cpu(qstats[qstat_pv_kick_wake], cpu);
> +			break;
> +		}
> +	}
> +
> +	if (counter == qstat_pv_hash_hops) {
> +		/*
> +		 * Return a X.XX decimal number
> +		 */
> +		len = snprintf(buf, sizeof(buf) - 1, "%llu.%02llu\n",
> +			       stat/kicks, ((stat%kicks)*100 + kicks/2)/kicks);
> +	} else {
> +		/*
> +		 * Round to the nearest ns
> +		 */
> +		if ((counter == qstat_pv_latency_kick) ||
> +		    (counter == qstat_pv_latency_wake))
> +			stat = kicks ? (stat + kicks/2)/kicks : 0;
> +		len = snprintf(buf, sizeof(buf) - 1, "%llu\n", stat);
> +	}
> +
> +	return simple_read_from_buffer(user_buf, count, ppos, buf, len);
> +}

That doesn't work on 32bit.

---
--- a/kernel/locking/qspinlock_stat.h
+++ b/kernel/locking/qspinlock_stat.h
@@ -127,18 +127,25 @@ static ssize_t qstat_read(struct file *f
 	}
 
 	if (counter == qstat_pv_hash_hops) {
+		u64 frac;
+
+		frac = 100ULL * do_div(stat, kicks);
+		frac = DIV_ROUND_CLOSEST_ULL(frac, kicks);
+
 		/*
 		 * Return a X.XX decimal number
 		 */
-		len = snprintf(buf, sizeof(buf) - 1, "%llu.%02llu\n",
-			       stat/kicks, ((stat%kicks)*100 + kicks/2)/kicks);
+		len = snprintf(buf, sizeof(buf) - 1, "%llu.%02llu\n", stat, frac);
 	} else {
 		/*
 		 * Round to the nearest ns
 		 */
 		if ((counter == qstat_pv_latency_kick) ||
-		    (counter == qstat_pv_latency_wake))
-			stat = kicks ? (stat + kicks/2)/kicks : 0;
+		    (counter == qstat_pv_latency_wake)) {
+			stat = 0;
+			if (kicks)
+				stat = DIV_ROUND_CLOSEST_ULL(stat, kicks);
+		}
 		len = snprintf(buf, sizeof(buf) - 1, "%llu\n", stat);
 	}
 
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/

[toc] | [next] | [standalone]

#1277707

[Multipart message — attachments visible in raw view] — view raw

On 11/23/2015 04:51 AM, Peter Zijlstra wrote:
> On Mon, Nov 09, 2015 at 07:09:25PM -0500, Waiman Long wrote:
>> +static ssize_t qstat_read(struct file *file, char __user *user_buf,
>> +			  size_t count, loff_t *ppos)
>> +{
>> +	char buf[64];
>> +	int cpu, counter, len;
>> +	u64 stat = 0, kicks = 0;
>> +
>> +	/*
>> +	 * Get the counter ID stored in file->f_inode->i_private
>> +	 */
>> +	if (!file->f_inode) {
>> +		WARN_ON_ONCE(1);
>> +		return -EBADF;
>> +	}
>> +	counter = (long)(file->f_inode->i_private);
>> +
>> +	if (counter>= qstat_num)
>> +		return -EBADF;
>> +
>> +	for_each_possible_cpu(cpu) {
>> +		stat += per_cpu(qstats[counter], cpu);
>> +		/*
>> +		 * Need to sum additional counter for some of them
>> +		 */
>> +		switch (counter) {
>> +
>> +		case qstat_pv_latency_kick:
>> +		case qstat_pv_hash_hops:
>> +			kicks += per_cpu(qstats[qstat_pv_kick_unlock], cpu);
>> +			break;
>> +
>> +		case qstat_pv_latency_wake:
>> +			kicks += per_cpu(qstats[qstat_pv_kick_wake], cpu);
>> +			break;
>> +		}
>> +	}
>> +
>> +	if (counter == qstat_pv_hash_hops) {
>> +		/*
>> +		 * Return a X.XX decimal number
>> +		 */
>> +		len = snprintf(buf, sizeof(buf) - 1, "%llu.%02llu\n",
>> +			       stat/kicks, ((stat%kicks)*100 + kicks/2)/kicks);
>> +	} else {
>> +		/*
>> +		 * Round to the nearest ns
>> +		 */
>> +		if ((counter == qstat_pv_latency_kick) ||
>> +		    (counter == qstat_pv_latency_wake))
>> +			stat = kicks ? (stat + kicks/2)/kicks : 0;
>> +		len = snprintf(buf, sizeof(buf) - 1, "%llu\n", stat);
>> +	}
>> +
>> +	return simple_read_from_buffer(user_buf, count, ppos, buf, len);
>> +}
> That doesn't work on 32bit.
>

Thanks for catching that. I should not use u64 for the stat and kick 
variables. I have attached the fixed version of the patch that should 
work on 32-bit.

Cheers,
Longman

[toc] | [prev] | [standalone]

Back to top | Article view | linux.kernel

csiph-web