Introduce a compile time flag to enable scheduler guidance of
frequency selection. This flag is also used to turn on or off
window-based load stats feature.
Having a compile time flag will let some platforms avoid any
overhead that may be present with this scheduler feature.
Change-Id: Id8dec9839f90dcac82f58ef7e2bd0ccd0b6bd16c
Signed-off-by: Srivatsa Vaddagiri <vatsa@codeaurora.org>
Following cleanups and improvements are made to window-based load
stats feature:
* Add sysctl to pick max, avg or most recent samples as task's
demand.
* Fix overflow possibility in calculation of sum for average policy.
* Use unscaled statistics when a task is running on a CPU which is
thermally throttled.
Change-Id: I8293565ca0c2a785dadf8adb6c67f579a445ed29
Signed-off-by: Srivatsa Vaddagiri <vatsa@codeaurora.org>
Some tasks can have a sporadic load pattern such that they can suddenly
start running for longer intervals of time after running for shorter
durations. To recognize such sharp increase in tasks' demands, max
between the average of 5 window load samples and the most recent sample
is chosen as the task demand.
Make the window size (sched_ravg_window) configurable at boot up
time. To prevent users from setting inappropriate values for window
size, min and max limits are defined. As 'ravg' struct tracks load for
both real-time and non real-time tasks it is moved out of sched_entity
struct.
In order to prevent changing function signatures for move_tasks() and
move_one_task() per-cpu variables are defined to track the total load
moved. In case multiple tasks are selected to migrate in one load
balance operation, loads > 100 could be sent through migration notifiers.
Prevent this scenario by setting mnd.load to 100 in such cases.
Define wrapper functions to compute cpu demands for tasks and to change
rq->cumulative_runnable_avg.
Change-Id: I9abfbf3b5fe23ae615a6acd3db9580cfdeb515b4
Signed-off-by: Srivatsa Vaddagiri <vatsa@codeaurora.org>
Signed-off-by: Rohit Gupta <rohgup@codeaurora.org>
Add a change to send notify_on_migrate hints on wakeups of
foreground tasks from scheduler if their load is above
wakeup_load_thresholds (default value is 60).
These hints can be used to choose an appropriate CPU frequency
corresponding to the load of the task being woken up.
Change-Id: Ieca413c1a8bd2b14a15a7591e8e15d22925c42ca
Signed-off-by: Rohit Gupta <rohgup@codeaurora.org>
Previously, on getting a migration notification cpu-boost changed
the scaling min of the destination frequency to match that of the
source frequency or sync_threshold whichever was minimum.
If the scheduler migration notification is extended with task load
(cpu demand) information, the cpu boost driver can use this load to
compute a suitable frequency for the migrating task. The required
frequency for the task is calculated by taking the load percentage
of the max frequency and no sync is performed if the load is less
than a particular value (migration_load_threshold).This change is
beneficial for both perf and power as demand of a task is taken into
consideration while making cpufreq decisions and unnecessary syncs
for lightweight tasks are avoided.
The task load information provided by scheduler comes from a
window-based load collection mechanism which also normalizes the
load collected by the scheduler to the max possible frequency
across all CPUs.
Change-Id: Id2ba91cc4139c90602557f9b3801fb06b3c38992
Signed-off-by: Rohit Gupta <rohgup@codeaurora.org>
Provide a metric per task that specifies how cpu bound a task is. Task
execution is monitored over several time windows and the fraction of
the window for which task was found to be executing or wanting to run
is recorded as task's demand. Windows over which task was sleeping are
ignored. We track last 5 recent windows for every task and the maximum
demand seen in any of the previous 5 windows (where task had some
activity) drives freq demand for every task.
A per-cpu metric (rq->cumulative_runnable_avg) is also provided which
is an aggregation of cpu demand of all tasks currently enqueued on it.
rq->cumulative_runnable_avg will be useful to know if cpu frequency
will need to be changed to match task demand.
Change-Id: Ib83207b9ba8683cd3304ee8a2290695c34f08fe2
Signed-off-by: Srivatsa Vaddagiri <vatsa@codeaurora.org>
While per-entity load-tracking is generally useful, beyond computing shares
distribution, e.g. runnable based load-balance (in progress), governors,
power-management, etc.
These facilities are not yet consumers of this data. This may be trivially
reverted when the information is required; but avoid paying the overhead for
calculations we will not use until then.
Change-Id: I459d52082af636d1181edb7acb3af973e95714f9
Signed-off-by: Paul Turner <pjt@google.com>
Reviewed-by: Ben Segall <bsegall@google.com>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Link: http://lkml.kernel.org/r/20120823141507.422162369@google.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Since we are now doing bottom up load accumulation we need explicit
notification when a task has been re-parented so that the old hierarchy can be
updated.
Adds: migrate_task_rq(struct task_struct *p, int next_cpu)
(The alternative is to do this out of __set_task_cpu, but it was suggested that
this would be a cleaner encapsulation.)
Change-Id: Icc4111cd4159f804186dd7c05f0c36f4e368288e
Signed-off-by: Paul Turner <pjt@google.com>
Reviewed-by: Ben Segall <bsegall@google.com>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Link: http://lkml.kernel.org/r/20120823141506.660023400@google.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
We are currently maintaining:
runnable_load(cfs_rq) = \Sum task_load(t)
For all running children t of cfs_rq. While this can be naturally updated for
tasks in a runnable state (as they are scheduled); this does not account for
the load contributed by blocked task entities.
This can be solved by introducing a separate accounting for blocked load:
blocked_load(cfs_rq) = \Sum runnable(b) * weight(b)
Obviously we do not want to iterate over all blocked entities to account for
their decay, we instead observe that:
runnable_load(t) = \Sum p_i*y^i
and that to account for an additional idle period we only need to compute:
y*runnable_load(t).
This means that we can compute all blocked entities at once by evaluating:
blocked_load(cfs_rq)` = y * blocked_load(cfs_rq)
Finally we maintain a decay counter so that when a sleeping entity re-awakens
we can determine how much of its load should be removed from the blocked sum.
Change-Id: Ib0b29b3d1ca872a13d7236ed142156ef0a35dc9a
Signed-off-by: Paul Turner <pjt@google.com>
Reviewed-by: Ben Segall <bsegall@google.com>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Link: http://lkml.kernel.org/r/20120823141506.585389902@google.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
For a given task t, we can compute its contribution to load as:
task_load(t) = runnable_avg(t) * weight(t)
On a parenting cfs_rq we can then aggregate:
runnable_load(cfs_rq) = \Sum task_load(t), for all runnable children t
Maintain this bottom up, with task entities adding their contributed load to
the parenting cfs_rq sum. When a task entity's load changes we add the same
delta to the maintained sum.
Change-Id: If3b74d345354a340836bbc356002982e25ab3244
Signed-off-by: Paul Turner <pjt@google.com>
Reviewed-by: Ben Segall <bsegall@google.com>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Link: http://lkml.kernel.org/r/20120823141506.514678907@google.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Instead of tracking averaging the load parented by a cfs_rq, we can track
entity load directly. With the load for a given cfs_rq then being the sum
of its children.
To do this we represent the historical contribution to runnable average
within each trailing 1024us of execution as the coefficients of a
geometric series.
We can express this for a given task t as:
runnable_sum(t) = \Sum u_i * y^i, runnable_avg_period(t) = \Sum 1024 * y^i
load(t) = weight_t * runnable_sum(t) / runnable_avg_period(t)
Where: u_i is the usage in the last i`th 1024us period (approximately 1ms)
~ms and y is chosen such that y^k = 1/2. We currently choose k to be 32 which
roughly translates to about a sched period.
Change-Id: Iafc090fe18ed1835cc501949e8ba1e4ed78c5de1
Signed-off-by: Paul Turner <pjt@google.com>
Reviewed-by: Ben Segall <bsegall@google.com>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Link: http://lkml.kernel.org/r/20120823141506.372695337@google.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
On no-so-small systems, it is possible for a single process to cause an
OOM condition by filling large pipes with data that are never read. A
typical process filling 4000 pipes with 1 MB of data will use 4 GB of
memory. On small systems it may be tricky to set the pipe max size to
prevent this from happening.
This patch makes it possible to enforce a per-user soft limit above
which new pipes will be limited to a single page, effectively limiting
them to 4 kB each, as well as a hard limit above which no new pipes may
be created for this user. This has the effect of protecting the system
against memory abuse without hurting other users, and still allowing
pipes to work correctly though with less data at once.
The limit are controlled by two new sysctls : pipe-user-pages-soft, and
pipe-user-pages-hard. Both may be disabled by setting them to zero. The
default soft limit allows the default number of FDs per process (1024)
to create pipes of the default size (64kB), thus reaching a limit of 64MB
before starting to create only smaller pipes. With 256 processes limited
to 1024 FDs each, this results in 1024*64kB + (256*1024 - 1024) * 4kB =
1084 MB of memory allocated for a user. The hard limit is disabled by
default to avoid breaking existing applications that make intensive use
of pipes (eg: for splicing).
Reported-by: socketpair@gmail.com
Reported-by: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp>
Mitigates: CVE-2013-4312 (Linux 2.0+)
Suggested-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Willy Tarreau <w@1wt.eu>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Conflicts:
Documentation/sysctl/fs.txt
fs/pipe.c
include/linux/sched.h
Change-Id: Ic7c678af18129943e16715fdaa64a97a7f0854be
commit 2ad654bc5e2b211e92f66da1d819e47d79a866f0 upstream.
When we change cpuset.memory_spread_{page,slab}, cpuset will flip
PF_SPREAD_{PAGE,SLAB} bit of tsk->flags for each task in that cpuset.
This should be done using atomic bitops, but currently we don't,
which is broken.
Tetsuo reported a hard-to-reproduce kernel crash on RHEL6, which happened
when one thread tried to clear PF_USED_MATH while at the same time another
thread tried to flip PF_SPREAD_PAGE/PF_SPREAD_SLAB. They both operate on
the same task.
Here's the full report:
https://lkml.org/lkml/2014/9/19/230
To fix this, we make PF_SPREAD_PAGE and PF_SPREAD_SLAB atomic flags.
v4:
- updated mm/slab.c. (Fengguang Wu)
- updated Documentation.
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Miao Xie <miaox@cn.fujitsu.com>
Cc: Kees Cook <keescook@chromium.org>
Fixes: 950592f7b9 ("cpusets: update tasks' page/slab spread flags in time")
Reported-by: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp>
Signed-off-by: Zefan Li <lizefan@huawei.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
[lizf: Backported to 3.4:
- adjust context
- check current->flags & PF_MEMPOLICY rather than current->mempolicy]
commit e0e5070b20e01f0321f97db4e4e174f3f6b49e50 upstream.
This will simplify code when we add new flags.
v3:
- Kees pointed out that no_new_privs should never be cleared, so we
shouldn't define task_clear_no_new_privs(). we define 3 macros instead
of a single one.
v2:
- updated scripts/tags.sh, suggested by Peter
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Miao Xie <miaox@cn.fujitsu.com>
Cc: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp>
Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Acked-by: Kees Cook <keescook@chromium.org>
Signed-off-by: Zefan Li <lizefan@huawei.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
[lizf: Backported to 3.4:
- adjust context
- remove no_new_priv code
- add atomic_flags to struct task_struct]
while_each_thread() and next_thread() should die, almost every lockless
usage is wrong.
1. Unless g == current, the lockless while_each_thread() is not safe.
while_each_thread(g, t) can loop forever if g exits, next_thread()
can't reach the unhashed thread in this case. Note that this can
happen even if g is the group leader, it can exec.
2. Even if while_each_thread() itself was correct, people often use
it wrongly.
It was never safe to just take rcu_read_lock() and loop unless
you verify that pid_alive(g) == T, even the first next_thread()
can point to the already freed/reused memory.
This patch adds signal_struct->thread_head and task->thread_node to
create the normal rcu-safe list with the stable head. The new
for_each_thread(g, t) helper is always safe under rcu_read_lock() as
long as this task_struct can't go away.
Note: of course it is ugly to have both task_struct->thread_node and the
old task_struct->thread_group, we will kill it later, after we change
the users of while_each_thread() to use for_each_thread().
Perhaps we can kill it even before we convert all users, we can
reimplement next_thread(t) using the new thread_head/thread_node. But
we can't do this right now because this will lead to subtle behavioural
changes. For example, do/while_each_thread() always sees at least one
task, while for_each_thread() can do nothing if the whole thread group
has died. Or thread_group_empty(), currently its semantics is not clear
unless thread_group_leader(p) and we need to audit the callers before we
can change it.
So this patch adds the new interface which has to coexist with the old
one for some time, hopefully the next changes will be more or less
straightforward and the old one will go away soon.
Signed-off-by: Oleg Nesterov <oleg@redhat.com>
Reviewed-by: Sergey Dyasly <dserrg@gmail.com>
Tested-by: Sergey Dyasly <dserrg@gmail.com>
Reviewed-by: Sameer Nanda <snanda@chromium.org>
Acked-by: David Rientjes <rientjes@google.com>
Cc: "Eric W. Biederman" <ebiederm@xmission.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Mandeep Singh Baines <msb@chromium.org>
Cc: "Ma, Xindong" <xindong.ma@intel.com>
Cc: Michal Hocko <mhocko@suse.cz>
Cc: "Tu, Xiaobing" <xiaobing.tu@intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Conflicts:
kernel/fork.c
Since seccomp transitions between threads requires updates to the
no_new_privs flag to be atomic, the flag must be part of an atomic flag
set. This moves the nnp flag into a separate task field, and introduces
accessors.
Signed-off-by: Kees Cook <keescook@chromium.org>
Reviewed-by: Oleg Nesterov <oleg@redhat.com>
Reviewed-by: Andy Lutomirski <luto@amacapital.net>
Conflicts:
fs/exec.c
include/linux/sched.h
kernel/sys.c
Replaces the seccomp_t typedef with struct seccomp to match modern
kernel style.
Signed-off-by: Will Drewry <wad@chromium.org>
Reviewed-by: James Morris <jmorris@namei.org>
Acked-by: Serge Hallyn <serge.hallyn@canonical.com>
Acked-by: Eric Paris <eparis@redhat.com>
v18: rebase
...
v14: rebase/nochanges
v13: rebase on to 88ebdda615
v12: rebase on to linux-next
v8-v11: no changes
v7: struct seccomp_struct -> struct seccomp
v6: original inclusion in this series.
With this change, calling
prctl(PR_SET_NO_NEW_PRIVS, 1, 0, 0, 0)
disables privilege granting operations at execve-time. For example, a
process will not be able to execute a setuid binary to change their uid
or gid if this bit is set. The same is true for file capabilities.
Additionally, LSM_UNSAFE_NO_NEW_PRIVS is defined to ensure that
LSMs respect the requested behavior.
To determine if the NO_NEW_PRIVS bit is set, a task may call
prctl(PR_GET_NO_NEW_PRIVS, 0, 0, 0, 0);
It returns 1 if set and 0 if it is not set. If any of the arguments are
non-zero, it will return -1 and set errno to -EINVAL.
(PR_SET_NO_NEW_PRIVS behaves similarly.)
This functionality is desired for the proposed seccomp filter patch
series. By using PR_SET_NO_NEW_PRIVS, it allows a task to modify the
system call behavior for itself and its child tasks without being
able to impact the behavior of a more privileged task.
Another potential use is making certain privileged operations
unprivileged. For example, chroot may be considered "safe" if it cannot
affect privileged tasks.
Note, this patch causes execve to fail when PR_SET_NO_NEW_PRIVS is
set and AppArmor is in use. It is fixed in a subsequent patch.
Signed-off-by: Andy Lutomirski <luto@amacapital.net>
Signed-off-by: Will Drewry <wad@chromium.org>
Acked-by: Eric Paris <eparis@redhat.com>
v18: updated change desc
v17: using new define values as per 3.4
Conflicts:
include/linux/prctl.h
kernel/sys.c
This reverts commit 616c310e83b872024271c915c1b9ab505b9efad9.
(Move PREEMPT_RCU preemption to switch_to() invocation).
Testing by Sasha Levin <levinsasha928@gmail.com> showed that this
can result in deadlock due to invoking the scheduler when one of
the runqueue locks is held. Because this commit was simply a
performance optimization, revert it.
CRs-fixed: 657837
Change-Id: Idc7a560cf2d1696d8cd7d24c99747cdca227bcc2
Reported-by: Sasha Levin <levinsasha928@gmail.com>
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Tested-by: Sasha Levin <levinsasha928@gmail.com>
Git-commit: cba6d0d64ee53772b285d0c0c288deefbeaf7775
Git-repo: git://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git
Signed-off-by: Ramesh Gupta Guntha <rggupt@codeaurora.org>
Currently, PREEMPT_RCU readers are enqueued upon entry to the scheduler.
This is inefficient because enqueuing is required only if there is a
context switch, and entry to the scheduler does not guarantee a context
switch.
The commit therefore moves the enqueuing to immediately precede the
call to switch_to() from the scheduler.
CRs-fixed: 657837
Change-Id: I7d824196483cafb82371b978f3d995a3ab1696a0
Signed-off-by: Paul E. McKenney <paul.mckenney@linaro.org>
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Tested-by: Linus Torvalds <torvalds@linux-foundation.org>
Git-commit: 616c310e83b872024271c915c1b9ab505b9efad9
Git-repo: git://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git
Signed-off-by: Ramesh Gupta Guntha <rggupt@codeaurora.org>
The wake-affine scheduler feature is currently always trying to pull
the wakee close to the waker. In theory this should be beneficial if
the waker's CPU caches hot data for the wakee, and it's also beneficial
in the extreme ping-pong high context switch rate case.
Testing shows it can benefit hackbench up to 15%.
However, the feature is somewhat blind, from which some workloads
such as pgbench suffer. It's also time-consuming algorithmically.
Testing shows it can damage pgbench up to 50% - far more than the
benefit it brings in the best case.
So wake-affine should be smarter and it should realize when to
stop its thankless effort at trying to find a suitable CPU to wake on.
This patch introduces 'wakee_flips', which will be increased each
time the task flips (switches) its wakee target.
So a high 'wakee_flips' value means the task has more than one
wakee, and the bigger the number, the higher the wakeup frequency.
Now when making the decision on whether to pull or not, pay attention to
the wakee with a high 'wakee_flips', pulling such a task may benefit
the wakee. Also imply that the waker will face cruel competition later,
it could be very cruel or very fast depends on the story behind
'wakee_flips', waker therefore suffers.
Furthermore, if waker also has a high 'wakee_flips', that implies that
multiple tasks rely on it, then waker's higher latency will damage all
of them, so pulling wakee seems to be a bad deal.
Thus, when 'waker->wakee_flips / wakee->wakee_flips' becomes
higher and higher, the cost of pulling seems to be worse and worse.
The patch therefore helps the wake-affine feature to stop its pulling
work when:
wakee->wakee_flips > factor &&
waker->wakee_flips > (factor * wakee->wakee_flips)
The 'factor' here is the number of CPUs in the current CPU's NUMA node,
so a bigger node will lead to more pulling since the trial becomes more
severe.
After applying the patch, pgbench shows up to 40% improvements and no regressions.
Tested with 12 cpu x86 server and tip 3.10.0-rc7.
The percentages in the final column highlight the areas with the biggest wins,
all other areas improved as well:
pgbench base smart
| db_size | clients | tps | | tps |
+---------+---------+-------+ +-------+
| 22 MB | 1 | 10598 | | 10796 |
| 22 MB | 2 | 21257 | | 21336 |
| 22 MB | 4 | 41386 | | 41622 |
| 22 MB | 8 | 51253 | | 57932 |
| 22 MB | 12 | 48570 | | 54000 |
| 22 MB | 16 | 46748 | | 55982 | +19.75%
| 22 MB | 24 | 44346 | | 55847 | +25.93%
| 22 MB | 32 | 43460 | | 54614 | +25.66%
| 7484 MB | 1 | 8951 | | 9193 |
| 7484 MB | 2 | 19233 | | 19240 |
| 7484 MB | 4 | 37239 | | 37302 |
| 7484 MB | 8 | 46087 | | 50018 |
| 7484 MB | 12 | 42054 | | 48763 |
| 7484 MB | 16 | 40765 | | 51633 | +26.66%
| 7484 MB | 24 | 37651 | | 52377 | +39.11%
| 7484 MB | 32 | 37056 | | 51108 | +37.92%
| 15 GB | 1 | 8845 | | 9104 |
| 15 GB | 2 | 19094 | | 19162 |
| 15 GB | 4 | 36979 | | 36983 |
| 15 GB | 8 | 46087 | | 49977 |
| 15 GB | 12 | 41901 | | 48591 |
| 15 GB | 16 | 40147 | | 50651 | +26.16%
| 15 GB | 24 | 37250 | | 52365 | +40.58%
| 15 GB | 32 | 36470 | | 50015 | +37.14%
Signed-off-by: Michael Wang <wangyun@linux.vnet.ibm.com>
Cc: Mike Galbraith <efault@gmx.de>
Signed-off-by: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/51D50057.9000809@linux.vnet.ibm.com
[ Improved the changelog. ]
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Signed-off-by: Paul Reioux <reioux@gmail.com>
Date Mon, 15 Apr 2013 10:37:58 -0400
The current mutex spinning code (with MUTEX_SPIN_ON_OWNER option turned
on) allow multiple tasks to spin on a single mutex concurrently. A
potential problem with the current approach is that when the mutex
becomes available, all the spinning tasks will try to acquire the
mutex more or less simultaneously. As a result, there will be a lot of
cacheline bouncing especially on systems with a large number of CPUs.
This patch tries to reduce this kind of contention by putting the
mutex spinners into a queue so that only the first one in the queue
will try to acquire the mutex. This will reduce contention and allow
all the tasks to move forward faster.
The queuing of mutex spinners is done using an MCS lock based
implementation which will further reduce contention on the mutex
cacheline than a similar ticket spinlock based implementation. This
patch will add a new field into the mutex data structure for holding
the MCS lock. This expands the mutex size by 8 bytes for 64-bit system
and 4 bytes for 32-bit system. This overhead will be avoid if the
MUTEX_SPIN_ON_OWNER option is turned off.
The following table shows the jobs per minute (JPM) scalability data
on an 8-node 80-core Westmere box with a 3.7.10 kernel. The numactl
command is used to restrict the running of the fserver workloads to
1/2/4/8 nodes with hyperthreading off.
+-----------------+-----------+-----------+-------------+----------+
| Configuration | Mean JPM | Mean JPM | Mean JPM | % Change |
| | w/o patch | patch 1 | patches 1&2 | 1->1&2 |
+-----------------+------------------------------------------------+
| | User Range 1100 - 2000 |
+-----------------+------------------------------------------------+
| 8 nodes, HT off | 227972 | 227237 | 305043 | +34.2% |
| 4 nodes, HT off | 393503 | 381558 | 394650 | +3.4% |
| 2 nodes, HT off | 334957 | 325240 | 338853 | +4.2% |
| 1 node , HT off | 198141 | 197972 | 198075 | +0.1% |
+-----------------+------------------------------------------------+
| | User Range 200 - 1000 |
+-----------------+------------------------------------------------+
| 8 nodes, HT off | 282325 | 312870 | 332185 | +6.2% |
| 4 nodes, HT off | 390698 | 378279 | 393419 | +4.0% |
| 2 nodes, HT off | 336986 | 326543 | 340260 | +4.2% |
| 1 node , HT off | 197588 | 197622 | 197582 | 0.0% |
+-----------------+-----------+-----------+-------------+----------+
At low user range 10-100, the JPM differences were within +/-1%. So
they are not that interesting.
The fserver workload uses mutex spinning extensively. With just
the mutex change in the first patch, there is no noticeable change
in performance. Rather, there is a slight drop in performance. This
mutex spinning patch more than recovers the lost performance and show
a significant increase of +30% at high user load with the full 8 nodes.
Similar improvements were also seen in a 3.8 kernel.
The table below shows the %time spent by different kernel functions
as reported by perf when running the fserver workload at 1500 users
with all 8 nodes.
+-----------------------+-----------+---------+-------------+
| Function | % time | % time | % time |
| | w/o patch | patch 1 | patches 1&2 |
+-----------------------+-----------+---------+-------------+
| __read_lock_failed | 34.96% | 34.91% | 29.14% |
| __write_lock_failed | 10.14% | 10.68% | 7.51% |
| mutex_spin_on_owner | 3.62% | 3.42% | 2.33% |
| mspin_lock | N/A | N/A | 9.90% |
| __mutex_lock_slowpath | 1.46% | 0.81% | 0.14% |
| _raw_spin_lock | 2.25% | 2.50% | 1.10% |
+-----------------------+-----------+---------+-------------+
The fserver workload for an 8-node system is dominated by the
contention in the read/write lock. Mutex contention also plays a
role. With the first patch only, mutex contention is down (as shown by
the __mutex_lock_slowpath figure) which help a little bit. We saw only
a few percents improvement with that.
By applying patch 2 as well, the single mutex_spin_on_owner figure is
now split out into an additional mspin_lock figure. The time increases
from 3.42% to 11.23%. It shows a great reduction in contention among
the spinners leading to a 30% improvement. The time ratio 9.9/2.33=4.3
indicates that there are on average 4+ spinners waiting in the spin_lock
loop for each spinner in the mutex_spin_on_owner loop. Contention in
other locking functions also go down by quite a lot.
The table below shows the performance change of both patches 1 & 2 over
patch 1 alone in other AIM7 workloads (at 8 nodes, hyperthreading off).
+--------------+---------------+----------------+-----------------+
| Workload | mean % change | mean % change | mean % change |
| | 10-100 users | 200-1000 users | 1100-2000 users |
+--------------+---------------+----------------+-----------------+
| alltests | 0.0% | -0.8% | +0.6% |
| five_sec | -0.3% | +0.8% | +0.8% |
| high_systime | +0.4% | +2.4% | +2.1% |
| new_fserver | +0.1% | +14.1% | +34.2% |
| shared | -0.5% | -0.3% | -0.4% |
| short | -1.7% | -9.8% | -8.3% |
+--------------+---------------+----------------+-----------------+
The short workload is the only one that shows a decline in performance
probably due to the spinner locking and queuing overhead.
Signed-off-by: Waiman Long <Waiman.Long@hp.com>
The smpboot threads rely on the park/unpark mechanism which binds per
cpu threads on a particular core. Though the functionality is racy:
CPU0 CPU1 CPU2
unpark(T) wake_up_process(T)
clear(SHOULD_PARK) T runs
leave parkme() due to !SHOULD_PARK
bind_to(CPU2) BUG_ON(wrong CPU)
We cannot let the tasks move themself to the target CPU as one of
those tasks is actually the migration thread itself, which requires
that it starts running on the target cpu right away.
The solution to this problem is to prevent wakeups in park mode which
are not from unpark(). That way we can guarantee that the association
of the task to the target cpu is working correctly.
Add a new task state (TASK_PARKED) which prevents other wakeups and
use this state explicitly for the unpark wakeup.
Peter noticed: Also, since the task state is visible to userspace and
all the parked tasks are still in the PID space, its a good hint in ps
and friends that these tasks aren't really there for the moment.
The migration thread has another related issue.
CPU0 CPU1
Bring up CPU2
create_thread(T)
park(T)
wait_for_completion()
parkme()
complete()
sched_set_stop_task()
schedule(TASK_PARKED)
The sched_set_stop_task() call is issued while the task is on the
runqueue of CPU1 and that confuses the hell out of the stop_task class
on that cpu. So we need the same synchronizaion before
sched_set_stop_task().
Reported-by: Dave Jones <davej@redhat.com>
Reported-and-tested-by: Dave Hansen <dave@sr71.net>
Reported-and-tested-by: Borislav Petkov <bp@alien8.de>
Acked-by: Peter Ziljstra <peterz@infradead.org>
Cc: Srivatsa S. Bhat <srivatsa.bhat@linux.vnet.ibm.com>
Cc: dhillf@gmail.com
Cc: Ingo Molnar <mingo@kernel.org>
Cc: stable@vger.kernel.org
Link: http://lkml.kernel.org/r/alpine.LFD.2.02.1304091635430.21884@ionos
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Change-Id: If1e9993951c4ad1f6f35ad0698f6ccd05a67e81f
On systems where CPUs may run asynchronously, task migrations
between CPUs running at grossly different speeds can cause
problems.
This change provides a mechanism to notify a subsystem
in the kernel if a task in a particular cgroup migrates to a
different CPU. Other subsystems (such as cpufreq) may then
register for this notifier to take appropriate action when
such a task is migrated.
The cgroup attribute to set for this behavior is
"notify_on_migrate" .
Change-Id: Ie1868249e53ef901b89c837fdc33b0ad0c0a4590
Signed-off-by: Steve Muckle <smuckle@codeaurora.org>
Signed-off-by: Mekala Natarajan <mekalan@codeaurora.org>
commit 5aaa0b7a2ed5b12692c9ffb5222182bd558d3146 upstream.
Follow up on commit 556061b00 ("sched/nohz: Fix rq->cpu_load[]
calculations") since while that fixed the busy case it regressed the
mostly idle case.
Add a callback from the nohz exit to also age the rq->cpu_load[]
array. This closes the hole where either there was no nohz load
balance pass during the nohz, or there was a 'significant' amount of
idle time between the last nohz balance and the nohz exit.
So we'll update unconditionally from the tick to not insert any
accidental 0 load periods while busy, and we try and catch up from
nohz idle balance and nohz exit. Both these are still prone to missing
a jiffy, but that has always been the case.
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: pjt@google.com
Cc: Venkatesh Pallipadi <venki@google.com>
Link: http://lkml.kernel.org/n/tip-kt0trz0apodbf84ucjfdbr1a@git.kernel.org
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Cc: Li Zefan <lizefan@huawei.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit 57d2aa00dcec67afa52478730f2b524521af14fb upstream.
The issue below was found in 2.6.34-rt rather than mainline rt
kernel, but the issue still exists upstream as well.
So please let me describe how it was noticed on 2.6.34-rt:
On this version, each softirq has its own thread, it means there
is at least one RT FIFO task per cpu. The priority of these
tasks is set to 49 by default. If user launches an RT FIFO task
with priority lower than 49 of softirq RT tasks, it's possible
there are two RT FIFO tasks enqueued one cpu runqueue at one
moment. By current strategy of balancing RT tasks, when it comes
to RT tasks, we really need to put them off to a CPU that they
can run on as soon as possible. Even if it means a bit of cache
line flushing, we want RT tasks to be run with the least latency.
When the user RT FIFO task which just launched before is
running, the sched timer tick of the current cpu happens. In this
tick period, the timeout value of the user RT task will be
updated once. Subsequently, we try to wake up one softirq RT
task on its local cpu. As the priority of current user RT task
is lower than the softirq RT task, the current task will be
preempted by the higher priority softirq RT task. Before
preemption, we check to see if current can readily move to a
different cpu. If so, we will reschedule to allow the RT push logic
to try to move current somewhere else. Whenever the woken
softirq RT task runs, it first tries to migrate the user FIFO RT
task over to a cpu that is running a task of lesser priority. If
migration is done, it will send a reschedule request to the found
cpu by IPI interrupt. Once the target cpu responds the IPI
interrupt, it will pick the migrated user RT task to preempt its
current task. When the user RT task is running on the new cpu,
the sched timer tick of the cpu fires. So it will tick the user
RT task again. This also means the RT task timeout value will be
updated again. As the migration may be done in one tick period,
it means the user RT task timeout value will be updated twice
within one tick.
If we set a limit on the amount of cpu time for the user RT task
by setrlimit(RLIMIT_RTTIME), the SIGXCPU signal should be posted
upon reaching the soft limit.
But exactly when the SIGXCPU signal should be sent depends on the
RT task timeout value. In fact the timeout mechanism of sending
the SIGXCPU signal assumes the RT task timeout is increased once
every tick.
However, currently the timeout value may be added twice per
tick. So it results in the SIGXCPU signal being sent earlier
than expected.
To solve this issue, we prevent the timeout value from increasing
twice within one tick time by remembering the jiffies value of
last updating the timeout. As long as the RT task's jiffies is
different with the global jiffies value, we allow its timeout to
be updated.
Signed-off-by: Ying Xue <ying.xue@windriver.com>
Signed-off-by: Fan Du <fan.du@windriver.com>
Reviewed-by: Yong Zhang <yong.zhang0@gmail.com>
Acked-by: Steven Rostedt <rostedt@goodmis.org>
Cc: <peterz@infradead.org>
Link: http://lkml.kernel.org/r/1342508623-2887-1-git-send-email-ying.xue@windriver.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
[ lizf: backported to 3.4: adjust context ]
Signed-off-by: Li Zefan <lizefan@huawei.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit d049f74f2dbe71354d43d393ac3a188947811348 upstream.
The get_dumpable() return value is not boolean. Most users of the
function actually want to be testing for non-SUID_DUMP_USER(1) rather than
SUID_DUMP_DISABLE(0). The SUID_DUMP_ROOT(2) is also considered a
protected state. Almost all places did this correctly, excepting the two
places fixed in this patch.
Wrong logic:
if (dumpable == SUID_DUMP_DISABLE) { /* be protective */ }
or
if (dumpable == 0) { /* be protective */ }
or
if (!dumpable) { /* be protective */ }
Correct logic:
if (dumpable != SUID_DUMP_USER) { /* be protective */ }
or
if (dumpable != 1) { /* be protective */ }
Without this patch, if the system had set the sysctl fs/suid_dumpable=2, a
user was able to ptrace attach to processes that had dropped privileges to
that user. (This may have been partially mitigated if Yama was enabled.)
The macros have been moved into the file that declares get/set_dumpable(),
which means things like the ia64 code can see them too.
CVE-2013-2929
Reported-by: Vasily Kulikov <segoon@openwall.com>
Signed-off-by: Kees Cook <keescook@chromium.org>
Cc: "Luck, Tony" <tony.luck@intel.com>
Cc: Oleg Nesterov <oleg@redhat.com>
Cc: "Eric W. Biederman" <ebiederm@xmission.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit e56fb2874015370e3b7f8d85051f6dce26051df9 upstream.
threadgroup_lock() takes signal->cred_guard_mutex to ensure that
thread_group_leader() is stable. This doesn't look nice, the scope of
this lock in do_execve() is huge.
And as Dave pointed out this can lead to deadlock, we have the
following dependencies:
do_execve: cred_guard_mutex -> i_mutex
cgroup_mount: i_mutex -> cgroup_mutex
attach_task_by_pid: cgroup_mutex -> cred_guard_mutex
Change de_thread() to take threadgroup_change_begin() around the
switch-the-leader code and change threadgroup_lock() to avoid
->cred_guard_mutex.
Note that de_thread() can't sleep with ->group_rwsem held, this can
obviously deadlock with the exiting leader if the writer is active, so it
does threadgroup_change_end() before schedule().
Reported-by: Dave Jones <davej@redhat.com>
Acked-by: Tejun Heo <tj@kernel.org>
Acked-by: Li Zefan <lizefan@huawei.com>
Signed-off-by: Oleg Nesterov <oleg@redhat.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
[ zhj: adjust context ]
Signed-off-by: Zhao Hongjiang <zhaohongjiang@huawei.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
A killed task can stay in the task list long after its
memory has been returned to the system, therefore
ignore any tasks whose mm struct has been freed.
Change-Id: I76394b203b4ab2312437c839976f0ecb7b6dde4e
CRs-fixed: 450383
Signed-off-by: Liam Mark <lmark@codeaurora.org>
commit 910ffdb18a6408e14febbb6e4b6840fd2c928c82 upstream.
Cleanup and preparation for the next change.
signal_wake_up(resume => true) is overused. None of ptrace/jctl callers
actually want to wakeup a TASK_WAKEKILL task, but they can't specify the
necessary mask.
Turn signal_wake_up() into signal_wake_up_state(state), reintroduce
signal_wake_up() as a trivial helper, and add ptrace_signal_wake_up()
which adds __TASK_TRACED.
This way ptrace_signal_wake_up() can work "inside" ptrace_request()
even if the tracee doesn't have the TASK_WAKEKILL bit set.
Signed-off-by: Oleg Nesterov <oleg@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The PF_WAKE_UP_IDLE per-task flag made it impossible to enable
the old behavior of SD_SHARE_PKG_RESOURCES, where every task
migrates to an idle CPU on wakeup.
The sched_wake_to_idle sysctl value, when made nonzero, will cause
all tasks to migrate to an idle CPU if one is available when the
task is woken up. This is regardless of how PF_WAKE_UP_IDLE is
configured for tasks in the system. Similar to PF_WAKE_UP_IDLE,
the SD_SHARE_PKG_RESOURCES scheduler domain flag must be enabled
for the sysctl value to have an effect.
Change-Id: I23bed846d26502c7aed600bfcf1c13053a7e5f61
Signed-off-by: Steve Muckle <smuckle@codeaurora.org>
(cherry picked from commit 9d5b38dc0025d19df5b756b16024b4269e73f282)
Certain workloads may benefit from the SD_SHARE_PKG_RESOURCES behavior
of waking their tasks up on idle CPUs. The feature has too much of a
negative impact on other workloads however to apply globally. The
PF_WAKE_UP_IDLE flag tells the scheduler to wake up tasks that have this
flag set, or tasks woken by tasks with this flag set, on an idle CPU
if one is available.
Signed-off-by: Steve Muckle <smuckle@codeaurora.org>
(cherry picked from commit 500988016c53b675eae02401bd97462fed1737d6)
Change-Id: I6996afacecaec2959b84794eb032fe0a1ef2b3d7
Signed-off-by: Sudhir Sharma <sudsha@codeaurora.org>
Add code to calculate the run queue depth of a cpu and iowait
depth of the cpu.
The scheduler calls in to sched_update_nr_prod whenever there
is a runqueue change. This function maintains the runqueue average
and the iowait of that cpu in that time interval.
Whoever wants to know the runqueue average is expected to call
sched_get_nr_running_avg periodically to get the accumulated
runqueue and iowait averages for all the cpus.
Signed-off-by: Jeff Ohlstein <johlstei@codeaurora.org>
(cherry picked from commit 0299fcaaad80e2c0ac9aa583c95107f6edc27750)
Signed-off-by: Ram Kumar Chakravarthy Chebathini <rcheba@codeaurora.org>
(cherry picked from commit 5325d8a44f8296170cd8f42e245dd79f7b7347bf)
Change-Id: I92875789886c3f7c2457a063de4b3b6d8c893718
Signed-off-by: Sudhir Sharma <sudsha@codeaurora.org>
commit 8323f26ce3425460769605a6aece7a174edaa7d1 upstream.
Stefan reported a crash on a kernel before a3e5d1091c ("sched:
Don't call task_group() too many times in set_task_rq()"), he
found the reason to be that the multiple task_group()
invocations in set_task_rq() returned different values.
Looking at all that I found a lack of serialization and plain
wrong comments.
The below tries to fix it using an extra pointer which is
updated under the appropriate scheduler locks. Its not pretty,
but I can't really see another way given how all the cgroup
stuff works.
Reported-and-tested-by: Stefan Bader <stefan.bader@canonical.com>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Link: http://lkml.kernel.org/r/1340364965.18025.71.camel@twins
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit 5167e8d5417bf5c322a703d2927daec727ea40dd upstream.
Thanks to Charles Wang for spotting the defects in the current code:
- If we go idle during the sample window -- after sampling, we get a
negative bias because we can negate our own sample.
- If we wake up during the sample window we get a positive bias
because we push the sample to a known active period.
So rewrite the entire nohz load-avg muck once again, now adding
copious documentation to the code.
Reported-and-tested-by: Doug Smythies <dsmythies@telus.net>
Reported-and-tested-by: Charles Wang <muming.wq@gmail.com>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Andrew Morton <akpm@linux-foundation.org>
Link: http://lkml.kernel.org/r/1340373782.18025.74.camel@twins
[ minor edits ]
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
This patch adds a notifier which can be used by subsystems that may
be interested in when a task has completely died and is about to
have it's last resource freed.
The Android lowmemory killer uses this to determine when a task
it has killed has finally given up its goods.
Signed-off-by: San Mehat <san@google.com>
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.12 (GNU/Linux)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=G9mT
-----END PGP SIGNATURE-----
Merge tag 'split-asm_system_h-for-linus-20120328' of git://git.kernel.org/pub/scm/linux/kernel/git/dhowells/linux-asm_system
Pull "Disintegrate and delete asm/system.h" from David Howells:
"Here are a bunch of patches to disintegrate asm/system.h into a set of
separate bits to relieve the problem of circular inclusion
dependencies.
I've built all the working defconfigs from all the arches that I can
and made sure that they don't break.
The reason for these patches is that I recently encountered a circular
dependency problem that came about when I produced some patches to
optimise get_order() by rewriting it to use ilog2().
This uses bitops - and on the SH arch asm/bitops.h drags in
asm-generic/get_order.h by a circuituous route involving asm/system.h.
The main difficulty seems to be asm/system.h. It holds a number of
low level bits with no/few dependencies that are commonly used (eg.
memory barriers) and a number of bits with more dependencies that
aren't used in many places (eg. switch_to()).
These patches break asm/system.h up into the following core pieces:
(1) asm/barrier.h
Move memory barriers here. This already done for MIPS and Alpha.
(2) asm/switch_to.h
Move switch_to() and related stuff here.
(3) asm/exec.h
Move arch_align_stack() here. Other process execution related bits
could perhaps go here from asm/processor.h.
(4) asm/cmpxchg.h
Move xchg() and cmpxchg() here as they're full word atomic ops and
frequently used by atomic_xchg() and atomic_cmpxchg().
(5) asm/bug.h
Move die() and related bits.
(6) asm/auxvec.h
Move AT_VECTOR_SIZE_ARCH here.
Other arch headers are created as needed on a per-arch basis."
Fixed up some conflicts from other header file cleanups and moving code
around that has happened in the meantime, so David's testing is somewhat
weakened by that. We'll find out anything that got broken and fix it..
* tag 'split-asm_system_h-for-linus-20120328' of git://git.kernel.org/pub/scm/linux/kernel/git/dhowells/linux-asm_system: (38 commits)
Delete all instances of asm/system.h
Remove all #inclusions of asm/system.h
Add #includes needed to permit the removal of asm/system.h
Move all declarations of free_initmem() to linux/mm.h
Disintegrate asm/system.h for OpenRISC
Split arch_align_stack() out from asm-generic/system.h
Split the switch_to() wrapper out of asm-generic/system.h
Move the asm-generic/system.h xchg() implementation to asm-generic/cmpxchg.h
Create asm-generic/barrier.h
Make asm-generic/cmpxchg.h #include asm-generic/cmpxchg-local.h
Disintegrate asm/system.h for Xtensa
Disintegrate asm/system.h for Unicore32 [based on ver #3, changed by gxt]
Disintegrate asm/system.h for Tile
Disintegrate asm/system.h for Sparc
Disintegrate asm/system.h for SH
Disintegrate asm/system.h for Score
Disintegrate asm/system.h for S390
Disintegrate asm/system.h for PowerPC
Disintegrate asm/system.h for PA-RISC
Disintegrate asm/system.h for MN10300
...
Remove all #inclusions of asm/system.h preparatory to splitting and killing
it. Performed with the following command:
perl -p -i -e 's!^#\s*include\s*<asm/system[.]h>.*\n!!' `grep -Irl '^#\s*include\s*<asm/system[.]h>' *`
Signed-off-by: David Howells <dhowells@redhat.com>
Userspace service managers/supervisors need to track their started
services. Many services daemonize by double-forking and get implicitly
re-parented to PID 1. The service manager will no longer be able to
receive the SIGCHLD signals for them, and is no longer in charge of
reaping the children with wait(). All information about the children is
lost at the moment PID 1 cleans up the re-parented processes.
With this prctl, a service manager process can mark itself as a sort of
'sub-init', able to stay as the parent for all orphaned processes
created by the started services. All SIGCHLD signals will be delivered
to the service manager.
Receiving SIGCHLD and doing wait() is in cases of a service-manager much
preferred over any possible asynchronous notification about specific
PIDs, because the service manager has full access to the child process
data in /proc and the PID can not be re-used until the wait(), the
service-manager itself is in charge of, has happened.
As a side effect, the relevant parent PID information does not get lost
by a double-fork, which results in a more elaborate process tree and
'ps' output:
before:
# ps afx
253 ? Ss 0:00 /bin/dbus-daemon --system --nofork
294 ? Sl 0:00 /usr/libexec/polkit-1/polkitd
328 ? S 0:00 /usr/sbin/modem-manager
608 ? Sl 0:00 /usr/libexec/colord
658 ? Sl 0:00 /usr/libexec/upowerd
819 ? Sl 0:00 /usr/libexec/imsettings-daemon
916 ? Sl 0:00 /usr/libexec/udisks-daemon
917 ? S 0:00 \_ udisks-daemon: not polling any devices
after:
# ps afx
294 ? Ss 0:00 /bin/dbus-daemon --system --nofork
426 ? Sl 0:00 \_ /usr/libexec/polkit-1/polkitd
449 ? S 0:00 \_ /usr/sbin/modem-manager
635 ? Sl 0:00 \_ /usr/libexec/colord
705 ? Sl 0:00 \_ /usr/libexec/upowerd
959 ? Sl 0:00 \_ /usr/libexec/udisks-daemon
960 ? S 0:00 | \_ udisks-daemon: not polling any devices
977 ? Sl 0:00 \_ /usr/libexec/packagekitd
This prctl is orthogonal to PID namespaces. PID namespaces are isolated
from each other, while a service management process usually requires the
services to live in the same namespace, to be able to talk to each
other.
Users of this will be the systemd per-user instance, which provides
init-like functionality for the user's login session and D-Bus, which
activates bus services on-demand. Both need init-like capabilities to
be able to properly keep track of the services they start.
Many thanks to Oleg for several rounds of review and insights.
[akpm@linux-foundation.org: fix comment layout and spelling]
[akpm@linux-foundation.org: add lengthy code comment from Oleg]
Reviewed-by: Oleg Nesterov <oleg@redhat.com>
Signed-off-by: Lennart Poettering <lennart@poettering.net>
Signed-off-by: Kay Sievers <kay.sievers@vrfy.org>
Acked-by: Valdis Kletnieks <Valdis.Kletnieks@vt.edu>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Commit c0ff7453bb ("cpuset,mm: fix no node to alloc memory when
changing cpuset's mems") wins a super prize for the largest number of
memory barriers entered into fast paths for one commit.
[get|put]_mems_allowed is incredibly heavy with pairs of full memory
barriers inserted into a number of hot paths. This was detected while
investigating at large page allocator slowdown introduced some time
after 2.6.32. The largest portion of this overhead was shown by
oprofile to be at an mfence introduced by this commit into the page
allocator hot path.
For extra style points, the commit introduced the use of yield() in an
implementation of what looks like a spinning mutex.
This patch replaces the full memory barriers on both read and write
sides with a sequence counter with just read barriers on the fast path
side. This is much cheaper on some architectures, including x86. The
main bulk of the patch is the retry logic if the nodemask changes in a
manner that can cause a false failure.
While updating the nodemask, a check is made to see if a false failure
is a risk. If it is, the sequence number gets bumped and parallel
allocators will briefly stall while the nodemask update takes place.
In a page fault test microbenchmark, oprofile samples from
__alloc_pages_nodemask went from 4.53% of all samples to 1.15%. The
actual results were
3.3.0-rc3 3.3.0-rc3
rc3-vanilla nobarrier-v2r1
Clients 1 UserTime 0.07 ( 0.00%) 0.08 (-14.19%)
Clients 2 UserTime 0.07 ( 0.00%) 0.07 ( 2.72%)
Clients 4 UserTime 0.08 ( 0.00%) 0.07 ( 3.29%)
Clients 1 SysTime 0.70 ( 0.00%) 0.65 ( 6.65%)
Clients 2 SysTime 0.85 ( 0.00%) 0.82 ( 3.65%)
Clients 4 SysTime 1.41 ( 0.00%) 1.41 ( 0.32%)
Clients 1 WallTime 0.77 ( 0.00%) 0.74 ( 4.19%)
Clients 2 WallTime 0.47 ( 0.00%) 0.45 ( 3.73%)
Clients 4 WallTime 0.38 ( 0.00%) 0.37 ( 1.58%)
Clients 1 Flt/sec/cpu 497620.28 ( 0.00%) 520294.53 ( 4.56%)
Clients 2 Flt/sec/cpu 414639.05 ( 0.00%) 429882.01 ( 3.68%)
Clients 4 Flt/sec/cpu 257959.16 ( 0.00%) 258761.48 ( 0.31%)
Clients 1 Flt/sec 495161.39 ( 0.00%) 517292.87 ( 4.47%)
Clients 2 Flt/sec 820325.95 ( 0.00%) 850289.77 ( 3.65%)
Clients 4 Flt/sec 1020068.93 ( 0.00%) 1022674.06 ( 0.26%)
MMTests Statistics: duration
Sys Time Running Test (seconds) 135.68 132.17
User+Sys Time Running Test (seconds) 164.2 160.13
Total Elapsed Time (seconds) 123.46 120.87
The overall improvement is small but the System CPU time is much
improved and roughly in correlation to what oprofile reported (these
performance figures are without profiling so skew is expected). The
actual number of page faults is noticeably improved.
For benchmarks like kernel builds, the overall benefit is marginal but
the system CPU time is slightly reduced.
To test the actual bug the commit fixed I opened two terminals. The
first ran within a cpuset and continually ran a small program that
faulted 100M of anonymous data. In a second window, the nodemask of the
cpuset was continually randomised in a loop.
Without the commit, the program would fail every so often (usually
within 10 seconds) and obviously with the commit everything worked fine.
With this patch applied, it also worked fine so the fix should be
functionally equivalent.
Signed-off-by: Mel Gorman <mgorman@suse.de>
Cc: Miao Xie <miaox@cn.fujitsu.com>
Cc: David Rientjes <rientjes@google.com>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Christoph Lameter <cl@linux.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Pull scheduler changes for v3.4 from Ingo Molnar
* 'sched-core-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip: (27 commits)
printk: Make it compile with !CONFIG_PRINTK
sched/x86: Fix overflow in cyc2ns_offset
sched: Fix nohz load accounting -- again!
sched: Update yield() docs
printk/sched: Introduce special printk_sched() for those awkward moments
sched/nohz: Correctly initialize 'next_balance' in 'nohz' idle balancer
sched: Cleanup cpu_active madness
sched: Fix load-balance wreckage
sched: Clean up parameter passing of proc_sched_autogroup_set_nice()
sched: Ditch per cgroup task lists for load-balancing
sched: Rename load-balancing fields
sched: Move load-balancing arguments into helper struct
sched/rt: Do not submit new work when PI-blocked
sched/rt: Prevent idle task boosting
sched/wait: Add __wake_up_all_locked() API
sched/rt: Document scheduler related skip-resched-check sites
sched/rt: Use schedule_preempt_disabled()
sched/rt: Add schedule_preempt_disabled()
sched/rt: Do not throttle when PI boosting
sched/rt: Keep period timer ticking when rt throttling is active
...
Pull irq/core changes for v3.4 from Ingo Molnar
* 'irq-core-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
genirq: Remove paranoid warnons and bogus fixups
genirq: Flush the irq thread on synchronization
genirq: Get rid of unnecessary IRQTF_DIED flag
genirq: No need to check IRQTF_DIED before stopping a thread handler
genirq: Get rid of unnecessary irqaction field in task_struct
genirq: Fix incorrect check for forced IRQ thread handler
softirq: Reduce invoke_softirq() code duplication
genirq: Fix long-term regression in genirq irq_set_irq_type() handling
x86-32/irq: Don't switch to irq stack for a user-mode irq
Pull RCU changes for v3.4 from Ingo Molnar. The major features of this
series are:
- making RCU more aggressive about entering dyntick-idle mode in order
to improve energy efficiency
- converting a few more call_rcu()s to kfree_rcu()s
- applying a number of rcutree fixes and cleanups to rcutiny
- removing CONFIG_SMP #ifdefs from treercu
- allowing RCU CPU stall times to be set via sysfs
- adding CPU-stall capability to rcutorture
- adding more RCU-abuse diagnostics
- updating documentation
- fixing yet more issues located by the still-ongoing top-to-bottom
inspection of RCU, this time with a special focus on the CPU-hotplug
code path.
* 'core-rcu-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip: (48 commits)
rcu: Stop spurious warnings from synchronize_sched_expedited
rcu: Hold off RCU_FAST_NO_HZ after timer posted
rcu: Eliminate softirq-mediated RCU_FAST_NO_HZ idle-entry loop
rcu: Add RCU_NONIDLE() for idle-loop RCU read-side critical sections
rcu: Allow nesting of rcu_idle_enter() and rcu_idle_exit()
rcu: Remove redundant check for rcu_head misalignment
PTR_ERR should be called before its argument is cleared.
rcu: Convert WARN_ON_ONCE() in rcu_lock_acquire() to lockdep
rcu: Trace only after NULL-pointer check
rcu: Call out dangers of expedited RCU primitives
rcu: Rework detection of use of RCU by offline CPUs
lockdep: Add CPU-idle/offline warning to lockdep-RCU splat
rcu: No interrupt disabling for rcu_prepare_for_idle()
rcu: Move synchronize_sched_expedited() to rcutree.c
rcu: Check for illegal use of RCU from offlined CPUs
rcu: Update stall-warning documentation
rcu: Add CPU-stall capability to rcutorture
rcu: Make documentation give more realistic rcutorture duration
rcutorture: Permit holding off CPU-hotplug operations during boot
rcu: Print scheduling-clock information on RCU CPU stall-warning messages
...
When a new thread handler is created, an irqaction is passed to it as
data. Not only that irqaction is stored in task_struct by the handler
for later use, but also a structure associated with the kernel thread
keeps this value as long as the thread exists.
This fix kicks irqaction out off task_struct. Yes, I introduce new bit
field. But it allows not only to eliminate the duplicate, but also
shortens size of task_struct.
Reported-by: Oleg Nesterov <oleg@redhat.com>
Signed-off-by: Alexander Gordeev <agordeev@redhat.com>
Link: http://lkml.kernel.org/r/20120309135925.GB2114@dhcp-26-207.brq.redhat.com
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Srivatsa Vaddagiri