| Age | Commit message (Collapse) | Author |
|---|
|
Secure Digital cards use a different algorithm to calculate the timeout
for data transfers. Using the MMC one works often, but not always.
Signed-off-by: Pierre Ossman <drzeus@drzeus.cx>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
|
|
Let drivers constify MMC host method operations tables,
moving them from ".data" to ".rodata".
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
Acked-by: Pierre Ossman <drzeus@drzeus.cx>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
|
|
SGI-PV: 955993
SGI-Modid: xfs-linux-melb:xfs-kern:26934a
Signed-off-by: Nathan Scott <nathans@sgi.com>
Signed-off-by: David Chatterton <chatz@sgi.com>
|
|
xfs_splice_write() failed to update the on disk inode size when extending
the so when the file was closed the range extended by splice was truncated
off. Hence any region of a file written to by splice would end up as a
hole full of zeros.
SGI-PV: 955939
SGI-Modid: xfs-linux-melb:xfs-kern:26920a
Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: David Chatterton <chatz@sgi.com>
|
|
__blockdev_direct_IO for the DIO_OWN_LOCKING case for direct I/O reads
since it drops and reacquires the i_mutex while holding the iolock and
this violates the locking order.
SGI-PV: 955696
SGI-Modid: xfs-linux-melb:xfs-kern:26898a
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
Signed-off-by: David Chatterton <chatz@sgi.com>
|
|
The fix for recent ENOSPC deadlocks introduced certain limitations on
allocations. The fix could cause xfssyncd to loop endlessly if we did not
leave some space free for the allocator to work correctly. Basically, we
needed to ensure that we had at least 4 blocks free for an AG free list
and a block for the inode bmap btree at all times.
However, this did not take into account the fact that each AG has a free
list that needs 4 blocks. Hence any filesystem with more than one AG could
cause oversubscription of free space and make xfssyncd spin forever trying
to allocate space needed for AG freelists that was not available in the
AG.
The following patch reserves space for the free lists in all AGs plus the
inode bmap btree which prevents oversubscription. It also prevents those
blocks from being reported as free space (as they can never be used) and
makes the SMP in-core superblock accounting code and the reserved block
ioctl respect this requirement.
SGI-PV: 955674
SGI-Modid: xfs-linux-melb:xfs-kern:26894a
Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: David Chatterton <chatz@sgi.com>
|
|
Patch from Ben Dooks
It turns out we have both SMDK2413 and S3C2413 for
the same board.
Signed-off-by: Ben Dooks <ben-linux@fluff.org>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
|
|
Patch from Catalin Marinas
The ARM926EJ-S CPU has the VFP coprocessor and therefore it should be shown
in the /proc/cpuinfo if CONFIG_VFP is enabled.
Signed-off-by: Catalin Marinas <catalin.marinas@arm.com>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
|
|
Patch from Martin Michlmayr
Include linux/irq.h in the nslu2 code in order to avoid the following
compiler error:
CC arch/arm/mach-ixp4xx/nslu2-power.o
arch/arm/mach-ixp4xx/nslu2-power.c: In function 'nslu2_power_init':
arch/arm/mach-ixp4xx/nslu2-power.c:53: warning: implicit declaration of function 'set_irq_type'
arch/arm/mach-ixp4xx/nslu2-power.c:53: error: 'IRQ_TYPE_LEVEL_LOW' undeclared (first use in this function)
arch/arm/mach-ixp4xx/nslu2-power.c:53: error: (Each undeclared identifier is reported only once
arch/arm/mach-ixp4xx/nslu2-power.c:53: error: for each function it appears in.)
arch/arm/mach-ixp4xx/nslu2-power.c:54: error: 'IRQ_TYPE_LEVEL_HIGH' undeclared (first use in this function)
make[5]: *** [arch/arm/mach-ixp4xx/nslu2-power.o] Error 1
Signed-off-by: Martin Michlmayr <tbm@cyrius.com>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
|
|
Patch from Ben Dooks
Remove the pointless changelog comments from
arch/arm/mach-s3c2410 files, as all this can
be found from the revision control system.
Signed-off-by: Ben Dooks <ben-linux@fluff.org>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
|
|
Patch from Catalin Marinas
A comma was missing between tmp and tmp2.
Signed-off-by: Catalin Marinas <catalin.marinas@arm.com>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
|
|
master.kernel.org:/pub/scm/linux/kernel/git/jgarzik/netdev-2.6
* 'upstream-linus' of master.kernel.org:/pub/scm/linux/kernel/git/jgarzik/netdev-2.6:
[PATCH] myri10ge: update the firmware download URL in Kconfig
|
|
Use the generic time stuff for FRV.
Signed-off-by: John Stultz <johnstul@us.ibm.com>
Signed-Off-By: David Howells <dhowells@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
|
|
Signed-off-by: David Woodhouse <dwmw2@infradead.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
|
|
section
linux/device.h header is not included in the David Woodhouse's
kernel-headers git tree which is used for userspace kernel headers. Which
results in compile errors when building iproute2. Attached patch moves
linux/device.h include under the #ifdef __KERNEL__ section.
Signed-off-by: Ismail Donmez <ismail@pardus.org.tr>
Signed-off-by: David Woodhouse <dwmw2@infradead.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
|
|
Remove an unintended console_verbose() side-effect from add_taint().
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
|
|
PAE + swsusp results in hard-to-debug crash about 50% of time during
resume. Cause is known, fix needs to be ported from x86-64 (but we can't
make it to 2.6.18, and I'd like this to be worked around in 2.6.18).
Signed-off-by: Pavel Machek <pavel@suse.cz>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
|
|
Frank v. Waveren pointed out that on 64bit machines the timespec to
ktime_t conversion might overflow. This is also true for timeval to
ktime_t conversions. This breaks a "sleep inf" on 64bit machines.
While a timespec/timeval with tx.sec = MAX_LONG is valid by specification
the internal representation of ktime_t is based on nanoseconds. The
conversion of seconds to nanoseconds overflows for seconds values >=
(MAX_LONG / NSEC_PER_SEC).
Check the seconds argument to the conversion and limit it to the maximum
time which can be represented by ktime_t.
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Frank v Waveren <fvw@var.cx>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
|
|
Adds kernel-doc for alloc_super() type in fs/super.c.
Signed-off-by: Henrik Kretzschmar <henne@nachtwindheim.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
|
|
Fixes an error message on make xmldocs.
Signed-off-by: Henrik Kretzschmar <henne@nachtwindheim.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
|
|
With
CONFIG_SMP=y
CONFIG_PREEMPT=y
CONFIG_LOCKDEP=y
CONFIG_DEBUG_LOCK_ALLOC=y
# CONFIG_PROVE_LOCKING is not set
spin_unlock_irqrestore() goes through lockdep but spin_lock_irqsave() doesn't.
Apparently, bad things happen.
Acked-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
|
|
Update the firmware download URL in Kconfig to match the header
in drivers/net/myri10ge/myri10ge.c.
Signed-off-by: Brice Goglin <brice@myri.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>
|
|
|
|
New sparse caught that typo which could have caused erratic hardware
behaviour on some machines if the platform functions are used by the
firmware to change bits in some FCR registers.
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
|
|
* master.kernel.org:/home/rmk/linux-2.6-serial:
[SERIAL] 8250: constify some serial structs
[SERIAL] Make uart_match_port() work with all memory mapped UARTs
|
|
* master.kernel.org:/home/rmk/linux-2.6-arm:
[ARM] 3762/1: Fix ptrace cache coherency bug for ARM1136 VIPT nonaliasing Harvard caches
[ARM] 3765/1: S3C24XX: cleanup include/asm-arm/arch-s3c2410/dma.h
[ARM] 3764/1: S3C24XX: change type naming to kernel style
[ARM] 3763/1: add both rtcs to csb337 defconfig
[ARM] Fix ARM __raw_read_trylock() implementation
[ARM] 3750/3: Fix double VFP emulation for EABI kernels
|
|
It is not possible to find a sub-thread in ->children/->ptrace_children
lists, ptrace_attach() does not allow to attach to sub-threads.
Even if it was possible to ptrace the task from the same thread group,
we can't allow to release ->group_leader while there are others (ptracer)
threads in the same group.
Signed-off-by: Oleg Nesterov <oleg@tv-sign.ru>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
|
|
Harvard caches
Patch from George G. Davis
Resolve ARM1136 VIPT non-aliasing cache coherency issues observed when
using ptrace to set breakpoints and cleanup copy_{to,from}_user_page()
while we're here as requested by Russell King because "it's also far
too heavy on non-v6 CPUs".
NOTES:
1. Only access_process_vm() calls copy_{to,from}_user_page().
2. access_process_vm() calls get_user_pages() to pin down the "page".
3. get_user_pages() calls flush_dcache_page(page) which ensures cache
coherency between kernel and userspace mappings of "page". However
flush_dcache_page(page) may not invalidate I-Cache over this range
for all cases, specifically, I-Cache is not invalidated for the VIPT
non-aliasing case. So memory is consistent between kernel and user
space mappings of "page" but I-Cache may still be hot over this
range. IOW, we don't have to worry about flush_cache_page() before
memcpy().
4. Now, for the copy_to_user_page() case, after memcpy(), we must flush
the caches so memory is consistent with kernel cache entries and
invalidate the I-Cache if this mm region is executable. We don't
need to do anything after memcpy() for the copy_from_user_page()
case since kernel cache entries will be invalidated via the same
process above if we access "page" again. The flush_ptrace_access()
function (borrowed from SPARC64 implementation) is added to handle
cache flushing after memcpy() for the copy_to_user_page() case.
Signed-off-by: George G. Davis <gdavis@mvista.com>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
|
|
* master.kernel.org:/pub/scm/linux/kernel/git/gregkh/usb-2.6:
uhci-hcd: fix list access bug
USB: Support for ELECOM LD-USB20 in pegasus
USB: Add VIA quirk fixup for VT8235 usb2
USB: rtl8150_disconnect() needs tasklet_kill()
USB Storage: unusual_devs.h for Sony Ericsson M600i
USB Storage: Remove the finecam3 unusual_devs entry
UHCI: don't stop at an Iso error
usb gadget: g_ether spinlock recursion fix
USB: add all wacom device to hid-core.c blacklist
hid-core.c: Adds all GTCO CalComp Digitizers and InterWrite School Products to blacklist
USB floppy drive SAMSUNG SFD-321U/EP detected 8 times
|
|
Fix some more problems (inverted use of semaphores in some places). He
also moved my checks into within the protected section which is better.
Signed-off-by: Michael Hanselmann <linux-kernel@hansmi.ch>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: "Antonino A. Daplas" <adaplas@pol.net>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
|
|
Missed a place where I forgot to convert kfree() to kmem_cache_free() as
part of jbd-manage-its-own-slab changes.
Signed-off-by: Badari Pulavarty <pbadari@us.ibm.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
|
|
Since vma->vm_pgoff is in units of smallpages, VMAs for huge pages have the
lower HPAGE_SHIFT - PAGE_SHIFT bits always cleared, which results in badd
offsets to the interleave functions. Take this difference from small pages
into account when calculating the offset. This does add a 0-bit shift into
the small-page path (via alloc_page_vma()), but I think that is negligible.
Also add a BUG_ON to prevent the offset from growing due to a negative
right-shift, which probably shouldn't be allowed anyways.
Tested on an 8-memory node ppc64 NUMA box and got the interleaving I
expected.
Signed-off-by: Nishanth Aravamudan <nacc@us.ibm.com>
Signed-off-by: Adam Litke <agl@us.ibm.com>
Cc: Andi Kleen <ak@muc.de>
Acked-by: Christoph Lameter <clameter@engr.sgi.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
|
|
Fix some bugs in the patch that converted the IOC4 driver from port IO ops to
memio ops.
http://marc.theaimsgroup.com/?l=linux-ide&m=114895892231438&w=2
Problems fixed are:
- Call to default_hwif_mmiops() was not being done until _after_
first IO operation, resulting in the first IO operation being
done as a port IO op, instead of memio.
- request_region() calls needed to be request_mem_region()
- Incomplete error case handling.
- Non-usage of ioremap() and __iomem.
Signed-off-by: John Keller <jpk@sgi.com>
Signed-off-by: Jeremy Higdon <jeremy@sgi.com>
Cc: Alan Cox <alan@lxorguk.ukuu.org.uk>
Cc: Bartlomiej Zolnierkiewicz <B.Zolnierkiewicz@elka.pw.edu.pl>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
|
|
Revert the mixer element names of some Mic controls to the state of
2.6.17. This should fix the name mismatch in alsactl.
Signed-off-by: Takashi Iwai <tiwai@suse.de>
Cc: Jaroslav Kysela <perex@suse.cz>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
|
|
The port to genirq & the new powerpc interrupt model in 2.6.18 introduced a
bug in the legacy PowerMac PIC code (used on older machines) because of a
typo potentially causing hangs due to interrupt storms. This fixes it,
along with a performance issue causing us to do spurrious retriggers after
masking an interrupt.
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: "Antonino A. Daplas" <adaplas@pol.net>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
|
|
The via-pmu backlight code (introduced in 2.6.18) has various design issues
causing crashes on machines using it like the old Wallstreet powerbook
(Michael, the author, never managed to test on these and I just got my hand
on one of those old beasts).
This fixes them by no longer trying to hijack the backlight device of the
frontmost framebuffer (causing that framebuffer to crash) but having it's
own local bits instead. Might look weird but it's better that way on those
old machines, at least as a last-minute fix for 2.6.18. We might rework
the whole thing later. This patch also changes the way it gets notified of
sleep and wakeup in order to properly shut the backlight down on sleep and
bring it back on wakeup.
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: "Antonino A. Daplas" <adaplas@pol.net>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
|
|
This patch works around a complex dm-related deadlock/livelock down in the
mempool allocator.
Alasdair said:
Several dm targets suffer from this.
Mempools are not yet used correctly everywhere in device-mapper: they can
get shared when devices are stacked, and some targets share them across
multiple instances. I made fixing this one of the prerequisites for this
patch:
md-dm-reduce-stack-usage-with-stacked-block-devices.patch
which in some cases makes people more likely to hit the problem.
There's been some progress on this recently with (unfinished) dm-crypt
patches at:
http://www.kernel.org/pub/linux/kernel/people/agk/patches/2.6/editing/
(dm-crypt-move-io-to-workqueue.patch plus dependencies)
and:
I've no problems with a temporary workaround like that, but Milan Broz (a
new Redhat developer in the Czech Republic) has started reviewing all the
mempool usage in device-mapper so I'm expecting we'll soon have a proper fix
for this associated problems. [He's back from holiday at the start of next
week.]
For now, this sad-but-safe little patch will allow the machine to recover.
[akpm@osdl.org: rewrote changelog]
Cc: Alasdair G Kergon <agk@redhat.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
|
|
This patch schedules obsolete OSS drivers (with ALSA drivers that support
the same hardware) for removal.
A rationale of the patch is in
http://lkml.org/lkml/2006/7/11/186
Signed-off-by: Adrian Bunk <bunk@stusta.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
|
|
Olaf Kirch of SuSE tracked down a problem where module unloads of the IPMI
driver would occasionally result in Oopses. He tracked that down to a
variable that wasn't always initialized properly in some situations. This
patch initializes that variable. Olaf sent a patch that kzalloc-ed the
data, but this structure is large enough that I would perfer to not do
that. Thanks Olaf!
Signed-off-by: Corey Minyard <minyard@acm.org>
Cc: Olaf Kirch <okir@suse.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
|
|
Adds the description of the parameters from handle_bad_irq().
Signed-off-by: Henrik Kretzschmar <henne@nachtwindheim.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
|
|
The last argument of module_param is permissions, not default value.
Acked-by: Wim Van Sebroeck <wim@iguana.be>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
|
|
modprobe -v floppy on a Apple G5 writes incorrect stuff to dmesg:
Floppy drive(s): fd0 is 2.88M
The reason is that the legacy io check happens very late,
when part of the floppy stuff is already initialized.
check_legacy_ioport() returns either -ENODEV right away, or it walks
the device-tree looking for a floppy node.
Signed-off-by: Olaf Hering <olaf@aepfle.de>
Acked-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
|
|
Vitezslav Samel <samel@mail.cz> reports that an HP DL380 g4 fails using the
default arch due to the ISA bus having an ID of 32.
It would have worked OK with the generic arch - for some reason the default
arch doesn't support as many busses.
So bump that up to support 256 busses, but leave it at 32 if we're building a
tiny system to save a bit of memory.
Cc: Vitezslav Samel <samel@mail.cz>
Acked-by: Andi Kleen <ak@muc.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
|
|
We're testing the wrong task_struct field.
Acked-by: Chris Zankel <chris@zankel.net>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
|
|
Cleanup allocation and freeing of tsk->delays used by delay accounting.
This solves two problems reported for delay accounting:
1. oops in __delayacct_blkio_ticks
http://www.uwsg.indiana.edu/hypermail/linux/kernel/0608.2/1844.html
Currently tsk->delays is getting freed too early in task exit which can
cause a NULL tsk->delays to get accessed via reading of /proc/<tgid>/stats.
The patch fixes this problem by freeing tsk->delays closer to when
task_struct itself is freed up. As a result, it also eliminates the use of
tsk->delays_lock which was only being used (inadequately) to safeguard
access to tsk->delays while a task was exiting.
2. Possible memory leak in kernel/delayacct.c
http://www.uwsg.indiana.edu/hypermail/linux/kernel/0608.2/1389.html
The patch cleans up tsk->delays allocations after a bad fork which was
missing earlier.
The patch has been tested to fix the problems listed above and stress
tested with rapid calls to delay accounting's taskstats command interface
(which is the other path that can access the same data, besides the /proc
interface causing the oops above).
Signed-off-by: Shailabh Nagar <nagar@watson.ibm.com>
Cc: Balbir Singh <balbir@in.ibm.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
|
|
Apparently some systems export valid HPET addresses, but hpet_enable()
fails. Then when the HPET clocksource starts up, it only checks for a
valid HPET address, and the result is a system where time does not advance.
See http://bugme.osdl.org/show_bug.cgi?id=7062 for details.
This patch just makes sure we better check that the HPET is functional
before registering the HPET clocksource.
Signed-off-by: John Stultz <johnstul@us.ibm.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
|
|
Fix receive tty error handling in synclink_gt driver. Adrian reported
compiler warning for incorrect bit test against char variable. I
determined these and other device specific error bits were incorrectly
defined.
Signed-off-by: Paul Fulghum <paulkf@microgate.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
|
|
We need to be careful when referencing mirrors[i].rdev. It can disappear
under us at various times.
So:
fix a couple of problem places.
comment a couple of non-problem places
move an 'atomic_add' which deferences rdev down a little
way to some where where it is sure to not be NULL.
Signed-off-by: Neil Brown <neilb@suse.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
|
|
The ZVC counter update threshold is currently set to a fixed value of 32.
This patch sets up the threshold depending on the number of processors and
the sizes of the zones in the system.
With the current threshold of 32, I was able to observe slight contention
when more than 130-140 processors concurrently updated the counters. The
contention vanished when I either increased the threshold to 64 or used
Andrew's idea of overstepping the interval (see ZVC overstep patch).
However, we saw contention again at 220-230 processors. So we need higher
values for larger systems.
But the current default is already a bit of an overkill for smaller
systems. Some systems have tiny zones where precision matters. For
example i386 and x86_64 have 16M DMA zones and either 900M ZONE_NORMAL or
ZONE_DMA32. These are even present on SMP and NUMA systems.
The patch here sets up a threshold based on the number of processors in the
system and the size of the zone that these counters are used for. The
threshold should grow logarithmically, so we use fls() as an easy
approximation.
Results of tests on a system with 1024 processors (4TB RAM)
The following output is from a test allocating 1GB of memory concurrently
on each processor (Forking the process. So contention on mmap_sem and the
pte locks is not a factor):
X MIN
TYPE: CPUS WALL WALL SYS USER TOTCPU
fork 1 0.552 0.552 0.540 0.012 0.552
fork 4 0.552 0.548 2.164 0.036 2.200
fork 16 0.564 0.548 8.812 0.164 8.976
fork 128 0.580 0.572 72.204 1.208 73.412
fork 256 1.300 0.660 310.400 2.160 312.560
fork 512 3.512 0.696 1526.836 4.816 1531.652
fork 1020 20.024 0.700 17243.176 6.688 17249.863
So a threshold of 32 is fine up to 128 processors. At 256 processors contention
becomes a factor.
Overstepping the counter (earlier patch) improves the numbers a bit:
fork 4 0.552 0.548 2.164 0.040 2.204
fork 16 0.552 0.548 8.640 0.148 8.788
fork 128 0.556 0.548 69.676 0.956 70.632
fork 256 0.876 0.636 212.468 2.108 214.576
fork 512 2.276 0.672 997.324 4.260 1001.584
fork 1020 13.564 0.680 11586.436 6.088 11592.523
Still contention at 512 and 1020. Contention at 1020 is down by a third.
256 still has a slight bit of contention.
After this patch the counter threshold will be set to 125 which reduces
contention significantly:
fork 128 0.560 0.548 69.776 0.932 70.708
fork 256 0.636 0.556 143.460 2.036 145.496
fork 512 0.640 0.548 284.244 4.236 288.480
fork 1020 1.500 0.588 1326.152 8.892 1335.044
[akpm@osdl.org: !SMP build fix]
Signed-off-by: Christoph Lameter <clameter@sgi.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
|
|
Increments and decrements are usually grouped rather than mixed. We can
optimize the inc and dec functions for that case.
Increment and decrement the counters by 50% more than the threshold in
those cases and set the differential accordingly. This decreases the need
to update the atomic counters.
The idea came originally from Andrew Morton. The overstepping alone was
sufficient to address the contention issue found when updating the global
and the per zone counters from 160 processors.
Also remove some code in dec_zone_page_state.
Signed-off-by: Christoph Lameter <clameter@sgi.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
|
