24 Jun, 2005
4 commits
-
sys_timer_settime/sys_timer_delete needs to delete k_itimer->real.timer
synchronously while holding ->it_lock, which is also locked in
posix_timer_fn.This patch removes timer_active/set_timer_inactive which plays with
timer_list's internals in favour of using try_to_del_timer_sync(), which
was introduced in the previous patch.Signed-off-by: Oleg Nesterov
Signed-off-by: Andrew Morton
Signed-off-by: Linus Torvalds -
This patch splits del_timer_sync() into 2 functions. The new one,
try_to_del_timer_sync(), returns -1 when it hits executing timer.It can be used in interrupt context, or when the caller hold locks which
can prevent completion of the timer's handler.NOTE. Currently it can't be used in interrupt context in UP case, because
->running_timer is used only with CONFIG_SMP.Should the need arise, it is possible to kill #ifdef CONFIG_SMP in
set_running_timer(), it is cheap.Signed-off-by: Oleg Nesterov
Signed-off-by: Andrew Morton
Signed-off-by: Linus Torvalds -
This patch tries to solve following problems:
1. del_timer_sync() is racy. The timer can be fired again after
del_timer_sync have checked all cpus and before it will recheck
timer_pending().2. It has scalability problems. All cpus are scanned to determine
if the timer is running on that cpu.With this patch del_timer_sync is O(1) and no slower than plain
del_timer(pending_timer), unless it has to actually wait for
completion of the currently running timer.The only restriction is that the recurring timer should not use
add_timer_on().3. The timers are not serialized wrt to itself.
If CPU_0 does mod_timer(jiffies+1) while the timer is currently
running on CPU 1, it is quite possible that local interrupt on
CPU_0 will start that timer before it finished on CPU_1.4. The timers locking is suboptimal. __mod_timer() takes 3 locks
at once and still requires wmb() in del_timer/run_timers.The new implementation takes 2 locks sequentially and does not
need memory barriers.Currently ->base != NULL means that the timer is pending. In that case
->base.lock is used to lock the timer. __mod_timer also takes timer->lock
because ->base can be == NULL.This patch uses timer->entry.next != NULL as indication that the timer is
pending. So it does __list_del(), entry->next = NULL instead of list_del()
when the timer is deleted.The ->base field is used for hashed locking only, it is initialized
in init_timer() which sets ->base = per_cpu(tvec_bases). When the
tvec_bases.lock is locked, it means that all timers which are tied
to this base via timer->base are locked, and the base itself is locked
too.So __run_timers/migrate_timers can safely modify all timers which could
be found on ->tvX lists (pending timers).When the timer's base is locked, and the timer removed from ->entry list
(which means that _run_timers/migrate_timers can't see this timer), it is
possible to set timer->base = NULL and drop the lock: the timer remains
locked.This patch adds lock_timer_base() helper, which waits for ->base != NULL,
locks the ->base, and checks it is still the same.__mod_timer() schedules the timer on the local CPU and changes it's base.
However, it does not lock both old and new bases at once. It locks the
timer via lock_timer_base(), deletes the timer, sets ->base = NULL, and
unlocks old base. Then __mod_timer() locks new_base, sets ->base = new_base,
and adds this timer. This simplifies the code, because AB-BA deadlock is not
possible. __mod_timer() also ensures that the timer's base is not changed
while the timer's handler is running on the old base.__run_timers(), del_timer() do not change ->base anymore, they only clear
pending flag.So del_timer_sync() can test timer->base->running_timer == timer to detect
whether it is running or not.We don't need timer_list->lock anymore, this patch kills it.
We also don't need barriers. del_timer() and __run_timers() used smp_wmb()
before clearing timer's pending flag. It was needed because __mod_timer()
did not lock old_base if the timer is not pending, so __mod_timer()->list_add()
could race with del_timer()->list_del(). With this patch these functions are
serialized through base->lock.One problem. TIMER_INITIALIZER can't use per_cpu(tvec_bases). So this patch
adds globalstruct timer_base_s {
spinlock_t lock;
struct timer_list *running_timer;
} __init_timer_base;which is used by TIMER_INITIALIZER. The corresponding fields in tvec_t_base_s
struct are replaced by struct timer_base_s t_base.It is indeed ugly. But this can't have scalability problems. The global
__init_timer_base.lock is used only when __mod_timer() is called for the first
time AND the timer was compile time initialized. After that the timer migrates
to the local CPU.Signed-off-by: Oleg Nesterov
Acked-by: Ingo Molnar
Signed-off-by: Renaud Lienhart
Signed-off-by: Andrew Morton
Signed-off-by: Linus Torvalds -
Make the timer frequency selectable. The timer interrupt may cause bus
and memory contention in large NUMA systems since the interrupt occurs
on each processor HZ times per second.Signed-off-by: Christoph Lameter
Signed-off-by: Shai Fultheim
Signed-off-by: Andrew Morton
Signed-off-by: Linus Torvalds
22 Jun, 2005
6 commits
-
With Chris Wedgwood
As suggested by Chris, we can make the "just added" method ->release
conditional to UML only (better: to archs requesting it, i.e. only UML
currently), so that other archs don't get this unneeded crud, and if UML
won't need it any more we can kill this.Signed-off-by: Paolo 'Blaisorblade' Giarrusso
CC: Ingo Molnar
Signed-off-by: Andrew Morton
Signed-off-by: Linus Torvalds -
With Chris Wedgwood
Currently UML must explicitly call the UML-specific
free_irq_by_irq_and_dev() for each free_irq call it's done.This is needed because ->shutdown and/or ->disable are only called when the
last "action" for that irq is removed.Instead, for UML shared IRQs (UML IRQs are very often, if not always,
shared), for each dev_id some setup is done, which must be cleared on the
release of that fd. For instance, for each open console a new instance
(i.e. new dev_id) of the same IRQ is requested().Exactly, a fd is stored in an array (pollfds), which is after read by a
host thread and passed to poll(). Each event registered by poll() triggers
an interrupt. So, for each free_irq() we must remove the corresponding
host fd from the table, which we do via this -release() method.In this patch we add an appropriate hook for this, and remove all uses of
it by pointing the hook to the said procedure; this is safe to do since the
said procedure.Also some cosmetic improvements are included.
This is heavily based on some work by Chris Wedgwood, which however didn't
get the patch merged for something I'd call a "misunderstanding" (the need
for this patch wasn't cleanly explained, thus adding the generic hook was
felt as undesirable).Signed-off-by: Paolo 'Blaisorblade' Giarrusso
CC: Ingo Molnar
Signed-off-by: Andrew Morton
Signed-off-by: Linus Torvalds -
Remove part of comment on linking new vma in dup_mmap: since anon_vma rmap
came in, try_to_unmap_one knows the vma without needing find_vma. But add
a comment to note that here vma is inserted without mmap_sem.Signed-off-by: Hugh Dickins
Signed-off-by: Andrew Morton
Signed-off-by: Linus Torvalds -
Ingo recently introduced a great speedup for allocating new mmaps using the
free_area_cache pointer which boosts the specweb SSL benchmark by 4-5% and
causes huge performance increases in thread creation.The downside of this patch is that it does lead to fragmentation in the
mmap-ed areas (visible via /proc/self/maps), such that some applications
that work fine under 2.4 kernels quickly run out of memory on any 2.6
kernel.The problem is twofold:
1) the free_area_cache is used to continue a search for memory where
the last search ended. Before the change new areas were always
searched from the base address on.So now new small areas are cluttering holes of all sizes
throughout the whole mmap-able region whereas before small holes
tended to close holes near the base leaving holes far from the base
large and available for larger requests.2) the free_area_cache also is set to the location of the last
munmap-ed area so in scenarios where we allocate e.g. five regions of
1K each, then free regions 4 2 3 in this order the next request for 1K
will be placed in the position of the old region 3, whereas before we
appended it to the still active region 1, placing it at the location
of the old region 2. Before we had 1 free region of 2K, now we only
get two free regions of 1K -> fragmentation.The patch addresses thes issues by introducing yet another cache descriptor
cached_hole_size that contains the largest known hole size below the
current free_area_cache. If a new request comes in the size is compared
against the cached_hole_size and if the request can be filled with a hole
below free_area_cache the search is started from the base instead.The results look promising: Whereas 2.6.12-rc4 fragments quickly and my
(earlier posted) leakme.c test program terminates after 50000+ iterations
with 96 distinct and fragmented maps in /proc/self/maps it performs nicely
(as expected) with thread creation, Ingo's test_str02 with 20000 threads
requires 0.7s system time.Taking out Ingo's patch (un-patch available per request) by basically
deleting all mentions of free_area_cache from the kernel and starting the
search for new memory always at the respective bases we observe: leakme
terminates successfully with 11 distinctive hardly fragmented areas in
/proc/self/maps but thread creating is gringdingly slow: 30+s(!) system
time for Ingo's test_str02 with 20000 threads.Now - drumroll ;-) the appended patch works fine with leakme: it ends with
only 7 distinct areas in /proc/self/maps and also thread creation seems
sufficiently fast with 0.71s for 20000 threads.Signed-off-by: Wolfgang Wander
Credit-to: "Richard Purdie"
Signed-off-by: Ken Chen
Acked-by: Ingo Molnar (partly)
Signed-off-by: Andrew Morton
Signed-off-by: Linus Torvalds -
This is the core of the (much simplified) early reclaim. The goal of this
patch is to reclaim some easily-freed pages from a zone before falling back
onto another zone.One of the major uses of this is NUMA machines. With the default allocator
behavior the allocator would look for memory in another zone, which might be
off-node, before trying to reclaim from the current zone.This adds a zone tuneable to enable early zone reclaim. It is selected on a
per-zone basis and is turned on/off via syscall.Adding some extra throttling on the reclaim was also required (patch
4/4). Without the machine would grind to a crawl when doing a "make -j"
kernel build. Even with this patch the System Time is higher on
average, but it seems tolerable. Here are some numbers for kernbench
runs on a 2-node, 4cpu, 8Gig RAM Altix in the "make -j" run:wall user sys %cpu ctx sw. sleeps
---- ---- --- ---- ------ ------
No patch 1009 1384 847 258 298170 504402
w/patch, no reclaim 880 1376 667 288 254064 396745
w/patch & reclaim 1079 1385 926 252 291625 548873These numbers are the average of 2 runs of 3 "make -j" runs done right
after system boot. Run-to-run variability for "make -j" is huge, so
these numbers aren't terribly useful except to seee that with reclaim
the benchmark still finishes in a reasonable amount of time.I also looked at the NUMA hit/miss stats for the "make -j" runs and the
reclaim doesn't make any difference when the machine is thrashing away.Doing a "make -j8" on a single node that is filled with page cache pages
takes 700 seconds with reclaim turned on and 735 seconds without reclaim
(due to remote memory accesses).The simple zone_reclaim syscall program is at
http://www.bork.org/~mort/sgi/zone_reclaim.cSigned-off-by: Martin Hicks
Signed-off-by: Andrew Morton
Signed-off-by: Linus Torvalds -
This patch implements a number of smp_processor_id() cleanup ideas that
Arjan van de Ven and I came up with.The previous __smp_processor_id/_smp_processor_id/smp_processor_id API
spaghetti was hard to follow both on the implementational and on the
usage side.Some of the complexity arose from picking wrong names, some of the
complexity comes from the fact that not all architectures defined
__smp_processor_id.In the new code, there are two externally visible symbols:
- smp_processor_id(): debug variant.
- raw_smp_processor_id(): nondebug variant. Replaces all existing
uses of _smp_processor_id() and __smp_processor_id(). Defined
by every SMP architecture in include/asm-*/smp.h.There is one new internal symbol, dependent on DEBUG_PREEMPT:
- debug_smp_processor_id(): internal debug variant, mapped to
smp_processor_id().Also, i moved debug_smp_processor_id() from lib/kernel_lock.c into a new
lib/smp_processor_id.c file. All related comments got updated and/or
clarified.I have build/boot tested the following 8 .config combinations on x86:
{SMP,UP} x {PREEMPT,!PREEMPT} x {DEBUG_PREEMPT,!DEBUG_PREEMPT}
I have also build/boot tested x64 on UP/PREEMPT/DEBUG_PREEMPT. (Other
architectures are untested, but should work just fine.)Signed-off-by: Ingo Molnar
Signed-off-by: Arjan van de Ven
Signed-off-by: Andrew Morton
Signed-off-by: Linus Torvalds
21 Jun, 2005
1 commit
-
sysfs: fix the rest of the kernel so if an attribute doesn't
implement show or store method read/write will return
-EIO instead of 0 or -EINVAL or -EPERM.Signed-off-by: Dmitry Torokhov
Signed-off-by: Greg Kroah-Hartman
18 Jun, 2005
2 commits
-
Do all timer zapping in exit_itimers.
Signed-off-by: Ingo Molnar
Signed-off-by: Linus Torvalds
14 Jun, 2005
1 commit
-
On one path, cond_resched_lock() fails to return true if it dropped the lock.
We think this might be causing the crashes in JBD's log_do_checkpoint().Cc: Ingo Molnar
Signed-off-by: Andrew Morton
Signed-off-by: Linus Torvalds
02 Jun, 2005
1 commit
01 Jun, 2005
1 commit
-
flush_icache_range() is used in two different situation - in binfmt_elf.c &
co for user space mappings and module.c for kernel modules. On m68k
flush_icache_range() doesn't know which data to flush, as it has separate
address spaces and the pointer argument can be valid in either address
space.First I considered splitting flush_icache_range(), but this patch is
simpler. Setting the correct context gives flush_icache_range() enough
information to flush the correct data.Signed-off-by: Andrew Morton
Signed-off-by: Linus Torvalds
29 May, 2005
1 commit
-
The "unhandled interrupts" catcher, note_interrupt(), increments a global
desc->irq_count and grossly damages scaling of very large systems, e.g.,
>192p ia64 Altix, because of this highly contented cacheline, especially
for timer interrupts. 384p is severely crippled, and 512p is unuseable.All calls to note_interrupt() can be disabled by booting with "noirqdebug",
but this disables the useful interrupt checking for all interrupts.I propose eliminating note_interrupt() for all per-CPU interrupts. This
was the behavior of linux-2.6.10 and earlier, but in 2.6.11 a code
restructuring added a call to note_interrupt() for per-CPU interrupts.
Besides, note_interrupt() is a bit racy for concurrent CPU calls anyway, as
the desc->irq_count++ increment isn't atomic (which, if done, would make
scaling even worse).Signed-off-by: John Hawkes
Signed-off-by: Andrew Morton
Signed-off-by: Linus Torvalds
27 May, 2005
2 commits
-
There is a race in the kernel cpuset code, between the code
to handle notify_on_release, and the code to remove a cpuset.
The notify_on_release code can end up trying to access a
cpuset that has been removed. In the most common case, this
causes a NULL pointer dereference from the routine cpuset_path.
However all manner of bad things are possible, in theory at least.The existing code decrements the cpuset use count, and if the
count goes to zero, processes the notify_on_release request,
if appropriate. However, once the count goes to zero, unless we
are holding the global cpuset_sem semaphore, there is nothing to
stop another task from immediately removing the cpuset entirely,
and recycling its memory.The obvious fix would be to always hold the cpuset_sem
semaphore while decrementing the use count and dealing with
notify_on_release. However we don't want to force a global
semaphore into the mainline task exit path, as that might create
a scaling problem.The actual fix is almost as easy - since this is only an issue
for cpusets using notify_on_release, which the top level big
cpusets don't normally need to use, only take the cpuset_sem
for cpusets using notify_on_release.This code has been run for hours without a hiccup, while running
a cpuset create/destroy stress test that could crash the existing
kernel in seconds. This patch applies to the current -linus
git kernel.Signed-off-by: Paul Jackson
Acked-by: Simon Derr
Acked-by: Dinakar Guniguntala
Signed-off-by: Linus Torvalds -
Signed-off-by: David Woodhouse
26 May, 2005
1 commit
-
While they were all just simple blobs it made sense to just free them
as we walked through and logged them. Now that there are pointers to
other objects which need refcounting, we might as well revert to
_only_ logging them in audit_log_exit(), and put the code to free them
properly in only one place -- in audit_free_aux().Signed-off-by: David Woodhouse
----------------------------------------------------------
25 May, 2005
1 commit
-
If SIGKILL does not have priority, we cannot instantly kill task before it
makes some unexpected job. It can be critical, but we were unable to
reproduce this easily until Heiko Carstens
reported this problem on LKML.Signed-Off-By: Kirill Korotaev
Signed-Off-By: Alexey Kuznetsov
Signed-off-by: Andrew Morton
Signed-off-by: Linus Torvalds
24 May, 2005
2 commits
-
It comes from the user; it needs to be escaped.
Signed-off-by: David Woodhouse
-
These changes make processing of audit logs easier. Based on a patch
from Steve GrubbSigned-off-by: David Woodhouse
22 May, 2005
2 commits
-
Move audit_serial() into audit.c and use it to generate serial numbers
on messages even when there is no audit context from syscall auditing.
This allows us to disambiguate audit records when more than one is
generated in the same millisecond.Based on a patch by Steve Grubb after he observed the problem.
Signed-off-by: David Woodhouse
-
In _spin_unlock_bh(lock):
do { \
_raw_spin_unlock(lock); \
preempt_enable(); \
local_bh_enable(); \
__release(lock); \
} while (0)there is no reason for using preempt_enable() instead of a simple
preempt_enable_no_resched()Since we know bottom halves are disabled, preempt_schedule() will always
return at once (preempt_count!=0), and hence preempt_check_resched() is
useless here...This fixes it by using "preempt_enable_no_resched()" instead of the
"preempt_enable()", and thus avoids the useless preempt_check_resched()
just before re-enabling bottom halves.Signed-off-by: Samuel Thibault
Signed-off-by: Linus Torvalds
21 May, 2005
4 commits
-
The attached patch changes all occurrences of loginuid to auid. It also
changes everything to %u that is an unsigned type.Signed-off-by: Steve Grubb
Signed-off-by: David Woodhouse -
The original AVC_USER message wasn't consolidated with the new range of
user messages. The attached patch fixes the kernel so the old messages
work again.Signed-off-by: Steve Grubb
Signed-off-by: David Woodhouse -
This patch changes the SELinux AVC to defer logging of paths to the audit
framework upon syscall exit, by saving a reference to the (dentry,vfsmount)
pair in an auxiliary audit item on the current audit context for processing
by audit_log_exit.Signed-off-by: Stephen Smalley
Signed-off-by: David Woodhouse -
This patch removes the entwining of cpusets and hotplug code in the "No
more Mr. Nice Guy" case of sched.c move_task_off_dead_cpu().Since the hotplug code is holding a spinlock at this point, we cannot take
the cpuset semaphore, cpuset_sem, as would seem to be required either to
update the tasks cpuset, or to scan up the nested cpuset chain, looking for
the nearest cpuset ancestor that still has some CPUs that are online. So
we just punt and blast the tasks cpus_allowed with all bits allowed.This reverts these lines of code to what they were before the cpuset patch.
And it updates the cpuset Doc file, to match.The one known alternative to this that seems to work came from Dinakar
Guniguntala, and required the hotplug code to take the cpuset_sem semaphore
much earlier in its processing. So far as we know, the increased locking
entanglement between cpusets and hot plug of this alternative approach is
not worth doing in this case.Signed-off-by: Paul Jackson
Acked-by: Nathan Lynch
Acked-by: Dinakar Guniguntala
Signed-off-by: Andrew Morton
Signed-off-by: Linus Torvalds
19 May, 2005
5 commits
-
The limit on the number of outstanding audit messages was inadvertently
removed with the switch to queuing skbs directly for sending by a kernel
thread. Put it back again.Signed-off-by: David Woodhouse
-
Nobody does. Really, it gets very silly if auditd is recording its
own actions.Signed-off-by: David Woodhouse
-
netlink_unicast() will attempt to reallocate and will free messages if
the socket's rcvbuf limit is reached unless we give it an infinite
timeout. So do that, from a kernel thread which is dedicated to spewing
stuff up the netlink socket.Signed-off-by: David Woodhouse
-
* If vsnprintf returns -1, it will mess up the sk buffer space accounting.
This is fixed by not calling skb_put with bogus len values.* audit_log_hex was a loop that called audit_log_vformat with %02X for each
character. This is very inefficient since conversion from unsigned character
to Ascii representation is essentially masking, shifting, and byte lookups.
Also, the length of the converted string is well known - it's twice the
original. Fixed by rewriting the function.* audit_log_untrustedstring had no comments. This makes it hard for
someone to understand what the string format will be.* audit_log_d_path was never fixed to use untrustedstring. This could mess
up user space parsers. This was fixed to make a temp buffer, call d_path,
and log temp buffer using untrustedstring.From: Steve Grubb
Signed-off-by: David Woodhouse
18 May, 2005
2 commits
-
It's silly to have to add explicit entries for new userspace messages
as we invent them. Just treat all messages in the user range the same.Signed-off-by: David Woodhouse
-
This patch includes various tweaks in the messaging that appears during
system pm state transitions:* Warn about certain illegal calls in the device tree, like resuming
child before parent or suspending parent before child. This could
happen easily enough through sysfs, or in some cases when drivers
use device_pm_set_parent().* Be more consistent about dev_dbg() tracing ... do it for resume() and
shutdown() too, and never if the driver doesn't have that method.* Say which type of system sleep state is being entered.
Except for the warnings, these only affect debug messaging.
Signed-off-by: David Brownell
Acked-by: Pavel Machek
Signed-off-by: Greg Kroah-Hartman
17 May, 2005
4 commits
-
profile=schedule parsing is not quite what it should be. First, str[7] is
'e', not ',', but then even if it did fall through, prof_on =
SCHED_PROFILING would be clobbered inside if (get_option(...)) So a small
amount of rearrangement is done in this patch to correct it.Signed-off-by: Andrew Morton
Signed-off-by: Linus Torvalds -
Move add_preferred_console out of CONFIG_PRINTK so serial console does the
right thing.Signed-off-by: Matt Mackall
Signed-off-by: Andrew Morton
Signed-off-by: Linus Torvalds -
On my IA64 machine, after kernel 2.6.12-rc3 boots, an edge-triggered
interrupt (IRQ 46) keeps triggered over and over again. There is no IRQ 46
interrupt action handler. It has lots of impact on performance.Kernel 2.6.10 and its prior versions have no the problem. Basically,
kernel 2.6.10 will mask the spurious edge interrupt if the interrupt is
triggered for the second time and its status includes
IRQ_DISABLE|IRQ_PENDING.Originally, IA64 kernel has its own specific _irq_desc definitions in file
arch/ia64/kernel/irq.c. The definition initiates _irq_desc[irq].status to
IRQ_DISABLE. Since kernel 2.6.11, it was moved to architecture independent
codes, i.e. kernel/irq/handle.c, but kernel/irq/handle.c initiates
_irq_desc[irq].status to 0 instead of IRQ_DISABLE.Signed-off-by: Zhang Yanmin
Signed-off-by: Andrew Morton
Signed-off-by: Linus Torvalds -
Signed-off-by: David Woodhouse
