[CRIU] [PATCH 3/4] mount: add support for external block devices (v3)

[Andrew Vagin]

On Fri, Apr 22, 2016 at 04:49:48PM +0300, Pavel Emelyanov wrote:
> On 04/22/2016 09:12 AM, Andrey Vagin wrote:
> > From: Andrew Vagin <avagin at virtuozzo.com>
> > 
> > The idea is to allow to dump and restore mounts of
> > a specified block device.
> > 
> > Options:
> > dump:		--ext-mount-map dev[MAJOR:MINOR]:VAL
> 
> Why dev[MAJOR:MINOR]? Isn't just path (which is a mountpoint) enough?

We don't know where this device is mounted, we know that this device
may be mounted somewhere in a container.

In OpenVZ we can grant permissions to mount a block device in a
container. And a user of CT decides where to mount this device.

> 
> > restore:	--ext-mount-map VAL:DEVPATH
> > 
> > If we find a mount with a specified block device, we
> > set its type into FSTYPE__AUTO and write VAL into
> > the "source" field.
> > 
> > VAL is replaced on DEVPATH on restore.
> > 
> > https://jira.sw.ru/browse/PSBM-39381
> > 
> > v2: use --ext-mount-map instead of --external on dump
> > v3: clean up
> > 
> > Signed-off-by: Andrew Vagin <avagin at virtuozzo.com>
> > ---
> >  Documentation/criu.txt |  6 +++---
> >  criu/mount.c           | 35 +++++++++++++++++++++++++++++------
> >  2 files changed, 32 insertions(+), 9 deletions(-)
> > 
> > diff --git a/Documentation/criu.txt b/Documentation/criu.txt
> > index 1982f5d..07da757 100644
> > --- a/Documentation/criu.txt
> > +++ b/Documentation/criu.txt
> > @@ -176,9 +176,9 @@ In other words, do not use it until really needed.
> >      where locks are not holed by someone outside of it.
> >  
> >  *-M*, *--ext-mount-map* '<KEY>'*:*'<VAL>'::
> > -    Setup mapping for external mounts. '<KEY>' is a mountpoint inside container
> > -    and corresponding '<VAL>' is a string that will be written into the image
> > -    as mountpoint\'s root value.
> > +    Setup mapping for external mounts. '<KEY>' is a mountpoint or a device
> > +    (dev[maj:min]]) inside container and corresponding '<VAL>' is a string
> > +    that will be written into the image as mountpoint\'s root or source value.
> >  
> >  *--link-remap*::
> >      Allow one to link unlinked files back when possible (modifies FS till *restore*).
> > diff --git a/criu/mount.c b/criu/mount.c
> > index f57b10c..f8d0cc2 100644
> > --- a/criu/mount.c
> > +++ b/criu/mount.c
> > @@ -40,6 +40,8 @@
> >  #undef	LOG_PREFIX
> >  #define LOG_PREFIX "mnt: "
> >  
> > +static struct fstype fstypes[];
> > +
> >  int ext_mount_add(char *key, char *val)
> >  {
> >  	struct ext_mount *em;
> > @@ -487,15 +489,31 @@ static void mnt_tree_show(struct mount_info *tree, int off)
> >  static int try_resolve_ext_mount(struct mount_info *info)
> >  {
> >  	struct ext_mount *em;
> > +	char devstr[64];
> >  
> >  	em = ext_mount_lookup(info->mountpoint + 1 /* trim the . */);
> > -	if (em == NULL)
> > -		return -ENOTSUP;
> > +	if (em) {
> > +		pr_info("Found %s mapping for %s mountpoint\n",
> > +				em->val, info->mountpoint);
> > +		info->external = em;
> > +		return 0;
> > +	}
> >  
> > -	pr_info("Found %s mapping for %s mountpoint\n",
> > -			em->val, info->mountpoint);
> > -	info->external = em;
> > -	return 0;
> > +	snprintf(devstr, sizeof(devstr), "dev[%d/%d]",
> > +			kdev_major(info->s_dev),  kdev_minor(info->s_dev));
> > +
> > +	if (info->fstype->code == FSTYPE__UNSUPPORTED) {
> > +		em = ext_mount_lookup(devstr);
> > +		if (em) {
> > +			info->fstype = &fstypes[1];
> > +			BUG_ON(info->fstype->code != FSTYPE__AUTO);
> > +			xfree(info->source);
> > +			info->source = xstrdup(em->val);
> > +			return 0;
> > +		}
> > +	}
> > +
> > +	return -ENOTSUP;
> >  }
> >  
> >  static struct mount_info *find_widest_shared(struct mount_info *m)
> > @@ -2091,6 +2109,11 @@ static char *resolve_source(struct mount_info *mi)
> >  
> >  	if (mi->fstype->code == FSTYPE__AUTO) {
> >  		struct stat st;
> > +		struct ext_mount *key;
> > +
> > +		key = ext_mount_lookup(mi->source);
> > +		if (key)
> > +			return key->val;
> >  
> >  		if (!stat(mi->source, &st) && S_ISBLK(st.st_mode) &&
> >  		    major(st.st_rdev) == kdev_major(mi->s_dev) &&
> > 
>