| Age | Commit message (Collapse) | Author |
|---|
|
Some new assertions pointed out that the existing code has nested rt_mutex wait
state in the futex code.
Specifically, the futex_lock_pi() cancel case uses spin_lock() while there
still is a rt_waiter enqueued for this task, resulting in a state where there
are two waiters for the same task (and task_struct::pi_blocked_on gets
scrambled).
The reason to take hb->lock at this point is to avoid the wake_futex_pi()
EAGAIN case.
This happens when futex_top_waiter() and rt_mutex_top_waiter() state becomes
inconsistent. The current rules are such that this inconsistency will not be
observed.
Notably the case that needs to be avoided is where futex_lock_pi() and
futex_unlock_pi() interleave such that unlock will fail to observe a new
waiter.
*However* the case at hand is where a waiter is leaving, in this case the race
means a waiter that is going away is not observed -- which is harmless,
provided this race is explicitly handled.
This is a somewhat dangerous proposition because the converse race is not
observing a new waiter, which must absolutely not happen. But since the race is
valid this cannot be asserted.
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Reviewed-by: Thomas Gleixner <tglx@linutronix.de>
Reviewed-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Tested-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Link: https://lkml.kernel.org/r/20230915151943.GD6743@noisy.programming.kicks-ass.net
|
|
There used to be a BUG_ON(current->pi_blocked_on) in the lock acquisition
functions, but that vanished in one of the rtmutex overhauls.
Bring it back in form of a lockdep assert to catch code paths which take
rtmutex based locks with current::pi_blocked_on != NULL.
Reported-by: Crystal Wood <swood@redhat.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: "Peter Zijlstra (Intel)" <peterz@infradead.org>
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Link: https://lkml.kernel.org/r/20230908162254.999499-7-bigeasy@linutronix.de
|
|
Have rt_mutex use the rt_mutex specific scheduler helpers to avoid
recursion vs rtlock on the PI state.
[[ peterz: adapted to new names ]]
Reported-by: Crystal Wood <swood@redhat.com>
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Link: https://lkml.kernel.org/r/20230908162254.999499-6-bigeasy@linutronix.de
|
|
With PREEMPT_RT there is a rt_mutex recursion problem where
sched_submit_work() can use an rtlock (aka spinlock_t). More
specifically what happens is:
mutex_lock() /* really rt_mutex */
...
__rt_mutex_slowlock_locked()
task_blocks_on_rt_mutex()
// enqueue current task as waiter
// do PI chain walk
rt_mutex_slowlock_block()
schedule()
sched_submit_work()
...
spin_lock() /* really rtlock */
...
__rt_mutex_slowlock_locked()
task_blocks_on_rt_mutex()
// enqueue current task as waiter *AGAIN*
// *CONFUSION*
Fix this by making rt_mutex do the sched_submit_work() early, before
it enqueues itself as a waiter -- before it even knows *if* it will
wait.
[[ basically Thomas' patch but with different naming and a few asserts
added ]]
Originally-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Link: https://lkml.kernel.org/r/20230908162254.999499-5-bigeasy@linutronix.de
|
|
There are currently two implementations of this basic __schedule()
loop, and there is soon to be a third.
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Link: https://lkml.kernel.org/r/20230908162254.999499-4-bigeasy@linutronix.de
|
|
With DEBUG_RT_MUTEXES enabled the fast-path rt_mutex_cmpxchg_acquire()
always fails and all lock operations take the slow path.
Provide a new helper inline rt_mutex_try_acquire() which maps to
rt_mutex_cmpxchg_acquire() in the non-debug case. For the debug case
it invokes rt_mutex_slowtrylock() which can acquire a non-contended
rtmutex under full debug coverage.
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Link: https://lkml.kernel.org/r/20230908162254.999499-3-bigeasy@linutronix.de
|
|
Even though sched_submit_work() is ran from preemptible context,
it is discouraged to have it use blocking locks due to the recursion
potential.
Enforce this.
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Link: https://lkml.kernel.org/r/20230908162254.999499-2-bigeasy@linutronix.de
|
|
Change the comment to match the function name that the SYSCALL_DEFINE()
macros generate to prevent a kernel-doc warning.
kernel/pid.c:628: warning: expecting prototype for pidfd_open(). Prototype was for sys_pidfd_open() instead
Link: https://lkml.kernel.org/r/20230912060822.2500-1-rdunlap@infradead.org
Signed-off-by: Randy Dunlap <rdunlap@infradead.org>
Cc: Christian Brauner <brauner@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
|
|
A previous commit changed the arguments to task_work_cancel_match(), but
didn't document all of them.
Link: https://lkml.kernel.org/r/93938bff-baa3-4091-85f5-784aae297a07@kernel.dk
Fixes: c7aab1a7c52b ("task_work: add helper for more targeted task_work canceling")
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Reported-by: kernel test robot <lkp@intel.com>
Closes: https://lore.kernel.org/oe-kbuild-all/202309120307.zis3yQGe-lkp@intel.com/
Acked-by: Oleg Nesterov <oleg@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
|
|
On PREEMPT_RT keeping preemption disabled during the invocation of
cgroup_enter_frozen() is a problem because the function acquires
css_set_lock which is a sleeping lock on PREEMPT_RT and must not be
acquired with disabled preemption.
The preempt-disabled section is only for performance optimisation reasons
and can be avoided.
Extend the comment and don't disable preemption before scheduling on
PREEMPT_RT.
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Acked-by: Oleg Nesterov <oleg@redhat.com>
Link: https://lore.kernel.org/r/20230803100932.325870-3-bigeasy@linutronix.de
|
|
Commit 53da1d9456fe7 ("fix ptrace slowness") added a preempt-disable section
between read_unlock() and the following schedule() invocation without
explaining why it is needed.
Replace the existing contentless comment with a proper explanation to
clarify that it is not needed for correctness but for performance reasons.
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Acked-by: Oleg Nesterov <oleg@redhat.com>
Link: https://lore.kernel.org/r/20230803100932.325870-2-bigeasy@linutronix.de
|
|
Remove unused prev_offset, min_size, krec_size variables.
Reported-by: kernel test robot <lkp@intel.com>
Closes: https://lore.kernel.org/oe-kbuild-all/202309190634.fL17FWoT-lkp@intel.com/
Fixes: aaa619ebccb2 ("bpf: Refactor check_btf_func and split into two phases")
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
|
|
On 32-bit architectures, the pointer width is 32-bit, while we try to
cast from a u64 down to it, the compiler complains on mismatch in
integer size. Fix this by first casting to long which should match
the pointer width on targets supported by Linux.
Fixes: ec5290a178b7 ("bpf: Prevent KASAN false positive with bpf_throw")
Reported-by: Matthieu Baerts <matthieu.baerts@tessares.net>
Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com>
Tested-by: Matthieu Baerts <matthieu.baerts@tessares.net>
Link: https://lore.kernel.org/r/20230918155233.297024-3-memxor@gmail.com
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
|
|
Initial booting is setting the task flag to idle (PF_IDLE) by the call
path sched_init() -> init_idle(). Having the task idle and calling
call_rcu() in kernel/rcu/tiny.c means that TIF_NEED_RESCHED will be
set. Subsequent calls to any cond_resched() will enable IRQs,
potentially earlier than the IRQ setup has completed. Recent changes
have caused just this scenario and IRQs have been enabled early.
This causes a warning later in start_kernel() as interrupts are enabled
before they are fully set up.
Fix this issue by setting the PF_IDLE flag later in the boot sequence.
Although the boot task was marked as idle since (at least) d80e4fda576d,
I am not sure that it is wrong to do so. The forced context-switch on
idle task was introduced in the tiny_rcu update, so I'm going to claim
this fixes 5f6130fa52ee.
Fixes: 5f6130fa52ee ("tiny_rcu: Directly force QS when call_rcu_[bh|sched]() on idle_task")
Signed-off-by: Liam R. Howlett <Liam.Howlett@oracle.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: stable@vger.kernel.org
Link: https://lore.kernel.org/linux-mm/CAMuHMdWpvpWoDa=Ox-do92czYRvkok6_x6pYUH+ZouMcJbXy+Q@mail.gmail.com/
|
|
The name is a bit opaque - make it clear that this is about wakeup
preemption.
Also rename the ->check_preempt_curr() methods similarly.
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
|
|
Other scheduling classes already postfix their similar methods
with the class name.
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
|
|
A user can pre-configure certain CPUs in an isolated state at boot time
with the "isolcpus" kernel boot command line option. Those CPUs will
not be in the housekeeping_cpumask(HK_TYPE_DOMAIN) and so will not
be in any sched domains. This may conflict with the partition setup
at runtime. Those boot time isolated CPUs should only be used in an
isolated partition.
This patch adds the necessary check and disallows partition setup if the
check fails.
Signed-off-by: Waiman Long <longman@redhat.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
|
|
One can use "cpuset.cpus.partition" to create multiple scheduling domains
or to produce a set of isolated CPUs where load balancing is disabled.
The former use case is less common but the latter one can be frequently
used especially for the Telco use cases like DPDK.
The existing "isolated" partition can be used to produce isolated
CPUs if the applications have full control of a system. However, in a
containerized environment where all the apps are run in a container,
it is hard to distribute out isolated CPUs from the root down given
the unified hierarchy nature of cgroup v2.
The container running on isolated CPUs can be several layers down from
the root. The current partition feature requires that all the ancestors
of a leaf partition root must be parititon roots themselves. This can
be hard to configure.
This patch introduces a new type of partition called remote partition.
A remote partition is a partition whose parent is not a partition root
itself and its CPUs are acquired directly from available CPUs in the
top cpuset through a hierachical distribution of exclusive CPUs down
from it.
By contrast, the existing type of partitions where their parents have
to be valid partition roots are referred to as local partitions as they
have to be clustered around a parent partition root.
Child local partitons can be created under a remote partition, but
a remote partition cannot be created under a local partition. We may
relax this limitation in the future if there are use cases for such
configuration.
Manually writing to the "cpuset.cpus.exclusive" file is not necessary
when creating local partitions. However, writing proper values to
"cpuset.cpus.exclusive" down the cgroup hierarchy before the target
remote partition root is mandatory for the creation of a remote
partition.
The value in "cpuset.cpus.exclusive.effective" may change if its
"cpuset.cpus" or its parent's "cpuset.cpus.exclusive.effective" changes.
Signed-off-by: Waiman Long <longman@redhat.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
|
|
This patch introduces a new writable "cpuset.cpus.exclusive" control
file for v2 which will be added to non-root cpuset enabled cgroups. This new
file enables user to set a smaller list of exclusive CPUs to be used in
the creation of a cpuset partition.
The value written to "cpuset.cpus.exclusive" may not be the effective
value being used for the creation of cpuset partition, the effective
value will show up in "cpuset.cpus.exclusive.effective" and it is
subject to the constraint that it must also be a subset of cpus_allowed
and parent's "cpuset.cpus.exclusive.effective".
By writing to "cpuset.cpus.exclusive", "cpuset.cpus.exclusive.effective"
may be set to a non-empty value even for cgroups that are not valid
partition roots yet.
Signed-off-by: Waiman Long <longman@redhat.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
|
|
The creation of a cpuset partition means dedicating a set of exclusive
CPUs to be used by a particular partition only. These exclusive CPUs
will not be used by any cpusets outside of that partition.
To enable more flexibility in creating partitions, we need a way to
distribute exclusive CPUs that can be used in new partitions. Currently,
we have a subparts_cpus cpumask in struct cpuset that tracks only
the exclusive CPUs used by all the sub-partitions underneath a given
cpuset.
This patch reworks the way we do exclusive CPUs tracking. The
subparts_cpus is now renamed to effective_xcpus which tracks the
exclusive CPUs allocated to a partition root including those that are
further distributed down to sub-partitions underneath it. IOW, it also
includes the exclusive CPUs used by the current partition root. Note
that effective_xcpus can contain offline CPUs and it will always be a
subset of cpus_allowed.
The renamed effective_xcpus is now exposed via a new read-only
"cpuset.cpus.exclusive.effective" control file. The new effective_xcpus
cpumask should be set to cpus_allowed when a cpuset becomes a partition
root and be cleared if it is not a valid partition root.
In the next patch, we will enable write to another new control file to
enable further control of what can get into effective_xcpus.
Signed-off-by: Waiman Long <longman@redhat.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
|
|
Commit a86ce68078b2 ("cgroup/cpuset: Extract out CS_CPU_EXCLUSIVE
& CS_SCHED_LOAD_BALANCE handling") adds a new helper function
update_partition_sd_lb() to update the load balance state of the
cpuset. However the new load balance is determined by just looking at
whether the cpuset is a valid isolated partition root or not. That is
not enough if the cpuset is not a valid partition root but its parent
is in the isolated state (load balance off). Update the function to
set the new state to be the same as its parent in this case like what
has been done in commit c8c926200c55 ("cgroup/cpuset: Inherit parent's
load balance state in v2").
Fixes: a86ce68078b2 ("cgroup/cpuset: Extract out CS_CPU_EXCLUSIVE & CS_SCHED_LOAD_BALANCE handling")
Signed-off-by: Waiman Long <longman@redhat.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
|
|
Signed-off-by: Ingo Molnar <mingo@kernel.org>
|
|
Use css directly instead of dereferencing it from &cgroup->self, while
adding the cgroup v2 cft base and psi files in css_populate_dir(). Both
points to the same css, when css->ss is NULL, this avoids extra deferences
and makes code consistent in usage across the function.
Signed-off-by: Kamalesh Babulal <kamalesh.babulal@oracle.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
|
|
There is no check for possible failure while populating
cgroup1_base_files cft in css_populate_dir(), like its cgroup v2 counter
parts cgroup_{base,psi}_files. In case of failure, the cgroup might not
be set up right. Add ret value check to return on failure.
Signed-off-by: Kamalesh Babulal <kamalesh.babulal@oracle.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
|
|
wq_cpu_intensive_thresh_init()
Currently, if the wq_cpu_intensive_thresh_us is set to specific
value, will cause the wq_cpu_intensive_thresh_init() early exit
and missed creation of pwq_release_worker. this commit therefore
create the pwq_release_worker in advance before checking the
wq_cpu_intensive_thresh_us.
Signed-off-by: Zqiang <qiang.zhang1211@gmail.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
Fixes: 967b494e2fd1 ("workqueue: Use a kthread_worker to release pool_workqueues")
|
|
First commit 2930155b2e272 ("workqueue: Initialize unbound CPU pods later in
the boot") added the initialization of wq_update_pod_attrs_buf to
workqueue_init_early(), and then latter on, commit 84193c07105c6
("workqueue: Generalize unbound CPU pods") added it as well. This appeared
in a kmemleak run where the second allocation made the first allocation
leak.
Fixes: 84193c07105c6 ("workqueue: Generalize unbound CPU pods")
Signed-off-by: Steven Rostedt (Google) <rostedt@goodmis.org>
Reviewed-by: Geert Uytterhoeven <geert+renesas@glider.be>
Signed-off-by: Tejun Heo <tj@kernel.org>
|
|
For the write_atomic callback, the console driver may have unsafe
regions that need to be appropriately marked. Provide functions
that accept the nbcon_write_context struct to allow for the driver
to enter and exit unsafe regions.
Also provide a function for drivers to check if they are still the
owner of the console.
Co-developed-by: John Ogness <john.ogness@linutronix.de>
Signed-off-by: John Ogness <john.ogness@linutronix.de>
Signed-off-by: Thomas Gleixner (Intel) <tglx@linutronix.de>
Reviewed-by: Petr Mladek <pmladek@suse.com>
Signed-off-by: Petr Mladek <pmladek@suse.com>
Link: https://lore.kernel.org/r/20230916192007.608398-9-john.ogness@linutronix.de
|
|
Implement an emit function for nbcon consoles to output printk
messages. It utilizes the lockless printk_get_next_message() and
console_prepend_dropped() functions to retrieve/build the output
message. The emit function includes the required safety points to
check for handover/takeover and calls a new write_atomic callback
of the console driver to output the message. It also includes
proper handling for updating the nbcon console sequence number.
A new nbcon_write_context struct is introduced. This is provided
to the write_atomic callback and includes only the information
necessary for performing atomic writes.
Co-developed-by: John Ogness <john.ogness@linutronix.de>
Signed-off-by: John Ogness <john.ogness@linutronix.de>
Signed-off-by: Thomas Gleixner (Intel) <tglx@linutronix.de>
Reviewed-by: Petr Mladek <pmladek@suse.com>
Signed-off-by: Petr Mladek <pmladek@suse.com>
Link: https://lore.kernel.org/r/20230916192007.608398-8-john.ogness@linutronix.de
|
|
Add an atomic_long_t field @nbcon_seq to the console struct to
store the sequence number for nbcon consoles. For nbcon consoles
this will be used instead of the non-atomic @seq field. The new
field allows for safe atomic sequence number updates without
requiring any locking.
On 64bit systems the new field stores the full sequence number.
On 32bit systems the new field stores the lower 32 bits of the
sequence number, which are expanded to 64bit as needed by
folding the values based on the sequence numbers available in
the ringbuffer.
For 32bit systems, having a 32bit representation in the console
is sufficient. If a console ever gets more than 2^31 records
behind the ringbuffer then this is the least of the problems.
Co-developed-by: John Ogness <john.ogness@linutronix.de>
Signed-off-by: John Ogness <john.ogness@linutronix.de>
Signed-off-by: Thomas Gleixner (Intel) <tglx@linutronix.de>
Signed-off-by: Petr Mladek <pmladek@suse.com>
Link: https://lore.kernel.org/r/20230916192007.608398-7-john.ogness@linutronix.de
|
|
Provide functions that are related to the safe handover mechanism
and allow console drivers to dynamically specify unsafe regions:
- nbcon_context_can_proceed()
Invoked by a console owner to check whether a handover request
is pending or whether the console has been taken over by another
context. If a handover request is pending, this function will
also perform the handover, thus cancelling its own ownership.
- nbcon_context_enter_unsafe()/nbcon_context_exit_unsafe()
Invoked by a console owner to denote that the driver is about
to enter or leave a critical region where a take over is unsafe.
The exit variant is the point where the current owner releases
the lock for a higher priority context which asked for the
friendly handover.
The unsafe state is stored in the console state and allows a
new context to make informed decisions whether to attempt a
takeover of such a console. The unsafe state is also available
to the driver so that it can make informed decisions about the
required actions and possibly take a special emergency path.
Co-developed-by: John Ogness <john.ogness@linutronix.de>
Signed-off-by: John Ogness <john.ogness@linutronix.de>
Signed-off-by: Thomas Gleixner (Intel) <tglx@linutronix.de>
Reviewed-by: Petr Mladek <pmladek@suse.com>
Signed-off-by: Petr Mladek <pmladek@suse.com>
Link: https://lore.kernel.org/r/20230916192007.608398-6-john.ogness@linutronix.de
|
|
In case of hostile takeovers it must be ensured that the previous
owner cannot scribble over the output buffer of the emergency/panic
context. This is achieved by:
- Adding a global output buffer instance for the panic context.
This is the only situation where hostile takeovers can occur and
there is always at most 1 panic context.
- Allocating an output buffer per non-boot console upon console
registration. This buffer is used by the console owner when not
in panic context. (For boot consoles, the existing shared global
legacy output buffer is used instead. Boot console printing will
be synchronized with legacy console printing.)
- Choosing the appropriate buffer is handled in the acquire/release
functions.
Co-developed-by: John Ogness <john.ogness@linutronix.de>
Signed-off-by: John Ogness <john.ogness@linutronix.de>
Signed-off-by: Thomas Gleixner (Intel) <tglx@linutronix.de>
Reviewed-by: Petr Mladek <pmladek@suse.com>
Signed-off-by: Petr Mladek <pmladek@suse.com>
Link: https://lore.kernel.org/r/20230916192007.608398-5-john.ogness@linutronix.de
|
|
The nbcon boot consoles also need printk buffers that are available
very early. Since the nbcon boot consoles will also be serialized
by the console_lock, they can use the same static printk buffers
that the legacy consoles are using.
Make the legacy static printk buffers available outside of printk.c
so they can be used by nbcon.c.
Signed-off-by: John Ogness <john.ogness@linutronix.de>
Reviewed-by: Petr Mladek <pmladek@suse.com>
Signed-off-by: Petr Mladek <pmladek@suse.com>
Link: https://lore.kernel.org/r/20230916192007.608398-4-john.ogness@linutronix.de
|
|
Add per console acquire/release functionality.
The state of the console is maintained in the "nbcon_state" atomic
variable.
The console is locked when:
- The 'prio' field contains the priority of the context that owns the
console. Only higher priority contexts are allowed to take over the
lock. A value of 0 (NBCON_PRIO_NONE) means the console is not locked.
- The 'cpu' field denotes on which CPU the console is locked. It is used
to prevent busy waiting on the same CPU. Also it informs the lock owner
that it has lost the lock in a more complex scenario when the lock was
taken over by a higher priority context, released, and taken on another
CPU with the same priority as the interrupted owner.
The acquire mechanism uses a few more fields:
- The 'req_prio' field is used by the handover approach to make the
current owner aware that there is a context with a higher priority
waiting for the friendly handover.
- The 'unsafe' field allows to take over the console in a safe way in the
middle of emitting a message. The field is set only when accessing some
shared resources or when the console device is manipulated. It can be
cleared, for example, after emitting one character when the console
device is in a consistent state.
- The 'unsafe_takeover' field is set when a hostile takeover took the
console in an unsafe state. The console will stay in the unsafe state
until re-initialized.
The acquire mechanism uses three approaches:
1) Direct acquire when the console is not owned or is owned by a lower
priority context and is in a safe state.
2) Friendly handover mechanism uses a request/grant handshake. It is used
when the current owner has lower priority and the console is in an
unsafe state.
The requesting context:
a) Sets its priority into the 'req_prio' field.
b) Waits (with a timeout) for the owning context to unlock the
console.
c) Takes the lock and clears the 'req_prio' field.
The owning context:
a) Observes the 'req_prio' field set on exit from the unsafe
console state.
b) Gives up console ownership by clearing the 'prio' field.
3) Unsafe hostile takeover allows to take over the lock even when the
console is an unsafe state. It is used only in panic() by the final
attempt to flush consoles in a try and hope mode.
Note that separate record buffers are used in panic(). As a result,
the messages can be read and formatted without any risk even after
using the hostile takeover in unsafe state.
The release function simply clears the 'prio' field.
All operations on @console::nbcon_state are atomic cmpxchg based to
handle concurrency.
The acquire/release functions implement only minimal policies:
- Preference for higher priority contexts.
- Protection of the panic CPU.
All other policy decisions must be made at the call sites:
- What is marked as an unsafe section.
- Whether to spin-wait if there is already an owner and the console is
in an unsafe state.
- Whether to attempt an unsafe hostile takeover.
The design allows to implement the well known:
acquire()
output_one_printk_record()
release()
The output of one printk record might be interrupted with a higher priority
context. The new owner is supposed to reprint the entire interrupted record
from scratch.
Co-developed-by: John Ogness <john.ogness@linutronix.de>
Signed-off-by: John Ogness <john.ogness@linutronix.de>
Signed-off-by: Thomas Gleixner (Intel) <tglx@linutronix.de>
Signed-off-by: Petr Mladek <pmladek@suse.com>
Link: https://lore.kernel.org/r/20230916192007.608398-3-john.ogness@linutronix.de
|
|
The current console/printk subsystem is protected by a Big Kernel Lock,
(aka console_lock) which has ill defined semantics and is more or less
stateless. This puts severe limitations on the console subsystem and
makes forced takeover and output in emergency and panic situations a
fragile endeavour that is based on try and pray.
The goal of non-BKL (nbcon) consoles is to break out of the console lock
jail and to provide a new infrastructure that avoids the pitfalls and
also allows console drivers to be gradually converted over.
The proposed infrastructure aims for the following properties:
- Per console locking instead of global locking
- Per console state that allows to make informed decisions
- Stateful handover and takeover
As a first step, state is added to struct console. The per console state
is an atomic_t using a 32bit bit field.
Reserve state bits, which will be populated later in the series. Wire
it up into the console register/unregister functionality.
It was decided to use a bitfield because using a plain u32 with
mask/shift operations resulted in uncomprehensible code.
Co-developed-by: John Ogness <john.ogness@linutronix.de>
Signed-off-by: John Ogness <john.ogness@linutronix.de>
Signed-off-by: Thomas Gleixner (Intel) <tglx@linutronix.de>
Reviewed-by: Petr Mladek <pmladek@suse.com>
Signed-off-by: Petr Mladek <pmladek@suse.com>
Link: https://lore.kernel.org/r/20230916192007.608398-2-john.ogness@linutronix.de
|
|
Remove duplicated includes of linux/cgroup.h and linux/psi.h. Both of
these includes are included regardless of the config and they are all
protected by ifndef, so no point including them again.
Signed-off-by: GUO Zihua <guozihua@huawei.com>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Link: https://lore.kernel.org/r/20230818015633.18370-1-guozihua@huawei.com
|
|
When using sysbench to benchmark Postgres in a single docker instance
with sysbench's nr_threads set to nr_cpu, it is observed there are times
update_cfs_group() and update_load_avg() shows noticeable overhead on
a 2sockets/112core/224cpu Intel Sapphire Rapids(SPR):
13.75% 13.74% [kernel.vmlinux] [k] update_cfs_group
10.63% 10.04% [kernel.vmlinux] [k] update_load_avg
Annotate shows the cycles are mostly spent on accessing tg->load_avg
with update_load_avg() being the write side and update_cfs_group() being
the read side. tg->load_avg is per task group and when different tasks
of the same taskgroup running on different CPUs frequently access
tg->load_avg, it can be heavily contended.
E.g. when running postgres_sysbench on a 2sockets/112cores/224cpus Intel
Sappire Rapids, during a 5s window, the wakeup number is 14millions and
migration number is 11millions and with each migration, the task's load
will transfer from src cfs_rq to target cfs_rq and each change involves
an update to tg->load_avg. Since the workload can trigger as many wakeups
and migrations, the access(both read and write) to tg->load_avg can be
unbound. As a result, the two mentioned functions showed noticeable
overhead. With netperf/nr_client=nr_cpu/UDP_RR, the problem is worse:
during a 5s window, wakeup number is 21millions and migration number is
14millions; update_cfs_group() costs ~25% and update_load_avg() costs ~16%.
Reduce the overhead by limiting updates to tg->load_avg to at most once
per ms. The update frequency is a tradeoff between tracking accuracy and
overhead. 1ms is chosen because PELT window is roughly 1ms and it
delivered good results for the tests that I've done. After this change,
the cost of accessing tg->load_avg is greatly reduced and performance
improved. Detailed test results below.
==============================
postgres_sysbench on SPR:
25%
base: 42382±19.8%
patch: 50174±9.5% (noise)
50%
base: 67626±1.3%
patch: 67365±3.1% (noise)
75%
base: 100216±1.2%
patch: 112470±0.1% +12.2%
100%
base: 93671±0.4%
patch: 113563±0.2% +21.2%
==============================
hackbench on ICL:
group=1
base: 114912±5.2%
patch: 117857±2.5% (noise)
group=4
base: 359902±1.6%
patch: 361685±2.7% (noise)
group=8
base: 461070±0.8%
patch: 491713±0.3% +6.6%
group=16
base: 309032±5.0%
patch: 378337±1.3% +22.4%
=============================
hackbench on SPR:
group=1
base: 100768±2.9%
patch: 103134±2.9% (noise)
group=4
base: 413830±12.5%
patch: 378660±16.6% (noise)
group=8
base: 436124±0.6%
patch: 490787±3.2% +12.5%
group=16
base: 457730±3.2%
patch: 680452±1.3% +48.8%
============================
netperf/udp_rr on ICL
25%
base: 114413±0.1%
patch: 115111±0.0% +0.6%
50%
base: 86803±0.5%
patch: 86611±0.0% (noise)
75%
base: 35959±5.3%
patch: 49801±0.6% +38.5%
100%
base: 61951±6.4%
patch: 70224±0.8% +13.4%
===========================
netperf/udp_rr on SPR
25%
base: 104954±1.3%
patch: 107312±2.8% (noise)
50%
base: 55394±4.6%
patch: 54940±7.4% (noise)
75%
base: 13779±3.1%
patch: 36105±1.1% +162%
100%
base: 9703±3.7%
patch: 28011±0.2% +189%
==============================================
netperf/tcp_stream on ICL (all in noise range)
25%
base: 43092±0.1%
patch: 42891±0.5%
50%
base: 19278±14.9%
patch: 22369±7.2%
75%
base: 16822±3.0%
patch: 17086±2.3%
100%
base: 18216±0.6%
patch: 18078±2.9%
===============================================
netperf/tcp_stream on SPR (all in noise range)
25%
base: 34491±0.3%
patch: 34886±0.5%
50%
base: 19278±14.9%
patch: 22369±7.2%
75%
base: 16822±3.0%
patch: 17086±2.3%
100%
base: 18216±0.6%
patch: 18078±2.9%
Reported-by: Nitin Tekchandani <nitin.tekchandani@intel.com>
Suggested-by: Vincent Guittot <vincent.guittot@linaro.org>
Signed-off-by: Aaron Lu <aaron.lu@intel.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Reviewed-by: Vincent Guittot <vincent.guittot@linaro.org>
Reviewed-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Reviewed-by: David Vernet <void@manifault.com>
Tested-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Tested-by: Swapnil Sapkal <Swapnil.Sapkal@amd.com>
Link: https://lkml.kernel.org/r/20230912065808.2530-2-aaron.lu@intel.com
|
|
After commit f5d39b020809 ("freezer,sched: Rewrite core freezer logic"),
tasks that transition directly from TASK_FREEZABLE to TASK_FROZEN are
always woken up on the thaw path. Prior to that commit, tasks could ask
freezer to consider them "frozen enough" via freezer_do_not_count(). The
commit replaced freezer_do_not_count() with a TASK_FREEZABLE state which
allows freezer to immediately mark the task as TASK_FROZEN without
waking up the task. This is efficient for the suspend path, but on the
thaw path, the task is always woken up even if the task didn't need to
wake up and goes back to its TASK_(UN)INTERRUPTIBLE state. Although
these tasks are capable of handling of the wakeup, we can observe a
power/perf impact from the extra wakeup.
We observed on Android many tasks wait in the TASK_FREEZABLE state
(particularly due to many of them being binder clients). We observed
nearly 4x the number of tasks and a corresponding linear increase in
latency and power consumption when thawing the system. The latency
increased from ~15ms to ~50ms.
Avoid the spurious wakeups by saving the state of TASK_FREEZABLE tasks.
If the task was running before entering TASK_FROZEN state
(__refrigerator()) or if the task received a wake up for the saved
state, then the task is woken on thaw. saved_state from PREEMPT_RT locks
can be re-used because freezer would not stomp on the rtlock wait flow:
TASK_RTLOCK_WAIT isn't considered freezable.
Reported-by: Prakash Viswalingam <quic_prakashv@quicinc.com>
Signed-off-by: Elliot Berman <quic_eberman@quicinc.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
|
|
In preparation for freezer to also use saved_state, remove the
CONFIG_PREEMPT_RT compilation guard around saved_state.
On the arm64 platform I tested which did not have CONFIG_PREEMPT_RT,
there was no statistically significant deviation by applying this patch.
Test methodology:
perf bench sched message -g 40 -l 40
Signed-off-by: Elliot Berman <quic_eberman@quicinc.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
|
|
git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip
Pull scheduler fixes from Ingo Molnar:
"Fix a performance regression on large SMT systems, an Intel SMT4
balancing bug, and a topology setup bug on (Intel) hybrid processors"
* tag 'sched-urgent-2023-09-17' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
x86/sched: Restore the SD_ASYM_PACKING flag in the DIE domain
sched/fair: Fix SMT4 group_smt_balance handling
sched/fair: Optimize should_we_balance() for large SMT systems
|
|
git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip
Pull WARN fix from Ingo Molnar:
"Fix a missing preempt-enable in the WARN() slowpath"
* tag 'core-urgent-2023-09-17' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
panic: Reenable preemption in WARN slowpath
|
|
Alexei Starovoitov says:
====================
The following pull-request contains BPF updates for your *net-next* tree.
We've added 73 non-merge commits during the last 9 day(s) which contain
a total of 79 files changed, 5275 insertions(+), 600 deletions(-).
The main changes are:
1) Basic BTF validation in libbpf, from Andrii Nakryiko.
2) bpf_assert(), bpf_throw(), exceptions in bpf progs, from Kumar Kartikeya Dwivedi.
3) next_thread cleanups, from Oleg Nesterov.
4) Add mcpu=v4 support to arm32, from Puranjay Mohan.
5) Add support for __percpu pointers in bpf progs, from Yonghong Song.
6) Fix bpf tailcall interaction with bpf trampoline, from Leon Hwang.
7) Raise irq_work in bpf_mem_alloc while irqs are disabled to improve refill probabablity, from Hou Tao.
Please consider pulling these changes from:
git://git.kernel.org/pub/scm/linux/kernel/git/bpf/bpf-next.git
Thanks a lot!
Also thanks to reporters, reviewers and testers of commits in this pull-request:
Alan Maguire, Andrey Konovalov, Dave Marchevsky, "Eric W. Biederman",
Jiri Olsa, Maciej Fijalkowski, Quentin Monnet, Russell King (Oracle),
Song Liu, Stanislav Fomichev, Yonghong Song
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
The kfunc code to handle KF_ARG_PTR_TO_CALLBACK does not check the reg
type before using reg->subprogno. This can accidently permit invalid
pointers from being passed into callback helpers (e.g. silently from
different paths). Likewise, reg->subprogno from the per-register type
union may not be meaningful either. We need to reject any other type
except PTR_TO_FUNC.
Acked-by: Dave Marchevsky <davemarchevsky@fb.com>
Fixes: 5d92ddc3de1b ("bpf: Add callback validation to kfunc verifier logic")
Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com>
Link: https://lore.kernel.org/r/20230912233214.1518551-14-memxor@gmail.com
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
|
|
During testing, it was discovered that extensions to exception callbacks
had no checks, upon running a testcase, the kernel ended up running off
the end of a program having final call as bpf_throw, and hitting int3
instructions.
The reason is that while the default exception callback would have reset
the stack frame to return back to the main program's caller, the
replacing extension program will simply return back to bpf_throw, which
will instead return back to the program and the program will continue
execution, now in an undefined state where anything could happen.
The way to support extensions to an exception callback would be to mark
the BPF_PROG_TYPE_EXT main subprog as an exception_cb, and prevent it
from calling bpf_throw. This would make the JIT produce a prologue that
restores saved registers and reset the stack frame. But let's not do
that until there is a concrete use case for this, and simply disallow
this for now.
Similar issues will exist for fentry and fexit cases, where trampoline
saves data on the stack when invoking exception callback, which however
will then end up resetting the stack frame, and on return, the fexit
program will never will invoked as the return address points to the main
program's caller in the kernel. Instead of additional complexity and
back and forth between the two stacks to enable such a use case, simply
forbid it.
One key point here to note is that currently X86_TAIL_CALL_OFFSET didn't
require any modifications, even though we emit instructions before the
corresponding endbr64 instruction. This is because we ensure that a main
subprog never serves as an exception callback, and therefore the
exception callback (which will be a global subprog) can never serve as
the tail call target, eliminating any discrepancies. However, once we
support a BPF_PROG_TYPE_EXT to also act as an exception callback, it
will end up requiring change to the tail call offset to account for the
extra instructions. For simplicitly, tail calls could be disabled for
such targets.
Noting the above, it appears better to wait for a concrete use case
before choosing to permit extension programs to replace exception
callbacks.
As a precaution, we disable fentry and fexit for exception callbacks as
well.
Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com>
Link: https://lore.kernel.org/r/20230912233214.1518551-13-memxor@gmail.com
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
|
|
Now that bpf_throw kfunc is the first such call instruction that has
noreturn semantics within the verifier, this also kicks in dead code
elimination in unprecedented ways. For one, any instruction following
a bpf_throw call will never be marked as seen. Moreover, if a callchain
ends up throwing, any instructions after the call instruction to the
eventually throwing subprog in callers will also never be marked as
seen.
The tempting way to fix this would be to emit extra 'int3' instructions
which bump the jited_len of a program, and ensure that during runtime
when a program throws, we can discover its boundaries even if the call
instruction to bpf_throw (or to subprogs that always throw) is emitted
as the final instruction in the program.
An example of such a program would be this:
do_something():
...
r0 = 0
exit
foo():
r1 = 0
call bpf_throw
r0 = 0
exit
bar(cond):
if r1 != 0 goto pc+2
call do_something
exit
call foo
r0 = 0 // Never seen by verifier
exit //
main(ctx):
r1 = ...
call bar
r0 = 0
exit
Here, if we do end up throwing, the stacktrace would be the following:
bpf_throw
foo
bar
main
In bar, the final instruction emitted will be the call to foo, as such,
the return address will be the subsequent instruction (which the JIT
emits as int3 on x86). This will end up lying outside the jited_len of
the program, thus, when unwinding, we will fail to discover the return
address as belonging to any program and end up in a panic due to the
unreliable stack unwinding of BPF programs that we never expect.
To remedy this case, make bpf_prog_ksym_find treat IP == ksym.end as
part of the BPF program, so that is_bpf_text_address returns true when
such a case occurs, and we are able to unwind reliably when the final
instruction ends up being a call instruction.
Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com>
Link: https://lore.kernel.org/r/20230912233214.1518551-12-memxor@gmail.com
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
|
|
The KASAN stack instrumentation when CONFIG_KASAN_STACK is true poisons
the stack of a function when it is entered and unpoisons it when
leaving. However, in the case of bpf_throw, we will never return as we
switch our stack frame to the BPF exception callback. Later, this
discrepancy will lead to confusing KASAN splats when kernel resumes
execution on return from the BPF program.
Fix this by unpoisoning everything below the stack pointer of the BPF
program, which should cover the range that would not be unpoisoned. An
example splat is below:
BUG: KASAN: stack-out-of-bounds in stack_trace_consume_entry+0x14e/0x170
Write of size 8 at addr ffffc900013af958 by task test_progs/227
CPU: 0 PID: 227 Comm: test_progs Not tainted 6.5.0-rc2-g43f1c6c9052a-dirty #26
Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS 1.16.2-2.fc39 04/01/2014
Call Trace:
<TASK>
dump_stack_lvl+0x4a/0x80
print_report+0xcf/0x670
? arch_stack_walk+0x79/0x100
kasan_report+0xda/0x110
? stack_trace_consume_entry+0x14e/0x170
? stack_trace_consume_entry+0x14e/0x170
? __pfx_stack_trace_consume_entry+0x10/0x10
stack_trace_consume_entry+0x14e/0x170
? __sys_bpf+0xf2e/0x41b0
arch_stack_walk+0x8b/0x100
? __sys_bpf+0xf2e/0x41b0
? bpf_prog_test_run_skb+0x341/0x1c70
? bpf_prog_test_run_skb+0x341/0x1c70
stack_trace_save+0x9b/0xd0
? __pfx_stack_trace_save+0x10/0x10
? __kasan_slab_free+0x109/0x180
? bpf_prog_test_run_skb+0x341/0x1c70
? __sys_bpf+0xf2e/0x41b0
? __x64_sys_bpf+0x78/0xc0
? do_syscall_64+0x3c/0x90
? entry_SYSCALL_64_after_hwframe+0x6e/0xd8
kasan_save_stack+0x33/0x60
? kasan_save_stack+0x33/0x60
? kasan_set_track+0x25/0x30
? kasan_save_free_info+0x2b/0x50
? __kasan_slab_free+0x109/0x180
? kmem_cache_free+0x191/0x460
? bpf_prog_test_run_skb+0x341/0x1c70
kasan_set_track+0x25/0x30
kasan_save_free_info+0x2b/0x50
__kasan_slab_free+0x109/0x180
kmem_cache_free+0x191/0x460
bpf_prog_test_run_skb+0x341/0x1c70
? __pfx_bpf_prog_test_run_skb+0x10/0x10
? __fget_light+0x51/0x220
__sys_bpf+0xf2e/0x41b0
? __might_fault+0xa2/0x170
? __pfx___sys_bpf+0x10/0x10
? lock_release+0x1de/0x620
? __might_fault+0xcd/0x170
? __pfx_lock_release+0x10/0x10
? __pfx_blkcg_maybe_throttle_current+0x10/0x10
__x64_sys_bpf+0x78/0xc0
? syscall_enter_from_user_mode+0x20/0x50
do_syscall_64+0x3c/0x90
entry_SYSCALL_64_after_hwframe+0x6e/0xd8
RIP: 0033:0x7f0fbb38880d
Code: ff c3 66 2e 0f 1f 84 00 00 00 00 00 90 f3 0f 1e fa 48 89 f8 48 89 f7 48 89 d6 48 89 ca 4d
89 c2 4d 89 c8 4c 8b 4c 24 08 0f 05 <48> 3d 01 f0 ff ff 73 01 c3 48 8b 0d f3 45 12 00 f7 d8 64
89 01 48
RSP: 002b:00007ffe13907de8 EFLAGS: 00000206 ORIG_RAX: 0000000000000141
RAX: ffffffffffffffda RBX: 00007ffe13908708 RCX: 00007f0fbb38880d
RDX: 0000000000000050 RSI: 00007ffe13907e20 RDI: 000000000000000a
RBP: 00007ffe13907e00 R08: 0000000000000000 R09: 00007ffe13907e20
R10: 0000000000000064 R11: 0000000000000206 R12: 0000000000000003
R13: 0000000000000000 R14: 00007f0fbb532000 R15: 0000000000cfbd90
</TASK>
The buggy address belongs to stack of task test_progs/227
KASAN internal error: frame info validation failed; invalid marker: 0
The buggy address belongs to the virtual mapping at
[ffffc900013a8000, ffffc900013b1000) created by:
kernel_clone+0xcd/0x600
The buggy address belongs to the physical page:
page:00000000b70f4332 refcount:1 mapcount:0 mapping:0000000000000000 index:0x0 pfn:0x11418f
flags: 0x2fffe0000000000(node=0|zone=2|lastcpupid=0x7fff)
page_type: 0xffffffff()
raw: 02fffe0000000000 0000000000000000 dead000000000122 0000000000000000
raw: 0000000000000000 0000000000000000 00000001ffffffff 0000000000000000
page dumped because: kasan: bad access detected
Memory state around the buggy address:
ffffc900013af800: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
ffffc900013af880: 00 00 00 f1 f1 f1 f1 00 00 00 f3 f3 f3 f3 f3 00
>ffffc900013af900: 00 00 00 00 00 00 00 00 00 00 00 f1 00 00 00 00
^
ffffc900013af980: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
ffffc900013afa00: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
==================================================================
Disabling lock debugging due to kernel taint
Cc: Andrey Ryabinin <ryabinin.a.a@gmail.com>
Cc: Alexander Potapenko <glider@google.com>
Cc: Andrey Konovalov <andreyknvl@gmail.com>
Cc: Dmitry Vyukov <dvyukov@google.com>
Cc: Vincenzo Frascino <vincenzo.frascino@arm.com>
Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com>
Acked-by: Andrey Konovalov <andreyknvl@gmail.com>
Link: https://lore.kernel.org/r/20230912233214.1518551-11-memxor@gmail.com
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
|
|
In case of the default exception callback, change the behavior of
bpf_throw, where the passed cookie value is no longer ignored, but
is instead the return value of the default exception callback. As
such, we need to place restrictions on the value being passed into
bpf_throw in such a case, only allowing those permitted by the
check_return_code function.
Thus, bpf_throw can now control the return value of the program from
each call site without having the user install a custom exception
callback just to override the return value when an exception is thrown.
We also modify the hidden subprog instructions to now move BPF_REG_1 to
BPF_REG_0, so as to set the return value before exit in the default
callback.
Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com>
Link: https://lore.kernel.org/r/20230912233214.1518551-9-memxor@gmail.com
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
|
|
Since exception callbacks are not referenced using bpf_pseudo_func and
bpf_pseudo_call instructions, check_cfg traversal will never explore
instructions of the exception callback. Even after adding the subprog,
the program will then fail with a 'unreachable insn' error.
We thus need to begin walking from the start of the exception callback
again in check_cfg after a complete CFG traversal finishes, so as to
explore the CFG rooted at the exception callback.
Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com>
Link: https://lore.kernel.org/r/20230912233214.1518551-8-memxor@gmail.com
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
|
|
By default, the subprog generated by the verifier to handle a thrown
exception hardcodes a return value of 0. To allow user-defined logic
and modification of the return value when an exception is thrown,
introduce the 'exception_callback:' declaration tag, which marks a
callback as the default exception handler for the program.
The format of the declaration tag is 'exception_callback:<value>', where
<value> is the name of the exception callback. Each main program can be
tagged using this BTF declaratiion tag to associate it with an exception
callback. In case the tag is absent, the default callback is used.
As such, the exception callback cannot be modified at runtime, only set
during verification.
Allowing modification of the callback for the current program execution
at runtime leads to issues when the programs begin to nest, as any
per-CPU state maintaing this information will have to be saved and
restored. We don't want it to stay in bpf_prog_aux as this takes a
global effect for all programs. An alternative solution is spilling
the callback pointer at a known location on the program stack on entry,
and then passing this location to bpf_throw as a parameter.
However, since exceptions are geared more towards a use case where they
are ideally never invoked, optimizing for this use case and adding to
the complexity has diminishing returns.
Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com>
Link: https://lore.kernel.org/r/20230912233214.1518551-7-memxor@gmail.com
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
|
|
This patch splits the check_btf_info's check_btf_func check into two
separate phases. The first phase sets up the BTF and prepares
func_info, but does not perform any validation of required invariants
for subprogs just yet. This is left to the second phase, which happens
where check_btf_info executes currently, and performs the line_info and
CO-RE relocation.
The reason to perform this split is to obtain the userspace supplied
func_info information before we perform the add_subprog call, where we
would now require finding and adding subprogs that may not have a
bpf_pseudo_call or bpf_pseudo_func instruction in the program.
We require this as we want to enable userspace to supply exception
callbacks that can override the default hidden subprogram generated by
the verifier (which performs a hardcoded action). In such a case, the
exception callback may never be referenced in an instruction, but will
still be suitably annotated (by way of BTF declaration tags). For
finding this exception callback, we would require the program's BTF
information, and the supplied func_info information which maps BTF type
IDs to subprograms.
Since the exception callback won't actually be referenced through
instructions, later checks in check_cfg and do_check_subprogs will not
verify the subprog. This means that add_subprog needs to add them in the
add_subprog_and_kfunc phase before we move forward, which is why the BTF
and func_info are required at that point.
Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com>
Link: https://lore.kernel.org/r/20230912233214.1518551-6-memxor@gmail.com
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
|
|
This patch implements BPF exceptions, and introduces a bpf_throw kfunc
to allow programs to throw exceptions during their execution at runtime.
A bpf_throw invocation is treated as an immediate termination of the
program, returning back to its caller within the kernel, unwinding all
stack frames.
This allows the program to simplify its implementation, by testing for
runtime conditions which the verifier has no visibility into, and assert
that they are true. In case they are not, the program can simply throw
an exception from the other branch.
BPF exceptions are explicitly *NOT* an unlikely slowpath error handling
primitive, and this objective has guided design choices of the
implementation of the them within the kernel (with the bulk of the cost
for unwinding the stack offloaded to the bpf_throw kfunc).
The implementation of this mechanism requires use of add_hidden_subprog
mechanism introduced in the previous patch, which generates a couple of
instructions to move R1 to R0 and exit. The JIT then rewrites the
prologue of this subprog to take the stack pointer and frame pointer as
inputs and reset the stack frame, popping all callee-saved registers
saved by the main subprog. The bpf_throw function then walks the stack
at runtime, and invokes this exception subprog with the stack and frame
pointers as parameters.
Reviewers must take note that currently the main program is made to save
all callee-saved registers on x86_64 during entry into the program. This
is because we must do an equivalent of a lightweight context switch when
unwinding the stack, therefore we need the callee-saved registers of the
caller of the BPF program to be able to return with a sane state.
Note that we have to additionally handle r12, even though it is not used
by the program, because when throwing the exception the program makes an
entry into the kernel which could clobber r12 after saving it on the
stack. To be able to preserve the value we received on program entry, we
push r12 and restore it from the generated subprogram when unwinding the
stack.
For now, bpf_throw invocation fails when lingering resources or locks
exist in that path of the program. In a future followup, bpf_throw will
be extended to perform frame-by-frame unwinding to release lingering
resources for each stack frame, removing this limitation.
Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com>
Link: https://lore.kernel.org/r/20230912233214.1518551-5-memxor@gmail.com
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
|
