We're going to split the process wide cpu accounting into two parts:
- clocks; which can take all the time they want since they run
from user context.
- timers; which need constant time tracing but can affort the overhead
because they're default off -- and rare.
The clock readout will go back to a full sum of the thread group, for this
we need to re-add the exit stats that were removed in the initial itimer
rework (f06febc9: timers: fix itimer/many thread hang).
Furthermore, since that full sum can be rather slow for large thread groups
and we have the complete dead task stats, revert the do_notify_parent time
computation.
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Reviewed-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Convert all system calls to return a long. This should be a NOP since all
converted types should have the same size anyway.
With the exception of sys_exit_group which returned void. But that doesn't
matter since the system call doesn't return.
Signed-off-by: Heiko Carstens <heiko.carstens@de.ibm.com>
xacct_add_tsk() relies on do_exit()->update_hiwater_xxx() and uses
mm->hiwater_xxx directly, this leads to 2 problems:
- taskstats_user_cmd() can call fill_pid()->xacct_add_tsk() at any
moment before the task exits, so we should check the current values of
rss/vm anyway.
- do_exit()->update_hiwater_xxx() calls are racy. An exiting thread can
be preempted right before mm->hiwater_xxx = new_val, and another thread
can use A_LOT of memory and exit in between. When the first thread
resumes it can be the last thread in the thread group, in that case we
report the wrong hiwater_xxx values which do not take A_LOT into
account.
Introduce get_mm_hiwater_rss() and get_mm_hiwater_vm() helpers and change
xacct_add_tsk() to use them. The first helper will also be used by
rusage->ru_maxrss accounting.
Kill do_exit()->update_hiwater_xxx() calls. Unless we are going to
decrease rss/vm there is no point to update mm->hiwater_xxx, and nobody
can look at this mm_struct when exit_mmap() actually unmaps the memory.
Signed-off-by: Oleg Nesterov <oleg@redhat.com>
Acked-by: Hugh Dickins <hugh@veritas.com>
Reviewed-by: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Acked-by: Balbir Singh <balbir@linux.vnet.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
cgroup_mm_owner_callbacks() was brought in to support the memrlimit
controller, but sneaked into mainline ahead of it. That controller has
now been shelved, and the mm_owner_changed() args were inadequate for it
anyway (they needed an mm pointer instead of a task pointer).
Remove the dead code, and restore mm_update_next_owner() locking to how it
was before: taking mmap_sem there does nothing for memcontrol.c, now the
only user of mm->owner.
Signed-off-by: Hugh Dickins <hugh@veritas.com>
Cc: Paul Menage <menage@google.com>
Cc: Balbir Singh <balbir@in.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
* 'for-2.6.29' of git://git.kernel.dk/linux-2.6-block: (43 commits)
bio: get rid of bio_vec clearing
bounce: don't rely on a zeroed bio_vec list
cciss: simplify parameters to deregister_disk function
cfq-iosched: fix race between exiting queue and exiting task
loop: Do not call loop_unplug for not configured loop device.
loop: Flush possible running bios when loop device is released.
alpha: remove dead BIO_VMERGE_BOUNDARY
Get rid of CONFIG_LSF
block: make blk_softirq_init() static
block: use min_not_zero in blk_queue_stack_limits
block: add one-hit cache for disk partition lookup
cfq-iosched: remove limit of dispatch depth of max 4 times quantum
nbd: tell the block layer that it is not a rotational device
block: get rid of elevator_t typedef
aio: make the lookup_ioctx() lockless
bio: add support for inlining a number of bio_vecs inside the bio
bio: allow individual slabs in the bio_set
bio: move the slab pointer inside the bio_set
bio: only mempool back the largest bio_vec slab cache
block: don't use plugging on SSD devices
...
* 'core-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip: (63 commits)
stacktrace: provide save_stack_trace_tsk() weak alias
rcu: provide RCU options on non-preempt architectures too
printk: fix discarding message when recursion_bug
futex: clean up futex_(un)lock_pi fault handling
"Tree RCU": scalable classic RCU implementation
futex: rename field in futex_q to clarify single waiter semantics
x86/swiotlb: add default swiotlb_arch_range_needs_mapping
x86/swiotlb: add default phys<->bus conversion
x86: unify pci iommu setup and allow swiotlb to compile for 32 bit
x86: add swiotlb allocation functions
swiotlb: consolidate swiotlb info message printing
swiotlb: support bouncing of HighMem pages
swiotlb: factor out copy to/from device
swiotlb: add arch hook to force mapping
swiotlb: allow architectures to override phys<->bus<->phys conversions
swiotlb: add comment where we handle the overflow of a dma mask on 32 bit
rcu: fix rcutorture behavior during reboot
resources: skip sanity check of busy resources
swiotlb: move some definitions to header
swiotlb: allow architectures to override swiotlb pool allocation
...
Fix up trivial conflicts in
arch/x86/kernel/Makefile
arch/x86/mm/init_32.c
include/linux/hardirq.h
as per Ingo's suggestions.
When taking recursive faults in do_exit, if the io_context is not null,
exit_io_context() is being called. But it might decrement the refcount
more than once. It is better to leave this task alone.
Signed-off-by: Nikanth Karthikesan <knikanth@suse.de>
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
* 'tracing-core-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip: (241 commits)
sched, trace: update trace_sched_wakeup()
tracing/ftrace: don't trace on early stage of a secondary cpu boot, v3
Revert "x86: disable X86_PTRACE_BTS"
ring-buffer: prevent false positive warning
ring-buffer: fix dangling commit race
ftrace: enable format arguments checking
x86, bts: memory accounting
x86, bts: add fork and exit handling
ftrace: introduce tracing_reset_online_cpus() helper
tracing: fix warnings in kernel/trace/trace_sched_switch.c
tracing: fix warning in kernel/trace/trace.c
tracing/ring-buffer: remove unused ring_buffer size
trace: fix task state printout
ftrace: add not to regex on filtering functions
trace: better use of stack_trace_enabled for boot up code
trace: add a way to enable or disable the stack tracer
x86: entry_64 - introduce FTRACE_ frame macro v2
tracing/ftrace: add the printk-msg-only option
tracing/ftrace: use preempt_enable_no_resched_notrace in ring_buffer_time_stamp()
x86, bts: correctly report invalid bts records
...
Fixed up trivial conflict in scripts/recordmcount.pl due to SH bits
being already partly merged by the SH merge.
Change counter inheritance from a 'push' to a 'pull' model: instead of
child tasks pushing their final counts to the parent, reuse the wait4
infrastructure to pull counters as child tasks are exit-processed,
much like how cutime/cstime is collected.
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Impact: implement new performance feature
Counter inheritance can be used to run performance counters in a workload,
transparently - and pipe back the counter results to the parent counter.
Inheritance for performance counters works the following way: when creating
a counter it can be marked with the .inherit=1 flag. Such counters are then
'inherited' by all child tasks (be they fork()-ed or clone()-ed). These
counters get inherited through exec() boundaries as well (except through
setuid boundaries).
The counter values get added back to the parent counter(s) when the child
task(s) exit - much like stime/utime statistics are gathered. So inherited
counters are ideal to gather summary statistics about an application's
behavior via shell commands, without having to modify that application.
The timec.c command utilizes counter inheritance:
http://redhat.com/~mingo/perfcounters/timec.c
Sample output:
$ ./timec -e 1 -e 3 -e 5 ls -lR /usr/include/ >/dev/null
Performance counter stats for 'ls':
163516953 instructions
2295 cache-misses
2855182 branch-misses
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Impact: avoid losing some traces when a task is freed
do_exit() is not the last function called when a task finishes.
There are still some functions which are to be called such as
ree_task(). So we delay the freeing of the return stack to the
last moment.
Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Impact: use deeper function tracing depth safely
Some tests showed that function return tracing needed a more deeper depth
of function calls. But it could be unsafe to store these return addresses
to the stack.
So these arrays will now be allocated dynamically into task_struct of current
only when the tracer is activated.
Typical scheme when tracer is activated:
- allocate a return stack for each task in global list.
- fork: allocate the return stack for the newly created task
- exit: free return stack of current
- idle init: same as fork
I chose a default depth of 50. I don't have overruns anymore.
Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Impact: relax the locking of cpu-time accounting calls
->siglock buys nothing for thread_group_cputime() in do_sys_times() and
wait_task_zombie() (which btw takes the unrelated parent's ->siglock).
Actually I think do_sys_times() doesn't need ->siglock at all.
Signed-off-by: Oleg Nesterov <oleg@redhat.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Impact: API *CHANGE*. Must update all tracepoint users.
Add DEFINE_TRACE() to tracepoints to let them declare the tracepoint
structure in a single spot for all the kernel. It helps reducing memory
consumption, especially when declaring a lot of tracepoints, e.g. for
kmalloc tracing.
*API CHANGE WARNING*: now, DECLARE_TRACE() must be used in headers for
tracepoint declarations rather than DEFINE_TRACE(). This is the sane way
to do it. The name previously used was misleading.
Updates scheduler instrumentation to follow this API change.
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@polymtl.ca>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
We don't want to get rid of the futexes just at exit() time, we want to
drop them when doing an execve() too, since that gets rid of the
previous VM image too.
Doing it at mm_release() time means that we automatically always do it
when we disassociate a VM map from the task.
Reported-by: pageexec@freemail.hu
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Nick Piggin <npiggin@suse.de>
Cc: Hugh Dickins <hugh@veritas.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Brad Spengler <spender@grsecurity.net>
Cc: Alex Efros <powerman@powerman.name>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Oleg Nesterov <oleg@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Conflicts:
security/keys/internal.h
security/keys/process_keys.c
security/keys/request_key.c
Fixed conflicts above by using the non 'tsk' versions.
Signed-off-by: James Morris <jmorris@namei.org>
Inaugurate copy-on-write credentials management. This uses RCU to manage the
credentials pointer in the task_struct with respect to accesses by other tasks.
A process may only modify its own credentials, and so does not need locking to
access or modify its own credentials.
A mutex (cred_replace_mutex) is added to the task_struct to control the effect
of PTRACE_ATTACHED on credential calculations, particularly with respect to
execve().
With this patch, the contents of an active credentials struct may not be
changed directly; rather a new set of credentials must be prepared, modified
and committed using something like the following sequence of events:
struct cred *new = prepare_creds();
int ret = blah(new);
if (ret < 0) {
abort_creds(new);
return ret;
}
return commit_creds(new);
There are some exceptions to this rule: the keyrings pointed to by the active
credentials may be instantiated - keyrings violate the COW rule as managing
COW keyrings is tricky, given that it is possible for a task to directly alter
the keys in a keyring in use by another task.
To help enforce this, various pointers to sets of credentials, such as those in
the task_struct, are declared const. The purpose of this is compile-time
discouragement of altering credentials through those pointers. Once a set of
credentials has been made public through one of these pointers, it may not be
modified, except under special circumstances:
(1) Its reference count may incremented and decremented.
(2) The keyrings to which it points may be modified, but not replaced.
The only safe way to modify anything else is to create a replacement and commit
using the functions described in Documentation/credentials.txt (which will be
added by a later patch).
This patch and the preceding patches have been tested with the LTP SELinux
testsuite.
This patch makes several logical sets of alteration:
(1) execve().
This now prepares and commits credentials in various places in the
security code rather than altering the current creds directly.
(2) Temporary credential overrides.
do_coredump() and sys_faccessat() now prepare their own credentials and
temporarily override the ones currently on the acting thread, whilst
preventing interference from other threads by holding cred_replace_mutex
on the thread being dumped.
This will be replaced in a future patch by something that hands down the
credentials directly to the functions being called, rather than altering
the task's objective credentials.
(3) LSM interface.
A number of functions have been changed, added or removed:
(*) security_capset_check(), ->capset_check()
(*) security_capset_set(), ->capset_set()
Removed in favour of security_capset().
(*) security_capset(), ->capset()
New. This is passed a pointer to the new creds, a pointer to the old
creds and the proposed capability sets. It should fill in the new
creds or return an error. All pointers, barring the pointer to the
new creds, are now const.
(*) security_bprm_apply_creds(), ->bprm_apply_creds()
Changed; now returns a value, which will cause the process to be
killed if it's an error.
(*) security_task_alloc(), ->task_alloc_security()
Removed in favour of security_prepare_creds().
(*) security_cred_free(), ->cred_free()
New. Free security data attached to cred->security.
(*) security_prepare_creds(), ->cred_prepare()
New. Duplicate any security data attached to cred->security.
(*) security_commit_creds(), ->cred_commit()
New. Apply any security effects for the upcoming installation of new
security by commit_creds().
(*) security_task_post_setuid(), ->task_post_setuid()
Removed in favour of security_task_fix_setuid().
(*) security_task_fix_setuid(), ->task_fix_setuid()
Fix up the proposed new credentials for setuid(). This is used by
cap_set_fix_setuid() to implicitly adjust capabilities in line with
setuid() changes. Changes are made to the new credentials, rather
than the task itself as in security_task_post_setuid().
(*) security_task_reparent_to_init(), ->task_reparent_to_init()
Removed. Instead the task being reparented to init is referred
directly to init's credentials.
NOTE! This results in the loss of some state: SELinux's osid no
longer records the sid of the thread that forked it.
(*) security_key_alloc(), ->key_alloc()
(*) security_key_permission(), ->key_permission()
Changed. These now take cred pointers rather than task pointers to
refer to the security context.
(4) sys_capset().
This has been simplified and uses less locking. The LSM functions it
calls have been merged.
(5) reparent_to_kthreadd().
This gives the current thread the same credentials as init by simply using
commit_thread() to point that way.
(6) __sigqueue_alloc() and switch_uid()
__sigqueue_alloc() can't stop the target task from changing its creds
beneath it, so this function gets a reference to the currently applicable
user_struct which it then passes into the sigqueue struct it returns if
successful.
switch_uid() is now called from commit_creds(), and possibly should be
folded into that. commit_creds() should take care of protecting
__sigqueue_alloc().
(7) [sg]et[ug]id() and co and [sg]et_current_groups.
The set functions now all use prepare_creds(), commit_creds() and
abort_creds() to build and check a new set of credentials before applying
it.
security_task_set[ug]id() is called inside the prepared section. This
guarantees that nothing else will affect the creds until we've finished.
The calling of set_dumpable() has been moved into commit_creds().
Much of the functionality of set_user() has been moved into
commit_creds().
The get functions all simply access the data directly.
(8) security_task_prctl() and cap_task_prctl().
security_task_prctl() has been modified to return -ENOSYS if it doesn't
want to handle a function, or otherwise return the return value directly
rather than through an argument.
Additionally, cap_task_prctl() now prepares a new set of credentials, even
if it doesn't end up using it.
(9) Keyrings.
A number of changes have been made to the keyrings code:
(a) switch_uid_keyring(), copy_keys(), exit_keys() and suid_keys() have
all been dropped and built in to the credentials functions directly.
They may want separating out again later.
(b) key_alloc() and search_process_keyrings() now take a cred pointer
rather than a task pointer to specify the security context.
(c) copy_creds() gives a new thread within the same thread group a new
thread keyring if its parent had one, otherwise it discards the thread
keyring.
(d) The authorisation key now points directly to the credentials to extend
the search into rather pointing to the task that carries them.
(e) Installing thread, process or session keyrings causes a new set of
credentials to be created, even though it's not strictly necessary for
process or session keyrings (they're shared).
(10) Usermode helper.
The usermode helper code now carries a cred struct pointer in its
subprocess_info struct instead of a new session keyring pointer. This set
of credentials is derived from init_cred and installed on the new process
after it has been cloned.
call_usermodehelper_setup() allocates the new credentials and
call_usermodehelper_freeinfo() discards them if they haven't been used. A
special cred function (prepare_usermodeinfo_creds()) is provided
specifically for call_usermodehelper_setup() to call.
call_usermodehelper_setkeys() adjusts the credentials to sport the
supplied keyring as the new session keyring.
(11) SELinux.
SELinux has a number of changes, in addition to those to support the LSM
interface changes mentioned above:
(a) selinux_setprocattr() no longer does its check for whether the
current ptracer can access processes with the new SID inside the lock
that covers getting the ptracer's SID. Whilst this lock ensures that
the check is done with the ptracer pinned, the result is only valid
until the lock is released, so there's no point doing it inside the
lock.
(12) is_single_threaded().
This function has been extracted from selinux_setprocattr() and put into
a file of its own in the lib/ directory as join_session_keyring() now
wants to use it too.
The code in SELinux just checked to see whether a task shared mm_structs
with other tasks (CLONE_VM), but that isn't good enough. We really want
to know if they're part of the same thread group (CLONE_THREAD).
(13) nfsd.
The NFS server daemon now has to use the COW credentials to set the
credentials it is going to use. It really needs to pass the credentials
down to the functions it calls, but it can't do that until other patches
in this series have been applied.
Signed-off-by: David Howells <dhowells@redhat.com>
Acked-by: James Morris <jmorris@namei.org>
Signed-off-by: James Morris <jmorris@namei.org>
Use RCU to access another task's creds and to release a task's own creds.
This means that it will be possible for the credentials of a task to be
replaced without another task (a) requiring a full lock to read them, and (b)
seeing deallocated memory.
Signed-off-by: David Howells <dhowells@redhat.com>
Acked-by: James Morris <jmorris@namei.org>
Acked-by: Serge Hallyn <serue@us.ibm.com>
Signed-off-by: James Morris <jmorris@namei.org>
Separate the task security context from task_struct. At this point, the
security data is temporarily embedded in the task_struct with two pointers
pointing to it.
Note that the Alpha arch is altered as it refers to (E)UID and (E)GID in
entry.S via asm-offsets.
With comment fixes Signed-off-by: Marc Dionne <marc.c.dionne@gmail.com>
Signed-off-by: David Howells <dhowells@redhat.com>
Acked-by: James Morris <jmorris@namei.org>
Acked-by: Serge Hallyn <serue@us.ibm.com>
Signed-off-by: James Morris <jmorris@namei.org>
Impact: fix hang/crash on ia64 under high load
This is ugly, but the simplest patch by far.
Unlike other similar routines, account_group_exec_runtime() could be
called "implicitly" from within scheduler after exit_notify(). This
means we can race with the parent doing release_task(), we can't just
check ->signal != NULL.
Change __exit_signal() to do spin_unlock_wait(&task_rq(tsk)->lock)
before __cleanup_signal() to make sure ->signal can't be freed under
task_rq(tsk)->lock. Note that task_rq_unlock_wait() doesn't care
about the case when tsk changes cpu/rq under us, this should be OK.
Thanks to Ingo who nacked my previous buggy patch.
Signed-off-by: Oleg Nesterov <oleg@redhat.com>
Acked-by: Peter Zijlstra <peterz@infradead.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Reported-by: Doug Chapman <doug.chapman@hp.com>
This patch adds an additional field to the mm_owner callbacks. This field
is required to get to the mm that changed. Hold mmap_sem in write mode
before calling the mm_owner_changed callback
[hugh@veritas.com: fix mmap_sem deadlock]
Signed-off-by: Balbir Singh <balbir@linux.vnet.ibm.com>
Cc: Sudhir Kumar <skumar@linux.vnet.ibm.com>
Cc: YAMAMOTO Takashi <yamamoto@valinux.co.jp>
Cc: Paul Menage <menage@google.com>
Cc: Li Zefan <lizf@cn.fujitsu.com>
Cc: Pavel Emelianov <xemul@openvz.org>
Cc: Balbir Singh <balbir@linux.vnet.ibm.com>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: David Rientjes <rientjes@google.com>
Cc: Vivek Goyal <vgoyal@redhat.com>
Signed-off-by: Hugh Dickins <hugh@veritas.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Instrument the scheduler activity (sched_switch, migration, wakeups,
wait for a task, signal delivery) and process/thread
creation/destruction (fork, exit, kthread stop). Actually, kthread
creation is not instrumented in this patch because it is architecture
dependent. It allows to connect tracers such as ftrace which detects
scheduling latencies, good/bad scheduler decisions. Tools like LTTng can
export this scheduler information along with instrumentation of the rest
of the kernel activity to perform post-mortem analysis on the scheduler
activity.
About the performance impact of tracepoints (which is comparable to
markers), even without immediate values optimizations, tests done by
Hideo Aoki on ia64 show no regression. His test case was using hackbench
on a kernel where scheduler instrumentation (about 5 events in code
scheduler code) was added. See the "Tracepoints" patch header for
performance result detail.
Changelog :
- Change instrumentation location and parameter to match ftrace
instrumentation, previously done with kernel markers.
[ mingo@elte.hu: conflict resolutions ]
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@polymtl.ca>
Acked-by: 'Peter Zijlstra' <peterz@infradead.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
There's a race between mm->owner assignment and swapoff, more easily
seen when task slab poisoning is turned on. The condition occurs when
try_to_unuse() runs in parallel with an exiting task. A similar race
can occur with callers of get_task_mm(), such as /proc/<pid>/<mmstats>
or ptrace or page migration.
CPU0 CPU1
try_to_unuse
looks at mm = task0->mm
increments mm->mm_users
task 0 exits
mm->owner needs to be updated, but no
new owner is found (mm_users > 1, but
no other task has task->mm = task0->mm)
mm_update_next_owner() leaves
mmput(mm) decrements mm->mm_users
task0 freed
dereferencing mm->owner fails
The fix is to notify the subsystem via mm_owner_changed callback(),
if no new owner is found, by specifying the new task as NULL.
Jiri Slaby:
mm->owner was set to NULL prior to calling cgroup_mm_owner_callbacks(), but
must be set after that, so as not to pass NULL as old owner causing oops.
Daisuke Nishimura:
mm_update_next_owner() may set mm->owner to NULL, but mem_cgroup_from_task()
and its callers need to take account of this situation to avoid oops.
Hugh Dickins:
Lockdep warning and hang below exec_mmap() when testing these patches.
exit_mm() up_reads mmap_sem before calling mm_update_next_owner(),
so exec_mmap() now needs to do the same. And with that repositioning,
there's now no point in mm_need_new_owner() allowing for NULL mm.
Reported-by: Hugh Dickins <hugh@veritas.com>
Signed-off-by: Balbir Singh <balbir@linux.vnet.ibm.com>
Signed-off-by: Jiri Slaby <jirislaby@gmail.com>
Signed-off-by: Daisuke Nishimura <nishimura@mxp.nes.nec.co.jp>
Signed-off-by: Hugh Dickins <hugh@veritas.com>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Paul Menage <menage@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Overview
This patch reworks the handling of POSIX CPU timers, including the
ITIMER_PROF, ITIMER_VIRT timers and rlimit handling. It was put together
with the help of Roland McGrath, the owner and original writer of this code.
The problem we ran into, and the reason for this rework, has to do with using
a profiling timer in a process with a large number of threads. It appears
that the performance of the old implementation of run_posix_cpu_timers() was
at least O(n*3) (where "n" is the number of threads in a process) or worse.
Everything is fine with an increasing number of threads until the time taken
for that routine to run becomes the same as or greater than the tick time, at
which point things degrade rather quickly.
This patch fixes bug 9906, "Weird hang with NPTL and SIGPROF."
Code Changes
This rework corrects the implementation of run_posix_cpu_timers() to make it
run in constant time for a particular machine. (Performance may vary between
one machine and another depending upon whether the kernel is built as single-
or multiprocessor and, in the latter case, depending upon the number of
running processors.) To do this, at each tick we now update fields in
signal_struct as well as task_struct. The run_posix_cpu_timers() function
uses those fields to make its decisions.
We define a new structure, "task_cputime," to contain user, system and
scheduler times and use these in appropriate places:
struct task_cputime {
cputime_t utime;
cputime_t stime;
unsigned long long sum_exec_runtime;
};
This is included in the structure "thread_group_cputime," which is a new
substructure of signal_struct and which varies for uniprocessor versus
multiprocessor kernels. For uniprocessor kernels, it uses "task_cputime" as
a simple substructure, while for multiprocessor kernels it is a pointer:
struct thread_group_cputime {
struct task_cputime totals;
};
struct thread_group_cputime {
struct task_cputime *totals;
};
We also add a new task_cputime substructure directly to signal_struct, to
cache the earliest expiration of process-wide timers, and task_cputime also
replaces the it_*_expires fields of task_struct (used for earliest expiration
of thread timers). The "thread_group_cputime" structure contains process-wide
timers that are updated via account_user_time() and friends. In the non-SMP
case the structure is a simple aggregator; unfortunately in the SMP case that
simplicity was not achievable due to cache-line contention between CPUs (in
one measured case performance was actually _worse_ on a 16-cpu system than
the same test on a 4-cpu system, due to this contention). For SMP, the
thread_group_cputime counters are maintained as a per-cpu structure allocated
using alloc_percpu(). The timer functions update only the timer field in
the structure corresponding to the running CPU, obtained using per_cpu_ptr().
We define a set of inline functions in sched.h that we use to maintain the
thread_group_cputime structure and hide the differences between UP and SMP
implementations from the rest of the kernel. The thread_group_cputime_init()
function initializes the thread_group_cputime structure for the given task.
The thread_group_cputime_alloc() is a no-op for UP; for SMP it calls the
out-of-line function thread_group_cputime_alloc_smp() to allocate and fill
in the per-cpu structures and fields. The thread_group_cputime_free()
function, also a no-op for UP, in SMP frees the per-cpu structures. The
thread_group_cputime_clone_thread() function (also a UP no-op) for SMP calls
thread_group_cputime_alloc() if the per-cpu structures haven't yet been
allocated. The thread_group_cputime() function fills the task_cputime
structure it is passed with the contents of the thread_group_cputime fields;
in UP it's that simple but in SMP it must also safely check that tsk->signal
is non-NULL (if it is it just uses the appropriate fields of task_struct) and,
if so, sums the per-cpu values for each online CPU. Finally, the three
functions account_group_user_time(), account_group_system_time() and
account_group_exec_runtime() are used by timer functions to update the
respective fields of the thread_group_cputime structure.
Non-SMP operation is trivial and will not be mentioned further.
The per-cpu structure is always allocated when a task creates its first new
thread, via a call to thread_group_cputime_clone_thread() from copy_signal().
It is freed at process exit via a call to thread_group_cputime_free() from
cleanup_signal().
All functions that formerly summed utime/stime/sum_sched_runtime values from
from all threads in the thread group now use thread_group_cputime() to
snapshot the values in the thread_group_cputime structure or the values in
the task structure itself if the per-cpu structure hasn't been allocated.
Finally, the code in kernel/posix-cpu-timers.c has changed quite a bit.
The run_posix_cpu_timers() function has been split into a fast path and a
slow path; the former safely checks whether there are any expired thread
timers and, if not, just returns, while the slow path does the heavy lifting.
With the dedicated thread group fields, timers are no longer "rebalanced" and
the process_timer_rebalance() function and related code has gone away. All
summing loops are gone and all code that used them now uses the
thread_group_cputime() inline. When process-wide timers are set, the new
task_cputime structure in signal_struct is used to cache the earliest
expiration; this is checked in the fast path.
Performance
The fix appears not to add significant overhead to existing operations. It
generally performs the same as the current code except in two cases, one in
which it performs slightly worse (Case 5 below) and one in which it performs
very significantly better (Case 2 below). Overall it's a wash except in those
two cases.
I've since done somewhat more involved testing on a dual-core Opteron system.
Case 1: With no itimer running, for a test with 100,000 threads, the fixed
kernel took 1428.5 seconds, 513 seconds more than the unfixed system,
all of which was spent in the system. There were twice as many
voluntary context switches with the fix as without it.
Case 2: With an itimer running at .01 second ticks and 4000 threads (the most
an unmodified kernel can handle), the fixed kernel ran the test in
eight percent of the time (5.8 seconds as opposed to 70 seconds) and
had better tick accuracy (.012 seconds per tick as opposed to .023
seconds per tick).
Case 3: A 4000-thread test with an initial timer tick of .01 second and an
interval of 10,000 seconds (i.e. a timer that ticks only once) had
very nearly the same performance in both cases: 6.3 seconds elapsed
for the fixed kernel versus 5.5 seconds for the unfixed kernel.
With fewer threads (eight in these tests), the Case 1 test ran in essentially
the same time on both the modified and unmodified kernels (5.2 seconds versus
5.8 seconds). The Case 2 test ran in about the same time as well, 5.9 seconds
versus 5.4 seconds but again with much better tick accuracy, .013 seconds per
tick versus .025 seconds per tick for the unmodified kernel.
Since the fix affected the rlimit code, I also tested soft and hard CPU limits.
Case 4: With a hard CPU limit of 20 seconds and eight threads (and an itimer
running), the modified kernel was very slightly favored in that while
it killed the process in 19.997 seconds of CPU time (5.002 seconds of
wall time), only .003 seconds of that was system time, the rest was
user time. The unmodified kernel killed the process in 20.001 seconds
of CPU (5.014 seconds of wall time) of which .016 seconds was system
time. Really, though, the results were too close to call. The results
were essentially the same with no itimer running.
Case 5: With a soft limit of 20 seconds and a hard limit of 2000 seconds
(where the hard limit would never be reached) and an itimer running,
the modified kernel exhibited worse tick accuracy than the unmodified
kernel: .050 seconds/tick versus .028 seconds/tick. Otherwise,
performance was almost indistinguishable. With no itimer running this
test exhibited virtually identical behavior and times in both cases.
In times past I did some limited performance testing. those results are below.
On a four-cpu Opteron system without this fix, a sixteen-thread test executed
in 3569.991 seconds, of which user was 3568.435s and system was 1.556s. On
the same system with the fix, user and elapsed time were about the same, but
system time dropped to 0.007 seconds. Performance with eight, four and one
thread were comparable. Interestingly, the timer ticks with the fix seemed
more accurate: The sixteen-thread test with the fix received 149543 ticks
for 0.024 seconds per tick, while the same test without the fix received 58720
for 0.061 seconds per tick. Both cases were configured for an interval of
0.01 seconds. Again, the other tests were comparable. Each thread in this
test computed the primes up to 25,000,000.
I also did a test with a large number of threads, 100,000 threads, which is
impossible without the fix. In this case each thread computed the primes only
up to 10,000 (to make the runtime manageable). System time dominated, at
1546.968 seconds out of a total 2176.906 seconds (giving a user time of
629.938s). It received 147651 ticks for 0.015 seconds per tick, still quite
accurate. There is obviously no comparable test without the fix.
Signed-off-by: Frank Mayhar <fmayhar@google.com>
Cc: Roland McGrath <roland@redhat.com>
Cc: Alexey Dobriyan <adobriyan@gmail.com>
Cc: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Spencer reported a problem where utime and stime were going negative despite
the fixes in commit b27f03d4bd. The suspected
reason for the problem is that signal_struct maintains it's own utime and
stime (of exited tasks), these are not updated using the new task_utime()
routine, hence sig->utime can go backwards and cause the same problem
to occur (sig->utime, adds tsk->utime and not task_utime()). This patch
fixes the problem
TODO: using max(task->prev_utime, derived utime) works for now, but a more
generic solution is to implement cputime_max() and use the cputime_gt()
function for comparison.
Reported-by: spencer@bluehost.com
Signed-off-by: Balbir Singh <balbir@linux.vnet.ibm.com>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
We don't change pid_ns->child_reaper when the main thread of the
subnamespace init exits. As Robert Rex <robert.rex@exasol.com> pointed
out this is wrong.
Yes, the re-parenting itself works correctly, but if the reparented task
exits it needs ->parent->nsproxy->pid_ns in do_notify_parent(), and if the
main thread is zombie its ->nsproxy was already cleared by
exit_task_namespaces().
Introduce the new function, find_new_reaper(), which finds the new
->parent for the re-parenting and changes ->child_reaper if needed. Kill
the now unneeded exit_child_reaper().
Also move the changing of ->child_reaper from zap_pid_ns_processes() to
find_new_reaper(), this consolidates the games with ->child_reaper and
makes it stable under tasklist_lock.
Addresses http://bugzilla.kernel.org/show_bug.cgi?id=11391
Reported-by: Robert Rex <robert.rex@exasol.com>
Signed-off-by: Oleg Nesterov <oleg@tv-sign.ru>
Acked-by: Serge Hallyn <serue@us.ibm.com>
Acked-by: Pavel Emelyanov <xemul@openvz.org>
Acked-by: Sukadev Bhattiprolu <sukadev@linux.vnet.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
task->signal->notify_count is only initialized if
task->signal->group_exit_task is not NULL. Reorder a conditional so
that uninitialised memory is not used. Found by Valgrind.
Signed-off-by: Steve VanDeBogart <vandebo-lkml@nerdbox.net>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
My commit 2b2a1ff64a introduced a regression
(sorry about that) for the odd case of exit_signal=0 (e.g. clone_flags=0).
This is not a normal use, but it's used by a case in the glibc test suite.
Dying with exit_signal=0 sends no signal, but it's supposed to wake up a
parent's blocked wait*() calls (unlike the delayed_group_leader case).
This fixes tracehook_notify_death() and its caller to distinguish a
"signal 0" wakeup from the delayed_group_leader case (with no wakeup).
Signed-off-by: Roland McGrath <roland@redhat.com>
Tested-by: Serge Hallyn <serue@us.ibm.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Put all i/o statistics in struct proc_io_accounting and use inline functions to
initialize and increment statistics, removing a lot of single variable
assignments.
This also reduces the kernel size as following (with CONFIG_TASK_XACCT=y and
CONFIG_TASK_IO_ACCOUNTING=y).
text data bss dec hex filename
11651 0 0 11651 2d83 kernel/exit.o.before
11619 0 0 11619 2d63 kernel/exit.o.after
10886 132 136 11154 2b92 kernel/fork.o.before
10758 132 136 11026 2b12 kernel/fork.o.after
3082029 807968 4818600 8708597 84e1f5 vmlinux.o.before
3081869 807968 4818600 8708437 84e155 vmlinux.o.after
Signed-off-by: Andrea Righi <righi.andrea@gmail.com>
Acked-by: Oleg Nesterov <oleg@tv-sign.ru>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
This moves the ptrace logic in task death (exit_notify) into tracehook.h
inlines. Some code is rearranged slightly to make things nicer. There is
no change, only cleanup.
There is one hook called with the tasklist_lock write-locked, as ptrace
needs. There is also a new hook called after exit_state changes and
without locks. This is a better place for tracing work to be in the
future, since it doesn't delay the whole system with locking.
Signed-off-by: Roland McGrath <roland@redhat.com>
Cc: Oleg Nesterov <oleg@tv-sign.ru>
Reviewed-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
This moves the ptrace-related logic from release_task into tracehook.h and
ptrace.h inlines. It provides clean hooks both before and after locking
tasklist_lock, for future tracing logic to do more cleanup without the
lock.
This also changes release_task() itself in the rare "zap_leader" case to
set the leader to EXIT_DEAD before iterating. This maintains the
invariant that release_task() only ever handles a task in EXIT_DEAD. This
is a common-sense invariant that is already always true except in this one
arcane case of zombie leader whose parent ignores SIGCHLD.
This change is harmless and only costs one store in this one rare case.
It keeps the expected state more consisently sane, which is nicer when
debugging weirdness in release_task(). It also lets some future code in
the tracehook entry points rely on this invariant for bookkeeping.
Signed-off-by: Roland McGrath <roland@redhat.com>
Cc: Oleg Nesterov <oleg@tv-sign.ru>
Reviewed-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
