Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) client-ip=23.128.96.18;
Date:   Sat, 5 Jun 2021 15:00:16 +0200
From:   Christian Brauner <christian.brauner@ubuntu.com>
To:     Giuseppe Scrivano <gscrivan@redhat.com>
Cc:     ebiederm@xmission.com, "Serge E. Hallyn" <serge@hallyn.com>,
        linux-kernel@vger.kernel.org
Subject: Re: [PATCH v2] userns: automatically split user namespace extent
Message-ID: <20210605130016.jdkkviwtuefocset@wittgenstein>
References: <20201203150252.1229077-1-gscrivan@redhat.com>
 <20210510172351.GA19918@mail.hallyn.com>
 <20210510185715.GA20897@mail.hallyn.com>
 <87h7idbskw.fsf@redhat.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=utf-8
Content-Disposition: inline
In-Reply-To: <87h7idbskw.fsf@redhat.com>
Precedence: bulk

On Fri, Jun 04, 2021 at 04:41:19PM +0200, Giuseppe Scrivano wrote:
> Christian, Eric,
> 
> are you fine with this patch or is there anything more you'd like me to
> change?

Before being a little bit of a party pooper thanks for your patches! I
appreciate the work you're doing!

So my concern is that this may cause silent regressions/security issues
for tools in userspace by making this work automagically.

For example we have a go library that calculates idmap ranges and
extents. Those idmappings are stored in the database and in the
container's config and for backups and so on.

The calculated extents match exactly with how these lines look in
/proc/<pid>/*id_map.
If we miscalculate the extents and we try to write them to
/proc/<pid>/*id_map we get told to go away and immediately recognize the
bug.
With this patch however we may succeed and then we record misleading
extents in the db or the config.

Turning this into a general concern, I think it is a non-trivial
semantic change to break up the 1:1 correspondence between mappings
written and mappings applied that we had for such a long time now.

In general I'm not sure it should be the kernel that has the idmapping
ranges smarts.

I'd rather see a generic userspace library that container runtimes make
use of that also breaks up idmapping ranges. We can certainly accomodate
this in
https://pkg.go.dev/github.com/lxc/lxd/shared/idmap

Is that a reasonable concern?

Christian

> 
> Thanks,
> Giuseppe
> 
> 
> 
> "Serge E. Hallyn" <serge@hallyn.com> writes:
> 
> > On Mon, May 10, 2021 at 12:23:51PM -0500, Serge E. Hallyn wrote:
> >> On Thu, Dec 03, 2020 at 04:02:52PM +0100, Giuseppe Scrivano wrote:
> >> > writing to the id map fails when an extent overlaps multiple mappings
> >> > in the parent user namespace, e.g.:
> >> > 
> >> > $ cat /proc/self/uid_map
> >> >          0       1000          1
> >> >          1     100000      65536
> >> > $ unshare -U sleep 100 &
> >> > [1] 1029703
> >> > $ printf "0 0 100\n" | tee /proc/$!/uid_map
> >> > 0 0 100
> >> > tee: /proc/1029703/uid_map: Operation not permitted
> >> > 
> >> > To prevent it from happening, automatically split an extent so that
> >> > each portion fits in one extent in the parent user namespace.
> >> > 
> >> > $ cat /proc/self/uid_map
> >> >          0       1000          1
> >> >          1     110000      65536
> >> > $ unshare -U sleep 100 &
> >> > [1] 1552
> >> > $ printf "0 0 100\n" | tee /proc/$!/uid_map
> >> > 0 0 100
> >> > $ cat /proc/$!/uid_map
> >> >          0          0          1
> >> >          1          1         99
> >> > 
> >> > Signed-off-by: Giuseppe Scrivano <gscrivan@redhat.com>
> >> 
> >> The patch on the whole looks great, easy to reason about.  But I have
> >> one question below:
> >
> > As you pointed out, I was misreading the variable name, thank you :)
> >
> > Reviewed-by: Serge Hallyn <serge@hallyn.com>
> >
> >> 
> >> > ---
> >> > v2:
> >> > - move the split logic when the extent are mapped to the parent map to
> >> >   reduce lookup complexity.
> >> > 
> >> > v1: https://lkml.kernel.org/lkml/20201126100839.381415-1-gscrivan@redhat.com
> >> > 
> >> >  kernel/user_namespace.c | 79 +++++++++++++++++++++++++++++++++++------
> >> >  1 file changed, 68 insertions(+), 11 deletions(-)
> >> > 
> >> > diff --git a/kernel/user_namespace.c b/kernel/user_namespace.c
> >> > index 87804e0371fe..550612c6e794 100644
> >> > --- a/kernel/user_namespace.c
> >> > +++ b/kernel/user_namespace.c
> >> > @@ -312,6 +312,55 @@ static u32 map_id_down(struct uid_gid_map *map, u32 id)
> >> >  	return map_id_range_down(map, id, 1);
> >> >  }
> >> >  
> >> > +/**
> >> > + * find_and_split_extent_down - Find lower_first for the target extent
> >> > + * using the specified map.
> >> > + * If the extent doesn't fit in a single lower extent, split target and
> >> > + * write the remaining IDs (first and count) to the overflow extent.
> >> > + * If no overflow happens, overflow->count is set to 0.
> >> > + */
> >> > +static int find_and_split_extent_down(struct uid_gid_map *map,
> >> > +				       struct uid_gid_extent *target,
> >> > +				       struct uid_gid_extent *overflow)
> >> > +{
> >> > +	unsigned int extents = map->nr_extents;
> >> > +	u32 lower_first = target->lower_first;
> >> > +	struct uid_gid_extent *extent;
> >> > +	u32 off, available;
> >> > +
> >> > +	overflow->count = 0;
> >> > +
> >> > +	/* Find the lower extent that includes the first ID.  */
> >> > +	if (extents <= UID_GID_MAP_MAX_BASE_EXTENTS)
> >> > +		extent = map_id_range_down_base(extents, map, lower_first, 1);
> >> > +	else
> >> > +		extent = map_id_range_down_max(extents, map, lower_first, 1);
> >> > +
> >> > +	/* Could not map the first ID in the extent.  */
> >> > +	if (extent == NULL)
> >> > +		return -EPERM;
> >> > +
> >> > +	/* Offset of lower_first in the lower extent.  */
> >> > +	off = target->lower_first - extent->first;
> >> > +
> >> > +	/* How many IDs are available in the lower extent?  */
> >> > +	available = extent->count - off;
> >> > +
> >> > +	/* Requesting more IDs than available in the lower extent.  */
> >> > +	if (target->count > available) {
> >> > +		/* Move the remaining IDs to the overflow extent.  */
> >> > +		overflow->first = target->first + available;
> >> > +		overflow->lower_first = target->lower_first + available;
> >> > +		overflow->count = target->count - available;
> >> > +
> >> > +		/* Shrink the initial extent to what is available.  */
> >> > +		target->count = available;
> >> > +	}
> >> > +
> >> > +	target->lower_first = extent->lower_first + off;
> >> > +	return 0;
> >> > +}
> >> > +
> >> >  /**
> >> >   * map_id_up_base - Find idmap via binary search in static extent array.
> >> >   * Can only be called if number of mappings is equal or less than
> >> > @@ -749,6 +798,7 @@ static bool mappings_overlap(struct uid_gid_map *new_map,
> >> >   * insert_extent - Safely insert a new idmap extent into struct uid_gid_map.
> >> >   * Takes care to allocate a 4K block of memory if the number of mappings exceeds
> >> >   * UID_GID_MAP_MAX_BASE_EXTENTS.
> >> > + * The extent is appended at the position map->nr_extents.
> >> >   */
> >> >  static int insert_extent(struct uid_gid_map *map, struct uid_gid_extent *extent)
> >> >  {
> >> > @@ -968,30 +1018,37 @@ static ssize_t map_write(struct file *file, const char __user *buf,
> >> >  	if (!new_idmap_permitted(file, ns, cap_setid, &new_map))
> >> >  		goto out;
> >> >  
> >> > -	ret = -EPERM;
> >> >  	/* Map the lower ids from the parent user namespace to the
> >> >  	 * kernel global id space.
> >> >  	 */
> >> >  	for (idx = 0; idx < new_map.nr_extents; idx++) {
> >> > +		struct uid_gid_extent overflow;
> >> >  		struct uid_gid_extent *e;
> >> > -		u32 lower_first;
> >> >  
> >> >  		if (new_map.nr_extents <= UID_GID_MAP_MAX_BASE_EXTENTS)
> >> >  			e = &new_map.extent[idx];
> >> >  		else
> >> >  			e = &new_map.forward[idx];
> >> >  
> >> > -		lower_first = map_id_range_down(parent_map,
> >> > -						e->lower_first,
> >> > -						e->count);
> >> > -
> >> > -		/* Fail if we can not map the specified extent to
> >> > -		 * the kernel global id space.
> >> > -		 */
> >> > -		if (lower_first == (u32) -1)
> >> > +		ret = find_and_split_extent_down(parent_map, e, &overflow);
> >> > +		if (ret < 0)
> >> >  			goto out;
> >> >  
> >> > -		e->lower_first = lower_first;
> >> > +		/* If the extent doesn't fit in a single lower extent,
> >> > +		 * move what could not be mapped to a new extent.
> >> > +		 * The new extent is appended to the existing ones in
> >> > +		 * new_map, it will be checked again and if necessary it
> >> > +		 * is split further.
> >> > +		 */
> >> > +		if (overflow.count > 0) {
> >> > +			if (new_map.nr_extents == UID_GID_MAP_MAX_EXTENTS) {
> >> 
> >> Why are you doing this?  The insert_extent() will automatically extend it
> >> if needed, right?  So this condition should be fine?
> >> 
> >> > +				ret = -EINVAL;
> >> > +				goto out;
> >> > +			}
> >> > +			ret = insert_extent(&new_map, &overflow);
> >> > +			if (ret < 0)
> >> > +				goto out;
> >> > +		}
> >> >  	}
> >> >  
> >> >  	/*
> >> > -- 
> >> > 2.28.0
> >> 
> >> Cheers,
> >> Balint
> >> 
> >> >
> >> > -serge
> >> 
> >> 
> >> 
> >> -- 
> >> Balint Reczey
> >> Ubuntu & Debian Developer
> >> > 
> >
>