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>
- 8-bit interface mode never worked properly. The only adapter I have
which supports the 8b mode (the Jmicron) had some problems with its
clock wiring and they discovered it only now. We also discovered that
ProHG media is more sensitive to the ordering of initialization
commands.
- Make the driver fall back to highest supported mode instead of always
falling back to serial. The driver will attempt the switch to 8b mode
for any new MSPro card, but not all of them support it. Previously,
these new cards ended up in serial mode, which is not the best idea
(they work fine with 4b, after all).
- Edit some macros for better conformance to Sony documentation
Signed-off-by: Alex Dubov <oakad@yahoo.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
The iterator for_each_zone_zonelist() uses a struct zoneref *z cursor when
scanning zonelists to keep track of where in the zonelist it is. The
zoneref that is returned corresponds to the the next zone that is to be
scanned, not the current one. It was intended to be treated as an opaque
list.
When the page allocator is scanning a zonelist, it marks elements in the
zonelist corresponding to zones that are temporarily full. As the
zonelist is being updated, it uses the cursor here;
if (NUMA_BUILD)
zlc_mark_zone_full(zonelist, z);
This is intended to prevent rescanning in the near future but the zoneref
cursor does not correspond to the zone that has been found to be full.
This is an easy misunderstanding to make so this patch corrects the
problem by changing zoneref cursor to be the current zone being scanned
instead of the next one.
Signed-off-by: Mel Gorman <mel@csn.ul.ie>
Cc: Andy Whitcroft <apw@shadowen.org>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: <stable@kernel.org> [2.6.26.x]
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
akpm: these have no callers at this time, but they shall soon, so let's
get them right.
[akpm@linux-foundation.org: coding-style fixes]
Signed-off-by: Hiroshi DOYU <Hiroshi.DOYU@nokia.com>
Cc: Tony Lindgren <tony@atomide.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
I recently bought 3 HGST P7K500-series 500GB SATA drives and
had trouble accessing the block right on the LBA28-LBA48 border.
Here's how it fails (same for all 3 drives):
# dd if=/dev/sdc bs=512 count=1 skip=268435455 > /dev/null
dd: reading `/dev/sdc': Input/output error
0+0 records in
0+0 records out
0 bytes (0 B) copied, 0.288033 seconds, 0.0 kB/s
# dmesg
ata1.00: exception Emask 0x0 SAct 0x0 SErr 0x0 action 0x0
ata1.00: BMDMA stat 0x25
ata1.00: cmd c8/00:08:f8:ff:ff/00:00:00:00:00/ef tag 0 dma 4096 in
res 51/04:08:f8:ff:ff/00:00:00:00:00/ef Emask 0x1 (device error)
ata1.00: status: { DRDY ERR }
ata1.00: error: { ABRT }
ata1.00: configured for UDMA/33
ata1: EH complete
...
After some investigations, it turned out this seems to be caused
by misinterpretation of the ATA specification on LBA28 access.
Following part is the code in question:
=== include/linux/ata.h ===
static inline int lba_28_ok(u64 block, u32 n_block)
{
/* check the ending block number */
return ((block + n_block - 1) < ((u64)1 << 28)) && (n_block <= 256);
}
HGST drive (sometimes) fails with LBA28 access of {block = 0xfffffff,
n_block = 1}, and this behavior seems to be comformant. Other drives,
including other HGST drives are not that strict, through.
>From the ATA specification:
(http://www.t13.org/Documents/UploadedDocuments/project/d1410r3b-ATA-ATAPI-6.pdf)
8.15.29 Word (61:60): Total number of user addressable sectors
This field contains a value that is one greater than the total number
of user addressable sectors (see 6.2). The maximum value that shall
be placed in this field is 0FFFFFFFh.
So the driver shouldn't use the value of 0xfffffff for LBA28 request
as this exceeds maximum user addressable sector. The logical maximum
value for LBA28 is 0xffffffe.
The obvious fix is to cut "- 1" part, and the patch attached just do
that. I've been using the patched kernel for about a month now, and
the same fix is also floating on the net for some time. So I believe
this fix works reliably.
Just FYI, many Windows/Intel platform users also seems to be struck
by this, and HGST has issued a note pointing to Intel ICH8/9 driver.
"28-bit LBA command is being used to access LBAs 29-bits in length"
http://www.hitachigst.com/hddt/knowtree.nsf/cffe836ed7c12018862565b000530c74/b531b8bce8745fb78825740f00580e23
Also, *BSDs seems to have similar fix included sometime around ~2004,
through I have not checked out exact portion of the code.
Signed-off-by: Taisuke Yamada <tai@rakugaki.org>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
This new action will have the ability to change the priority and/or
queue_mapping fields on an sk_buff.
Signed-off-by: Alexander Duyck <alexander.h.duyck@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch is intended to add a qdisc to support the new tx multiqueue
architecture by providing a band for each hardware queue. By doing
this it is possible to support a different qdisc per physical hardware
queue.
This qdisc uses the skb->queue_mapping to select which band to place
the traffic onto. It then uses a round robin w/ a check to see if the
subqueue is stopped to determine which band to dequeue the packet from.
Signed-off-by: Alexander Duyck <alexander.h.duyck@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
in:
> http://bugzilla.kernel.org/show_bug.cgi?id=11543
The panic code called the kexec code which called mutex_trylock() which
called spin_lock_mutex() which then stupidly went and blurted a load of
debug stuff because of in_interrupt().
Keep the lock debug code from escallating an already crappy situation.
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Some of the HT code in mlme.c is misplaced:
* constants/definitions belong to the ieee80211.h header
* code being used in other modes as well shouldn't be there
Signed-off-by: Johannes Berg <johannes@sipsolutions.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
This patch follows 11n spec naming more rigorously replacing MIMO_PS
with SM_PS (Spatial Multiplexing Power Save).
(Originally submitted as 4 patches, "mac80211: change MIMO_PS to SM_PS",
"iwlwifi: change MIMO_PS to SM_PS", "ath9k: change MIMO_PS to SM_PS",
and "iwlwifi: remove double definition of SM PS". -- JWL)
Signed-off-by: Ron Rindjunsky <ron.rindjunsky@intel.com>
Signed-off-by: Tomas Winkler <tomas.winkler@intel.com>
Signed-off-by: Zhu Yi <yi.zhu@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
As part of going idle, we already look at the time of the next timer event to determine
which C-state to select etc.
This patch adds functionality that causes the timers that are past their
soft expire time, to fire at this time, before we calculate the next wakeup
time. This functionality will thus avoid wakeups by running timers before
going idle rather than specially waking up for it.
Signed-off-by: Arjan van de Ven <arjan@linux.intel.com>
We still have life time issues with the sysfs command filter kobject,
so disable it for 2.6.27 release. We can revisit this and make it work
properly for 2.6.28, for 2.6.27 release it's too risky.
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
The GPIO connected to ADS7846 nPENIRQ signal is usually used to get
the pendown state as well. Introduce a .gpio_pendown, and use this
to decide the pendown state if .get_pendown_state is NULL.
Signed-off-by: Eric Miao <eric.miao@marvell.com>
Signed-off-by: Dmitry Torokhov <dtor@mail.ru>
Define USE_SPLIT_PTLOCKS as a constant expression rather than repeating
"NR_CPUS >= CONFIG_SPLIT_PTLOCK_CPUS" all over the place.
Signed-off-by: Jeremy Fitzhardinge <jeremy.fitzhardinge@citrix.com>
Acked-by: Hugh Dickins <hugh@veritas.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
useful to establish a lock dependency in case the actual dependency is
rare or hard to trigger.
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Acked-by: Nick Piggin <nickpiggin@yahoo.com.au>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
is_buffer_dma_capable helper function is to see if a memory region is
DMA-capable or not. The arugments are the dma_mask (or
coherent_dma_mask) of a device and the address and size of a memory
region.
Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
Acked-by: Joerg Roedel <joerg.roedel@amd.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
- unbreak ia64 (and powerpc) where function pointers dont
point at code but at data (reported by Tony Luck)
[ mingo@elte.hu: various cleanups ]
Signed-off-by: Arjan van de Ven <arjan@linux.intel.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Now that we can configure smc91x leds from its platform data,
it seems rather useful to move the led definitions to the
externally visible header file.
Signed-off-by: Marc Zyngier <marc.zyngier@altran.com>
* 'sched-fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip:
sched: arch_reinit_sched_domains() must destroy domains to force rebuild
sched, cpuset: rework sched domains and CPU hotplug handling (v4)
Fix compile error:
arch/x86/mm/init_32.c: In function 'mem_init':
arch/x86/mm/init_32.c:908: error: implicit declaration of function 'start_periodic_check_for_corruption'
Signed-off-by: Jeremy Fitzhardinge <jeremy.fitzhardinge@citrix.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Right now, there is no notifier that is called on a new cpu, before the new
cpu begins processing interrupts/softirqs.
Various kernel function would need that notification, e.g. kvm works around
by calling smp_call_function_single(), rcu polls cpu_online_map.
The patch adds a CPU_STARTING notification. It also adds a helper function
that sends the message to all cpu_chain handlers.
Tested on x86-64.
All other archs are untested. Especially on sparc, I'm not sure if I got
it right.
Signed-off-by: Manfred Spraul <manfred@colorfullife.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Calculate the journal device name once and stash it away in the
journal_s structure. This avoids needing to call bdevname()
everywhere and reduces stack usage by not needing to allocate an
on-stack buffer. In addition, we eliminate the '/' that can appear in
device names (e.g. "cciss/c0d0p9" --- see kernel bugzilla #11321) that
can cause problems when creating proc directory names, and include the
inode number to support ocfs2 which creates multiple journals with
different inode numbers.
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
this patch adds a _range version of hrtimer_start() so that range timers
can be created; the hrtimer_start() function is just a wrapper around this.
In addition, hrtimer_start_expires() will now preserve existing ranges.
Signed-off-by: Arjan van de Ven <arjan@linux.intel.com>
This patch provides a mechanism for platforms to be able to supply the
LED configuration via platform data, rather than having to hard code
it in smc91x.h.
Acked-by: Eric Miao <eric.y.miao@gmail.com>
Acked-by: Nicolas Pitre <nico@cam.org>
Acked-by: Jeff Garzik <jgarzik@pobox.com>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
Perodically check for corruption in low phusical memory. Don't bother
checking at fault time, since it won't show anything useful.
Signed-off-by: Hugh Dickins <hugh@veritas.com>
Signed-off-by: Jeremy Fitzhardinge <jeremy.fitzhardinge@citrix.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Some BIOSes have been observed to corrupt memory in the low 64k. This
change:
- Reserves all memory which does not have to be in that area, to
prevent it from being used as general memory by the kernel. Things
like the SMP trampoline are still in the memory, however.
- Clears the reserved memory so we can observe changes to it.
- Adds a function check_for_bios_corruption() which checks and reports on
memory becoming unexpectedly non-zero. Currently it's called in the
x86 fault handler, and the powermanagement debug output.
Signed-off-by: Jeremy Fitzhardinge <jeremy@goop.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
We don't need whole 32 of them, only NR_SOFTIRQS.
Signed-off-by: Alexey Dobriyan <adobriyan@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
What I realized recently is that calling rebuild_sched_domains() in
arch_reinit_sched_domains() by itself is not enough when cpusets are enabled.
partition_sched_domains() code is trying to avoid unnecessary domain rebuilds
and will not actually rebuild anything if new domain masks match the old ones.
What this means is that doing
echo 1 > /sys/devices/system/cpu/sched_mc_power_savings
on a system with cpusets enabled will not take affect untill something changes
in the cpuset setup (ie new sets created or deleted).
This patch fixes restore correct behaviour where domains must be rebuilt in
order to enable MC powersaving flags.
Test on quad-core Core2 box with both CONFIG_CPUSETS and !CONFIG_CPUSETS.
Also tested on dual-core Core2 laptop. Lockdep is happy and things are working
as expected.
Signed-off-by: Max Krasnyansky <maxk@qualcomm.com>
Tested-by: Vaidyanathan Srinivasan <svaidy@linux.vnet.ibm.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
