[Patch 7/8] Enhanced partition statistics: procfs

[Patch 7/8] Enhanced partition statistics: procfs

[Jerome Marchand]

Reports enhanced partition statistics in /proc/diskstats.

Signed-off-by: Jerome Marchand <jmarchan@redhat.com>
---
 genhd.c |   27 +++++++++++++++++++++------
 1 file changed, 21 insertions(+), 6 deletions(-)
diff -urNp -X linux-2.6/Documentation/dontdiff linux-2.6.orig/block/genhd.c linux-2.6/block/genhd.c
--- linux-2.6.orig/block/genhd.c	2007-12-04 17:37:31.000000000 +0100
+++ linux-2.6/block/genhd.c	2007-12-05 13:52:55.000000000 +0100
@@ -655,12 +655,27 @@ static int diskstats_show(struct seq_fil
 	for (n = 0; n < gp->minors - 1; n++) {
 		struct hd_struct *hd = gp->part[n];
 
-		if (hd && hd->nr_sects)
-			seq_printf(s, "%4d %4d %s %u %u %u %u\n",
-				gp->major, n + gp->first_minor + 1,
-				disk_name(gp, n + 1, buf),
-				hd->ios[0], hd->sectors[0],
-				hd->ios[1], hd->sectors[1]);
+		if (!hd || !hd->nr_sects) continue;
+
+		preempt_disable();
+		part_round_stats(hd);
+		preempt_enable();
+		seq_printf(s, "%4d %4d %s %lu %lu %llu "
+			   "%u %lu %lu %llu %u %u %u %u\n",
+			   gp->major, n + gp->first_minor + 1,
+			   disk_name(gp, n + 1, buf),
+			   part_stat_read(hd, ios[0]),
+			   part_stat_read(hd, merges[0]),
+			   (unsigned long long)part_stat_read(hd, sectors[0]),
+			   jiffies_to_msecs(part_stat_read(hd, ticks[0])),
+			   part_stat_read(hd, ios[1]),
+			   part_stat_read(hd, merges[1]),
+			   (unsigned long long)part_stat_read(hd, sectors[1]),
+			   jiffies_to_msecs(part_stat_read(hd, ticks[1])),
+			   hd->in_flight,
+			   jiffies_to_msecs(part_stat_read(hd, io_ticks)),
+			   jiffies_to_msecs(part_stat_read(hd, time_in_queue))
+			);
 	}
  
 	return 0;

Re: [Patch 7/8] Enhanced partition statistics: procfs

[Randy Dunlap]

On Thu, 13 Dec 2007 17:18:04 +0100 Jerome Marchand wrote:

> Reports enhanced partition statistics in /proc/diskstats.

> Signed-off-by: Jerome Marchand <jmarchan@redhat.com>
> ---
>  genhd.c |   27 +++++++++++++++++++++------
>  1 file changed, 21 insertions(+), 6 deletions(-)
> diff -urNp -X linux-2.6/Documentation/dontdiff linux-2.6.orig/block/genhd.c linux-2.6/block/genhd.c
> --- linux-2.6.orig/block/genhd.c	2007-12-04 17:37:31.000000000 +0100
> +++ linux-2.6/block/genhd.c	2007-12-05 13:52:55.000000000 +0100
> @@ -655,12 +655,27 @@ static int diskstats_show(struct seq_fil
>  	for (n = 0; n < gp->minors - 1; n++) {
>  		struct hd_struct *hd = gp->part[n];
>  
> -		if (hd && hd->nr_sects)
> -			seq_printf(s, "%4d %4d %s %u %u %u %u\n",
> -				gp->major, n + gp->first_minor + 1,
> -				disk_name(gp, n + 1, buf),
> -				hd->ios[0], hd->sectors[0],
> -				hd->ios[1], hd->sectors[1]);
> +		if (!hd || !hd->nr_sects) continue;

Don't put if () and following statement on same line.

> +
> +		preempt_disable();
> +		part_round_stats(hd);
> +		preempt_enable();
> +		seq_printf(s, "%4d %4d %s %lu %lu %llu "
> +			   "%u %lu %lu %llu %u %u %u %u\n",
> +			   gp->major, n + gp->first_minor + 1,
> +			   disk_name(gp, n + 1, buf),
> +			   part_stat_read(hd, ios[0]),
> +			   part_stat_read(hd, merges[0]),
> +			   (unsigned long long)part_stat_read(hd, sectors[0]),
> +			   jiffies_to_msecs(part_stat_read(hd, ticks[0])),
> +			   part_stat_read(hd, ios[1]),
> +			   part_stat_read(hd, merges[1]),
> +			   (unsigned long long)part_stat_read(hd, sectors[1]),
> +			   jiffies_to_msecs(part_stat_read(hd, ticks[1])),
> +			   hd->in_flight,
> +			   jiffies_to_msecs(part_stat_read(hd, io_ticks)),
> +			   jiffies_to_msecs(part_stat_read(hd, time_in_queue))
> +			);
>  	}
>   
>  	return 0;
> 
> --

The format of the /proc file is not changed except to add data
at the end of each line.  Is that correct?

---
~Randy

Re: [Patch 7/8] Enhanced partition statistics: procfs

[Jerome Marchand]

Randy Dunlap wrote:
> 
> The format of the /proc file is not changed except to add data
> at the end of each line.  Is that correct?
> 

No. The line concerning partitions will display different data. The new format match the current format
concerning disks.

Current partition format:
major minor name reads_issued sectors_read writes_issued sectors_written

New partition format (current disk format):
major minor name reads_issued reads_merged sector_read ms_reading writes_completed writes_merged sectors_written ms_writing I/Os_running ms_I/O weighted_ms_I/O

Jérôme

Re: [Patch 7/8] Enhanced partition statistics: procfs

[Randy Dunlap]

On Thu, 13 Dec 2007 18:36:45 +0100 Jerome Marchand wrote:

> Randy Dunlap wrote:
> > 
> > The format of the /proc file is not changed except to add data
> > at the end of each line.  Is that correct?
> > 
> 
> No. The line concerning partitions will display different data. The new format match the current format
> concerning disks.
> 
> Current partition format:
> major minor name reads_issued sectors_read writes_issued sectors_written
> 
> New partition format (current disk format):
> major minor name reads_issued reads_merged sector_read ms_reading writes_completed writes_merged sectors_written ms_writing I/Os_running ms_I/O weighted_ms_I/O


I would have expected the /proc file contents to represent a user
interface that we shouldn't change, other than by adding to the end
of each line.


---
~Randy

Re: [Patch 7/8] Enhanced partition statistics: procfs

[Jerome Marchand]

Randy Dunlap wrote:
> On Thu, 13 Dec 2007 18:36:45 +0100 Jerome Marchand wrote:
> 
>> Randy Dunlap wrote:
>>> The format of the /proc file is not changed except to add data
>>> at the end of each line.  Is that correct?
>>>
>> No. The line concerning partitions will display different data. The new format match the current format
>> concerning disks.
>>
>> Current partition format:
>> major minor name reads_issued sectors_read writes_issued sectors_written
>>
>> New partition format (current disk format):
>> major minor name reads_issued reads_merged sector_read ms_reading writes_completed writes_merged sectors_written ms_writing I/Os_running ms_I/O weighted_ms_I/O
> 
> 
> I would have expected the /proc file contents to represent a user
> interface that we shouldn't change, other than by adding to the end
> of each line.
> 

I could eventually keep the old data and add the new ones at the end
of the line, but it will leads to a less coherent interface. Moreover
all the users of /proc/diskstats I know about use the number of field
to make the distinction between current disk and partitions
format. They all will be confused if we add new fields.

Jérôme