Completely separate the signal delivery path from the notify handler as
they have different semantics versus the event handling.
The signal delivery only needs to ensure that the interrupted user context
was not in a critical section or the section is aborted before it switches
to the signal frame context. The signal frame context does not have the
original instruction pointer anymore, so that can't be handled on exit to
user space.
No point in updating the CPU/CID ids as they might change again before the
task returns to user space for real.
The fast path optimization, which checks for the 'entry from user via
interrupt' condition is only available for architectures which use the
generic entry code.
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Reviewed-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Link: https://patch.msgid.link/20251027084307.455429038@linutronix.de
Provide a new and straight forward implementation to set the IDs (CPU ID,
Node ID and MM CID), which can be later inlined into the fast path.
It does all operations in one scoped_user_rw_access() section and retrieves
also the critical section member (rseq::cs_rseq) from user space to avoid
another user..begin/end() pair. This is in preparation for optimizing the
fast path to avoid extra work when not required.
On rseq registration set the CPU ID fields to RSEQ_CPU_ID_UNINITIALIZED and
node and MM CID to zero. That's the same as the kernel internal reset
values. That makes the debug validation in the exit code work correctly on
the first exit to user space.
Use it to replace the whole related zoo in rseq.c
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Reviewed-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Link: https://patch.msgid.link/20251027084307.393972266@linutronix.de
Provide a straight forward implementation to check for and eventually
clear/fixup critical sections in user space.
The non-debug version does only the minimal sanity checks and aims for
efficiency.
There are two attack vectors, which are checked for:
1) An abort IP which is in the kernel address space. That would cause at
least x86 to return to kernel space via IRET.
2) A rogue critical section descriptor with an abort IP pointing to some
arbitrary address, which is not preceded by the RSEQ signature.
If the section descriptors are invalid then the resulting misbehaviour of
the user space application is not the kernels problem.
The kernel provides a run-time switchable debug slow path, which implements
the full zoo of checks including termination of the task when one of the
gazillion conditions is not met.
Replace the zoo in rseq.c with it and invoke it from the TIF_NOTIFY_RESUME
handler. Move the remainders into the CONFIG_DEBUG_RSEQ section, which will
be replaced and removed in a subsequent step.
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Reviewed-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Link: https://patch.msgid.link/20251027084307.151465632@linutronix.de
In preparation for a major rewrite of this code, provide a data structure
for rseq management.
Put all the rseq related data into it (except for the debug part), which
allows to simplify fork/execve by using memset() and memcpy() instead of
adding new fields to initialize over and over.
Create a storage struct for event management as well and put the
sched_switch event and a indicator for RSEQ on a task into it as a
start. That uses a union, which allows to mask and clear the whole lot
efficiently.
The indicators are explicitly not a bit field. Bit fields generate abysmal
code.
The boolean members are defined as u8 as that actually guarantees that it
fits. There seem to be strange architecture ABIs which need more than 8
bits for a boolean.
The has_rseq member is redundant vs. task::rseq, but it turns out that
boolean operations and quick checks on the union generate better code than
fiddling with separate entities and data types.
This struct will be extended over time to carry more information.
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Reviewed-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Link: https://patch.msgid.link/20251027084306.527086690@linutronix.de
Hypervisors invoke resume_user_mode_work() before entering the guest, which
clears TIF_NOTIFY_RESUME. The @regs argument is NULL as there is no user
space context available to them, so the rseq notify handler skips
inspecting the critical section, but updates the CPU/MM CID values
unconditionally so that the eventual pending rseq event is not lost on the
way to user space.
This is a pointless exercise as the task might be rescheduled before
actually returning to user space and it creates unnecessary work in the
vcpu_run() loops.
It's way more efficient to ignore that invocation based on @regs == NULL
and let the hypervisors re-raise TIF_NOTIFY_RESUME after returning from the
vcpu_run() loop before returning from the ioctl().
This ensures that a pending RSEQ update is not lost and the IDs are updated
before returning to user space.
Once the RSEQ handling is decoupled from TIF_NOTIFY_RESUME, this turns into
a NOOP.
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Reviewed-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Acked-by: Sean Christopherson <seanjc@google.com>
Link: https://patch.msgid.link/20251027084306.399495855@linutronix.de
Since commit 0190e4198e ("rseq: Deprecate RSEQ_CS_FLAG_NO_RESTART_ON_*
flags") the bits in task::rseq_event_mask are meaningless and just extra
work in terms of setting them individually.
Aside of that the only relevant point where an event has to be raised is
context switch. Neither the CPU nor MM CID can change without going through
a context switch.
Collapse them all into a single boolean which simplifies the code a lot and
remove the pointless invocations which have been sprinkled all over the
place for no value.
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Reviewed-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Link: https://patch.msgid.link/20251027084306.336978188@linutronix.de
The RSEQ critical section mechanism only clears the event mask when a
critical section is registered, otherwise it is stale and collects
bits.
That means once a critical section is installed the first invocation of
that code when TIF_NOTIFY_RESUME is set will abort the critical section,
even when the TIF bit was not raised by the rseq preempt/migrate/signal
helpers.
This also has a performance implication because TIF_NOTIFY_RESUME is a
multiplexing TIF bit, which is utilized by quite some infrastructure. That
means every invocation of __rseq_notify_resume() goes unconditionally
through the heavy lifting of user space access and consistency checks even
if there is no reason to do so.
Keeping the stale event mask around when exiting to user space also
prevents it from being utilized by the upcoming time slice extension
mechanism.
Avoid this by reading and clearing the event mask before doing the user
space critical section access with interrupts or preemption disabled, which
ensures that the read and clear operation is CPU local atomic versus
scheduling and the membarrier IPI.
This is correct as after re-enabling interrupts/preemption any relevant
event will set the bit again and raise TIF_NOTIFY_RESUME, which makes the
user space exit code take another round of TIF bit clearing.
If the event mask was non-zero, invoke the slow path. On debug kernels the
slow path is invoked unconditionally and the result of the event mask
evaluation is handed in.
Add a exit path check after the TIF bit loop, which validates on debug
kernels that the event mask is zero before exiting to user space.
While at it reword the convoluted comment why the pt_regs pointer can be
NULL under certain circumstances.
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Reviewed-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Link: https://patch.msgid.link/20251027084306.022571576@linutronix.de
When conditional jumps are performed on the same scalar register
(e.g., r0 <= r0, r0 > r0, r0 < r0), the BPF verifier incorrectly
attempts to adjust the register's min/max bounds. This leads to
invalid range bounds and triggers a BUG warning.
The problematic BPF program:
0: call bpf_get_prandom_u32
1: w8 = 0x80000000
2: r0 &= r8
3: if r0 > r0 goto <exit>
The instruction 3 triggers kernel warning:
3: if r0 > r0 goto <exit>
true_reg1: range bounds violation u64=[0x1, 0x0] s64=[0x1, 0x0] u32=[0x1, 0x0] s32=[0x1, 0x0] var_off=(0x0, 0x0)
true_reg2: const tnum out of sync with range bounds u64=[0x0, 0xffffffffffffffff] s64=[0x8000000000000000, 0x7fffffffffffffff] var_off=(0x0, 0x0)
Comparing a register with itself should not change its bounds and
for most comparison operations, comparing a register with itself has
a known result (e.g., r0 == r0 is always true, r0 < r0 is always false).
Fix this by:
1. Enhance is_scalar_branch_taken() to properly handle branch direction
computation for same register comparisons across all BPF jump operations
2. Adds early return in reg_set_min_max() to avoid bounds adjustment
for unknown branch directions (e.g., BPF_JSET) on the same register
The fix ensures that unnecessary bounds adjustments are skipped, preventing
the verifier bug while maintaining correct branch direction analysis.
Reported-by: Kaiyan Mei <M202472210@hust.edu.cn>
Reported-by: Yinhao Hu <dddddd@hust.edu.cn>
Closes: https://lore.kernel.org/all/1881f0f5.300df.199f2576a01.Coremail.kaiyanm@hust.edu.cn/
Signed-off-by: KaFai Wan <kafai.wan@linux.dev>
Acked-by: Eduard Zingerman <eddyz87@gmail.com>
Link: https://lore.kernel.org/r/20251103063108.1111764-2-kafai.wan@linux.dev
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
ftrace_hash_ipmodify_enable() checks IPMODIFY and DIRECT ftrace_ops on
the same kernel function. When needed, ftrace_hash_ipmodify_enable()
calls ops->ops_func() to prepare the direct ftrace (BPF trampoline) to
share the same function as the IPMODIFY ftrace (livepatch).
ftrace_hash_ipmodify_enable() is called in register_ftrace_direct() path,
but not called in modify_ftrace_direct() path. As a result, the following
operations will break livepatch:
1. Load livepatch to a kernel function;
2. Attach fentry program to the kernel function;
3. Attach fexit program to the kernel function.
After 3, the kernel function being used will not be the livepatched
version, but the original version.
Fix this by adding __ftrace_hash_update_ipmodify() to
__modify_ftrace_direct() and adjust some logic around the call.
Signed-off-by: Song Liu <song@kernel.org>
Reviewed-by: Jiri Olsa <jolsa@kernel.org>
Link: https://lore.kernel.org/r/20251027175023.1521602-3-song@kernel.org
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Acked-by: Steven Rostedt (Google) <rostedt@goodmis.org>
When livepatch is attached to the same function as bpf trampoline with
a fexit program, bpf trampoline code calls register_ftrace_direct()
twice. The first time will fail with -EAGAIN, and the second time it
will succeed. This requires register_ftrace_direct() to unregister
the address on the first attempt. Otherwise, the bpf trampoline cannot
attach. Here is an easy way to reproduce this issue:
insmod samples/livepatch/livepatch-sample.ko
bpftrace -e 'fexit:cmdline_proc_show {}'
ERROR: Unable to attach probe: fexit:vmlinux:cmdline_proc_show...
Fix this by cleaning up the hash when register_ftrace_function_nolock hits
errors.
Also, move the code that resets ops->func and ops->trampoline to the error
path of register_ftrace_direct(); and add a helper function reset_direct()
in register_ftrace_direct() and unregister_ftrace_direct().
Fixes: d05cb47066 ("ftrace: Fix modification of direct_function hash while in use")
Cc: stable@vger.kernel.org # v6.6+
Reported-by: Andrey Grodzovsky <andrey.grodzovsky@crowdstrike.com>
Closes: https://lore.kernel.org/live-patching/c5058315a39d4615b333e485893345be@crowdstrike.com/
Cc: Steven Rostedt (Google) <rostedt@goodmis.org>
Cc: Masami Hiramatsu (Google) <mhiramat@kernel.org>
Acked-and-tested-by: Andrey Grodzovsky <andrey.grodzovsky@crowdstrike.com>
Signed-off-by: Song Liu <song@kernel.org>
Reviewed-by: Jiri Olsa <jolsa@kernel.org>
Link: https://lore.kernel.org/r/20251027175023.1521602-2-song@kernel.org
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Acked-by: Steven Rostedt (Google) <rostedt@goodmis.org>
sched_ext_free() was called from __put_task_struct() when the last reference
to the task is dropped, which could be long after the task has finished
running. This causes cgroup-related problems:
- ops.init_task() can be called on a cgroup which didn't get ops.cgroup_init()'d
during scheduler load, because the cgroup might be destroyed/unlinked
while the zombie or dead task is still lingering on the scx_tasks list.
- ops.cgroup_exit() could be called before ops.exit_task() is called on all
member tasks, leading to incorrect exit ordering.
Fix by moving it to finish_task_switch() to be called right after the final
context switch away from the dying task, matching when sched_class->task_dead()
is called. Rename it to sched_ext_dead() to match the new calling context.
By calling sched_ext_dead() before cgroup_task_dead(), we ensure that:
- Tasks visible on scx_tasks list have valid cgroups during scheduler load,
as cgroup_mutex prevents cgroup destruction while the task is still linked.
- All member tasks have ops.exit_task() called and are removed from scx_tasks
before the cgroup can be destroyed and trigger ops.cgroup_exit().
This fix is made possible by the cgroup_task_dead() split in the previous patch.
This also makes more sense resource-wise as there's no point in keeping
scheduler side resources around for dead tasks.
Reported-by: Dan Schatzberg <dschatzberg@meta.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Reviewed-by: Andrea Righi <arighi@nvidia.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
Pull cgroup/for-6.19 to receive:
16dad7801a ("cgroup: Rename cgroup lifecycle hooks to cgroup_task_*()")
260fbcb92b ("cgroup: Move dying_tasks cleanup from cgroup_task_release() to cgroup_task_free()")
d245698d72 ("cgroup: Defer task cgroup unlink until after the task is done switching out")
These are needed for the sched_ext cgroup exit ordering fix.
Signed-off-by: Tejun Heo <tj@kernel.org>
When a task exits, css_set_move_task(tsk, cset, NULL, false) unlinks the task
from its cgroup. From the cgroup's perspective, the task is now gone. If this
makes the cgroup empty, it can be removed, triggering ->css_offline() callbacks
that notify controllers the cgroup is going offline resource-wise.
However, the exiting task can still run, perform memory operations, and schedule
until the final context switch in finish_task_switch(). This creates a confusing
situation where controllers are told a cgroup is offline while resource
activities are still happening in it. While this hasn't broken existing
controllers, it has caused direct confusion for sched_ext schedulers.
Split cgroup_task_exit() into two functions. cgroup_task_exit() now only calls
the subsystem exit callbacks and continues to be called from do_exit(). The
css_set cleanup is moved to the new cgroup_task_dead() which is called from
finish_task_switch() after the final context switch, so that the cgroup only
appears empty after the task is truly done running.
This also reorders operations so that subsys->exit() is now called before
unlinking from the cgroup, which shouldn't break anything.
Cc: Dan Schatzberg <dschatzberg@meta.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Signed-off-by: Tejun Heo <tj@kernel.org>
Currently, cgroup_task_exit() adds thread group leaders with live member
threads to their css_set's dying_tasks list (so cgroup.procs iteration can
still see the leader), and cgroup_task_release() later removes them with
list_del_init(&task->cg_list).
An upcoming patch will defer the dying_tasks list addition, moving it from
cgroup_task_exit() (called from do_exit()) to a new function called from
finish_task_switch(). However, release_task() (which calls
cgroup_task_release()) can run either before or after finish_task_switch(),
creating a race where cgroup_task_release() might try to remove the task from
dying_tasks before or while it's being added.
Move the list_del_init() from cgroup_task_release() to cgroup_task_free() to
fix this race. cgroup_task_free() runs from __put_task_struct(), which is
always after both paths, making the cleanup safe.
Cc: Dan Schatzberg <dschatzberg@meta.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Signed-off-by: Tejun Heo <tj@kernel.org>
The current names cgroup_exit(), cgroup_release(), and cgroup_free() are
confusing because they look like they're operating on cgroups themselves when
they're actually task lifecycle hooks. For example, cgroup_init() initializes
the cgroup subsystem while cgroup_exit() is a task exit notification to
cgroup. Rename them to cgroup_task_exit(), cgroup_task_release(), and
cgroup_task_free() to make it clear that these operate on tasks.
Cc: Dan Schatzberg <dschatzberg@meta.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Reviewed-by: Chen Ridong <chenridong@huawei.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
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 */
}
Example 2: List network namespaces only
void list_network_namespaces(void)
{
struct ns_id_req req = {
.size = sizeof(req),
.ns_id = 0,
.ns_type = NET_NS, /* Only network namespaces */
.user_ns_id = 0,
};
uint64_t ids[100];
ssize_t ret;
ret = listns(&req, ids, 100, 0);
if (ret < 0) {
perror("listns");
return;
}
printf("Network namespaces: %zd found\n", ret);
for (ssize_t i = 0; i < ret; i++)
printf(" netns ID: %llu\n", (unsigned long long)ids[i]);
}
Example 3: List namespaces owned by current user namespace
void list_owned_namespaces(void)
{
struct ns_id_req req = {
.size = sizeof(req),
.ns_id = 0,
.ns_type = 0, /* All types */
.user_ns_id = LISTNS_CURRENT_USER, /* Current userns */
};
uint64_t ids[100];
ssize_t ret;
ret = listns(&req, ids, 100, 0);
if (ret < 0) {
perror("listns");
return;
}
printf("Namespaces owned by my user namespace: %zd\n", ret);
for (ssize_t i = 0; i < ret; i++)
printf(" ns ID: %llu\n", (unsigned long long)ids[i]);
}
Example 4: List multiple namespace types
void list_network_and_mount_namespaces(void)
{
struct ns_id_req req = {
.size = sizeof(req),
.ns_id = 0,
.ns_type = NET_NS | MNT_NS, /* Network and mount */
.user_ns_id = 0,
};
uint64_t ids[100];
ssize_t ret;
ret = listns(&req, ids, 100, 0);
printf("Network and mount namespaces: %zd found\n", ret);
}
Example 5: Pagination through large namespace sets
void list_all_with_pagination(void)
{
struct ns_id_req req = {
.size = sizeof(req),
.ns_id = 0,
.ns_type = 0,
.user_ns_id = 0,
};
uint64_t ids[50];
size_t total = 0;
ssize_t ret;
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()
Link: https://patch.msgid.link/20251029-work-namespace-nstree-listns-v4-19-2e6f823ebdc0@kernel.org
Signed-off-by: Christian Brauner <brauner@kernel.org>
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.
Link: https://patch.msgid.link/20251029-work-namespace-nstree-listns-v4-16-2e6f823ebdc0@kernel.org
Signed-off-by: Christian Brauner <brauner@kernel.org>
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.
Link: https://patch.msgid.link/20251029-work-namespace-nstree-listns-v4-15-2e6f823ebdc0@kernel.org
Signed-off-by: Christian Brauner <brauner@kernel.org>
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.
Link: https://patch.msgid.link/20251029-work-namespace-nstree-listns-v4-11-2e6f823ebdc0@kernel.org
Signed-off-by: Christian Brauner <brauner@kernel.org>
cpu-clock usage by the async-profiler tool can trigger a system hang,
which got bisected back to the following commit by Octavia Togami:
18dbcbfabf ("perf: Fix the POLL_HUP delivery breakage") causes this issue
The root cause of the hang is that cpu-clock is a special type of SW
event which relies on hrtimers. The __perf_event_overflow() callback
is invoked from the hrtimer handler for cpu-clock events, and
__perf_event_overflow() tries to call cpu_clock_event_stop()
to stop the event, which calls htimer_cancel() to cancel the hrtimer.
But that's a recursion into the hrtimer code from a hrtimer handler,
which (unsurprisingly) deadlocks.
To fix this bug, use hrtimer_try_to_cancel() instead, and set
the PERF_HES_STOPPED flag, which causes perf_swevent_hrtimer()
to stop the event once it sees the PERF_HES_STOPPED flag.
[ mingo: Fixed the comments and improved the changelog. ]
Closes: https://lore.kernel.org/all/CAHPNGSQpXEopYreir+uDDEbtXTBvBvi8c6fYXJvceqtgTPao3Q@mail.gmail.com/
Fixes: 18dbcbfabf ("perf: Fix the POLL_HUP delivery breakage")
Reported-by: Octavia Togami <octavia.togami@gmail.com>
Suggested-by: Peter Zijlstra <peterz@infradead.org>
Signed-off-by: Dapeng Mi <dapeng1.mi@linux.intel.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Tested-by: Octavia Togami <octavia.togami@gmail.com>
Cc: stable@vger.kernel.org
Link: https://github.com/lucko/spark/issues/530
Link: https://patch.msgid.link/20251015051828.12809-1-dapeng1.mi@linux.intel.com
Crystal reports that the PCIe Advanced Error Reporting driver gets stuck
in an infinite loop on PREEMPT_RT:
Both the primary interrupt handler aer_irq() as well as the secondary
handler aer_isr() are forced into threads with identical priority.
Crystal writes that on the ARM system in question, the primary handler
has to clear an error in the Root Error Status register...
"before the next error happens, or else the hardware will set the
Multiple ERR_COR Received bit. If that bit is set, then aer_isr()
can't rely on the Error Source Identification register, so it scans
through all devices looking for errors -- and for some reason, on
this system, accessing the AER registers (or any Config Space above
0x400, even though there are capabilities located there) generates
an Unsupported Request Error (but returns valid data). Since this
happens more than once, without aer_irq() preempting, it causes
another multi error and we get stuck in a loop."
The issue does not show on non-PREEMPT_RT because the primary handler
runs in hardirq context and thus can preempt the threaded secondary
handler, clear the Root Error Status register and prevent the secondary
handler from getting stuck.
Emulate the same behavior on PREEMPT_RT by assigning a lower default
priority to the secondary handler if the primary handler is forced into
a thread.
Reported-by: Crystal Wood <crwood@redhat.com>
Signed-off-by: Lukas Wunner <lukas@wunner.de>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Tested-by: Crystal Wood <crwood@redhat.com>
Reviewed-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Link: https://patch.msgid.link/f6dcdb41be2694886b8dbf4fe7b3ab89e9d5114c.1761569303.git.lukas@wunner.de
Closes: https://lore.kernel.org/r/20250902224441.368483-1-crwood@redhat.com/
Idle migrators don't walk the whole tree in order to find out if there
are timers to migrate because they recorded the next deadline to be
verified within a single check in tmigr_requires_handle_remote().
Remove the related dead code and data.
Signed-off-by: Frederic Weisbecker <frederic@kernel.org>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Link: https://patch.msgid.link/20251024132536.39841-7-frederic@kernel.org
The CPU doing the prepare work for a remote target must be online from
the tree point of view and its hierarchy must be active, otherwise
propagating its active state up to the new root branch would be either
incorrect or racy.
Assert those conditions with more sanity checks.
Signed-off-by: Frederic Weisbecker <frederic@kernel.org>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Link: https://patch.msgid.link/20251024132536.39841-5-frederic@kernel.org
When a CPU from a new node boots, the old root may happen to be
connected to the new root even if their node mismatch, as depicted in
the following scenario:
1) CPU 0 boots and creates the first group for node 0.
[GRP0:0]
node 0
|
CPU 0
2) CPU 1 from node 1 boots and creates a new top that corresponds to
node 1, but it also connects the old root from node 0 to the new root
from node 1 by mistake.
[GRP1:0]
node 1
/ \
/ \
[GRP0:0] [GRP0:1]
node 0 node 1
| |
CPU 0 CPU 1
3) This eventually leads to an imbalanced tree where some node 0 CPUs
migrate node 1 timers (and vice versa) way before reaching the
crossnode groups, resulting in more frequent remote memory accesses
than expected.
[GRP2:0]
NUMA_NO_NODE
/ \
[GRP1:0] [GRP1:1]
node 1 node 0
/ \ |
/ \ [...]
[GRP0:0] [GRP0:1]
node 0 node 1
| |
CPU 0... CPU 1...
A balanced tree should only contain groups having children that belong
to the same node:
[GRP2:0]
NUMA_NO_NODE
/ \
[GRP1:0] [GRP1:0]
node 0 node 1
/ \ / \
/ \ / \
[GRP0:0] [...] [...] [GRP0:1]
node 0 node 1
| |
CPU 0... CPU 1...
In order to fix this, the hierarchy must be unfolded up to the crossnode
level as soon as a node mismatch is detected. For example the stage 2
above should lead to this layout:
[GRP2:0]
NUMA_NO_NODE
/ \
[GRP1:0] [GRP1:1]
node 0 node 1
/ \
/ \
[GRP0:0] [GRP0:1]
node 0 node 1
| |
CPU 0 CPU 1
This means that not only GRP1:0 must be created but also GRP1:1 and
GRP2:0 in order to prepare a balanced tree for next CPUs to boot.
Fixes: 7ee9887703 ("timers: Implement the hierarchical pull model")
Signed-off-by: Frederic Weisbecker <frederic@kernel.org>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Link: https://patch.msgid.link/20251024132536.39841-4-frederic@kernel.org
Initializing the tmc's group, the group's number of children and the
group's parent can all be done without locking because:
1) Reading the group's parent and its group mask is done locklessly.
2) The connections prepared for a given CPU hierarchy are visible to the
target CPU once online, thanks to the CPU hotplug enforced memory
ordering.
3) In case of a newly created upper level, the new root and its
connections and initialization are made visible by the CPU which made
the connections. When that CPUs goes idle in the future, the new link
is published by tmigr_inactive_up() through the atomic RmW on
->migr_state.
4) If CPUs were still walking up the active hierarchy, they could observe
the new root earlier. In this case the ordering is enforced by an
early initialization of the group mask and by barriers that maintain
address dependency as explained in:
b729cc1ec2 ("timers/migration: Fix another race between hotplug and idle entry/exit")
de3ced72a7 ("timers/migration: Enforce group initialization visibility to tree walkers")
5) Timers are propagated by a chain of group locking from the bottom to
the top. And while doing so, the tree also propagates groups links
and initialization. Therefore remote expiration, which also relies
on group locking, will observe those links and initialization while
holding the root lock before walking the tree remotely and update
remote timers. This is especially important for migrators in the
active hierarchy that may observe the new root early.
Therefore the locking is unnecessary at initialization. If anything, it
just brings confusion. Remove it.
Signed-off-by: Frederic Weisbecker <frederic@kernel.org>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Link: https://patch.msgid.link/20251024132536.39841-3-frederic@kernel.org
On large NUMA systems, while running a test program that saturates the
inter-processor and inter-NUMA links, acquiring the jiffies_lock can be
very expensive.
If the cpu designated to do jiffies updates (tick_do_timer_cpu) gets
delayed and other cpus decide to do the jiffies update themselves, a large
number of them decide to do so at the same time.
The inexpensive check against tick_next_period is far quicker than actually
acquiring the lock, so most of these get in line to obtain the lock. If
obtaining the lock is slow enough, this spirals into the vast majority of
CPUs continuously being stuck waiting for this lock, just to obtain it and
find out that time has already been updated by another cpu. For example, on
one random entry to kdb by manually-injected NMI, 2912 of 3840 CPUs were
observed to be stuck there.
To avoid this, allow only one non-timekeeper CPU to call
tick_do_update_jiffies64() at any given time, resetting ts->stalled jiffies
only if the jiffies update function is actually called.
With this change, manually interrupting the test at most two CPUs are
observed to invoke tick_do_update_jiffies64() - the timekeeper and one
other.
Signed-off-by: Steve Wahl <steve.wahl@hpe.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Acked-by: Shrikanth Hegde <sshegde@linux.ibm.com>
Link: https://patch.msgid.link/20251027183456.343407-1-steve.wahl@hpe.com
Pull bpf fixes from Alexei Starovoitov:
- Mark migrate_disable/enable() as always_inline to avoid issues with
partial inlining (Yonghong Song)
- Fix powerpc stack register definition in libbpf bpf_tracing.h (Andrii
Nakryiko)
- Reject negative head_room in __bpf_skb_change_head (Daniel Borkmann)
- Conditionally include dynptr copy kfuncs (Malin Jonsson)
- Sync pending IRQ work before freeing BPF ring buffer (Noorain Eqbal)
- Do not audit capability check in x86 do_jit() (Ondrej Mosnacek)
- Fix arm64 JIT of BPF_ST insn when it writes into arena memory
(Puranjay Mohan)
* tag 'bpf-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/bpf/bpf:
bpf/arm64: Fix BPF_ST into arena memory
bpf: Make migrate_disable always inline to avoid partial inlining
bpf: Reject negative head_room in __bpf_skb_change_head
bpf: Conditionally include dynptr copy kfuncs
libbpf: Fix powerpc's stack register definition in bpf_tracing.h
bpf: Do not audit capability check in do_jit()
bpf: Sync pending IRQ work before freeing ring buffer
Reading /proc/irq/N/smp_affinity* races with irq_set_affinity() and
irq_move_masked_irq(), leading to old or torn output for users.
After a user writes a new CPU mask to /proc/irq/N/affinity*, the syscall
returns success, yet a subsequent read of the same file immediately returns
a value different from what was just written.
That's due to a race between show_irq_affinity() and irq_move_masked_irq()
which lets the read observe a transient, inconsistent affinity mask.
Cure it by guarding the read with irq_desc::lock.
[ tglx: Massaged change log ]
Signed-off-by: Muchun Song <songmuchun@bytedance.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Link: https://patch.msgid.link/20251028090408.76331-1-songmuchun@bytedance.com
