Pull scheduler updates from Ingo Molnar:
"The main changes in this cycle were:
- Optimized support for Intel "Cluster-on-Die" (CoD) topologies (Dave
Hansen)
- Various sched/idle refinements for better idle handling (Nicolas
Pitre, Daniel Lezcano, Chuansheng Liu, Vincent Guittot)
- sched/numa updates and optimizations (Rik van Riel)
- sysbench speedup (Vincent Guittot)
- capacity calculation cleanups/refactoring (Vincent Guittot)
- Various cleanups to thread group iteration (Oleg Nesterov)
- Double-rq-lock removal optimization and various refactorings
(Kirill Tkhai)
- various sched/deadline fixes
... and lots of other changes"
* 'sched-core-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip: (72 commits)
sched/dl: Use dl_bw_of() under rcu_read_lock_sched()
sched/fair: Delete resched_cpu() from idle_balance()
sched, time: Fix build error with 64 bit cputime_t on 32 bit systems
sched: Improve sysbench performance by fixing spurious active migration
sched/x86: Fix up typo in topology detection
x86, sched: Add new topology for multi-NUMA-node CPUs
sched/rt: Use resched_curr() in task_tick_rt()
sched: Use rq->rd in sched_setaffinity() under RCU read lock
sched: cleanup: Rename 'out_unlock' to 'out_free_new_mask'
sched: Use dl_bw_of() under RCU read lock
sched/fair: Remove duplicate code from can_migrate_task()
sched, mips, ia64: Remove __ARCH_WANT_UNLOCKED_CTXSW
sched: print_rq(): Don't use tasklist_lock
sched: normalize_rt_tasks(): Don't use _irqsave for tasklist_lock, use task_rq_lock()
sched: Fix the task-group check in tg_has_rt_tasks()
sched/fair: Leverage the idle state info when choosing the "idlest" cpu
sched: Let the scheduler see CPU idle states
sched/deadline: Fix inter- exclusive cpusets migrations
sched/deadline: Clear dl_entity params when setscheduling to different class
sched/numa: Kill the wrong/dead TASK_DEAD check in task_numa_fault()
...
Pull tracing fixes from Steven Rostedt:
"Seems that Peter Zijlstra added a new check that is making old code
scream nasty warnings:
WARNING: CPU: 0 PID: 91 at kernel/sched/core.c:7253 __might_sleep+0x9a/0x378()
do not call blocking ops when !TASK_RUNNING; state=1 set at [<ffffffff8d79b511>] event_test_thread+0x48/0x93
Call Trace:
__might_sleep+0x9a/0x378
down_read+0x26/0x98
exit_signals+0x27/0x1c2
do_exit+0x193/0x10bd
kthread+0x156/0x156
ret_from_fork+0x7a/0xb0
These are triggered by some self tests that run at start up when
configure in. Although the code is technically correct, they are a
little sloppy and not very robust. They work now because it runs at
boot up and the tests do not call anything that might trigger a
spurious wake up. But that doesn't mean those tests wont change in
the future.
It's best to clean them now to make sure the tests used to test the
internal workings of the system don't cause breakage themselves.
This also quiets the warnings made by the new checks"
* tag 'trace-3.18-2' of git://git.kernel.org/pub/scm/linux/kernel/git/rostedt/linux-trace:
tracing: Clean up scheduling in trace_wakeup_test_thread()
tracing: Robustify wait loop
Pull tracing updates from Steven Rostedt:
"This set has a few minor updates, but the big change is the redesign
of the trampoline logic.
The trampoline logic of 3.17 required a descriptor for every function
that is registered to be traced and uses a trampoline. Currently,
only the function graph tracer uses a trampoline, but if you were to
trace all 32,000 (give or take a few thousand) functions with the
function graph tracer, it would create 32,000 descriptors to let us
know that there's a trampoline associated with it. This takes up a
bit of memory when there's a better way to do it.
The redesign now reuses the ftrace_ops' (what registers the function
graph tracer) hash tables. The hash tables tell ftrace what the
tracer wants to trace or doesn't want to trace. There's two of them:
one that tells us what to trace, the other tells us what not to trace.
If the first one is empty, it means all functions should be traced,
otherwise only the ones that are listed should be. The second hash
table tells us what not to trace, and if it is empty, all functions
may be traced, and if there's any listed, then those should not be
traced even if they exist in the first hash table.
It took a bit of massaging, but now these hashes can be used to keep
track of what has a trampoline and what does not, and allows the
ftrace accounting to work. Now we can trace all functions when using
the function graph trampoline, and avoid needing to create any special
descriptors to hold all the functions that are being traced"
* tag 'trace-3.18' of git://git.kernel.org/pub/scm/linux/kernel/git/rostedt/linux-trace:
ftrace: Only disable ftrace_enabled to test buffer in selftest
ftrace: Add sanity check when unregistering last ftrace_ops
kernel: trace_syscalls: Replace rcu_assign_pointer() with RCU_INIT_POINTER()
tracing: generate RCU warnings even when tracepoints are disabled
ftrace: Replace tramp_hash with old_*_hash to save space
ftrace: Annotate the ops operation on update
ftrace: Grab any ops for a rec for enabled_functions output
ftrace: Remove freeing of old_hash from ftrace_hash_move()
ftrace: Set callback to ftrace_stub when no ops are registered
ftrace: Add helper function ftrace_ops_get_func()
ftrace: Add separate function for non recursive callbacks
Peter's new debugging tool triggers when tasks exit with !TASK_RUNNING.
The code in trace_wakeup_test_thread() also has a single schedule() call
that should be encompassed by a loop.
This cleans up the code a little to make it a bit more robust and
also makes the return exit properly with TASK_RUNNING.
Link: http://lkml.kernel.org/p/20141008135216.76142204@gandalf.local.home
Reported-by: Peter Zijlstra <peterz@infradead.org>
Acked-by: Peter Zijlstra <peterz@infreadead.org>
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
Commit 651e22f270 "ring-buffer: Always reset iterator to reader page"
fixed one bug but in the process caused another one. The reset is to
update the header page, but that fix also changed the way the cached
reads were updated. The cache reads are used to test if an iterator
needs to be updated or not.
A ring buffer iterator, when created, disables writes to the ring buffer
but does not stop other readers or consuming reads from happening.
Although all readers are synchronized via a lock, they are only
synchronized when in the ring buffer functions. Those functions may
be called by any number of readers. The iterator continues down when
its not interrupted by a consuming reader. If a consuming read
occurs, the iterator starts from the beginning of the buffer.
The way the iterator sees that a consuming read has happened since
its last read is by checking the reader "cache". The cache holds the
last counts of the read and the reader page itself.
Commit 651e22f270 changed what was saved by the cache_read when
the rb_iter_reset() occurred, making the iterator never match the cache.
Then if the iterator calls rb_iter_reset(), it will go into an
infinite loop by checking if the cache doesn't match, doing the reset
and retrying, just to see that the cache still doesn't match! Which
should never happen as the reset is suppose to set the cache to the
current value and there's locks that keep a consuming reader from
having access to the data.
Fixes: 651e22f270 "ring-buffer: Always reset iterator to reader page"
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
The ftrace_enabled variable is set to zero in the self tests to keep
delayed functions from being traced and messing with the checks. This
only needs to be done when the checks are being performed, otherwise,
if ftrace_enabled is off when calls back to the utility that is being
tested, it can cause errors to happen and the tests can fail with
false positives.
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
When the last ftrace_ops is unregistered, all the function records should
have a zeroed flags value. Make sure that is the case when the last ftrace_ops
is unregistered.
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
The uses of "rcu_assign_pointer()" are NULLing out the pointers.
According to RCU_INIT_POINTER()'s block comment:
"1. This use of RCU_INIT_POINTER() is NULLing out the pointer"
it is better to use it instead of rcu_assign_pointer() because it has a
smaller overhead.
The following Coccinelle semantic patch was used:
@@
@@
- rcu_assign_pointer
+ RCU_INIT_POINTER
(..., NULL)
Link: http://lkml.kernel.org/p/20140822142822.GA32391@ada
Signed-off-by: Andreea-Cristina Bernat <bernat.ada@gmail.com>
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
Allowing function callbacks to declare their own trampolines requires
that each ftrace_ops that has a trampoline must have some sort of
accounting that keeps track of which ops has a trampoline attached
to a record.
The easy way to solve this was to add a "tramp_hash" that created a
hash entry for every function that a ops uses with a trampoline.
But since we can have literally tens of thousands of functions being
traced, that means we need tens of thousands of descriptors to map
the ops to the function in the hash. This is quite expensive and
can cause enabling and disabling the function graph tracer to take
some time to start and stop. It can take up to several seconds to
disable or enable all functions in the function graph tracer for this
reason.
The better approach albeit more complex, is to keep track of how ops
are being enabled and disabled, and use that along with the counting
of the number of ops attached to records, to determive what ops has
a trampoline attached to a record at enabling and disabling of
tracing.
To do this, the tramp_hash has been replaced with an old_filter_hash
and old_notrace_hash, which get the copy of the ops filter_hash and
notrace_hash respectively. The old hashes is kept until the ops has
been modified or removed and the old hashes are used with the logic
of the accounting to determine the ops that have the trampoline of
a record. The reason this has less of a footprint is due to the trick
that an "empty" hash in the filter_hash means "all functions" and
an empty hash in the notrace hash means "no functions" in the hash.
This is much more efficienct, doesn't have the delay, and takes up
much less memory, as we do not need to map all the functions but
just figure out which functions are mapped at the time it is
enabled or disabled.
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
Add three new flags for ftrace_ops:
FTRACE_OPS_FL_ADDING
FTRACE_OPS_FL_REMOVING
FTRACE_OPS_FL_MODIFYING
These will be set for the ftrace_ops when they are first added
to the function tracing, being removed from function tracing
or just having their functions changed from function tracing,
respectively.
This will be needed to remove the tramp_hash, which can grow quite
big. The tramp_hash is used to note what functions a ftrace_ops
is using a trampoline for. Denoting which ftrace_ops is being
modified, will allow us to use the ftrace_ops hashes themselves,
which are much smaller as they have a global flag to denote if
a ftrace_ops is tracing all functions, as well as a notrace hash
if the ftrace_ops is tracing all but a few. The tramp_hash just
creates a hash item for every function, which can go into the 10s
of thousands if all functions are using the ftrace_ops trampoline.
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
When dumping the enabled_functions, use the first op that is
found with a trampoline to the record, as there should only be
one, as only one ops can be registered to a function that has
a trampoline.
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
ftrace_hash_move() currently frees the old hash that is passed to it
after replacing the pointer with the new hash. Instead of having the
function do that chore, have the caller perform the free.
This lets the ftrace_hash_move() be used a bit more freely, which
is needed for changing the way the trampoline logic is done.
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
The clean up that adds the helper function ftrace_ops_get_func()
caused the default function to not change when DYNAMIC_FTRACE was not
set and no ftrace_ops were registered. Although static tracing is
not very useful (not having DYNAMIC_FTRACE set), it is still supported
and we don't want to break it.
Clean up the if statement even more to specifically have the default
function call ftrace_stub when no ftrace_ops are registered. This
fixes the small bug for static tracing as well as makes the code a
bit more understandable.
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
Add the helper function to what the mcount trampoline is to call
for a ftrace_ops function. This helper will be used by arch code
in the future to set up dynamic trampolines. But as this does the
same tests that are performed in choosing what function to call for
the default mcount trampoline, might as well use it to clean up
the existing code.
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
Instead of using the generic list function for callbacks that
are not recursive, call a new helper function from the mcount
trampoline called ftrace_ops_recur_func() that will do the recursion
checking for the callback.
This eliminates an indirection as well as will help in future code
that will use dynamically allocated trampolines.
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
Epoll on trace_pipe can sometimes hang in a weird case. If the ring buffer is
empty when we set waiters_pending but an event shows up exactly at that moment
we can miss being woken up by the ring buffers irq work. Since
ring_buffer_empty() is inherently racey we will sometimes think that the buffer
is not empty. So we don't get woken up and we don't think there are any events
even though there were some ready when we added the watch, which makes us hang.
This patch fixes this by making sure that we are actually on the wait list
before we set waiters_pending, and add a memory barrier to make sure
ring_buffer_empty() is going to be correct.
Link: http://lkml.kernel.org/p/1408989581-23727-1-git-send-email-jbacik@fb.com
Cc: stable@vger.kernel.org # 3.10+
Cc: Martin Lau <kafai@fb.com>
Signed-off-by: Josef Bacik <jbacik@fb.com>
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
In __ftrace_replace_code(), when converting the call to a nop in a function
it needs to compare against the "curr" (current) value of the ftrace ops, and
not the "new" one. It currently does not affect x86 which is the only arch
to do the trampolines with function graph tracer, but when other archs that do
depend on this code implement the function graph trampoline, it can crash.
Here's an example when ARM uses the trampolines (in the future):
------------[ cut here ]------------
WARNING: CPU: 0 PID: 9 at kernel/trace/ftrace.c:1716 ftrace_bug+0x17c/0x1f4()
Modules linked in: omap_rng rng_core ipv6
CPU: 0 PID: 9 Comm: migration/0 Not tainted 3.16.0-test-10959-gf0094b28f303-dirty #52
[<c02188f4>] (unwind_backtrace) from [<c021343c>] (show_stack+0x20/0x24)
[<c021343c>] (show_stack) from [<c095a674>] (dump_stack+0x78/0x94)
[<c095a674>] (dump_stack) from [<c02532a0>] (warn_slowpath_common+0x7c/0x9c)
[<c02532a0>] (warn_slowpath_common) from [<c02532ec>] (warn_slowpath_null+0x2c/0x34)
[<c02532ec>] (warn_slowpath_null) from [<c02cbac4>] (ftrace_bug+0x17c/0x1f4)
[<c02cbac4>] (ftrace_bug) from [<c02cc44c>] (ftrace_replace_code+0x80/0x9c)
[<c02cc44c>] (ftrace_replace_code) from [<c02cc658>] (ftrace_modify_all_code+0xb8/0x164)
[<c02cc658>] (ftrace_modify_all_code) from [<c02cc718>] (__ftrace_modify_code+0x14/0x1c)
[<c02cc718>] (__ftrace_modify_code) from [<c02c7244>] (multi_cpu_stop+0xf4/0x134)
[<c02c7244>] (multi_cpu_stop) from [<c02c6e90>] (cpu_stopper_thread+0x54/0x130)
[<c02c6e90>] (cpu_stopper_thread) from [<c0271cd4>] (smpboot_thread_fn+0x1ac/0x1bc)
[<c0271cd4>] (smpboot_thread_fn) from [<c026ddf0>] (kthread+0xe0/0xfc)
[<c026ddf0>] (kthread) from [<c020f318>] (ret_from_fork+0x14/0x20)
---[ end trace dc9ce72c5b617d8f ]---
[ 65.047264] ftrace failed to modify [<c0208580>] asm_do_IRQ+0x10/0x1c
[ 65.054070] actual: 85:1b:00:eb
Fixes: 7413af1fb7 "ftrace: Make get_ftrace_addr() and get_ftrace_addr_old() global"
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
The latest rewrite of ftrace removed the separate ftrace_ops of
the function tracer and the function graph tracer and had them
share the same ftrace_ops. This simplified the accounting by removing
the multiple layers of functions called, where the global_ops func
would call a special list that would iterate over the other ops that
were registered within it (like function and function graph), which
itself was registered to the ftrace ops list of all functions
currently active. If that sounds confusing, the code that implemented
it was also confusing and its removal is a good thing.
The problem with this change was that it assumed that the function
and function graph tracer can never be used at the same time.
This is mostly true, but there is an exception. That is when the
function profiler uses the function graph tracer to profile.
The function profiler can be activated the same time as the function
tracer, and this breaks the assumption and the result is that ftrace
will crash (it detects the error and shuts itself down, it does not
cause a kernel oops).
To solve this issue, a previous change allowed the hash tables
for the functions traced by a ftrace_ops to be a pointer and let
multiple ftrace_ops share the same hash. This allows the function
and function_graph tracer to have separate ftrace_ops, but still
share the hash, which is what is done.
Now the function and function graph tracers have separate ftrace_ops
again, and the function tracer can be run while the function_profile
is active.
Cc: stable@vger.kernel.org # 3.16 (apply after 3.17-rc4 is out)
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
Now that a ftrace_hash can be shared by multiple ftrace_ops, they can dec
the rec->flags by more than once (one per those that share the ftrace_hash).
This means that the tramp_hash may not have a hash item when it was added.
For example, if two ftrace_ops share a hash for a ftrace record, and the
first ops has a trampoline, when it adds itself it will set the rec->flags
TRAMP flag and increments its nr_trampolines counter. When the second ops
is added, it must clear that tramp flag but also decrement the other ops
that shares its hash. As the update to the function callbacks has not yet
been performed, the other ops will not have the tramp hash set yet and it
can not be used to know to decrement its nr_trampolines.
Luckily, the tramp_hash does not need to be used. As the ftrace_mutex is
held, a ops with a trampoline to a record during an update of another ops
that shares the record will have its func_hash pointing to it. Since a
trampoline can only be set for a record if only one ops is attached to it,
we can just check if the record has a trampoline (the FTRACE_FL_TRAMP flag
is set) and then find the ops that has this record in its hashes.
Also added some output to help debug when things go wrong.
Cc: stable@vger.kernel.org # 3.16+ (apply after 3.17-rc4 is out)
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
When updating what an ftrace_ops traces, if it is registered (that is,
actively tracing), and that ftrace_ops uses the shared global_ops
local_hash, then we need to update all tracers that are active and
also share the global_ops' ftrace_hash_ops.
Cc: stable@vger.kernel.org # 3.16 (apply after 3.17-rc4 is out)
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
Currently the top level debug file system function tracer shares its
ftrace_ops with the function graph tracer. This was thought to be fine
because the tracers are not used together, as one can only enable
function or function_graph tracer in the current_tracer file.
But that assumption proved to be incorrect. The function profiler
can use the function graph tracer when function tracing is enabled.
Since all function graph users uses the function tracing ftrace_ops
this causes a conflict and when a user enables both function profiling
as well as the function tracer it will crash ftrace and disable it.
The quick solution so far is to move them as separate ftrace_ops like
it was earlier. The problem though is to synchronize the functions that
are traced because both function and function_graph tracer are limited
by the selections made in the set_ftrace_filter and set_ftrace_notrace
files.
To handle this, a new structure is made called ftrace_ops_hash. This
structure will now hold the filter_hash and notrace_hash, and the
ftrace_ops will point to this structure. That will allow two ftrace_ops
to share the same hashes.
Since most ftrace_ops do not share the hashes, and to keep allocation
simple, the ftrace_ops structure will include both a pointer to the
ftrace_ops_hash called func_hash, as well as the structure itself,
called local_hash. When the ops are registered, the func_hash pointer
will be initialized to point to the local_hash within the ftrace_ops
structure. Some of the ftrace internal ftrace_ops will be initialized
statically. This will allow for the function and function_graph tracer
to have separate ops but still share the same hash tables that determine
what functions they trace.
Cc: stable@vger.kernel.org # 3.16 (apply after 3.17-rc4 is out)
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
