Commit 4be2c95d ("taskstats: pad taskstats netlink response for aligment
issues on ia64") added a null field to align the taskstats structure but
the discussion centered around ia64. The issue exists on other platforms
with inefficient unaligned access and adding them piecemeal would be an
unmaintainable mess.
This patch uses Dave Miller's suggestion of using a combination of
CONFIG_64BIT && !CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS to determine
whether alignment is needed.
Note that this will cause breakage on those platforms with applications
like iotop which had hard-coded offsets into the packet to access the
taskstats structure.
The message seen on systems without the alignment fixes looks like: kernel
unaligned access to 0xe000023879dca9bc, ip=0xa000000100133d10
The addresses may vary but resolve to locations inside __delayacct_add_tsk.
iotop makes what I'd call unreasonable assumptions about the contents of a
netlink genetlink packet containing generic attributes. They're typed and
have headers that specify value lengths, so the client can (should)
identify and skip the ones the client doesn't understand.
The kernel, as of version 2.6.36, presented a packet like so:
+--------------------------------+
| genlmsghdr - 4 bytes |
+--------------------------------+
| NLA header - 4 bytes | /* Aggregate header */
+-+------------------------------+
| | NLA header - 4 bytes | /* PID header */
| +------------------------------+
| | pid/tgid - 4 bytes |
| +------------------------------+
| | NLA header - 4 bytes | /* stats header */
| + -----------------------------+ <- oops. aligned on 4 byte boundary
| | struct taskstats - 328 bytes |
+-+------------------------------+
The iotop code expects that the kernel will behave as it did then,
assuming that the packet format is set in stone. The format is set in
stone, but the packet offsets are not. There's nothing in the packet
format that guarantees that the packet will be sent in exactly the same
way. The attribute contents are set (or versioned) and the aggregate
contents are set but they can be anywhere in the packet.
The issue here isn't that an unaligned structure gets passed to userspace,
it's that the NLA infrastructure has something of a weakness: The 4 byte
attribute header may force the payload to be unaligned. The taskstats
structure is created at an unaligned location and then 64-bit values are
operated on inside the kernel, so the unaligned access warnings gets
spewed everywhere.
It's possible to use the unaligned access API to operate on the structure
in the kernel but it seems like a wasted effort to work around userspace
code that isn't following the packet format. Any new additions would also
need the be worked around. It's a maintenance nightmare.
The conclusion of the earlier discussion seemed to be "ok fine, if we have
to break it, don't break it on arches that don't have the problem." Dave
pointed out that the unaligned access problem doesn't only exist on ia64,
but also on other 64-bit arches that don't have efficient unaligned access
and it should be fixed there as well. The committed version of the patch
and this addition keep with the conclusion of that discussion not to break
it unnecessarily, which the pid padding and the packet padding fixes did
do. x86_64 and powerpc don't suffer this problem so they shouldn't suffer
the solution. Other 64-bit architectures do and will, though.
Signed-off-by: Jeff Mahoney <jeffm@suse.com>
Reported-by: David S. Miller <davem@davemloft.net>
Acked-by: David S. Miller <davem@davemloft.net>
Cc: Dan Carpenter <error27@gmail.com>
Cc: Balbir Singh <balbir@in.ibm.com>
Cc: Florian Mickler <florian@mickler.org>
Cc: Guillaume Chazarain <guichaz@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
* 'for-2.6.38' of git://git.kernel.org/pub/scm/linux/kernel/git/tj/percpu: (30 commits)
gameport: use this_cpu_read instead of lookup
x86: udelay: Use this_cpu_read to avoid address calculation
x86: Use this_cpu_inc_return for nmi counter
x86: Replace uses of current_cpu_data with this_cpu ops
x86: Use this_cpu_ops to optimize code
vmstat: User per cpu atomics to avoid interrupt disable / enable
irq_work: Use per cpu atomics instead of regular atomics
cpuops: Use cmpxchg for xchg to avoid lock semantics
x86: this_cpu_cmpxchg and this_cpu_xchg operations
percpu: Generic this_cpu_cmpxchg() and this_cpu_xchg support
percpu,x86: relocate this_cpu_add_return() and friends
connector: Use this_cpu operations
xen: Use this_cpu_inc_return
taskstats: Use this_cpu_ops
random: Use this_cpu_inc_return
fs: Use this_cpu_inc_return in buffer.c
highmem: Use this_cpu_xx_return() operations
vmstat: Use this_cpu_inc_return for vm statistics
x86: Support for this_cpu_add, sub, dec, inc_return
percpu: Generic support for this_cpu_add, sub, dec, inc_return
...
Fixed up conflicts: in arch/x86/kernel/{apic/nmi.c, apic/x2apic_uv_x.c, process.c}
as per Tejun.
The taskstats structure is internally aligned on 8 byte boundaries but the
layout of the aggregrate reply, with two NLA headers and the pid (each 4
bytes), actually force the entire structure to be unaligned. This causes
the kernel to issue unaligned access warnings on some architectures like
ia64. Unfortunately, some software out there doesn't properly unroll the
NLA packet and assumes that the start of the taskstats structure will
always be 20 bytes from the start of the netlink payload. Aligning the
start of the taskstats structure breaks this software, which we don't
want. So, for now the alignment only happens on architectures that
require it and those users will have to update to fixed versions of those
packages. Space is reserved in the packet only when needed. This ifdef
should be removed in several years e.g. 2012 once we can be confident
that fixed versions are installed on most systems. We add the padding
before the aggregate since the aggregate is already a defined type.
Commit 85893120 ("delayacct: align to 8 byte boundary on 64-bit systems")
previously addressed the alignment issues by padding out the pid field.
This was supposed to be a compatible change but the circumstances
described above mean that it wasn't. This patch backs out that change,
since it was a hack, and introduces a new NULL attribute type to provide
the padding. Padding the response with 4 bytes avoids allocating an
aligned taskstats structure and copying it back. Since the structure
weighs in at 328 bytes, it's too big to do it on the stack.
Signed-off-by: Jeff Mahoney <jeffm@suse.com>
Reported-by: Brian Rogers <brian@xyzw.org>
Cc: Jeff Mahoney <jeffm@suse.com>
Cc: Guillaume Chazarain <guichaz@gmail.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>
Use this_cpu_inc_return in one place and avoid ugly __raw_get_cpu in
another.
V3->V4:
- Fix off by one.
V4-V4f:
- Use &listener_array
Cc: Michael Holzheu <holzheu@linux.vnet.ibm.com>
Acked-by: H. Peter Anvin <hpa@zytor.com>
Signed-off-by: Christoph Lameter <cl@linux.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
Separate the finding of a task_struct by pid or tgid from filling the
taskstats data. This makes the code more readable.
Signed-off-by: Michael Holzheu <holzheu@linux.vnet.ibm.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>
Move each taskstats command into a single function. This makes the code
more readable and makes it easier to add new commands.
Signed-off-by: Michael Holzheu <holzheu@linux.vnet.ibm.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>
prepare_reply() sets up an skb for the response. The payload contains:
+--------------------------------+
| genlmsghdr - 4 bytes |
+--------------------------------+
| NLA header - 4 bytes | /* Aggregate header */
+-+------------------------------+
| | NLA header - 4 bytes | /* PID header */
| +------------------------------+
| | pid/tgid - 4 bytes |
| +------------------------------+
| | NLA header - 4 bytes | /* stats header */
| + -----------------------------+ <- oops. aligned on 4 byte boundary
| | struct taskstats - 328 bytes |
+-+------------------------------+
The start of the taskstats struct must be 8 byte aligned on IA64 (and
other systems with 8 byte alignment rules for 64-bit types) or runtime
alignment warnings will be issued.
This patch pads the pid/tgid field out to sizeof(long), which forces the
alignment of taskstats. The getdelays userspace code is ok with this
since it assumes 32-bit pid/tgid and then honors that header's length
field.
An array is used to avoid exposing kernel memory contents to userspace in
the response.
Signed-off-by: Jeff Mahoney <jeffm@suse.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>
percpu.h is included by sched.h and module.h and thus ends up being
included when building most .c files. percpu.h includes slab.h which
in turn includes gfp.h making everything defined by the two files
universally available and complicating inclusion dependencies.
percpu.h -> slab.h dependency is about to be removed. Prepare for
this change by updating users of gfp and slab facilities include those
headers directly instead of assuming availability. As this conversion
needs to touch large number of source files, the following script is
used as the basis of conversion.
http://userweb.kernel.org/~tj/misc/slabh-sweep.py
The script does the followings.
* Scan files for gfp and slab usages and update includes such that
only the necessary includes are there. ie. if only gfp is used,
gfp.h, if slab is used, slab.h.
* When the script inserts a new include, it looks at the include
blocks and try to put the new include such that its order conforms
to its surrounding. It's put in the include block which contains
core kernel includes, in the same order that the rest are ordered -
alphabetical, Christmas tree, rev-Xmas-tree or at the end if there
doesn't seem to be any matching order.
* If the script can't find a place to put a new include (mostly
because the file doesn't have fitting include block), it prints out
an error message indicating which .h file needs to be added to the
file.
The conversion was done in the following steps.
1. The initial automatic conversion of all .c files updated slightly
over 4000 files, deleting around 700 includes and adding ~480 gfp.h
and ~3000 slab.h inclusions. The script emitted errors for ~400
files.
2. Each error was manually checked. Some didn't need the inclusion,
some needed manual addition while adding it to implementation .h or
embedding .c file was more appropriate for others. This step added
inclusions to around 150 files.
3. The script was run again and the output was compared to the edits
from #2 to make sure no file was left behind.
4. Several build tests were done and a couple of problems were fixed.
e.g. lib/decompress_*.c used malloc/free() wrappers around slab
APIs requiring slab.h to be added manually.
5. The script was run on all .h files but without automatically
editing them as sprinkling gfp.h and slab.h inclusions around .h
files could easily lead to inclusion dependency hell. Most gfp.h
inclusion directives were ignored as stuff from gfp.h was usually
wildly available and often used in preprocessor macros. Each
slab.h inclusion directive was examined and added manually as
necessary.
6. percpu.h was updated not to include slab.h.
7. Build test were done on the following configurations and failures
were fixed. CONFIG_GCOV_KERNEL was turned off for all tests (as my
distributed build env didn't work with gcov compiles) and a few
more options had to be turned off depending on archs to make things
build (like ipr on powerpc/64 which failed due to missing writeq).
* x86 and x86_64 UP and SMP allmodconfig and a custom test config.
* powerpc and powerpc64 SMP allmodconfig
* sparc and sparc64 SMP allmodconfig
* ia64 SMP allmodconfig
* s390 SMP allmodconfig
* alpha SMP allmodconfig
* um on x86_64 SMP allmodconfig
8. percpu.h modifications were reverted so that it could be applied as
a separate patch and serve as bisection point.
Given the fact that I had only a couple of failures from tests on step
6, I'm fairly confident about the coverage of this conversion patch.
If there is a breakage, it's likely to be something in one of the arch
headers which should be easily discoverable easily on most builds of
the specific arch.
Signed-off-by: Tejun Heo <tj@kernel.org>
Guess-its-ok-by: Christoph Lameter <cl@linux-foundation.org>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Lee Schermerhorn <Lee.Schermerhorn@hp.com>
Make remaining netlink policies as const.
Fixup coding style where needed.
Signed-off-by: Alexey Dobriyan <adobriyan@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This makes generic netlink network namespace aware. No
generic netlink families except for the controller family
are made namespace aware, they need to be checked one by
one and then set the family->netnsok member to true.
A new function genlmsg_multicast_netns() is introduced to
allow sending a multicast message in a given namespace,
for example when it applies to an object that lives in
that namespace, a new function genlmsg_multicast_allns()
to send a message to all network namespaces (for objects
that do not have an associated netns).
The function genlmsg_multicast() is changed to multicast
the message in just init_net, which is currently correct
for all generic netlink families since they only work in
init_net right now. Some will later want to work in all
net namespaces because they do not care about the netns
at all -- those will have to be converted to use one of
the new functions genlmsg_multicast_allns() or
genlmsg_multicast_netns() whenever they are made netns
aware in some way.
After this patch families can easily decide whether or
not they should be available in all net namespaces. Many
genl families us it for objects not related to networking
and should therefore be available in all namespaces, but
that will have to be done on a per family basis.
Note that this doesn't touch on the checkpoint/restart
problem where network namespaces could be used, genl
families and multicast groups are numbered globally and
I see no easy way of changing that, especially since it
must be possible to multicast to all network namespaces
for those families that do not care about netns.
Signed-off-by: Johannes Berg <johannes@sipsolutions.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
Impact: Reduce stack usage, use new cpumask API.
Mainly changing cpumask_t to 'struct cpumask' and similar simple API
conversion. Two conversions worth mentioning:
1) we use cpumask_any_but to avoid a temporary in kernel/softlockup.c,
2) Use cpumask_var_t in taskstats_user_cmd().
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
Signed-off-by: Mike Travis <travis@sgi.com>
Cc: Balbir Singh <balbir@linux.vnet.ibm.com>
Cc: Ingo Molnar <mingo@redhat.com>
Impact: change calling convention of existing cpumask APIs
Most cpumask functions started with cpus_: these have been replaced by
cpumask_ ones which take struct cpumask pointers as expected.
These four functions don't have good replacement names; fortunately
they're rarely used, so we just change them over.
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
Signed-off-by: Mike Travis <travis@sgi.com>
Acked-by: Ingo Molnar <mingo@elte.hu>
Cc: paulus@samba.org
Cc: mingo@redhat.com
Cc: tony.luck@intel.com
Cc: ralf@linux-mips.org
Cc: Greg Kroah-Hartman <gregkh@suse.de>
Cc: cl@linux-foundation.org
Cc: srostedt@redhat.com
Change references from for_each_cpu_mask to for_each_cpu_mask_nr
where appropriate
Reviewed-by: Paul Jackson <pj@sgi.com>
Reviewed-by: Christoph Lameter <clameter@sgi.com>
Signed-off-by: Mike Travis <travis@sgi.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
The pid to lookup a task by is passed inside taskstats code via genetlink
message.
Since netlink packets are now processed in the context of the sending task,
this is correct to lookup the task with find_task_by_vpid() here.
Besides, I fix the call to fill_pid() from taskstats_exit(), since the
tsk->pid is not required in fill_pid() in this case, and the pid field on
task_struct is going to be deprecated as well.
Signed-off-by: Pavel Emelyanov <xemul@openvz.org>
Cc: "Eric W. Biederman" <ebiederm@xmission.com>
Cc: Balbir Singh <balbir@in.ibm.com>
Cc: Jay Lan <jlan@engr.sgi.com>
Cc: Jonathan Lim <jlim@sgi.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
We'd better not nlmsg_free on a pointer containing an undefined value
(and without having anything allocated).
Spotted by the Coverity checker.
Signed-off-by: Adrian Bunk <bunk@kernel.org>
Acked-by: Balbir Singh <balbir@linux.vnet.ibm>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Fix the various misspellings of "system", controller", "interrupt" and
"[un]necessary".
Signed-off-by: Robert P. J. Day <rpjday@mindspring.com>
Signed-off-by: Adrian Bunk <bunk@kernel.org>
This patch is inspired by the discussion at
http://lkml.org/lkml/2007/4/11/187 and implements per cgroup statistics
as suggested by Andrew Morton in http://lkml.org/lkml/2007/4/11/263. The
patch is on top of 2.6.21-mm1 with Paul's cgroups v9 patches (forward
ported)
This patch implements per cgroup statistics infrastructure and re-uses
code from the taskstats interface. A new set of cgroup operations are
registered with commands and attributes. It should be very easy to
*extend* per cgroup statistics, by adding members to the cgroupstats
structure.
The current model for cgroupstats is a pull, a push model (to post
statistics on interesting events), should be very easy to add. Currently
user space requests for statistics by passing the cgroup file
descriptor. Statistics about the state of all the tasks in the cgroup
is returned to user space.
TODO's/NOTE:
This patch provides an infrastructure for implementing cgroup statistics.
Based on the needs of each controller, we can incrementally add more statistics,
event based support for notification of statistics, accumulation of taskstats
into cgroup statistics in the future.
Sample output
# ./cgroupstats -C /cgroup/a
sleeping 2, blocked 0, running 1, stopped 0, uninterruptible 0
# ./cgroupstats -C /cgroup/
sleeping 154, blocked 0, running 0, stopped 0, uninterruptible 0
If the approach looks good, I'll enhance and post the user space utility for
the same
Feedback, comments, test results are always welcome!
[akpm@linux-foundation.org: build fix]
Signed-off-by: Balbir Singh <balbir@linux.vnet.ibm.com>
Cc: Paul Menage <menage@google.com>
Cc: Jay Lan <jlan@engr.sgi.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
This patch cleans up duplicate includes in
kernel/
Signed-off-by: Jesper Juhl <jesper.juhl@gmail.com>
Acked-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Reviewed-by: Satyam Sharma <ssatyam@cse.iitk.ac.in>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Make available to the user the following task and process performance
statistics:
* Involuntary Context Switches (task_struct->nivcsw)
* Voluntary Context Switches (task_struct->nvcsw)
Statistics information is available from:
1. taskstats interface (Documentation/accounting/)
2. /proc/PID/status (task only).
This data is useful for detecting hyperactivity patterns between processes.
[akpm@linux-foundation.org: cleanup]
Signed-off-by: Maxim Uvarov <muvarov@ru.mvista.com>
Cc: Shailabh Nagar <nagar@watson.ibm.com>
Cc: Balbir Singh <balbir@in.ibm.com>
Cc: Jay Lan <jlan@engr.sgi.com>
Cc: Jonathan Lim <jlim@sgi.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
This patch provides a new macro
KMEM_CACHE(<struct>, <flags>)
to simplify slab creation. KMEM_CACHE creates a slab with the name of the
struct, with the size of the struct and with the alignment of the struct.
Additional slab flags may be specified if necessary.
Example
struct test_slab {
int a,b,c;
struct list_head;
} __cacheline_aligned_in_smp;
test_slab_cache = KMEM_CACHE(test_slab, SLAB_PANIC)
will create a new slab named "test_slab" of the size sizeof(struct
test_slab) and aligned to the alignment of test slab. If it fails then we
panic.
Signed-off-by: Christoph Lameter <clameter@sgi.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
For the common "(struct nlmsghdr *)skb->data" sequence, so that we reduce the
number of direct accesses to skb->data and for consistency with all the other
cast skb member helpers.
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Thomas Graf wrote:
>
> nla_nest_start() may return NULL, either rely on prepare_reply() to be
> correct and BUG() on failure or do proper error handling for all
> functions.
nla_put() in taskstat.c can fail only if the 'size' argument of alloc_skb()
was not right. This is a kernel bug, we should not hide it. So add 'BUG()'
on error path and check for 'na == NULL'.
> genlmsg_cancel() is only required in error paths for dumping
> procedures.
So we can remove 'genlmsg_cancel()' calls and 'void *reply' (saves 227 bytes).
Signed-off-by: Oleg Nesterov <oleg@tv-sign.ru>
Cc: Thomas Graf <tgraf@suug.ch>
Cc: Shailabh Nagar <nagar@watson.ibm.com>
Cc: Balbir Singh <balbir@in.ibm.com>
Cc: Jay Lan <jlan@sgi.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Currently taskstats_user_cmd()/taskstats_exit() do:
1) allocate stats
2) fill stats
3) make a temporary copy on stack (236 bytes)
4) copy that copy to skb
5) free stats
With the help of nla_reserve() we can operate on skb->data directly,
thus avoiding all these steps except 2).
So, before this patch:
// copy *stats to skb->data
int mk_reply(skb, ..., struct taskstats *stats);
fill_pid(stats);
mk_reply(skb, ..., stats);
After:
// return a pointer to skb->data
struct taskstats *mk_reply(skb, ...);
stat = mk_reply(skb, ...);
fill_pid(stats);
Shrinks taskatsks.o by 162 bytes.
A stupid benchmark (send one million TASKSTATS_CMD_ATTR_PID) shows the
real user sys
before:
4.02 0.06 3.96
4.02 0.04 3.98
4.02 0.04 3.97
after:
3.86 0.08 3.78
3.88 0.10 3.77
3.89 0.09 3.80
but this looks suspiciously good.
Signed-off-by: Oleg Nesterov <oleg@tv-sign.ru>
Acked-by: Shailabh Nagar <nagar@watson.ibm.com>
Cc: Balbir Singh <balbir@in.ibm.com>
Cc: Jay Lan <jlan@sgi.com>
Cc: Thomas Graf <tgraf@suug.ch>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Introduce mk_reply() helper which does all nla_put()s on reply.
Saves 453 bytes and a preparation for the next patch.
Signed-off-by: Oleg Nesterov <oleg@tv-sign.ru>
Acked-by: Shailabh Nagar <nagar@watson.ibm.com>
Cc: Balbir Singh <balbir@in.ibm.com>
Cc: Jay Lan <jlan@sgi.com>
Cc: Thomas Graf <tgraf@suug.ch>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Allocate ->signal->stats on demand in taskstats_exit(), this allows us to
remove taskstats_tgid_alloc() (the last non-trivial inline) from taskstat's
public interface.
Signed-off-by: Oleg Nesterov <oleg@tv-sign.ru>
Cc: Balbir Singh <balbir@in.ibm.com>
Cc: Shailabh Nagar <nagar@watson.ibm.com>
Cc: Jay Lan <jlan@engr.sgi.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
do_exit:
taskstats_exit_alloc()
...
taskstats_exit_send()
taskstats_exit_free()
I think this is not good, let it be a single function exported to the core
kernel, taskstats_exit(), which does alloc + send + free itself.
Signed-off-by: Oleg Nesterov <oleg@tv-sign.ru>
Cc: Balbir Singh <balbir@in.ibm.com>
Cc: Shailabh Nagar <nagar@watson.ibm.com>
Cc: Jay Lan <jlan@engr.sgi.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
If there are no listeners, every task does unneeded kmem_cache alloc/free on
exit. We don't need listeners->sem for 'if (!list_empty())' check. Yes, we may
have a false positive, but this doesn't differ from the case when the listener
is unregistered after we drop the semaphore. So we don't need to do allocation
beforehand.
Signed-off-by: Oleg Nesterov <oleg@tv-sign.ru>
Cc: Balbir Singh <balbir@in.ibm.com>
Acked-by: Shailabh Nagar <nagar@watson.ibm.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Replace all uses of kmem_cache_t with struct kmem_cache.
The patch was generated using the following script:
#!/bin/sh
#
# Replace one string by another in all the kernel sources.
#
set -e
for file in `find * -name "*.c" -o -name "*.h"|xargs grep -l $1`; do
quilt add $file
sed -e "1,\$s/$1/$2/g" $file >/tmp/$$
mv /tmp/$$ $file
quilt refresh
done
The script was run like this
sh replace kmem_cache_t "struct kmem_cache"
Signed-off-by: Christoph Lameter <clameter@sgi.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
SLAB_KERNEL is an alias of GFP_KERNEL.
Signed-off-by: Christoph Lameter <clameter@sgi.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
By modyfing genlmsg_put() to take a genl_family and by adding
genlmsg_put_reply() the process of constructing the netlink
and generic netlink headers is simplified.
Signed-off-by: Thomas Graf <tgraf@suug.ch>
Acked-by: Paul Moore <paul.moore@hp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Account for the netlink message header size directly in nlmsg_new()
instead of relying on the caller calculate it correctly.
Replaces error handling of message construction functions when
constructing notifications with bug traps since a failure implies
a bug in calculating the size of the skb.
Signed-off-by: Thomas Graf <tgraf@suug.ch>
Acked-by: Paul Moore <paul.moore@hp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
If there are no listeners, taskstats_exit_send() just returns because
taskstats_exit_alloc() didn't allocate *tidstats. This is wrong, each
sub-thread should do fill_tgid_exit() on exit, otherwise its ->delays is
not recorded in ->signal->stats and lost.
Q: We don't send TASKSTATS_TYPE_AGGR_TGID when single-threaded process
exits. Is it good? How can the listener figure out that it was actually a
process exit, not sub-thread?
Signed-off-by: Oleg Nesterov <oleg@tv-sign.ru>
Cc: Balbir Singh <balbir@in.ibm.com>
Acked-by: Shailabh Nagar <nagar@watson.ibm.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
prepare_reply() adds GENL_HDRLEN to the payload (genlmsg_total_size()),
but then it does genlmsg_put()->nlmsg_put(). This means we forget to
reserve a room for 'struct nlmsghdr'.
Signed-off-by: Oleg Nesterov <oleg@tv-sign.ru>
Cc: Thomas Graf <tgraf@suug.ch>
Cc: Andrew Morton <akpm@osdl.org>
Cc: Shailabh Nagar <nagar@watson.ibm.com>
Cc: Balbir Singh <balbir@in.ibm.com>
Cc: Jay Lan <jlan@sgi.com>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
'return genlmsg_cancel()' in taskstats_user_cmd/taskstats_exit_send
potentially leaks a skb. Unless we pass 'rep_skb' to the netlink layer
we own sk_buff. This means we should always do kfree_skb() on failure.
[ Thomas acked and pointed out missing return value in original version ]
Signed-off-by: Oleg Nesterov <oleg@tv-sign.ru>
Acked-by: Thomas Graf <tgraf@suug.ch>
Cc: Andrew Morton <akpm@osdl.org>
Cc: Shailabh Nagar <nagar@watson.ibm.com>
Cc: Balbir Singh <balbir@in.ibm.com>
Cc: Jay Lan <jlan@sgi.com>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
fill_tgid() should skip not only an already exited group leader. If the
task has ->exit_state != 0 it already did exit_notify(), so it also did
fill_tgid_exit()->delayacct_add_tsk(->signal->stats) and we should skip it
to avoid a double accounting.
This patch doesn't close the race completely, but it cleanups the code.
Signed-off-by: Oleg Nesterov <oleg@tv-sign.ru>
Cc: Shailabh Nagar <nagar@watson.ibm.com>
Cc: Balbir Singh <balbir@in.ibm.com>
Cc: Jay Lan <jlan@sgi.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Remove tasklist_lock from taskstats.c. find_task_by_pid() is rcu-safe.
->siglock allows us to traverse subthread without tasklist.
Q: delay accounting looks wrong to me. If sub-thread has already called
taskstats_exit_send() but didn't call release_task(self) yet it will be
accounted twice. The window is big. No?
Signed-off-by: Oleg Nesterov <oleg@tv-sign.ru>
Cc: Shailabh Nagar <nagar@watson.ibm.com>
Cc: Balbir Singh <balbir@in.ibm.com>
Cc: Jay Lan <jlan@sgi.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
signal_struct is (mostly) protected by ->sighand->siglock, I think we don't
need ->taskstats_lock to protect ->stats. This also allows us to simplify the
locking in fill_tgid().
Signed-off-by: Oleg Nesterov <oleg@tv-sign.ru>
Cc: Shailabh Nagar <nagar@watson.ibm.com>
Cc: Balbir Singh <balbir@in.ibm.com>
Cc: Jay Lan <jlan@sgi.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
1. fill_tgid() forgets to do put_task_struct(first).
2. release_task(first) can happen after fill_tgid() drops tasklist_lock,
it is unsafe to dereference first->signal.
This is a temporary fix, imho the locking should be reworked.
Signed-off-by: Oleg Nesterov <oleg@tv-sign.ru>
Cc: Shailabh Nagar <nagar@watson.ibm.com>
Cc: Balbir Singh <balbir@in.ibm.com>
Cc: Jay Lan <jlan@sgi.com>
Cc: <stable@kernel.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Add extended system accounting handling over taskstats interface. A
CONFIG_TASK_XACCT flag is created to enable the extended accounting code.
Signed-off-by: Jay Lan <jlan@sgi.com>
Cc: Shailabh Nagar <nagar@watson.ibm.com>
Cc: Balbir Singh <balbir@in.ibm.com>
Cc: Jes Sorensen <jes@sgi.com>
Cc: Chris Sturtivant <csturtiv@sgi.com>
Cc: Tony Ernst <tee@sgi.com>
Cc: Guillaume Thouvenin <guillaume.thouvenin@bull.net>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Add some basic accounting fields to the taskstats struct, add a new
kernel/tsacct.c to handle basic accounting data handling upon exit. A handle
is added to taskstats.c to invoke the basic accounting data handling.
Signed-off-by: Jay Lan <jlan@sgi.com>
Cc: Shailabh Nagar <nagar@watson.ibm.com>
Cc: Balbir Singh <balbir@in.ibm.com>
Cc: Jes Sorensen <jes@sgi.com>
Cc: Chris Sturtivant <csturtiv@sgi.com>
Cc: Tony Ernst <tee@sgi.com>
Cc: Guillaume Thouvenin <guillaume.thouvenin@bull.net>
Cc: "Michal Piotrowski" <michal.k.k.piotrowski@gmail.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
The addition of the CSA patch pushed the size of struct taskstats to 256
bytes. This exposed a problem with prepare_reply(), we were not allocating
space for the netlink and genetlink header. It worked earlier because
alloc_skb() would align the skb to SMP_CACHE_BYTES, which added some additonal
bytes.
Signed-off-by: Balbir Singh <balbir@in.ibm.com>
Cc: Jamal Hadi <hadi@cyberus.ca>
Cc: Shailabh Nagar <nagar@watson.ibm.com>
Cc: Thomas Graf <tgraf@suug.ch>
Cc: "David S. Miller" <davem@davemloft.net>
Cc: Jay Lan <jlan@engr.sgi.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Adds:
nlmsg_get_pos() return current position in message
nlmsg_trim() trim part of message
nla_reserve_nohdr(skb, len) reserve room for an attribute w/o hdr
nla_put_nohdr(skb, len, data) add attribute w/o hdr
nla_find_nested() find attribute in nested attributes
Fixes nlmsg_new() to take allocation flags and consider size.
Signed-off-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
Add a missing freeing of skb in the case there are no listeners at all.
Also remove the returning of error values by the function as it is unused
by the sole caller.
Signed-off-by: Shailabh Nagar <nagar@watson.ibm.com>
Signed-off-by: Chandra Seetharaman <sekharan@us.ibm.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Complete the separation of delay accounting and taskstats by ignoring the
return value of delay accounting functions that fill in parts of taskstats
before it is sent out (either in response to a command or as part of a task
exit).
Also make delayacct_add_tsk return silently when delay accounting is turned
off rather than treat it as an error.
Signed-off-by: Shailabh Nagar <nagar@watson.ibm.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
In send_cpu_listeners(), which is called on the exit path, a down_write()
was protecting operations like skb_clone() and genlmsg_unicast() that do
GFP_KERNEL allocations. If the oom-killer decides to kill tasks to satisfy
the allocations,the exit of those tasks could block on the same semphore.
The down_write() was only needed to allow removal of invalid listeners from
the listener list. The patch converts the down_write to a down_read and
defers the removal to a separate critical region. This ensures that even
if the oom-killer is called, no other task's exit is blocked as it can
still acquire another down_read.
Thanks to Andrew Morton & Herbert Xu for pointing out the oom related
pitfalls, and to Chandra Seetharaman for suggesting this fix instead of
using something more complex like RCU.
Signed-off-by: Chandra Seetharaman <sekharan@us.ibm.com>
Signed-off-by: Shailabh Nagar <nagar@watson.ibm.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
On systems with a large number of cpus, with even a modest rate of tasks
exiting per cpu, the volume of taskstats data sent on thread exit can
overflow a userspace listener's buffers.
One approach to avoiding overflow is to allow listeners to get data for a
limited and specific set of cpus. By scaling the number of listeners
and/or the cpus they monitor, userspace can handle the statistical data
overload more gracefully.
In this patch, each listener registers to listen to a specific set of cpus
by specifying a cpumask. The interest is recorded per-cpu. When a task
exits on a cpu, its taskstats data is unicast to each listener interested
in that cpu.
Thanks to Andrew Morton for pointing out the various scalability and
general concerns of previous attempts and for suggesting this design.
[akpm@osdl.org: build fix]
Signed-off-by: Shailabh Nagar <nagar@watson.ibm.com>
Signed-off-by: Balbir Singh <balbir@in.ibm.com>
Signed-off-by: Chandra Seetharaman <sekharan@us.ibm.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Send per-tgid data only once during exit of a thread group instead of once
with each member thread exit.
Currently, when a thread exits, besides its per-tid data, the per-tgid data
of its thread group is also sent out, if its thread group is non-empty.
The per-tgid data sent consists of the sum of per-tid stats for all
*remaining* threads of the thread group.
This patch modifies this sending in two ways:
- the per-tgid data is sent only when the last thread of a thread group
exits. This cuts down heavily on the overhead of sending/receiving
per-tgid data, especially when other exploiters of the taskstats
interface aren't interested in per-tgid stats
- the semantics of the per-tgid data sent are changed. Instead of being
the sum of per-tid data for remaining threads, the value now sent is the
true total accumalated statistics for all threads that are/were part of
the thread group.
The patch also addresses a minor issue where failure of one accounting
subsystem to fill in the taskstats structure was causing the send of
taskstats to not be sent at all.
The patch has been tested for stability and run cerberus for over 4 hours
on an SMP.
[akpm@osdl.org: bugfixes]
Signed-off-by: Shailabh Nagar <nagar@watson.ibm.com>
Signed-off-by: Balbir Singh <balbir@in.ibm.com>
Cc: Jay Lan <jlan@engr.sgi.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
