From: Steve Dickson <SteveD@redhat.com>
To: linux-nfs@vger.kernel.org
Subject: Re: [PATCH] Get normalized paths for comparing NFS export paths
Date: Fri, 02 Mar 2012 14:10:25 -0500 [thread overview]
Message-ID: <4F511B21.1020907@RedHat.com> (raw)
In-Reply-To: <20120207204401.GA31752@us.ibm.com>
Hello,
First of all my apologizes for dropping this... I saw Boaz's
NACK and never got back to it.. More comments in-line.
On 02/07/2012 03:44 PM, Malahal Naineni wrote:
> NFSv4 gladly accepts and mounts "hostname:path" instead of
> "hostname:/path". NFS also accepts other forms for pathname, but it
> doesn't use the exact pathname string used for generating /proc/mounts.
> This causes mount entry mistmatch between /etc/mtab and /proc/mounts
> files. The former will have the exact given pathname string but the
> latter will have a modified path name.
>
> Signed-off-by: Malahal Naineni <malahal@us.ibm.com>
> ---
> utils/mount/nfsumount.c | 94 +++++++++++++++++++++++++++++++++++++++++-----
> 1 files changed, 83 insertions(+), 11 deletions(-)
>
> diff --git a/utils/mount/nfsumount.c b/utils/mount/nfsumount.c
> index 3538d88..54542a5 100644
> --- a/utils/mount/nfsumount.c
> +++ b/utils/mount/nfsumount.c
> @@ -139,6 +139,88 @@ static int del_mtab(const char *spec, const char *node)
> return EX_FILEIO;
> }
>
> +/*
> + * Return normalized path.
> + *
> + * Resolve "." and ".." components. Replace multiple slashes with one
> + * slash. realpath is close but doesn't work for us as the path won't
> + * exist on the client.
> + *
> + * The return string must be freed by the caller.
> + */
> +static char *normpath(const char *path)
> +{
> + const char *ptr, *next, *end;
> + char *norm; /* result */
> +
> + if (!path)
> + return NULL;
> +
> + norm = malloc(strlen(path)+1);
> + if (!norm)
> + return NULL;
> +
> + end = path+strlen(path);
> + *norm = '\0'; /* Make it a NULL string */
> + for (ptr = path; ptr < end; ptr = next+1) {
> + next = strchr(ptr, '/');
> + if (!next)
> + next = end;
> + int pclen = next - ptr; /* path component length */
> + if (strncmp(ptr, ".", pclen) == 0)
> + continue;
> + if (strncmp(ptr, "..", pclen) == 0) {
> + char *tmp = strrchr(norm, '/');
> + if (tmp)
> + *tmp = '\0';
> + continue;
> + }
> +
> + /* Add the new component */
> + strncat(norm, "/", 1);
> + strncat(norm, ptr, pclen);
> + }
> +
> + /*
> + * If there was no component to copy, norm would be null.
> + * Return "/" in that case
> + */
> + if (*norm == '\0')
> + strcpy(norm, "/");
> +
> + return norm;
> +}
> +
> +/*
> + * Detect if the given two entries refer to the same mount entry.
> + * Usually one entry is from /etc/mtab and the other is from
> + * /proc/mounts.
> + */
> +static int nfs_same_mount_entry(const struct mntentchn *mc1,
> + const struct mntentchn *mc2)
> +{
> + char *host1, *host2;
> + char *path1, *path2;
> + char *norm1, *norm2;
> + int retval;
> +
> + nfs_parse_devname(mc1->m.mnt_fsname, &host1, &path1);
> + nfs_parse_devname(mc2->m.mnt_fsname, &host2, &path2);
> + norm1 = normpath(path1);
> + norm2 = normpath(path2);
> +
> + retval = strcmp(host1, host2) == 0 && strcmp(norm1, norm2) == 0;
> +
> + free(host1);
> + free(host2);
> + free(path1);
> + free(path2);
> + free(norm1);
> + free(norm2);
> +
> + return retval;
> +}
> +
> /*
> * Detect NFSv4 mounts.
> *
> @@ -161,17 +243,7 @@ static int nfs_umount_is_vers4(const struct mntentchn *mc)
> goto not_found;
>
> do {
> - size_t nlen = strlen(pmc->m.mnt_fsname);
> -
> - /*
> - * It's possible the mount location string in /proc/mounts
> - * ends with a '/'. In this case, if the entry came from
> - * /etc/mtab, it won't have the trailing '/' so deal with
> - * it.
> - */
> - while (pmc->m.mnt_fsname[nlen - 1] == '/')
> - nlen--;
> - if (strncmp(pmc->m.mnt_fsname, mc->m.mnt_fsname, nlen) != 0)
> + if (!nfs_same_mount_entry(pmc, mc))
> continue;
Question, How does this patch stop the mtab from getting polluted with
the device name with multiple "/"?
steved.
next prev parent reply other threads:[~2012-03-02 19:10 UTC|newest]
Thread overview: 27+ messages / expand[flat|nested] mbox.gz Atom feed top
2012-02-03 1:42 [PATCH] Check for beginning '/' in the mount path Malahal Naineni
2012-02-03 11:15 ` Boaz Harrosh
2012-02-03 12:16 ` NeilBrown
2012-02-03 14:29 ` Malahal Naineni
2012-02-05 11:03 ` Boaz Harrosh
2012-02-06 18:11 ` Malahal Naineni
2012-02-07 20:44 ` [PATCH] Get normalized paths for comparing NFS export paths Malahal Naineni
2012-02-16 18:09 ` Malahal Naineni
2012-03-02 19:10 ` Steve Dickson [this message]
2012-03-02 19:27 ` Malahal Naineni
2012-03-02 20:57 ` Steve Dickson
2012-03-02 22:01 ` Malahal Naineni
2012-03-03 17:39 ` Steve Dickson
2012-03-03 19:12 ` Myklebust, Trond
2012-03-04 22:31 ` Steve Dickson
2012-03-04 22:46 ` Myklebust, Trond
2012-03-04 23:08 ` Steve Dickson
2012-03-05 4:46 ` Malahal Naineni
2012-03-05 12:03 ` Steve Dickson
2012-03-04 22:58 ` Steve Dickson
2012-03-04 23:26 ` Myklebust, Trond
2012-03-05 0:03 ` Steve Dickson
2012-03-05 2:04 ` Myklebust, Trond
2012-03-05 4:53 ` Malahal Naineni
2012-03-05 11:55 ` Steve Dickson
2012-03-05 14:47 ` Malahal Naineni
2012-03-05 15:03 ` Steve Dickson
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=4F511B21.1020907@RedHat.com \
--to=steved@redhat.com \
--cc=linux-nfs@vger.kernel.org \
/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).