linux-nfs.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
From: "J. Bruce Fields" <bfields@fieldses.org>
To: Karel Zak <kzak@redhat.com>
Cc: Sander Klein <roedie@roedie.nl>, linux-nfs@vger.kernel.org
Subject: Re: rpc.mountd high cpu usage
Date: Mon, 6 Jan 2014 16:25:14 -0500	[thread overview]
Message-ID: <20140106212514.GG31764@fieldses.org> (raw)
In-Reply-To: <20140106111805.GF4435@x2.net.home>

On Mon, Jan 06, 2014 at 12:18:05PM +0100, Karel Zak wrote:
> On Thu, Dec 12, 2013 at 10:46:42AM -0500, J. Bruce Fields wrote:
> > That's done by match_fsid().  Which does do a stat of the export path,
> > but not of all the devices....  That's probably happening in one of the
> > libblkid calls in uuid_by_path()?  I wonder if there's something wrong
> > with libblkid configuration or with the way we're using it?
> 
> The blkid EVALUATE=<methods> configuration matters for UUID to
> devname conversion, not vice versa. 
>  
> If you want to get filesystem UUID you have to ask udev (link with
> libudev) or read it from the device (link with libblkid).
> 
> > > statfs("/some/export", {f_type=0x2fc12fc1, f_bsize=131072,
> > > f_blocks=50331648, f_bfree=8355553, f_bavail=8355553,
> > > f_files=2139121087, f_ffree=2139021753, f_fsid={1912623216,
> > > 10933642}, f_namelen=255, f_frsize=131072}) = 0
> > > stat("/some/export", {st_mode=S_IFDIR|0775, st_size=15, ...}) = 0
> > > lstat("/some/export", {st_mode=S_IFDIR|0775, st_size=15, ...}) = 0
> > > lstat("/some/export/..", {st_mode=S_IFDIR|0755, st_size=12288, ...}) = 0
> > > stat("/some/export", {st_mode=S_IFDIR|0775, st_size=15, ...}) = 0
> > > open("/sys/dev/block/0:187", O_RDONLY)  = -1 ENOENT (No such file or
> > > directory)
> 
> This is blkid_devno_to_devname(), the first attempt is /sys, but it
> failed, then it tries to scan /dev (which is pretty expensive method).
> 
> BTW, what is device 0:187? 
> 
> It seems like btrfs... maybe we can  optimize the code to ignore such
> devices especially when get_uuid_blkdev() filters out btrfs :-)
> 
> The another problem is poorly designed relationship between match_fsid()
> and uuid_by_path().  The function uuid_by_path() is called in loop and 
> blkid is *always* requested, but the blkid_val is used when type==0. 
> It would be better to call get_uuid_blkdev() only when type==0.
> 
> The patch below is just untested suggestion :-)

Thanks!

It's too bad we need to special-case these filesystems.

--b.


> 
>     Karel
> 
> >From e4f5b38c87ef8d713c6d0a3169afaf1acc7e22c6 Mon Sep 17 00:00:00 2001
> From: Karel Zak <kzak@redhat.com>
> Date: Mon, 6 Jan 2014 12:03:01 +0100
> Subject: [PATCH] mountd: optimize libblkid usage
> 
>  * use get_uuid_blkdev() only first time for the path (it means
>    that uuid_by_path() is called with type==0)
> 
>  * don't use libblkid for btrfs, network or pseudo filesystems
> 
> Note that the patch defines the fs type ID rather than include
> <linux/magic.h> as this file seems incomplete and libc specific).
> 
> Signed-off-by: Karel Zak <kzak@redhat.com>
> ---
>  utils/mountd/cache.c | 38 +++++++++++++++++++++++++++++++++++---
>  1 file changed, 35 insertions(+), 3 deletions(-)
> 
> diff --git a/utils/mountd/cache.c b/utils/mountd/cache.c
> index e04b86e..73c0be6 100644
> --- a/utils/mountd/cache.c
> +++ b/utils/mountd/cache.c
> @@ -266,6 +266,26 @@ static int get_uuid(const char *val, size_t uuidlen, char *u)
>  	return 1;
>  }
>  
> +
> +/*
> + * Don't ask libblkid for these filesystems. Note that BTRF is ignored, because
> + * we generate the identifier from statfs->f_fsid. The rest are network or
> + * pseudo filesystems. (See <linux/magic.h> for the basic IDs.)
> + */
> +static const long int nonblkid_filesystems[] = {
> +	0x9123683E,	/* BTRFS_SUPER_MAGIC */
> +	0xFF534D42,	/* CIFS_MAGIC_NUMBER */
> +	0x1373,		/* DEVFS_SUPER_MAGIC */
> +	0x73757245,	/* CODA_SUPER_MAGIC */
> +	0x564C,		/* NCP_SUPER_MAGIC */
> +	0x6969,		/* NFS_SUPER_MAGIC */
> +	0x9FA0,		/* PROC_SUPER_MAGIC */
> +	0x62656572,	/* SYSFS_MAGIC */
> +	0x517B,		/* SMB_SUPER_MAGIC */
> +	0x01021994,	/* TMPFS_SUPER_MAGIC */
> +	0		/* last */
> +};
> +
>  static int uuid_by_path(char *path, int type, size_t uuidlen, char *uuid)
>  {
>  	/* get a uuid for the filesystem found at 'path'.
> @@ -297,12 +317,24 @@ static int uuid_by_path(char *path, int type, size_t uuidlen, char *uuid)
>  	 */
>  	struct statfs64 st;
>  	char fsid_val[17];
> -	const char *blkid_val;
> +	const char *blkid_val = NULL;
>  	const char *val;
> +	int rc;
>  
> -	blkid_val = get_uuid_blkdev(path);
> +	rc = statfs64(path, &st);
> +
> +	if (type == 0 && rc == 0) {
> +		const long int *bad;
> +
> +		for (bad = nonblkid_filesystems; *bad; bad++) {
> +			if (*bad == st.f_type)
> +				break;
> +		}
> +		if (*bad == 0)
> +			blkid_val = get_uuid_blkdev(path);
> +	}
>  
> -	if (statfs64(path, &st) == 0 &&
> +	if (rc == 0 &&
>  	    (st.f_fsid.__val[0] || st.f_fsid.__val[1]))
>  		snprintf(fsid_val, 17, "%08x%08x",
>  			 st.f_fsid.__val[0], st.f_fsid.__val[1]);
> -- 
> 1.8.4.2
> 

  parent reply	other threads:[~2014-01-06 21:25 UTC|newest]

Thread overview: 15+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2013-12-12 10:13 rpc.mountd high cpu usage Sander Klein
2013-12-12 15:46 ` J. Bruce Fields
2013-12-12 16:42   ` Sander Klein
2013-12-12 21:22     ` J. Bruce Fields
2013-12-13 19:32       ` Sander Klein
2013-12-19 17:09         ` J. Bruce Fields
2013-12-23 11:01           ` Sander Klein
2014-01-06 11:18   ` Karel Zak
2014-01-06 16:20     ` Sander Klein
2014-01-06 16:31       ` Karel Zak
2014-01-06 21:25     ` J. Bruce Fields [this message]
2014-01-07 19:55       ` Sander Klein
2014-01-07 20:49         ` J. Bruce Fields
2014-01-07 21:01         ` Steve Dickson
2014-01-08  9:09           ` Sander Klein

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20140106212514.GG31764@fieldses.org \
    --to=bfields@fieldses.org \
    --cc=kzak@redhat.com \
    --cc=linux-nfs@vger.kernel.org \
    --cc=roedie@roedie.nl \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).