[Cluster-devel] [PATCH 2/2] mkfs.gfs2: Open the target device with O_EXCL

[Bob Peterson <rpeterso@redhat.com>]

----- Original Message -----
> O_EXCL will let local mounters know that the device is busy while
> mkfs.gfs2 is running so that they don't try to access it.
> 
> Before:
>   # mount /dev/vdc /mnt/test
>   mount: /dev/vdc: can't read superblock
> 
> With O_EXCL:
>   # mount /dev/vdc /mnt/test
>   mount: /dev/vdc is already mounted or /mnt/test busy
> 
> Signed-off-by: Andrew Price <anprice@redhat.com>
> ---
>  gfs2/mkfs/main_mkfs.c | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/gfs2/mkfs/main_mkfs.c b/gfs2/mkfs/main_mkfs.c
> index 107d3cc..4436f93 100644
> --- a/gfs2/mkfs/main_mkfs.c
> +++ b/gfs2/mkfs/main_mkfs.c
> @@ -854,7 +854,7 @@ static void open_dev(struct mkfs_dev *dev)
>  {
>  	int error;
>  
> -	dev->fd = open(dev->path, O_RDWR | O_CLOEXEC);
> +	dev->fd = open(dev->path, O_RDWR|O_CLOEXEC|O_EXCL);
>  	if (dev->fd < 0) {
>  		perror(dev->path);
>  		exit(1);
> --
> 2.4.11
> 
> 
Hi Andy,

In fact, we used to have O_EXCL:

https://git.fedorahosted.org/cgit/cluster.git/commit/gfs2/mkfs/main_mkfs.c?h=RHEL6&id=b0db0bad3e6a8996fcaecbb93bfe9f8de97b7508

I remember Ryan doing this, but I'm not sure why or how it disappeared.
We need to chase down why O_EXCL was removed last time; whether it was just
an oversight/mistake or whether it caused problems elsewhere.

Regards,

Bob Peterson
Red Hat File Systems