Message-ID: <655a8ee6-dd94-effd-738a-9ce8db8ebed7@redhat.com>
Date:   Sat, 11 Mar 2023 11:19:19 -0500
MIME-Version: 1.0
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:102.0) Gecko/20100101
 Thunderbird/102.8.0
Subject: Re: mountd: Possible bug in next_mnt()
Content-Language: en-US
To:     Richard Weinberger <richard@nod.at>,
        linux-nfs <linux-nfs@vger.kernel.org>
Cc:     trond.myklebust@hammerspace.com,
        chris chilvers <chris.chilvers@appsbroker.com>
References: <1497292229.221220.1678287959937.JavaMail.zimbra@nod.at>
From:   Steve Dickson <steved@redhat.com>
In-Reply-To: <1497292229.221220.1678287959937.JavaMail.zimbra@nod.at>
Content-Type: text/plain; charset=UTF-8; format=flowed
Content-Transfer-Encoding: 7bit
Precedence: bulk

Hello,

My apologies... Eating some PTO...

On 3/8/23 10:05 AM, Richard Weinberger wrote:
> Hi!
> 
> next_mnt() finds submounts below a given path p.
> While investigating into an issue in my crossmount patches for nfs-utils I noticed
> that it does not work when fsid=root, rootdir=/some/path/ and then "/" is being exported.
> In this case next_mnt() is asked to find submounts of "/" but returns none.
I'm not clear as what you are saying... "rootdir=/some/path/" is not an
export option.

> 
> In my opinion this wrong because every mount is a submount of "/".
> 
> The following change fixes the problem on my side but I'm not sure whether
> "/" is a special case in mountd where next_mnt() has to bail out.
> 
> diff --git a/support/export/cache.c b/support/export/cache.c
> index 2497d4f48df3..be20cb34adcb 100644
> --- a/support/export/cache.c
> +++ b/support/export/cache.c
> @@ -410,13 +410,13 @@ static char *next_mnt(void **v, char *p)
>                  *v = f;
>          } else
>                  f = *v;
> -       while ((me = getmntent(f)) != NULL && l > 1) {
> +       while ((me = getmntent(f)) != NULL && l >= 1) {
>                  char *mnt_dir = nfsd_path_strip_root(me->mnt_dir);
>   
>                  if (!mnt_dir)
>                          continue;
>   
> -               if (strncmp(mnt_dir, p, l) == 0 && mnt_dir[l] == '/')
> +               if (strncmp(mnt_dir, p, l) == 0 && (l == 1 || mnt_dir[l] == '/'))
>                          return mnt_dir;
>          }
>          endmntent(f);
> 
> Comments? :-)
Putting this is the correct patch format including a Signed-off-by
signature... would help.

steved.
> 
> Thanks,
> //richard
>