Now that cgroup_unload_subsys() can tell whether the root css is
online or not, we can safely call cgroup_unload_subsys() after idr
init failure in cgroup_load_subsys().
Replace the manual unrolling and invoke cgroup_unload_subsys() on
failure. This drops cgroup_mutex inbetween but should be safe as the
subsystem will fail try_module_get() and thus can't be mounted
inbetween. As this means that cgroup_unload_subsys() can be called
before css_sets are rehashed, remove BUG_ON() on %NULL
css_set->subsys[] from cgroup_unload_subsys().
Signed-off-by: Tejun Heo <tj@kernel.org>
Acked-by: Li Zefan <lizefan@huawei.com>
New helpers on/offline_css() respectively wrap ->post_create() and
->pre_destroy() invocations. online_css() sets CSS_ONLINE after
->post_create() is complete and offline_css() invokes ->pre_destroy()
iff CSS_ONLINE is set and clears it while also handling the temporary
dropping of cgroup_mutex.
This patch doesn't introduce any behavior change at the moment but
will be used to improve cgroup_create() failure path and allow
->post_create() to fail.
Signed-off-by: Tejun Heo <tj@kernel.org>
Acked-by: Li Zefan <lizefan@huawei.com>
Separate out cgroup_destroy_locked() from cgroup_destroy(). This will
be later used in cgroup_create() failure path.
While at it, add lockdep asserts on i_mutex and cgroup_mutex, and move
@d and @parent assignments to their declarations.
This patch doesn't introduce any functional difference.
Signed-off-by: Tejun Heo <tj@kernel.org>
Acked-by: Li Zefan <lizefan@huawei.com>
* If idr init fails, cgroup_load_subsys() cleared dummytop->subsys[]
before calilng ->destroy() making CSS inaccessible to the callback,
and didn't unlink ss->sibling. As no modular controller uses
->use_id, this doesn't cause any actual problems.
* cgroup_unload_subsys() was forgetting to free idr, call
->pre_destroy() and clear ->active. As there currently is no
modular controller which uses ->use_id, ->pre_destroy() or ->active,
this doesn't cause any actual problems.
Signed-off-by: Tejun Heo <tj@kernel.org>
Acked-by: Li Zefan <lizefan@huawei.com>
Make cgroup_init_subsys() grab cgroup_mutex while initializing a
subsystem so that all helpers and callbacks are called under the
context they expect. This isn't strictly necessary as
cgroup_init_subsys() doesn't race with anybody but will allow adding
lockdep assertions.
Signed-off-by: Tejun Heo <tj@kernel.org>
Acked-by: Li Zefan <lizefan@huawei.com>
Consistently use @css and @dummytop in these two functions instead of
referring to them indirectly.
Signed-off-by: Tejun Heo <tj@kernel.org>
Acked-by: Li Zefan <lizefan@huawei.com>
Currently, CSS_* flags are defined as bit positions and manipulated
using atomic bitops. There's no reason to use atomic bitops for them
and bit positions are clunkier to deal with than bit masks. Make
CSS_* bit masks instead and use the usual C bitwise operators to
access them.
Signed-off-by: Tejun Heo <tj@kernel.org>
Acked-by: Li Zefan <lizefan@huawei.com>
cgroup->dentry is marked and used as a RCU pointer; however, it isn't
one - the final dentry put doesn't go through call_rcu(). cgroup and
dentry share the same RCU freeing rule via synchronize_rcu() in
cgroup_diput() (kfree_rcu() used on cgrp is unnecessary). If cgrp is
accessible under RCU read lock, so is its dentry and dereferencing
cgrp->dentry doesn't need any further RCU protection or annotation.
While not being accurate, before the previous patch, the RCU accessors
served a purpose as memory barriers - cgroup->dentry used to be
assigned after the cgroup was made visible to cgroup_path(), so the
assignment and dereferencing in cgroup_path() needed the memory
barrier pair. Now that list_add_tail_rcu() happens after
cgroup->dentry is assigned, this no longer is necessary.
Remove the now unnecessary and misleading RCU annotations from
cgroup->dentry. To make up for the removal of rcu_dereference_check()
in cgroup_path(), add an explicit rcu_lockdep_assert(), which asserts
the dereference rule of @cgrp, not cgrp->dentry.
Signed-off-by: Tejun Heo <tj@kernel.org>
Acked-by: Li Zefan <lizefan@huawei.com>
While creating a new cgroup, cgroup_create() links the newly allocated
cgroup into various places before trying to create its directory.
Because cgroup life-cycle is tied to the vfs objects, this makes it
impossible to use cgroup_rmdir() for rolling back creation - the
removal logic depends on having full vfs objects.
This patch moves directory creation above linking and collect linking
operations to one place. This allows directory creation failure to
share error exit path with css allocation failures and any failure
sites afterwards (to be added later) can use cgroup_rmdir() logic to
undo creation.
Note that this also makes the memory barriers around cgroup->dentry,
which currently is misleadingly using RCU operations, unnecessary.
This will be handled in the next patch.
While at it, locking BUG_ON() on i_mutex is converted to
lockdep_assert_held().
v2: Patch originally removed %NULL dentry check in cgroup_path();
however, Li pointed out that this patch doesn't make it
unnecessary as ->create() may call cgroup_path(). Drop the
change for now.
Signed-off-by: Tejun Heo <tj@kernel.org>
Acked-by: Li Zefan <lizefan@huawei.com>
The operation order of cgroup creation is about to change and
cgroup_create_dir() is more of a hindrance than a proper abstraction.
Open-code it by moving the parent nlink adjustment next to self nlink
adjustment in cgroup_create_file() and the rest to cgroup_create().
This patch doesn't introduce any behavior change.
Signed-off-by: Tejun Heo <tj@kernel.org>
Acked-by: Li Zefan <lizefan@huawei.com>
Not strictly necessary but it's annoying to have uninitialized
list_head around.
Signed-off-by: Tejun Heo <tj@kernel.org>
Acked-by: Li Zefan <lizefan@huawei.com>
Currently, cgroup doesn't provide any generic helper for walking a
given cgroup's children or descendants. This patch adds the following
three macros.
* cgroup_for_each_child() - walk immediate children of a cgroup.
* cgroup_for_each_descendant_pre() - visit all descendants of a cgroup
in pre-order tree traversal.
* cgroup_for_each_descendant_post() - visit all descendants of a
cgroup in post-order tree traversal.
All three only require the user to hold RCU read lock during
traversal. Verifying that each iterated cgroup is online is the
responsibility of the user. When used with proper synchronization,
cgroup_for_each_descendant_pre() can be used to propagate state
updates to descendants in reliable way. See comments for details.
v2: s/config/state/ in commit message and comments per Michal. More
documentation on synchronization rules.
Signed-off-by: Tejun Heo <tj@kernel.org>
Reviewed-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujisu.com>
Reviewed-by: Michal Hocko <mhocko@suse.cz>
Acked-by: Li Zefan <lizefan@huawei.com>
Use RCU safe list operations for cgroup->children. This will be used
to implement cgroup children / descendant walking which can be used by
controllers.
Note that cgroup_create() now puts a new cgroup at the end of the
->children list instead of head. This isn't strictly necessary but is
done so that the iteration order is more conventional.
Signed-off-by: Tejun Heo <tj@kernel.org>
Reviewed-by: Michal Hocko <mhocko@suse.cz>
Reviewed-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Acked-by: Li Zefan <lizefan@huawei.com>
Currently, there's no way for a controller to find out whether a new
cgroup finished all ->create() allocatinos successfully and is
considered "live" by cgroup.
This becomes a problem later when we add generic descendants walking
to cgroup which can be used by controllers as controllers don't have a
synchronization point where it can synchronize against new cgroups
appearing in such walks.
This patch adds ->post_create(). It's called after all ->create()
succeeded and the cgroup is linked into the generic cgroup hierarchy.
This plays the counterpart of ->pre_destroy().
When used in combination with the to-be-added generic descendant
iterators, ->post_create() can be used to implement reliable state
inheritance. It will be explained with the descendant iterators.
v2: Added a paragraph about its future use w/ descendant iterators per
Michal.
v3: Forgot to add ->post_create() invocation to cgroup_load_subsys().
Fixed.
Signed-off-by: Tejun Heo <tj@kernel.org>
Acked-by: Michal Hocko <mhocko@suse.cz>
Reviewed-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Acked-by: Li Zefan <lizefan@huawei.com>
Cc: Glauber Costa <glommer@parallels.com>
'start' is set to buf + buflen and do the '--' immediately.
Just set it to 'buf + buflen - 1' directly.
Signed-off-by: Tao Ma <boyu.mt@taobao.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
Cc: Li Zefan <lizefan@huawei.com>
Pull rmdir updates into for-3.8 so that further callback updates can
be put on top. This pull created a trivial conflict between the
following two commits.
8c7f6edbda ("cgroup: mark subsystems with broken hierarchy support and whine if cgroups are nested for them")
ed95779340 ("cgroup: kill cgroup_subsys->__DEPRECATED_clear_css_refs")
The former added a field to cgroup_subsys and the latter removed one
from it. They happen to be colocated causing the conflict. Keeping
what's added and removing what's removed resolves the conflict.
Signed-off-by: Tejun Heo <tj@kernel.org>
All ->pre_destory() implementations return 0 now, which is the only
allowed return value. Make it return void.
Signed-off-by: Tejun Heo <tj@kernel.org>
Reviewed-by: Michal Hocko <mhocko@suse.cz>
Acked-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Acked-by: Li Zefan <lizefan@huawei.com>
Cc: Balbir Singh <bsingharora@gmail.com>
Cc: Vivek Goyal <vgoyal@redhat.com>
CGRP_WAIT_ON_RMDIR is another kludge which was added to make cgroup
destruction rollback somewhat working. cgroup_rmdir() used to drain
CSS references and CGRP_WAIT_ON_RMDIR and the associated waitqueue and
helpers were used to allow the task performing rmdir to wait for the
next relevant event.
Unfortunately, the wait is visible to controllers too and the
mechanism got exposed to memcg by 887032670d ("cgroup avoid permanent
sleep at rmdir").
Now that the draining and retries are gone, CGRP_WAIT_ON_RMDIR is
unnecessary. Remove it and all the mechanisms supporting it. Note
that memcontrol.c changes are essentially revert of 887032670d
("cgroup avoid permanent sleep at rmdir").
Signed-off-by: Tejun Heo <tj@kernel.org>
Reviewed-by: Michal Hocko <mhocko@suse.cz>
Reviewed-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Acked-by: Li Zefan <lizefan@huawei.com>
Cc: Balbir Singh <bsingharora@gmail.com>
Because ->pre_destroy() could fail and can't be called under
cgroup_mutex, cgroup destruction did something very ugly.
1. Grab cgroup_mutex and verify it can be destroyed; fail otherwise.
2. Release cgroup_mutex and call ->pre_destroy().
3. Re-grab cgroup_mutex and verify it can still be destroyed; fail
otherwise.
4. Continue destroying.
In addition to being ugly, it has been always broken in various ways.
For example, memcg ->pre_destroy() expects the cgroup to be inactive
after it's done but tasks can be attached and detached between #2 and
#3 and the conditions that memcg verified in ->pre_destroy() might no
longer hold by the time control reaches #3.
Now that ->pre_destroy() is no longer allowed to fail. We can switch
to the following.
1. Grab cgroup_mutex and verify it can be destroyed; fail otherwise.
2. Deactivate CSS's and mark the cgroup removed thus preventing any
further operations which can invalidate the verification from #1.
3. Release cgroup_mutex and call ->pre_destroy().
4. Re-grab cgroup_mutex and continue destroying.
After this change, controllers can safely assume that ->pre_destroy()
will only be called only once for a given cgroup and, once
->pre_destroy() is called, the cgroup will stay dormant till it's
destroyed.
This removes the only reason ->pre_destroy() can fail - new task being
attached or child cgroup being created inbetween. Error out path is
removed and ->pre_destroy() invocation is open coded in
cgroup_rmdir().
v2: cgroup_call_pre_destroy() removal moved to this patch per Michal.
Commit message updated per Glauber.
Signed-off-by: Tejun Heo <tj@kernel.org>
Reviewed-by: Michal Hocko <mhocko@suse.cz>
Reviewed-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Acked-by: Li Zefan <lizefan@huawei.com>
Cc: Glauber Costa <glommer@parallels.com>
This patch makes cgroup_create() fail if @parent is marked removed.
This is to prepare for further updates to cgroup_rmdir() path.
Note that this change isn't strictly necessary. cgroup can only be
created via mkdir and the removed marking and dentry removal happen
without releasing cgroup_mutex, so cgroup_create() can never race with
cgroup_rmdir(). Even after the scheduled updates to cgroup_rmdir(),
cgroup_mkdir() and cgroup_rmdir() are synchronized by i_mutex
rendering the added liveliness check unnecessary.
Do it anyway such that locking is contained inside cgroup proper and
we don't get nasty surprises if we ever grow another caller of
cgroup_create().
Signed-off-by: Tejun Heo <tj@kernel.org>
Reviewed-by: Michal Hocko <mhocko@suse.cz>
Reviewed-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Acked-by: Li Zefan <lizefan@huawei.com>
CSS_REMOVED is one of the several contortions which were necessary to
support css reference draining on cgroup removal. All css->refcnts
which need draining should be deactivated and verified to equal zero
atomically w.r.t. css_tryget(). If any one isn't zero, all refcnts
needed to be re-activated and css_tryget() shouldn't fail in the
process.
This was achieved by letting css_tryget() busy-loop until either the
refcnt is reactivated (failed removal attempt) or CSS_REMOVED is set
(committing to removal).
Now that css refcnt draining is no longer used, there's no need for
atomic rollback mechanism. css_tryget() simply can look at the
reference count and fail if it's deactivated - it's never getting
re-activated.
This patch removes CSS_REMOVED and updates __css_tryget() to fail if
the refcnt is deactivated. As deactivation and removal are a single
step now, they no longer need to be protected against css_tryget()
happening from irq context. Remove local_irq_disable/enable() from
cgroup_rmdir().
Note that this removes css_is_removed() whose only user is VM_BUG_ON()
in memcontrol.c. We can replace it with a check on the refcnt but
given that the only use case is a debug assert, I think it's better to
simply unexport it.
v2: Comment updated and explanation on local_irq_disable/enable()
added per Michal Hocko.
Signed-off-by: Tejun Heo <tj@kernel.org>
Reviewed-by: Michal Hocko <mhocko@suse.cz>
Reviewed-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Acked-by: Li Zefan <lizefan@huawei.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Balbir Singh <bsingharora@gmail.com>
2ef37d3fe4 ("memcg: Simplify mem_cgroup_force_empty_list error
handling") removed the last user of __DEPRECATED_clear_css_refs. This
patch removes __DEPRECATED_clear_css_refs and mechanisms to support
it.
* Conditionals dependent on __DEPRECATED_clear_css_refs removed.
* cgroup_clear_css_refs() can no longer fail. All that needs to be
done are deactivating refcnts, setting CSS_REMOVED and putting the
base reference on each css. Remove cgroup_clear_css_refs() and the
failure path, and open-code the loops into cgroup_rmdir().
This patch keeps the two for_each_subsys() loops separate while open
coding them. They can be merged now but there are scheduled changes
which need them to be separate, so keep them separate to reduce the
amount of churn.
local_irq_save/restore() from cgroup_clear_css_refs() are replaced
with local_irq_disable/enable() for simplicity. This is safe as
cgroup_rmdir() is always called with IRQ enabled. Note that this IRQ
switching is necessary to ensure that css_tryget() isn't called from
IRQ context on the same CPU while lower context is between CSS
deactivation and setting CSS_REMOVED as css_tryget() would hang
forever in such cases waiting for CSS to be re-activated or
CSS_REMOVED set. This will go away soon.
v2: cgroup_call_pre_destroy() removal dropped per Michal. Commit
message updated to explain local_irq_disable/enable() conversion.
Signed-off-by: Tejun Heo <tj@kernel.org>
Reviewed-by: Michal Hocko <mhocko@suse.cz>
Reviewed-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Acked-by: Li Zefan <lizefan@huawei.com>
This reverts commit 7e3aa30ac8.
The commit incorrectly assumed that fork path always performed
threadgroup_change_begin/end() and depended on that for
synchronization against task exit and cgroup migration paths instead
of explicitly grabbing task_lock().
threadgroup_change is not locked when forking a new process (as
opposed to a new thread in the same process) and even if it were it
wouldn't be effective as different processes use different threadgroup
locks.
Revert the incorrect optimization.
Signed-off-by: Tejun Heo <tj@kernel.org>
LKML-Reference: <20121008020000.GB2575@localhost>
Acked-by: Li Zefan <lizefan@huawei.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: stable@vger.kernel.org
This reverts commit 7e381b0eb1.
The commit incorrectly assumed that fork path always performed
threadgroup_change_begin/end() and depended on that for
synchronization against task exit and cgroup migration paths instead
of explicitly grabbing task_lock().
threadgroup_change is not locked when forking a new process (as
opposed to a new thread in the same process) and even if it were it
wouldn't be effective as different processes use different threadgroup
locks.
Revert the incorrect optimization.
Signed-off-by: Tejun Heo <tj@kernel.org>
LKML-Reference: <20121008020000.GB2575@localhost>
Acked-by: Li Zefan <lizefan@huawei.com>
Bitterly-Acked-by: Frederic Weisbecker <fweisbec@gmail.com>
Cc: stable@vger.kernel.org
notify_on_release must be triggered when the last process in a cgroup is
move to another. But if the first(and only) process in a cgroup is moved to
another, notify_on_release is not triggered.
# mkdir /cgroup/cpu/SRC
# mkdir /cgroup/cpu/DST
#
# echo 1 >/cgroup/cpu/SRC/notify_on_release
# echo 1 >/cgroup/cpu/DST/notify_on_release
#
# sleep 300 &
[1] 8629
#
# echo 8629 >/cgroup/cpu/SRC/tasks
# echo 8629 >/cgroup/cpu/DST/tasks
-> notify_on_release for /SRC must be triggered at this point,
but it isn't.
This is because put_css_set() is called before setting CGRP_RELEASABLE
in cgroup_task_migrate(), and is a regression introduce by the
commit:74a1166d(cgroups: make procs file writable), which was merged
into v3.0.
Cc: Ben Blum <bblum@andrew.cmu.edu>
Cc: <stable@vger.kernel.org> # v3.0.x and later
Acked-by: Li Zefan <lizefan@huawei.com>
Signed-off-by: Daisuke Nishimura <nishimura@mxp.nes.nec.co.jp>
Signed-off-by: Tejun Heo <tj@kernel.org>
cgroup core has a bug which violates a basic rule about event
notifications - when a new entity needs to be added, you add that to
the notification list first and then make the new entity conform to
the current state. If done in the reverse order, an event happening
inbetween will be lost.
cgroup_subsys->fork() is invoked way before the new task is added to
the css_set. Currently, cgroup_freezer is the only user of ->fork()
and uses it to make new tasks conform to the current state of the
freezer. If FROZEN state is requested while fork is in progress
between cgroup_fork_callbacks() and cgroup_post_fork(), the child
could escape freezing - the cgroup isn't frozen when ->fork() is
called and the freezer couldn't see the new task on the css_set.
This patch moves cgroup_subsys->fork() invocation to
cgroup_post_fork() after the new task is added to the css_set.
cgroup_fork_callbacks() is removed.
Because now a task may be migrated during cgroup_subsys->fork(),
freezer_fork() is updated so that it adheres to the usual RCU locking
and the rather pointless comment on why locking can be different there
is removed (if it doesn't make anything simpler, why even bother?).
Signed-off-by: Tejun Heo <tj@kernel.org>
Cc: Oleg Nesterov <oleg@redhat.com>
Cc: Rafael J. Wysocki <rjw@sisk.pl>
Cc: stable@vger.kernel.org
Pull cgroup hierarchy update from Tejun Heo:
"Currently, different cgroup subsystems handle nested cgroups
completely differently. There's no consistency among subsystems and
the behaviors often are outright broken.
People at least seem to agree that the broken hierarhcy behaviors need
to be weeded out if any progress is gonna be made on this front and
that the fallouts from deprecating the broken behaviors should be
acceptable especially given that the current behaviors don't make much
sense when nested.
This patch makes cgroup emit warning messages if cgroups for
subsystems with broken hierarchy behavior are nested to prepare for
fixing them in the future. This was put in a separate branch because
more related changes were expected (didn't make it this round) and the
memory cgroup wanted to pull in this and make changes on top."
* 'for-3.7-hierarchy' of git://git.kernel.org/pub/scm/linux/kernel/git/tj/cgroup:
cgroup: mark subsystems with broken hierarchy support and whine if cgroups are nested for them
Currently, cgroup hierarchy support is a mess. cpu related subsystems
behave correctly - configuration, accounting and control on a parent
properly cover its children. blkio and freezer completely ignore
hierarchy and treat all cgroups as if they're directly under the root
cgroup. Others show yet different behaviors.
These differing interpretations of cgroup hierarchy make using cgroup
confusing and it impossible to co-mount controllers into the same
hierarchy and obtain sane behavior.
Eventually, we want full hierarchy support from all subsystems and
probably a unified hierarchy. Users using separate hierarchies
expecting completely different behaviors depending on the mounted
subsystem is deterimental to making any progress on this front.
This patch adds cgroup_subsys.broken_hierarchy and sets it to %true
for controllers which are lacking in hierarchy support. The goal of
this patch is two-fold.
* Move users away from using hierarchy on currently non-hierarchical
subsystems, so that implementing proper hierarchy support on those
doesn't surprise them.
* Keep track of which controllers are broken how and nudge the
subsystems to implement proper hierarchy support.
For now, start with a single warning message. We can whine louder
later on.
v2: Fixed a typo spotted by Michal. Warning message updated.
v3: Updated memcg part so that it doesn't generate warning in the
cases where .use_hierarchy=false doesn't make the behavior
different from root.use_hierarchy=true. Fixed a typo spotted by
Glauber.
v4: Check ->broken_hierarchy after cgroup creation is complete so that
->create() can affect the result per Michal. Dropped unnecessary
memcg root handling per Michal.
Signed-off-by: Tejun Heo <tj@kernel.org>
Acked-by: Michal Hocko <mhocko@suse.cz>
Acked-by: Li Zefan <lizefan@huawei.com>
Acked-by: Serge E. Hallyn <serue@us.ibm.com>
Cc: Glauber Costa <glommer@parallels.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Paul Turner <pjt@google.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Thomas Graf <tgraf@suug.ch>
Cc: Vivek Goyal <vgoyal@redhat.com>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Arnaldo Carvalho de Melo <acme@ghostprotocols.net>
Cc: Neil Horman <nhorman@tuxdriver.com>
Cc: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
WARNING: With this change it is impossible to load external built
controllers anymore.
In case where CONFIG_NETPRIO_CGROUP=m and CONFIG_NET_CLS_CGROUP=m is
set, corresponding subsys_id should also be a constant. Up to now,
net_prio_subsys_id and net_cls_subsys_id would be of the type int and
the value would be assigned during runtime.
By switching the macro definition IS_SUBSYS_ENABLED from IS_BUILTIN
to IS_ENABLED, all *_subsys_id will have constant value. That means we
need to remove all the code which assumes a value can be assigned to
net_prio_subsys_id and net_cls_subsys_id.
A close look is necessary on the RCU part which was introduces by
following patch:
commit f845172531
Author: Herbert Xu <herbert@gondor.apana.org.au> Mon May 24 09:12:34 2010
Committer: David S. Miller <davem@davemloft.net> Mon May 24 09:12:34 2010
cls_cgroup: Store classid in struct sock
Tis code was added to init_cgroup_cls()
/* We can't use rcu_assign_pointer because this is an int. */
smp_wmb();
net_cls_subsys_id = net_cls_subsys.subsys_id;
respectively to exit_cgroup_cls()
net_cls_subsys_id = -1;
synchronize_rcu();
and in module version of task_cls_classid()
rcu_read_lock();
id = rcu_dereference(net_cls_subsys_id);
if (id >= 0)
classid = container_of(task_subsys_state(p, id),
struct cgroup_cls_state, css)->classid;
rcu_read_unlock();
Without an explicit explaination why the RCU part is needed. (The
rcu_deference was fixed by exchanging it to rcu_derefence_index_check()
in a later commit, but that is a minor detail.)
So here is my pondering why it was introduced and why it safe to
remove it now. Note that this code was copied over to net_prio the
reasoning holds for that subsystem too.
The idea behind the RCU use for net_cls_subsys_id is to make sure we
get a valid pointer back from task_subsys_state(). task_subsys_state()
is just blindly accessing the subsys array and returning the
pointer. Obviously, passing in -1 as id into task_subsys_state()
returns an invalid value (out of lower bound).
So this code makes sure that only after module is loaded and the
subsystem registered, the id is assigned.
Before unregistering the module all old readers must have left the
critical section. This is done by assigning -1 to the id and issuing a
synchronized_rcu(). Any new readers wont call task_subsys_state()
anymore and therefore it is safe to unregister the subsystem.
The new code relies on the same trick, but it looks at the subsys
pointer return by task_subsys_state() (remember the id is constant
and therefore we allways have a valid index into the subsys
array).
No precautions need to be taken during module loading
module. Eventually, all CPUs will get a valid pointer back from
task_subsys_state() because rebind_subsystem() which is called after
the module init() function will assigned subsys[net_cls_subsys_id] the
newly loaded module subsystem pointer.
When the subsystem is about to be removed, rebind_subsystem() will
called before the module exit() function. In this case,
rebind_subsys() will assign subsys[net_cls_subsys_id] a NULL pointer
and then it calls synchronize_rcu(). All old readers have left by then
the critical section. Any new reader wont access the subsystem
anymore. At this point we are safe to unregister the subsystem. No
synchronize_rcu() call is needed.
Signed-off-by: Daniel Wagner <daniel.wagner@bmw-carit.de>
Signed-off-by: Tejun Heo <tj@kernel.org>
Acked-by: Li Zefan <lizefan@huawei.com>
Acked-by: Neil Horman <nhorman@tuxdriver.com>
Cc: "David S. Miller" <davem@davemloft.net>
Cc: "Paul E. McKenney" <paulmck@linux.vnet.ibm.com>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Eric Dumazet <edumazet@google.com>
Cc: Gao feng <gaofeng@cn.fujitsu.com>
Cc: Glauber Costa <glommer@parallels.com>
Cc: Herbert Xu <herbert@gondor.apana.org.au>
Cc: Jamal Hadi Salim <jhs@mojatatu.com>
Cc: John Fastabend <john.r.fastabend@intel.com>
Cc: Kamezawa Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: netdev@vger.kernel.org
Cc: cgroups@vger.kernel.org
The *_subsys_id will be used as index to access the subsys. Therefore
we need to care we populate the subsystem at the correct position by
using designated initialization.
With this change we are able to interleave builtin and modules in the subsys
array.
Signed-off-by: Daniel Wagner <daniel.wagner@bmw-carit.de>
Signed-off-by: Tejun Heo <tj@kernel.org>
Acked-by: Li Zefan <lizefan@huawei.com>
Acked-by: Neil Horman <nhorman@tuxdriver.com>
Cc: Gao feng <gaofeng@cn.fujitsu.com>
Cc: Jamal Hadi Salim <jhs@mojatatu.com>
Cc: John Fastabend <john.r.fastabend@intel.com>
Cc: netdev@vger.kernel.org
Cc: cgroups@vger.kernel.org
Before we are able to define all subsystem ids at compile time we need
a more fine grained control what gets defined when we include
cgroup_subsys.h. For example we define the enums for the subsystems or
to declare for struct cgroup_subsys (builtin subsystem) by including
cgroup_subsys.h and defining SUBSYS accordingly.
Currently, the decision if a subsys is used is defined inside the
header by testing if CONFIG_*=y is true. By moving this test outside
of cgroup_subsys.h we are able to control it on the include level.
This is done by introducing IS_SUBSYS_ENABLED which then is defined
according the task, e.g. is CONFIG_*=y or CONFIG_*=m.
Signed-off-by: Daniel Wagner <daniel.wagner@bmw-carit.de>
Signed-off-by: Tejun Heo <tj@kernel.org>
Acked-by: Li Zefan <lizefan@huawei.com>
Acked-by: Neil Horman <nhorman@tuxdriver.com>
Cc: Gao feng <gaofeng@cn.fujitsu.com>
Cc: Jamal Hadi Salim <jhs@mojatatu.com>
Cc: John Fastabend <john.r.fastabend@intel.com>
Cc: netdev@vger.kernel.org
Cc: cgroups@vger.kernel.org
CGROUP_BUILTIN_SUBSYS_COUNT is used as start index or stop index when
looping over the subsys array looking either at the builtin or the
module subsystems. Since all the builtin subsystems have an id which
is lower then CGROUP_BUILTIN_SUBSYS_COUNT we know that any module will
have an id larger than CGROUP_BUILTIN_SUBSYS_COUNT. In short the ids
are sorted.
We are about to change id assignment to happen only at compile time
later in this series. That means we can't rely on the above trick
since all ids will always be defined at compile time. Furthermore,
ordering the builtin subsystems and the module subsystems is not
really necessary.
So we need a different way to know which subsystem is a builtin or a
module one. We can use the subsys[]->module pointer for this. Any
place where we need to know if a subsys is module we just check for
the pointer. If it is NULL then the subsystem is a builtin one.
With this we are able to drop the CGROUP_BUILTIN_SUBSYS_COUNT
enum. Though we need to introduce a temporary placeholder so that we
don't get a compilation error when only CONFIG_CGROUP is selected and
no single controller. An empty enum definition is not valid. Later in
this series we are able to remove the placeholder again.
And with this change we get a fix for this:
kernel/cgroup.c: In function ‘cgroup_load_subsys’:
kernel/cgroup.c:4326:38: warning: array subscript is below array bounds [-Warray-bounds]
when CONFIG_CGROUP=y and no built in controller was enabled.
Signed-off-by: Daniel Wagner <daniel.wagner@bmw-carit.de>
Signed-off-by: Tejun Heo <tj@kernel.org>
Acked-by: Li Zefan <lizefan@huawei.com>
Acked-by: Neil Horman <nhorman@tuxdriver.com>
Cc: Gao feng <gaofeng@cn.fujitsu.com>
Cc: Jamal Hadi Salim <jhs@mojatatu.com>
Cc: John Fastabend <john.r.fastabend@intel.com>
Cc: netdev@vger.kernel.org
Cc: cgroups@vger.kernel.org
This is one of the items in the plumber's wish list.
For use cases:
>> What would the use case be for this?
>
> Attaching meta information to services, in an easily discoverable
> way. For example, in systemd we create one cgroup for each service, and
> could then store data like the main pid of the specific service as an
> xattr on the cgroup itself. That way we'd have almost all service state
> in the cgroupfs, which would make it possible to terminate systemd and
> later restart it without losing any state information. But there's more:
> for example, some very peculiar services cannot be terminated on
> shutdown (i.e. fakeraid DM stuff) and it would be really nice if the
> services in question could just mark that on their cgroup, by setting an
> xattr. On the more desktopy side of things there are other
> possibilities: for example there are plans defining what an application
> is along the lines of a cgroup (i.e. an app being a collection of
> processes). With xattrs one could then attach an icon or human readable
> program name on the cgroup.
>
> The key idea is that this would allow attaching runtime meta information
> to cgroups and everything they model (services, apps, vms), that doesn't
> need any complex userspace infrastructure, has good access control
> (i.e. because the file system enforces that anyway, and there's the
> "trusted." xattr namespace), notifications (inotify), and can easily be
> shared among applications.
>
> Lennart
v7:
- no changes
v6:
- remove user xattr namespace, only allow trusted and security
v5:
- check for capabilities before setting/removing xattrs
v4:
- no changes
v3:
- instead of config option, use mount option to enable xattr support
Original-patch-by: Li Zefan <lizefan@huawei.com>
Cc: Li Zefan <lizefan@huawei.com>
Cc: Tejun Heo <tj@kernel.org>
Cc: Hugh Dickins <hughd@google.com>
Cc: Hillf Danton <dhillf@gmail.com>
Cc: Lennart Poettering <lpoetter@redhat.com>
Signed-off-by: Li Zefan <lizefan@huawei.com>
Signed-off-by: Aristeu Rozanski <aris@redhat.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
When remounting cgroupfs with some subsystems added to it and some
removed, cgroup will remove all the files in root directory and then
re-popluate it.
What I'm doing here is, only remove files which belong to subsystems that
are to be unbinded, and only create files for newly-added subsystems.
The purpose is to have all other files untouched.
This is a preparation for cgroup xattr support.
v7:
- checkpatch warnings fixed
v6:
- no changes
v5:
- no changes
v4:
- refactored cgroup_clear_directory() to not use cgroup_rm_file()
- instead of going thru the list of files, get the file list using the
subsystems
- use 'subsys_mask' instead of {added,removed}_bits and made
cgroup_populate_dir() to match the parameters with cgroup_clear_directory()
v3:
- refresh patches after recent refactoring
Original-patch-by: Li Zefan <lizefan@huawei.com>
Cc: Li Zefan <lizefan@huawei.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Hillf Danton <dhillf@gmail.com>
Cc: Lennart Poettering <lpoetter@redhat.com>
Signed-off-by: Li Zefan <lizefan@huawei.com>
Signed-off-by: Aristeu Rozanski <aris@redhat.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
Pull cgroup changes from Tejun Heo:
"Nothing too interesting. A minor bug fix and some cleanups."
* 'for-3.6' of git://git.kernel.org/pub/scm/linux/kernel/git/tj/cgroup:
cgroup: Update remount documentation
cgroup: cgroup_rm_files() was calling simple_unlink() with the wrong inode
cgroup: Remove populate() documentation
cgroup: remove hierarchy_mutex
Pass mount flags to sget() so that it can use them in initialising a new
superblock before the set function is called. They could also be passed to the
compare function.
Signed-off-by: David Howells <dhowells@redhat.com>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Just the flags; only NFS cares even about that, but there are
legitimate uses for such argument. And getting rid of that
completely would require splitting ->lookup() into a couple
of methods (at least), so let's leave that alone for now...
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
While refactoring cgroup file removal path, 05ef1d7c4a "cgroup:
introduce struct cfent" incorrectly changed the @dir argument of
simple_unlink() to the inode of the file being deleted instead of that
of the containing directory.
The effect of this bug is minor - ctime and mtime of the parent
weren't properly updated on file deletion.
Fix it by using @cgrp->dentry->d_inode instead.
Signed-off-by: Tejun Heo <tj@kernel.org>
Reported-by: Al Viro <viro@ZenIV.linux.org.uk>
Acked-by: Li Zefan <lizefan@huawei.com>
Cc: stable@vger.kernel.org
48ddbe1946 "cgroup: make css->refcnt clearing on cgroup removal
optional" allowed a css to linger after the associated cgroup is
removed. As a css holds a reference on the cgroup's dentry, it means
that cgroup dentries may linger for a while.
Destroying a superblock which has dentries with positive refcnts is a
critical bug and triggers BUG() in vfs code. As each cgroup dentry
holds an s_active reference, any lingering cgroup has both its dentry
and the superblock pinned and thus preventing premature release of
superblock.
Unfortunately, after 48ddbe1946, there's a small window while
releasing a cgroup which is directly under the root of the hierarchy.
When a cgroup directory is released, vfs layer first deletes the
corresponding dentry and then invokes dput() on the parent, which may
recurse further, so when a cgroup directly below root cgroup is
released, the cgroup is first destroyed - which releases the s_active
it was holding - and then the dentry for the root cgroup is dput().
This creates a window where the root dentry's refcnt isn't zero but
superblock's s_active is. If umount happens before or during this
window, vfs will see the root dentry with non-zero refcnt and trigger
BUG().
Before 48ddbe1946, this problem didn't exist because the last dentry
reference was guaranteed to be put synchronously from rmdir(2)
invocation which holds s_active around the whole process.
Fix it by holding an extra superblock->s_active reference across
dput() from css release, which is the dput() path added by 48ddbe1946
and the only one which doesn't hold an extra s_active ref across the
final cgroup dput().
Signed-off-by: Tejun Heo <tj@kernel.org>
LKML-Reference: <4FEEA5CB.8070809@huawei.com>
Reported-by: shyju pv <shyju.pv@huawei.com>
Tested-by: shyju pv <shyju.pv@huawei.com>
Cc: Sasha Levin <levinsasha928@gmail.com>
Acked-by: Li Zefan <lizefan@huawei.com>
This reverts commit fa980ca87d. The
commit was an attempt to fix a race condition where a cgroup hierarchy
may be unmounted with positive dentry reference on root cgroup. While
the commit made the race condition slightly more difficult to trigger,
the race was still there and could be reliably triggered using a
different test case.
Revert the incorrect fix. The next commit will describe the race and
fix it correctly.
Signed-off-by: Tejun Heo <tj@kernel.org>
LKML-Reference: <4FEEA5CB.8070809@huawei.com>
Reported-by: shyju pv <shyju.pv@huawei.com>
Cc: Sasha Levin <levinsasha928@gmail.com>
Acked-by: Li Zefan <lizefan@huawei.com>
When we fixed the race between atomic_dec and css_refcnt, we missed
the fact that css_refcnt internally subtracts CSS_DEACT_BIAS to get
the actual reference count. This can potentially cause a refcount leak
if __css_put races with cgroup_clear_css_refs.
Signed-off-by: Salman Qazi <sqazi@google.com>
Acked-by: Li Zefan <lizefan@huawei.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
It was introduced for memcg to iterate cgroup hierarchy without
holding cgroup_mutex, but soon after that it was replaced with
a lockless way in memcg.
No one used hierarchy_mutex since that, so remove it.
Signed-off-by: Li Zefan <lizefan@huawei.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
__css_put is using atomic_dec on the ref count, and then
looking at the ref count to make decisions. This is prone
to races, as someone else may decrement ref count between
our decrement and our decision. Instead, we should base our
decisions on the value that we decremented the ref count to.
(This results in an actual race on Google's kernel which I
haven't been able to reproduce on the upstream kernel. Having
said that, it's still incorrect by inspection).
Signed-off-by: Salman Qazi <sqazi@google.com>
Acked-by: Li Zefan <lizefan@huawei.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
Cc: stable@vger.kernel.org
Pull cgroup fix from Tejun Heo:
"This fixes the possible premature superblock release on umount bug
mentioned during v3.5-rc1 pull request.
Originally, cgroup dentry destruction path assumed that cgroup dentry
didn't have any reference left after cgroup removal thus put super
during dentry removal. Now that there can be lingering dentry
references, this led to super being put with live dentries. This
patch fixes the problem by putting super ref on dentry release instead
of removal."
* 'for-3.5-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/tj/cgroup:
cgroup: superblock can't be released with active dentries
Library functions should not grab locks when the callsites can do it,
even if the lock nests like the rcu read-side lock does.
Push the rcu_read_lock() from css_is_ancestor() to its single user,
mem_cgroup_same_or_subtree() in preparation for another user that may
already hold the rcu read-side lock.
Signed-off-by: Johannes Weiner <hannes@cmpxchg.org>
Cc: Konstantin Khlebnikov <khlebnikov@openvz.org>
Acked-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Acked-by: Michal Hocko <mhocko@suse.cz>
Acked-by: Li Zefan <lizf@cn.fujitsu.com>
Cc: Li Zefan <lizf@cn.fujitsu.com>
Cc: Tejun Heo <tj@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
48ddbe1946 "cgroup: make css->refcnt clearing on cgroup removal
optional" allowed a css to linger after the associated cgroup is
removed. As a css holds a reference on the cgroup's dentry, it means
that cgroup dentries may linger for a while.
cgroup_create() does grab an active reference on the superblock to
prevent it from going away while there are !root cgroups; however, the
reference is put from cgroup_diput() which is invoked on cgroup
removal, so cgroup dentries which are removed but persisting due to
lingering csses already have released their superblock active refs
allowing superblock to be killed while those dentries are around.
Given the right condition, this makes cgroup_kill_sb() call
kill_litter_super() with dentries with non-zero d_count leading to
BUG() in shrink_dcache_for_umount_subtree().
Fix it by adding cgroup_dops->d_release() operation and moving
deactivate_super() to it. cgroup_diput() now marks dentry->d_fsdata
with itself if superblock should be deactivated and cgroup_d_release()
deactivates the superblock on dentry release.
Signed-off-by: Tejun Heo <tj@kernel.org>
Reported-by: Sasha Levin <levinsasha928@gmail.com>
Tested-by: Sasha Levin <levinsasha928@gmail.com>
LKML-Reference: <CA+1xoqe5hMuxzCRhMy7J0XchDk2ZnuxOHJKikROk1-ReAzcT6g@mail.gmail.com>
Acked-by: Li Zefan <lizefan@huawei.com>
Pull user namespace enhancements from Eric Biederman:
"This is a course correction for the user namespace, so that we can
reach an inexpensive, maintainable, and reasonably complete
implementation.
Highlights:
- Config guards make it impossible to enable the user namespace and
code that has not been converted to be user namespace safe.
- Use of the new kuid_t type ensures the if you somehow get past the
config guards the kernel will encounter type errors if you enable
user namespaces and attempt to compile in code whose permission
checks have not been updated to be user namespace safe.
- All uids from child user namespaces are mapped into the initial
user namespace before they are processed. Removing the need to add
an additional check to see if the user namespace of the compared
uids remains the same.
- With the user namespaces compiled out the performance is as good or
better than it is today.
- For most operations absolutely nothing changes performance or
operationally with the user namespace enabled.
- The worst case performance I could come up with was timing 1
billion cache cold stat operations with the user namespace code
enabled. This went from 156s to 164s on my laptop (or 156ns to
164ns per stat operation).
- (uid_t)-1 and (gid_t)-1 are reserved as an internal error value.
Most uid/gid setting system calls treat these value specially
anyway so attempting to use -1 as a uid would likely cause
entertaining failures in userspace.
- If setuid is called with a uid that can not be mapped setuid fails.
I have looked at sendmail, login, ssh and every other program I
could think of that would call setuid and they all check for and
handle the case where setuid fails.
- If stat or a similar system call is called from a context in which
we can not map a uid we lie and return overflowuid. The LFS
experience suggests not lying and returning an error code might be
better, but the historical precedent with uids is different and I
can not think of anything that would break by lying about a uid we
can't map.
- Capabilities are localized to the current user namespace making it
safe to give the initial user in a user namespace all capabilities.
My git tree covers all of the modifications needed to convert the core
kernel and enough changes to make a system bootable to runlevel 1."
Fix up trivial conflicts due to nearby independent changes in fs/stat.c
* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/ebiederm/user-namespace: (46 commits)
userns: Silence silly gcc warning.
cred: use correct cred accessor with regards to rcu read lock
userns: Convert the move_pages, and migrate_pages permission checks to use uid_eq
userns: Convert cgroup permission checks to use uid_eq
userns: Convert tmpfs to use kuid and kgid where appropriate
userns: Convert sysfs to use kgid/kuid where appropriate
userns: Convert sysctl permission checks to use kuid and kgids.
userns: Convert proc to use kuid/kgid where appropriate
userns: Convert ext4 to user kuid/kgid where appropriate
userns: Convert ext3 to use kuid/kgid where appropriate
userns: Convert ext2 to use kuid/kgid where appropriate.
userns: Convert devpts to use kuid/kgid where appropriate
userns: Convert binary formats to use kuid/kgid where appropriate
userns: Add negative depends on entries to avoid building code that is userns unsafe
userns: signal remove unnecessary map_cred_ns
userns: Teach inode_capable to understand inodes whose uids map to other namespaces.
userns: Fail exec for suid and sgid binaries with ids outside our user namespace.
userns: Convert stat to return values mapped from kuids and kgids
userns: Convert user specfied uids and gids in chown into kuids and kgid
userns: Use uid_eq gid_eq helpers when comparing kuids and kgids in the vfs
...
