Now that we changed the generic reference counting mechanism for all
namespaces to never manipulate reference counts of initial namespaces we
can drop the special handling for pid namespaces.
The ipc namespace may call put_ipc_ns() and get_ipc_ns() before it is
added to the namespace tree. Assign the id early like we do for a some
other namespaces.
fs: use boolean to indicate anonymous mount namespace
Stop playing games with the namespace id and use a boolean instead:
* This will remove the special-casing we need to do everywhere for mount
namespaces.
* It will allow us to use asserts on the namespace id for initial
namespaces everywhere.
* It will allow us to put anonymous mount namespaces on the namespaces
trees in the future and thus make them available to statmount() and
listmount().
Foward declare struct ns_common and remove the include of ns_common.h.
We want ns_common.h to possibly include nstree structures but not the
other way around.
Merge patch series "ns: fixes for namespace iteration and active reference counting"
Christian Brauner <brauner@kernel.org> says:
* Make sure to initialize the active reference count for the initial
network namespace and prevent __ns_common_init() from returning too
early.
* Make sure that passive reference counts are dropped outside of rcu
read locks as some namespaces such as the mount namespace do in fact
sleep when putting the last reference.
* The setns() system call supports:
(1) namespace file descriptors (nsfd)
(2) process file descriptors (pidfd)
When using nsfds the namespaces will remain active because they are
pinned by the vfs. However, when pidfds are used things are more
complicated.
When the target task exits and passes through exit_nsproxy_namespaces()
or is reaped and thus also passes through exit_cred_namespaces() after
the setns()'ing task has called prepare_nsset() but before the active
reference count of the set of namespaces it wants to setns() to might
have been dropped already:
// ns_active_ref_put() will also put
// the reference on the owner of the
// namespace. If the only reason the
// owning namespace was alive was
// because it was a parent of @ns
// it's active reference count now goes
// to zero... --------------------------------
// |
// ns->__ns_active_ref == 0 |
// parent_ns->__ns_active_ref == 0 |
| commit_nsset()
-----------------> // If setns()
// now manages to install the namespaces
// it will call ns_active_ref_get()
// on them thus bumping the active reference
// count from zero again but without also
// taking the required reference on the owner.
// Thus we get:
//
// ns->__ns_active_ref == 1
// parent_ns->__ns_active_ref == 0
When later someone does ns_active_ref_put() on @ns it will underflow
parent_ns->__ns_active_ref leading to a splat from our asserts
thinking there are still active references when in fact the counter
just underflowed.
So resurrect the ownership chain if necessary as well. If the caller
succeeded to grab passive references to the set of namespaces the
setns() should simply succeed even if the target task exists or gets
reaped in the meantime.
The race is rare and can only be triggered when using pidfs to setns()
to namespaces. Also note that active reference on initial namespaces are
nops.
Since we now always handle parent references directly we can drop
ns_ref_active_get_owner() when adding a namespace to a namespace tree.
This is now all handled uniformly in the places where the new namespaces
actually become active.
* patches from https://patch.msgid.link/20251109-namespace-6-19-fixes-v1-0-ae8a4ad5a3b3@kernel.org:
selftests/namespaces: test for efault
selftests/namespaces: add active reference count regression test
ns: add asserts for active refcount underflow
ns: handle setns(pidfd, ...) cleanly
ns: return EFAULT on put_user() error
ns: make sure reference are dropped outside of rcu lock
ns: don't increment or decrement initial namespaces
ns: don't skip active reference count initialization
Merge patch "kbuild: Add '-fms-extensions' to areas with dedicated CFLAGS"
Nathan Chancellor <nathan@kernel.org> says:
Shared branch between Kbuild and other trees for enabling
'-fms-extensions' for 6.19.
* tag 'kbuild-ms-extensions-6.19' of ssh://gitolite.kernel.org/pub/scm/linux/kernel/git/kbuild/linux:
kbuild: Add '-fms-extensions' to areas with dedicated CFLAGS
Kbuild: enable -fms-extensions
jfs: Rename _inline to avoid conflict with clang's '-fms-extensions'
(1) namespace file descriptors (nsfd)
(2) process file descriptors (pidfd)
When using nsfds the namespaces will remain active because they are
pinned by the vfs. However, when pidfds are used things are more
complicated.
When the target task exits and passes through exit_nsproxy_namespaces()
or is reaped and thus also passes through exit_cred_namespaces() after
the setns()'ing task has called prepare_nsset() but before the active
reference count of the set of namespaces it wants to setns() to might
have been dropped already:
// ns_active_ref_put() will also put
// the reference on the owner of the
// namespace. If the only reason the
// owning namespace was alive was
// because it was a parent of @ns
// it's active reference count now goes
// to zero... --------------------------------
// |
// ns->__ns_active_ref == 0 |
// parent_ns->__ns_active_ref == 0 |
| commit_nsset()
-----------------> // If setns()
// now manages to install the namespaces
// it will call ns_active_ref_get()
// on them thus bumping the active reference
// count from zero again but without also
// taking the required reference on the owner.
// Thus we get:
//
// ns->__ns_active_ref == 1
// parent_ns->__ns_active_ref == 0
When later someone does ns_active_ref_put() on @ns it will underflow
parent_ns->__ns_active_ref leading to a splat from our asserts
thinking there are still active references when in fact the counter
just underflowed.
So resurrect the ownership chain if necessary as well. If the caller
succeeded to grab passive references to the set of namespaces the
setns() should simply succeed even if the target task exists or gets
reaped in the meantime and thus has dropped all active references to its
namespaces.
The race is rare and can only be triggered when using pidfs to setns()
to namespaces. Also note that active reference on initial namespaces are
nops.
Since we now always handle parent references directly we can drop
ns_ref_active_get_owner() when adding a namespace to a namespace tree.
This is now all handled uniformly in the places where the new namespaces
actually become active.
ns: make sure reference are dropped outside of rcu lock
The mount namespace may in fact sleep when putting the last passive
reference so we need to drop the namespace reference outside of the rcu
read lock. Do this by delaying the put until the next iteration where
we've already moved on to the next namespace and legitimized it. Once we
drop the rcu read lock to call put_user() we will also drop the
reference to the previous namespace in the tree.
As announced a while ago this is the next step building on the nstree
work from prior cycles. There's a bunch of fixes and semantic cleanups
in here and a ton of tests.
Currently listns() is relying on active namespace reference counts which
are introduced alongside this series.
While a namespace is on the namespace trees with a valid reference count
it is possible to reopen it through a namespace file handle. This is all
fine but has some issues that should be addressed.
On current kernels a namespace is visible to userspace in the
following cases:
(1) The namespace is in use by a task.
(2) The namespace is persisted through a VFS object (namespace file
descriptor or bind-mount).
Note that (2) only cares about direct persistence of the namespace
itself not indirectly via e.g., file->f_cred file references or
similar.
(3) The namespace is a hierarchical namespace type and is the parent of
a single or multiple child namespaces.
Case (3) is interesting because it is possible that a parent namespace
might not fulfill any of (1) or (2), i.e., is invisible to userspace but
it may still be resurrected through the NS_GET_PARENT ioctl().
Currently namespace file handles allow much broader access to namespaces
than what is currently possible via (1)-(3). The reason is that
namespaces may remain pinned for completely internal reasons yet are
inaccessible to userspace.
For example, a user namespace my remain pinned by get_cred() calls to
stash the opener's credentials into file->f_cred. As it stands file
handles allow to resurrect such a users namespace even though this
should not be possible via (1)-(3). This is a fundamental uapi change
that we shouldn't do if we don't have to.
Consider the following insane case: Various architectures support the
CONFIG_MMU_LAZY_TLB_REFCOUNT option which uses lazy TLB destruction.
When this option is set a userspace task's struct mm_struct may be used
for kernel threads such as the idle task and will only be destroyed once
the cpu's runqueue switches back to another task. But because of ptrace()
permission checks struct mm_struct stashes the user namespace of the
task that struct mm_struct originally belonged to. The kernel thread
will take a reference on the struct mm_struct and thus pin it.
So on an idle system user namespaces can be persisted for arbitrary
amounts of time which also means that they can be resurrected using
namespace file handles. That makes no sense whatsoever. The problem is
of course excarabted on large systems with a huge number of cpus.
To handle this nicely we introduce an active reference count which
tracks (1)-(3). This is easy to do as all of these things are already
managed centrally. Only (1)-(3) will count towards the active reference
count and only namespaces which are active may be opened via namespace
file handles.
The problem is that namespaces may be resurrected. Which means that they
can become temporarily inactive and will be reactived some time later.
Currently the only example of this is the SIOGCSKNS socket ioctl. The
SIOCGSKNS ioctl allows to open a network namespace file descriptor based
on a socket file descriptor.
If a socket is tied to a network namespace that subsequently becomes
inactive but that socket is persisted by another process in another
network namespace (e.g., via SCM_RIGHTS of pidfd_getfd()) then the
SIOCGSKNS ioctl will resurrect this network namespace.
So calls to open_related_ns() and open_namespace() will end up
resurrecting the corresponding namespace tree.
Note that the active reference count does not regulate the lifetime of
the namespace itself. This is still done by the normal reference count.
The active reference count can only be elevated if the regular reference
count is elevated.
The active reference count also doesn't regulate the presence of a
namespace on the namespace trees. It only regulates its visiblity to
namespace file handles (and in later patches to listns()).
A namespace remains on the namespace trees from creation until its
actual destruction. This will allow the kernel to always reach any
namespace trivially and it will also enable subsystems like bpf to walk
the namespace lists on the system for tracing or general introspection
purposes.
Note that different namespaces have different visibility lifetimes on
current kernels. While most namespace are immediately released when the
last task using them exits, the user- and pid namespace are persisted
and thus both remain accessible via /proc/<pid>/ns/<ns_type>.
The user namespace lifetime is aliged with struct cred and is only
released through exit_creds(). However, it becomes inaccessible to
userspace once the last task using it is reaped, i.e., when
release_task() is called and all proc entries are flushed. Similarly,
the pid namespace is also visible until the last task using it has been
reaped and the associated pid numbers are freed.
The active reference counts of the user- and pid namespace are
decremented once the task is reaped.
Based on the namespace trees and the active reference count, a new
listns() system call that allows userspace to iterate through namespaces
in the system. This provides a programmatic interface to discover and
inspect namespaces, enhancing existing namespace apis.
Currently, there is no direct way for userspace to enumerate namespaces
in the system. Applications must resort to scanning /proc/<pid>/ns/
across all processes, which is:
1. Inefficient - requires iterating over all processes
2. Incomplete - misses inactive namespaces that aren't attached to any
running process but are kept alive by file descriptors, bind mounts,
or parent namespace references
3. Permission-heavy - requires access to /proc for many processes
4. No ordering or ownership.
5. No filtering per namespace type: Must always iterate and check all
namespaces.
The list goes on. The listns() system call solves these problems by
providing direct kernel-level enumeration of namespaces. It is similar
to listmount() but obviously tailored to namespaces.
/*
* @req: Pointer to struct ns_id_req specifying search parameters
* @ns_ids: User buffer to receive namespace IDs
* @nr_ns_ids: Size of ns_ids buffer (maximum number of IDs to return)
* @flags: Reserved for future use (must be 0)
*/
ssize_t listns(const struct ns_id_req *req, u64 *ns_ids,
size_t nr_ns_ids, unsigned int flags);
Returns:
- On success: Number of namespace IDs written to ns_ids
- On error: Negative error code
/*
* @size: Structure size
* @ns_id: Starting point for iteration; use 0 for first call, then
* use the last returned ID for subsequent calls to paginate
* @ns_type: Bitmask of namespace types to include (from enum ns_type):
* 0: Return all namespace types
* MNT_NS: Mount namespaces
* NET_NS: Network namespaces
* USER_NS: User namespaces
* etc. Can be OR'd together
* @user_ns_id: Filter results to namespaces owned by this user namespace:
* 0: Return all namespaces (subject to permission checks)
* LISTNS_CURRENT_USER: Namespaces owned by caller's user namespace
* Other value: Namespaces owned by the specified user namespace ID
*/
struct ns_id_req {
__u32 size; /* sizeof(struct ns_id_req) */
__u32 spare; /* Reserved, must be 0 */
__u64 ns_id; /* Last seen namespace ID (for pagination) */
__u32 ns_type; /* Filter by namespace type(s) */
__u32 spare2; /* Reserved, must be 0 */
__u64 user_ns_id; /* Filter by owning user namespace */
};
Example 1: List all namespaces
void list_all_namespaces(void)
{
struct ns_id_req req = {
.size = sizeof(req),
.ns_id = 0, /* Start from beginning */
.ns_type = 0, /* All types */
.user_ns_id = 0, /* All user namespaces */
};
uint64_t ids[100];
ssize_t ret;
printf("All namespaces in the system:\n");
do {
ret = listns(&req, ids, 100, 0);
if (ret < 0) {
perror("listns");
break;
}
for (ssize_t i = 0; i < ret; i++)
printf(" Namespace ID: %llu\n", (unsigned long long)ids[i]);
/* Continue from last seen ID */
if (ret > 0)
req.ns_id = ids[ret - 1];
} while (ret == 100); /* Buffer was full, more may exist */
}
printf("Enumerating all namespaces with pagination:\n");
while (1) {
ret = listns(&req, ids, 50, 0);
if (ret < 0) {
perror("listns");
break;
}
if (ret == 0)
break; /* No more namespaces */
total += ret;
printf(" Batch: %zd namespaces\n", ret);
/* Last ID in this batch becomes start of next batch */
req.ns_id = ids[ret - 1];
if (ret < 50)
break; /* Partial batch = end of results */
}
printf("Total: %zu namespaces\n", total);
}
listns() respects namespace isolation and capabilities:
(1) Global listing (user_ns_id = 0):
- Requires CAP_SYS_ADMIN in the namespace's owning user namespace
- OR the namespace must be in the caller's namespace context (e.g.,
a namespace the caller is currently using)
- User namespaces additionally allow listing if the caller has
CAP_SYS_ADMIN in that user namespace itself
(2) Owner-filtered listing (user_ns_id != 0):
- Requires CAP_SYS_ADMIN in the specified owner user namespace
- OR the namespace must be in the caller's namespace context
- This allows unprivileged processes to enumerate namespaces they own
(3) Visibility:
- Only "active" namespaces are listed
- A namespace is active if it has a non-zero __ns_ref_active count
- This includes namespaces used by running processes, held by open
file descriptors, or kept active by bind mounts
- Inactive namespaces (kept alive only by internal kernel
references) are not visible via listns()
* patches from https://patch.msgid.link/20251029-work-namespace-nstree-listns-v4-0-2e6f823ebdc0@kernel.org: (74 commits)
selftests/namespace: test listns() pagination
selftests/namespace: add stress test
selftests/namespace: commit_creds() active reference tests
selftests/namespace: third threaded active reference count test
selftests/namespace: second threaded active reference count test
selftests/namespace: first threaded active reference count test
selftests/namespaces: twelth inactive namespace resurrection test
selftests/namespaces: eleventh inactive namespace resurrection test
selftests/namespaces: tenth inactive namespace resurrection test
selftests/namespaces: ninth inactive namespace resurrection test
selftests/namespaces: eigth inactive namespace resurrection test
selftests/namespaces: seventh inactive namespace resurrection test
selftests/namespaces: sixth inactive namespace resurrection test
selftests/namespaces: fifth inactive namespace resurrection test
selftests/namespaces: fourth inactive namespace resurrection test
selftests/namespaces: third inactive namespace resurrection test
selftests/namespaces: second inactive namespace resurrection test
selftests/namespaces: first inactive namespace resurrection test
selftests/namespaces: seventh listns() permission test
selftests/namespaces: sixth listns() permission test
...
Stress tests for namespace active reference counting.
These tests validate that the active reference counting system can
handle high load scenarios including rapid namespace
creation/destruction, large numbers of concurrent namespaces, and
various edge cases under stress.
selftests/namespace: third threaded active reference count test
Test that namespaces become inactive after subprocess with multiple
threads exits. Create a subprocess that unshares user and network
namespaces, then creates two threads that share those namespaces. Verify
that after all threads and subprocess exit, the namespaces are no longer
listed by listns() and cannot be opened by open_by_handle_at().
selftests/namespace: second threaded active reference count test
Test that a namespace remains active while a thread holds an fd to it.
Even after the thread exits, the namespace should remain active as long
as another thread holds a file descriptor to it.
selftests/namespaces: twelth inactive namespace resurrection test
Test multi-level namespace resurrection across three user namespace levels.
This test creates a complex namespace hierarchy with three levels of user
namespaces and a network namespace at the deepest level. It verifies that
the resurrection semantics work correctly when SIOCGSKNS is called on a
socket from an inactive namespace tree, and that listns() and
open_by_handle_at() correctly respect visibility rules.
Hierarchy after child processes exit (all with 0 active refcount):
net_L3A (0) <- Level 3 network namespace
|
+
userns_L3 (0) <- Level 3 user namespace
|
+
userns_L2 (0) <- Level 2 user namespace
|
+
userns_L1 (0) <- Level 1 user namespace
|
x
init_user_ns
The test verifies:
1. SIOCGSKNS on a socket from inactive net_L3A resurrects the entire chain
2. After resurrection, all namespaces are visible in listns()
3. Resurrected namespaces can be reopened via file handles
4. Closing the netns FD cascades down: the entire ownership chain
(userns_L3 -> userns_L2 -> userns_L1) becomes inactive again
5. Inactive namespaces disappear from listns() and cannot be reopened
6. Calling SIOCGSKNS again on the same socket resurrects the tree again
7. After second resurrection, namespaces are visible and can be reopened
selftests/namespaces: eleventh inactive namespace resurrection test
Test combined listns() and file handle operations with socket-kept
netns. Create a netns, keep it alive with a socket, verify it appears in
listns(), then reopen it via file handle obtained from listns() entry.
selftests/namespaces: tenth inactive namespace resurrection test
Test that socket-kept netns can be reopened via file handle.
Verify that a network namespace kept alive by a socket FD can be
reopened using file handles even after the creating process exits.
selftests/namespaces: ninth inactive namespace resurrection test
Test that socket-kept netns appears in listns() output.
Verify that a network namespace kept alive by a socket FD appears in
listns() output even after the creating process exits, and that it
disappears when the socket is closed.
selftests/namespaces: second inactive namespace resurrection test
Test that socket file descriptors keep network namespaces active. Create
a network namespace, create a socket in it, then exit the namespace. The
namespace should remain active while the socket FD is held.
selftests/namespaces: third listns() permission test
Test that users cannot see namespaces from unrelated user namespaces.
Create two sibling user namespaces, verify they can't see each other's
owned namespaces.
selftests/namespaces: first listns() permission test
Test that unprivileged users can only see namespaces they're currently
in. Create a namespace, drop privileges, verify we can only see our own
namespaces.
selftests/namespaces: fourteenth active reference count tests
Test that user namespace as a child also propagates correctly.
Create user_A -> user_B, verify when user_B is active that user_A
is also active. This is different from non-user namespace children.
selftests/namespaces: thirteenth active reference count tests
Test that parent stays active as long as ANY child is active.
Create parent user namespace with two child net namespaces.
Parent should remain active until BOTH children are inactive.
selftests/namespaces: twelth active reference count tests
Test hierarchical propagation with deep namespace hierarchy.
Create: init_user_ns -> user_A -> user_B -> net_ns
When net_ns is active, both user_A and user_B should be active.
This verifies the conditional recursion in __ns_ref_active_put() works.
selftests/namespaces: seventh active reference count tests
Test hierarchical active reference propagation.
When a child namespace is active, its owning user namespace should also
be active automatically due to hierarchical active reference propagation.
This ensures parents are always reachable when children are active.
selftests/namespaces: sixth active reference count tests
Test that an open file descriptor keeps a namespace active.
Even after the creating process exits, the namespace should remain
active as long as an fd is held open.
selftests/namespaces: second active reference count tests
Test namespace lifecycle: create a namespace in a child process, get a
file handle while it's active, then try to reopen after the process
exits (namespace becomes inactive).
Add a new listns() system call that allows userspace to iterate through
namespaces in the system. This provides a programmatic interface to
discover and inspect namespaces, enhancing existing namespace apis.
Currently, there is no direct way for userspace to enumerate namespaces
in the system. Applications must resort to scanning /proc/<pid>/ns/
across all processes, which is:
1. Inefficient - requires iterating over all processes
2. Incomplete - misses inactive namespaces that aren't attached to any
running process but are kept alive by file descriptors, bind mounts,
or parent namespace references
3. Permission-heavy - requires access to /proc for many processes
4. No ordering or ownership.
5. No filtering per namespace type: Must always iterate and check all
namespaces.
The list goes on. The listns() system call solves these problems by
providing direct kernel-level enumeration of namespaces. It is similar
to listmount() but obviously tailored to namespaces.
/*
* @req: Pointer to struct ns_id_req specifying search parameters
* @ns_ids: User buffer to receive namespace IDs
* @nr_ns_ids: Size of ns_ids buffer (maximum number of IDs to return)
* @flags: Reserved for future use (must be 0)
*/
ssize_t listns(const struct ns_id_req *req, u64 *ns_ids,
size_t nr_ns_ids, unsigned int flags);
Returns:
- On success: Number of namespace IDs written to ns_ids
- On error: Negative error code
/*
* @size: Structure size
* @ns_id: Starting point for iteration; use 0 for first call, then
* use the last returned ID for subsequent calls to paginate
* @ns_type: Bitmask of namespace types to include (from enum ns_type):
* 0: Return all namespace types
* MNT_NS: Mount namespaces
* NET_NS: Network namespaces
* USER_NS: User namespaces
* etc. Can be OR'd together
* @user_ns_id: Filter results to namespaces owned by this user namespace:
* 0: Return all namespaces (subject to permission checks)
* LISTNS_CURRENT_USER: Namespaces owned by caller's user namespace
* Other value: Namespaces owned by the specified user namespace ID
*/
struct ns_id_req {
__u32 size; /* sizeof(struct ns_id_req) */
__u32 spare; /* Reserved, must be 0 */
__u64 ns_id; /* Last seen namespace ID (for pagination) */
__u32 ns_type; /* Filter by namespace type(s) */
__u32 spare2; /* Reserved, must be 0 */
__u64 user_ns_id; /* Filter by owning user namespace */
};
Example 1: List all namespaces
void list_all_namespaces(void)
{
struct ns_id_req req = {
.size = sizeof(req),
.ns_id = 0, /* Start from beginning */
.ns_type = 0, /* All types */
.user_ns_id = 0, /* All user namespaces */
};
uint64_t ids[100];
ssize_t ret;
printf("All namespaces in the system:\n");
do {
ret = listns(&req, ids, 100, 0);
if (ret < 0) {
perror("listns");
break;
}
for (ssize_t i = 0; i < ret; i++)
printf(" Namespace ID: %llu\n", (unsigned long long)ids[i]);
/* Continue from last seen ID */
if (ret > 0)
req.ns_id = ids[ret - 1];
} while (ret == 100); /* Buffer was full, more may exist */
}
printf("Enumerating all namespaces with pagination:\n");
while (1) {
ret = listns(&req, ids, 50, 0);
if (ret < 0) {
perror("listns");
break;
}
if (ret == 0)
break; /* No more namespaces */
total += ret;
printf(" Batch: %zd namespaces\n", ret);
/* Last ID in this batch becomes start of next batch */
req.ns_id = ids[ret - 1];
if (ret < 50)
break; /* Partial batch = end of results */
}
printf("Total: %zu namespaces\n", total);
}
Permission Model
listns() respects namespace isolation and capabilities:
(1) Global listing (user_ns_id = 0):
- Requires CAP_SYS_ADMIN in the namespace's owning user namespace
- OR the namespace must be in the caller's namespace context (e.g.,
a namespace the caller is currently using)
- User namespaces additionally allow listing if the caller has
CAP_SYS_ADMIN in that user namespace itself
(2) Owner-filtered listing (user_ns_id != 0):
- Requires CAP_SYS_ADMIN in the specified owner user namespace
- OR the namespace must be in the caller's namespace context
- This allows unprivileged processes to enumerate namespaces they own
(3) Visibility:
- Only "active" namespaces are listed
- A namespace is active if it has a non-zero __ns_ref_active count
- This includes namespaces used by running processes, held by open
file descriptors, or kept active by bind mounts
- Inactive namespaces (kept alive only by internal kernel
references) are not visible via listns()
The namespace tree doesn't express the ownership concept of namespace
appropriately. Maintain a list of directly owned namespaces per user
namespace. This will allow userspace and the kernel to use the listns()
system call to walk the namespace tree by owning user namespace. The
rbtree is used to find the relevant namespace entry point which allows
to continue iteration and the owner list can be used to walk the tree
completely lock free.
nstree: assign fixed ids to the initial namespaces
The initial set of namespace comes with fixed inode numbers making it
easy for userspace to identify them solely based on that information.
This has long preceeded anything here.
Similarly, let's assign fixed namespace ids for the initial namespaces.
Kill the cookie and use a sequentially increasing number. This has the
nice side-effect that the owning user namespace will always have a
namespace id that is smaller than any of it's descendant namespaces.
The namespace file handle struct nsfs_file_handle is uapi and userspace
is expressly allowed to generate file handles without going through
name_to_handle_at().
Allow userspace to generate a file handle where both the inode number
and the namespace type are zero and just pass in the unique namespace
id. The kernel uses the unified namespace tree to find the namespace and
open the file handle.
When the kernel creates a file handle via name_to_handle_at() it will
always fill in the type and the inode number allowing userspace to
retrieve core information.
The namespace tree is, among other things, currently used to support
file handles for namespaces. When a namespace is created it is placed on
the namespace trees and when it is destroyed it is removed from the
namespace trees.
While a namespace is on the namespace trees with a valid reference count
it is possible to reopen it through a namespace file handle. This is all
fine but has some issues that should be addressed.
On current kernels a namespace is visible to userspace in the
following cases:
(1) The namespace is in use by a task.
(2) The namespace is persisted through a VFS object (namespace file
descriptor or bind-mount).
Note that (2) only cares about direct persistence of the namespace
itself not indirectly via e.g., file->f_cred file references or
similar.
(3) The namespace is a hierarchical namespace type and is the parent of
a single or multiple child namespaces.
Case (3) is interesting because it is possible that a parent namespace
might not fulfill any of (1) or (2), i.e., is invisible to userspace but
it may still be resurrected through the NS_GET_PARENT ioctl().
Currently namespace file handles allow much broader access to namespaces
than what is currently possible via (1)-(3). The reason is that
namespaces may remain pinned for completely internal reasons yet are
inaccessible to userspace.
For example, a user namespace my remain pinned by get_cred() calls to
stash the opener's credentials into file->f_cred. As it stands file
handles allow to resurrect such a users namespace even though this
should not be possible via (1)-(3). This is a fundamental uapi change
that we shouldn't do if we don't have to.
Consider the following insane case: Various architectures support the
CONFIG_MMU_LAZY_TLB_REFCOUNT option which uses lazy TLB destruction.
When this option is set a userspace task's struct mm_struct may be used
for kernel threads such as the idle task and will only be destroyed once
the cpu's runqueue switches back to another task. But because of ptrace()
permission checks struct mm_struct stashes the user namespace of the
task that struct mm_struct originally belonged to. The kernel thread
will take a reference on the struct mm_struct and thus pin it.
So on an idle system user namespaces can be persisted for arbitrary
amounts of time which also means that they can be resurrected using
namespace file handles. That makes no sense whatsoever. The problem is
of course excarabted on large systems with a huge number of cpus.
To handle this nicely we introduce an active reference count which
tracks (1)-(3). This is easy to do as all of these things are already
managed centrally. Only (1)-(3) will count towards the active reference
count and only namespaces which are active may be opened via namespace
file handles.
The problem is that namespaces may be resurrected. Which means that they
can become temporarily inactive and will be reactived some time later.
Currently the only example of this is the SIOGCSKNS socket ioctl. The
SIOCGSKNS ioctl allows to open a network namespace file descriptor based
on a socket file descriptor.
If a socket is tied to a network namespace that subsequently becomes
inactive but that socket is persisted by another process in another
network namespace (e.g., via SCM_RIGHTS of pidfd_getfd()) then the
SIOCGSKNS ioctl will resurrect this network namespace.
So calls to open_related_ns() and open_namespace() will end up
resurrecting the corresponding namespace tree.
Note that the active reference count does not regulate the lifetime of
the namespace itself. This is still done by the normal reference count.
The active reference count can only be elevated if the regular reference
count is elevated.
The active reference count also doesn't regulate the presence of a
namespace on the namespace trees. It only regulates its visiblity to
namespace file handles (and in later patches to listns()).
A namespace remains on the namespace trees from creation until its
actual destruction. This will allow the kernel to always reach any
namespace trivially and it will also enable subsystems like bpf to walk
the namespace lists on the system for tracing or general introspection
purposes.
Note that different namespaces have different visibility lifetimes on
current kernels. While most namespace are immediately released when the
last task using them exits, the user- and pid namespace are persisted
and thus both remain accessible via /proc/<pid>/ns/<ns_type>.
The user namespace lifetime is aliged with struct cred and is only
released through exit_creds(). However, it becomes inaccessible to
userspace once the last task using it is reaped, i.e., when
release_task() is called and all proc entries are flushed. Similarly,
the pid namespace is also visible until the last task using it has been
reaped and the associated pid numbers are freed.
The active reference counts of the user- and pid namespace are
decremented once the task is reaped.
The current naming is very misleading as this really isn't exiting all
of the task's namespaces. It is only exiting the namespaces that hang of
off nsproxy. Reflect that in the name.
While pidfs dentries are never hashed and thus retain_dentry() will never
consider them for placing them on the LRU it isn't great to always have
to go and remember that. Raise DCACHE_DONTCACHE explicitly as a visual
marker that dentries aren't kept but freed immediately instead.
projects / thirdparty / linux.git / log
