PSI accounts stalls for each cgroup separately and aggregates it
at each level of the hierarchy. This may cause non-negligible overhead
for some workloads when under deep level of the hierarchy.
commit 3958e2d0c3 ("cgroup: make per-cgroup pressure stall tracking configurable")
make PSI to skip per-cgroup stall accounting, only account system-wide
to avoid this each level overhead.
But for our use case, we also want leaf cgroup PSI stats accounted for
userspace adjustment on that cgroup, apart from only system-wide adjustment.
So this patch introduce a per-cgroup PSI accounting disable/re-enable
interface "cgroup.pressure", which is a read-write single value file that
allowed values are "0" and "1", the defaults is "1" so per-cgroup
PSI stats is enabled by default.
Implementation details:
It should be relatively straight-forward to disable and re-enable
state aggregation, time tracking, averaging on a per-cgroup level,
if we can live with losing history from while it was disabled.
I.e. the avgs will restart from 0, total= will have gaps.
But it's hard or complex to stop/restart groupc->tasks[] updates,
which is not implemented in this patch. So we always update
groupc->tasks[] and PSI_ONCPU bit in psi_group_change() even when
the cgroup PSI stats is disabled.
Suggested-by: Johannes Weiner <hannes@cmpxchg.org>
Suggested-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Chengming Zhou <zhouchengming@bytedance.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Link: https://lkml.kernel.org/r/20220907090332.2078-1-zhouchengming@bytedance.com
We use iterate_groups() to iterate each level psi_group to update
PSI stats, which is a very hot path.
In current code, iterate_groups() have to use multiple branches and
cgroup_parent() to get parent psi_group for each level, which is not
very efficient.
This patch cache parent psi_group in struct psi_group, only need to get
psi_group of task itself first, then just use group->parent to iterate.
Signed-off-by: Chengming Zhou <zhouchengming@bytedance.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Link: https://lore.kernel.org/r/20220825164111.29534-10-zhouchengming@bytedance.com
Now PSI already tracked workload pressure stall information for
CPU, memory and IO. Apart from these, IRQ/SOFTIRQ could have
obvious impact on some workload productivity, such as web service
workload.
When CONFIG_IRQ_TIME_ACCOUNTING, we can get IRQ/SOFTIRQ delta time
from update_rq_clock_task(), in which we can record that delta
to CPU curr task's cgroups as PSI_IRQ_FULL status.
Note we don't use PSI_IRQ_SOME since IRQ/SOFTIRQ always happen in
the current task on the CPU, make nothing productive could run
even if it were runnable, so we only use PSI_IRQ_FULL.
Signed-off-by: Chengming Zhou <zhouchengming@bytedance.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Link: https://lore.kernel.org/r/20220825164111.29534-8-zhouchengming@bytedance.com
Way back when PSI_MEM_FULL was accounted from the timer tick, task
switching could simply iterate next and prev to the common ancestor to
update TSK_ONCPU and be done.
Then memstall ticks were replaced with checking curr->in_memstall
directly in psi_group_change(). That meant that now if the task switch
was between a memstall and a !memstall task, we had to iterate through
the common ancestors at least ONCE to fix up their state_masks.
We added the identical_state filter to make sure the common ancestor
elimination was skipped in that case. It seems that was always a
little too eager, because it caused us to walk the common ancestors
*twice* instead of the required once: the iteration for next could
have stopped at the common ancestor; prev could have updated TSK_ONCPU
up to the common ancestor, then finish to the root without changing
any flags, just to get the new curr->in_memstall into the state_masks.
This patch recognizes this and makes it so that we walk to the root
exactly once if state_mask needs updating, which is simply catching up
on a missed optimization that could have been done in commit 7fae6c8171
("psi: Use ONCPU state tracking machinery to detect reclaim") directly.
Apart from this, it's also necessary for the next patch "sched/psi: remove
NR_ONCPU task accounting". Suppose we walk the common ancestors twice:
(1) psi_group_change(.clear = 0, .set = TSK_ONCPU)
(2) psi_group_change(.clear = TSK_ONCPU, .set = 0)
We previously used tasks[NR_ONCPU] to record TSK_ONCPU, tasks[NR_ONCPU]++
in (1) then tasks[NR_ONCPU]-- in (2), so tasks[NR_ONCPU] still be correct.
The next patch change to use one bit in state mask to record TSK_ONCPU,
PSI_ONCPU bit will be set in (1), but then be cleared in (2), which cause
the psi_group_cpu has task running on CPU but without PSI_ONCPU bit set!
With this patch, we will never walk the common ancestors twice, so won't
have above problem.
Suggested-by: Johannes Weiner <hannes@cmpxchg.org>
Signed-off-by: Chengming Zhou <zhouchengming@bytedance.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Link: https://lore.kernel.org/r/20220825164111.29534-6-zhouchengming@bytedance.com
This patch move psi_task_change/psi_task_switch declarations out of
PSI public header, since they are only needed for implementing the
PSI stats tracking in sched/stats.h
psi_task_switch is obvious, psi_task_change can't be public helper
since it doesn't check psi_disabled static key. And there is no
any user now, so put it in sched/stats.h too.
Signed-off-by: Chengming Zhou <zhouchengming@bytedance.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Link: https://lore.kernel.org/r/20220825164111.29534-5-zhouchengming@bytedance.com
We don't want to wake periodic aggregation work back up if the
task change is the aggregation worker itself going to sleep, or
we'll ping-pong forever.
Previously, we would use psi_task_change() in psi_dequeue() when
task going to sleep, so this check was put in psi_task_change().
But commit 4117cebf1a ("psi: Optimize task switch inside shared cgroups")
defer task sleep handling to psi_task_switch(), won't go through
psi_task_change() anymore.
So this patch move this check to psi_task_switch().
Fixes: 4117cebf1a ("psi: Optimize task switch inside shared cgroups")
Signed-off-by: Chengming Zhou <zhouchengming@bytedance.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Link: https://lore.kernel.org/r/20220825164111.29534-2-zhouchengming@bytedance.com
It is better to use SYSCTL_ZERO and SYSCTL_ONE_HUNDRED instead of &i_zero
and &i_one_hundred, and then we can remove these two local variable.
No functional change.
Signed-off-by: Liu Shixin <liushixin2@huawei.com>
Signed-off-by: Luis Chamberlain <mcgrof@kernel.org>
sysctl_vals and sysctl_long_vals are declared even if sysctl is disabled.
Move its definition to sysctl.c to make sure their integrity in any case.
Signed-off-by: Liu Shixin <liushixin2@huawei.com>
Signed-off-by: Luis Chamberlain <mcgrof@kernel.org>
Remove max_extfrag_threshold and replace by SYSCTL_ONE_THOUSAND.
No functional change.
Signed-off-by: Liu Shixin <liushixin2@huawei.com>
Signed-off-by: Luis Chamberlain <mcgrof@kernel.org>
This patch provides debug/modules/unloaded_tainted file to see a
record of unloaded tainted modules.
Signed-off-by: Aaron Tomlin <atomlin@redhat.com>
Signed-off-by: Luis Chamberlain <mcgrof@kernel.org>
For a lot of use cases in future patches, we will want to modify the
state of registers part of some same 'group' (e.g. same ref_obj_id). It
won't just be limited to releasing reference state, but setting a type
flag dynamically based on certain actions, etc.
Hence, we need a way to easily pass a callback to the function that
iterates over all registers in current bpf_verifier_state in all frames
upto (and including) the curframe.
While in C++ we would be able to easily use a lambda to pass state and
the callback together, sadly we aren't using C++ in the kernel. The next
best thing to avoid defining a function for each case seems like
statement expressions in GNU C. The kernel already uses them heavily,
hence they can passed to the macro in the style of a lambda. The
statement expression will then be substituted in the for loop bodies.
Variables __state and __reg are set to current bpf_func_state and reg
for each invocation of the expression inside the passed in verifier
state.
Then, convert mark_ptr_or_null_regs, clear_all_pkt_pointers,
release_reference, find_good_pkt_pointers, find_equal_scalars to
use bpf_for_each_reg_in_vstate.
Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com>
Link: https://lore.kernel.org/r/20220904204145.3089-16-memxor@gmail.com
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
While auditing 6b959ba22d ("perf/core: Fix reentry problem in
perf_output_read_group()") a few spots were found that wanted
assertions.
Notable for_each_sibling_event() relies on exclusion from
modification. This would normally be holding either ctx->lock or
ctx->mutex, however due to how things are constructed disabling IRQs
is a valid and sufficient substitute for ctx->lock.
Another possible site to add assertions would be the various
pmu::{add,del,read,..}() methods, but that's not trivially expressable
in C -- the best option is wrappers, but those are easy enough to
forget.
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Besides the branch type filtering requests, 'event.attr.branch_sample_type'
also contains various flags indicating which additional information should
be captured, along with the base branch record. These flags help configure
the underlying hardware, and capture the branch records appropriately when
required e.g after PMU interrupt. But first, this moves an existing helper
perf_sample_save_hw_index() into the header before adding some more helpers
for other branch sample filter flags.
Signed-off-by: Anshuman Khandual <anshuman.khandual@arm.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Link: https://lkml.kernel.org/r/20220906084414.396220-1-anshuman.khandual@arm.com
Rewrite the core freezer to behave better wrt thawing and be simpler
in general.
By replacing PF_FROZEN with TASK_FROZEN, a special block state, it is
ensured frozen tasks stay frozen until thawed and don't randomly wake
up early, as is currently possible.
As such, it does away with PF_FROZEN and PF_FREEZER_SKIP, freeing up
two PF_flags (yay!).
Specifically; the current scheme works a little like:
freezer_do_not_count();
schedule();
freezer_count();
And either the task is blocked, or it lands in try_to_freezer()
through freezer_count(). Now, when it is blocked, the freezer
considers it frozen and continues.
However, on thawing, once pm_freezing is cleared, freezer_count()
stops working, and any random/spurious wakeup will let a task run
before its time.
That is, thawing tries to thaw things in explicit order; kernel
threads and workqueues before doing bringing SMP back before userspace
etc.. However due to the above mentioned races it is entirely possible
for userspace tasks to thaw (by accident) before SMP is back.
This can be a fatal problem in asymmetric ISA architectures (eg ARMv9)
where the userspace task requires a special CPU to run.
As said; replace this with a special task state TASK_FROZEN and add
the following state transitions:
TASK_FREEZABLE -> TASK_FROZEN
__TASK_STOPPED -> TASK_FROZEN
__TASK_TRACED -> TASK_FROZEN
The new TASK_FREEZABLE can be set on any state part of TASK_NORMAL
(IOW. TASK_INTERRUPTIBLE and TASK_UNINTERRUPTIBLE) -- any such state
is already required to deal with spurious wakeups and the freezer
causes one such when thawing the task (since the original state is
lost).
The special __TASK_{STOPPED,TRACED} states *can* be restored since
their canonical state is in ->jobctl.
With this, frozen tasks need an explicit TASK_FROZEN wakeup and are
free of undue (early / spurious) wakeups.
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Reviewed-by: Ingo Molnar <mingo@kernel.org>
Acked-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Link: https://lore.kernel.org/r/20220822114649.055452969@infradead.org
Make wait_task_inactive()'s @match_state work like ttwu()'s @state.
That is, instead of an equal comparison, use it as a mask. This allows
matching multiple block conditions.
(removes the unlikely; it doesn't make sense how it's only part of the
condition)
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Link: https://lore.kernel.org/r/20220822114648.856734578@infradead.org
Rafael explained that the reason for having both PF_NOFREEZE and
PF_FREEZER_SKIP is that {,un}lock_system_sleep() is callable from
kthread context that has previously called set_freezable().
In preparation of merging the flags, have {,un}lock_system_slee() save
and restore current->flags.
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Acked-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Link: https://lore.kernel.org/r/20220822114648.725003428@infradead.org
It's uncertain whether idle cores exist or not if shared sched-
domains are not ready, so returning "no idle cores" usually
makes sense.
While __update_idle_core() is an exception, it checks status
of this core and set hint to shared sched-domain if necessary.
So the whole logic of this function depends on the existence
of shared sched-domain, and can certainly bail out early if
it is not available.
It's somehow a little tricky, and as Josh suggested that it
should be transient while the domain isn't ready. So remove
the self-defined default value to make things more clearer.
Signed-off-by: Abel Wu <wuyun.abel@bytedance.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Reviewed-by: Josh Don <joshdon@google.com>
Acked-by: Mel Gorman <mgorman@techsingularity.net>
Link: https://lore.kernel.org/r/20220907112000.1854-5-wuyun.abel@bytedance.com
Sparse reported a warning at bpf_map_free_kptrs()
"warning: Using plain integer as NULL pointer"
During the process of fixing this warning, it was discovered that the current
code erroneously writes to the pointer variable instead of deferencing and
writing to the actual kptr. Hence, Sparse tool accidentally helped to uncover
this problem. Fix this by doing WRITE_ONCE(*p, 0) instead of WRITE_ONCE(p, 0).
Note that the effect of this bug is that unreferenced kptrs will not be cleared
during check_and_free_fields. It is not a problem if the clearing is not done
during map_free stage, as there is nothing to free for them.
Fixes: 14a324f6a6 ("bpf: Wire up freeing of referenced kptr")
Signed-off-by: Jules Irenge <jbi.octave@gmail.com>
Link: https://lore.kernel.org/r/Yxi3pJaK6UDjVJSy@playground
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
For drivers (outside of network), the incoming data is not statically
defined in a struct. Most of the time the data buffer is kzalloc-ed
and thus we can not rely on eBPF and BTF to explore the data.
This commit allows to return an arbitrary memory, previously allocated by
the driver.
An interesting extra point is that the kfunc can mark the exported
memory region as read only or read/write.
So, when a kfunc is not returning a pointer to a struct but to a plain
type, we can consider it is a valid allocated memory assuming that:
- one of the arguments is either called rdonly_buf_size or
rdwr_buf_size
- and this argument is a const from the caller point of view
We can then use this parameter as the size of the allocated memory.
The memory is either read-only or read-write based on the name
of the size parameter.
Acked-by: Kumar Kartikeya Dwivedi <memxor@gmail.com>
Signed-off-by: Benjamin Tissoires <benjamin.tissoires@redhat.com>
Link: https://lore.kernel.org/r/20220906151303.2780789-7-benjamin.tissoires@redhat.com
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
When a function was trying to access data from context in a syscall eBPF
program, the verifier was rejecting the call unless it was accessing the
first element.
This is because the syscall context is not known at compile time, and
so we need to check this when actually accessing it.
Check for the valid memory access if there is no convert_ctx callback,
and allow such situation to happen.
Acked-by: Kumar Kartikeya Dwivedi <memxor@gmail.com>
Signed-off-by: Benjamin Tissoires <benjamin.tissoires@redhat.com>
Link: https://lore.kernel.org/r/20220906151303.2780789-4-benjamin.tissoires@redhat.com
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
btf_check_subprog_arg_match() was used twice in verifier.c:
- when checking for the type mismatches between a (sub)prog declaration
and BTF
- when checking the call of a subprog to see if the provided arguments
are correct and valid
This is problematic when we check if the first argument of a program
(pointer to ctx) is correctly accessed:
To be able to ensure we access a valid memory in the ctx, the verifier
assumes the pointer to context is not null.
This has the side effect of marking the program accessing the entire
context, even if the context is never dereferenced.
For example, by checking the context access with the current code, the
following eBPF program would fail with -EINVAL if the ctx is set to null
from the userspace:
```
SEC("syscall")
int prog(struct my_ctx *args) {
return 0;
}
```
In that particular case, we do not want to actually check that the memory
is correct while checking for the BTF validity, but we just want to
ensure that the (sub)prog definition matches the BTF we have.
So split btf_check_subprog_arg_match() in two so we can actually check
for the memory used when in a call, and ignore that part when not.
Note that a further patch is in preparation to disentangled
btf_check_func_arg_match() from these two purposes, and so right now we
just add a new hack around that by adding a boolean to this function.
Signed-off-by: Benjamin Tissoires <benjamin.tissoires@redhat.com>
Acked-by: Kumar Kartikeya Dwivedi <memxor@gmail.com>
Link: https://lore.kernel.org/r/20220906151303.2780789-3-benjamin.tissoires@redhat.com
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
selinux_audit_rule_update() has been renamed to audit_update_lsm_rules()
since commit d7a96f3a1a ("Audit: internally use the new LSM audit
hooks"), so remove it.
Signed-off-by: Xiu Jianfeng <xiujianfeng@huawei.com>
Signed-off-by: Paul Moore <paul@paul-moore.com>
Add __dyndbg_classes section, using __dyndbg as a model. Use it:
vmlinux.lds.h:
KEEP the new section, which also silences orphan section warning on
loadable modules. Add (__start_/__stop_)__dyndbg_classes linker
symbols for the c externs (below).
kernel/module/main.c:
- fill new fields in find_module_sections(), using section_objs()
- extend callchain prototypes
to pass classes, length
load_module(): pass new info to dynamic_debug_setup()
dynamic_debug_setup(): new params, pass through to ddebug_add_module()
dynamic_debug.c:
- add externs to the linker symbols.
ddebug_add_module():
- It currently builds a debug_table, and *will* find and attach classes.
dynamic_debug_init():
- add class fields to the _ddebug_info cursor var: di.
Signed-off-by: Jim Cromie <jim.cromie@gmail.com>
Link: https://lore.kernel.org/r/20220904214134.408619-16-jim.cromie@gmail.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
This new struct composes the linker provided (vector,len) section,
and provides a place to add other __dyndbg[] state-data later:
descs - the vector of descriptors in __dyndbg section.
num_descs - length of the data/section.
Use it, in several different ways, as follows:
In lib/dynamic_debug.c:
ddebug_add_module(): Alter params-list, replacing 2 args (array,index)
with a struct _ddebug_info * containing them both, with room for
expansion. This helps future-proof the function prototype against the
looming addition of class-map info into the dyndbg-state, by providing
a place to add more member fields later.
NB: later add static struct _ddebug_info builtins_state declaration,
not needed yet.
ddebug_add_module() is called in 2 contexts:
In dynamic_debug_init(), declare, init a struct _ddebug_info di
auto-var to use as a cursor. Then iterate over the prdbg blocks of
the builtin modules, and update the di cursor before calling
_add_module for each.
Its called from kernel/module/main.c:load_info() for each loaded
module:
In internal.h, alter struct load_info, replacing the dyndbg array,len
fields with an embedded _ddebug_info containing them both; and
populate its members in find_module_sections().
The 2 calling contexts differ in that _init deals with contiguous
subranges of __dyndbgs[] section, packed together, while loadable
modules are added one at a time.
So rename ddebug_add_module() into outer/__inner fns, call __inner
from _init, and provide the offset into the builtin __dyndbgs[] where
the module's prdbgs reside. The cursor provides start, len of the
subrange for each. The offset will be used later to pack the results
of builtin __dyndbg_sites[] de-duplication, and is 0 and unneeded for
loadable modules,
Note:
kernel/module/main.c includes <dynamic_debug.h> for struct
_ddeubg_info. This might be prone to include loops, since its also
included by printk.h. Nothing has broken in robot-land on this.
cc: Luis Chamberlain <mcgrof@kernel.org>
Signed-off-by: Jim Cromie <jim.cromie@gmail.com>
Link: https://lore.kernel.org/r/20220904214134.408619-12-jim.cromie@gmail.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The second operand passed to slot_addr() is declared as int or unsigned int
in all call sites. The left-shift to get the offset of a slot can overflow
if swiotlb size is larger than 4G.
Convert the macro to an inline function and declare the second argument as
phys_addr_t to avoid the potential overflow.
Fixes: 26a7e09478 ("swiotlb: refactor swiotlb_tbl_map_single")
Signed-off-by: Chao Gao <chao.gao@intel.com>
Reviewed-by: Dongli Zhang <dongli.zhang@oracle.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
When bucket_find_contains() tries to find the original entry for a
partial sync, it manages to constrain its search in a way that is both
too restrictive and not restrictive enough. A driver which only uses
single mappings rather than scatterlists might not set max_seg_size, but
could still technically perform a partial sync at an offset of more than
64KB into a sufficiently large mapping, so we could stop searching too
early before reaching a legitimate entry. Conversely, if no valid entry
is present and max_range is large enough, we can pointlessly search
buckets that we've already searched, or that represent an impossible
wrapping around the bottom of the address space. At worst, the
(legitimate) case of max_seg_size == UINT_MAX can make the loop
infinite.
Replace the fragile and frankly hard-to-follow "range" logic with a
simple counted loop for the number of possible hash buckets below the
given address.
Reported-by: Yunfei Wang <yf.wang@mediatek.com>
Signed-off-by: Robin Murphy <robin.murphy@arm.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Allow struct argument in trampoline based programs where
the struct size should be <= 16 bytes. In such cases, the argument
will be put into up to 2 registers for bpf, x86_64 and arm64
architectures.
To support arch-specific trampoline manipulation,
add arg_flags for additional struct information about arguments
in btf_func_model. Such information will be used in arch specific
function arch_prepare_bpf_trampoline() to prepare argument access
properly in trampoline.
Signed-off-by: Yonghong Song <yhs@fb.com>
Link: https://lore.kernel.org/r/20220831152646.2078089-1-yhs@fb.com
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
__ksize() was made private. Use ksize() instead.
Reported-by: Stephen Rothwell <sfr@canb.auug.org.au>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
