2002-07-04 19:47:54

by Adrian Bunk

[permalink] [raw]
Subject: [2.4 patch] remove obsolete disk statistics header from /proc/partitions


ChangeSet 1.579 removes the disk statistics in /proc/partitions.

But the header line is still present:

<-- snip -->

# cat /proc/partitions
major minor #blocks name rio rmerge rsect ruse wio wmerge wsect wuse
running use aveq

22 64 20039544 hdd
22 65 20039512 hdd1
3 0 20010312 hda
3 1 1028128 hda1
3 2 1020127 hda2
3 3 1 hda3
3 5 3076416 hda5
3 6 2867571 hda6
3 7 1959898 hda7
3 8 10056658 hda8
#

<-- snip -->


I suggest the following patch to remove it:


--- drivers/block/genhd.c.old Thu Jul 4 20:32:35 2002
+++ drivers/block/genhd.c Thu Jul 4 20:33:26 2002
@@ -163,9 +163,7 @@
char buf[64];
int len, n;

- len = sprintf(page, "major minor #blocks name "
- "rio rmerge rsect ruse wio wmerge "
- "wsect wuse running use aveq\n\n");
+ len = sprintf(page, "major minor #blocks name\n\n");


read_lock(&gendisk_lock);


cu
Adrian

--

You only think this is a free country. Like the US the UK spends a lot of
time explaining its a free country because its a police state.
Alan Cox



2002-07-05 15:05:05

by Marcelo Tosatti

[permalink] [raw]
Subject: Re: [2.4 patch] remove obsolete disk statistics header from /proc/partitions


Applied.

Thanks

On Thu, 4 Jul 2002, Adrian Bunk wrote:

>
> ChangeSet 1.579 removes the disk statistics in /proc/partitions.

2002-07-05 15:31:35

by Thomas Tonino

[permalink] [raw]
Subject: Re: [2.4 patch] remove obsolete disk statistics header from /proc/partitions

Adrian Bunk wrote:

> Marcelos' BK repository (that will become 2.4.19-rc2) includes a patch to
> remove these statistics completely from /proc/partitions...

I certainly hope they move elsewhere; totally removing these is bad news
for anyone running a server that does disk I/O.

If the bugginess of part of the data is a problem, remove that data for
the time being. I can also imagine people feel it is bloated (blocks and
KB/sec), and maybe it should be reorganized to move elsewhere.

But measurement is important - the saying here is "meten is weten" - and
should not be removed from the kernel.

As for correcting: I can imagine a really dirty fix decrementing the
number of running requests once a second, on top of the fix in 2.5 that
prevents the number from going negative. The extra decrement makes disk
performance appear about 1% higher than it should: not a big falsification.



Thomas