improve the sysbench ramp-up phase and its peak throughput on
a 16way NUMA box, by turning on WAKE_AFFINE:
tip/sched tip/sched+wake-affine
-------------------------------------------------
1: 700 830 +15.65%
2: 1465 1391 -5.28%
4: 3017 3105 +2.81%
8: 5100 6021 +15.30%
16: 10725 10745 +0.19%
32: 10135 10150 +0.16%
64: 9338 9240 -1.06%
128: 8599 8252 -4.21%
256: 8475 8144 -4.07%
-------------------------------------------------
SUM: 57558 57882 +0.56%
this change also improves lat_ctx from 6.69 usecs to 1.11 usec:
$ ./lat_ctx -s 0 2
"size=0k ovr=1.19
2 1.11
$ ./lat_ctx -s 0 2
"size=0k ovr=1.22
2 6.69
in sysbench it's an overall win with some weakness at the lots-of-clients
side. That happens because we now under-balance this workload
a bit. To counter that effect, turn on NEWIDLE:
wake-idle wake-idle+newidle
-------------------------------------------------
1: 830 834 +0.43%
2: 1391 1401 +0.65%
4: 3105 3091 -0.43%
8: 6021 6046 +0.42%
16: 10745 10736 -0.08%
32: 10150 10206 +0.55%
64: 9240 9533 +3.08%
128: 8252 8355 +1.24%
256: 8144 8384 +2.87%
-------------------------------------------------
SUM: 57882 58591 +1.21%
as a bonus this not only improves the many-clients case but
also improves the (more important) rampup phase.
sysbench is a workload that quickly breaks down if the
scheduler over-balances, so since it showed an improvement
under NEWIDLE this change is definitely good.
Yanmin Zhang reported:
Comparing with 2.6.25, volanoMark has big regression with kernel 2.6.26-rc1.
It's about 50% on my 8-core stoakley, 16-core tigerton, and Itanium Montecito.
With bisect, I located the following patch:
| 18d95a2832 is first bad commit
| commit 18d95a2832
| Author: Peter Zijlstra <a.p.zijlstra@chello.nl>
| Date: Sat Apr 19 19:45:00 2008 +0200
|
| sched: fair-group: SMP-nice for group scheduling
Revert it so that we get v2.6.25 behavior.
Bisected-by: Yanmin Zhang <yanmin_zhang@linux.intel.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
-tip testing found the following build breakage:
drivers/built-in.o: In function `xen_suspend':
manage.c:(.text+0x4390f): undefined reference to `xen_console_resume'
with this config:
http://redhat.com/~mingo/misc/config-Thu_May_29_09_23_16_CEST_2008.bad
i have bisected it down to:
| commit 0e91398f2a
| Author: Jeremy Fitzhardinge <jeremy@goop.org>
| Date: Mon May 26 23:31:27 2008 +0100
|
| xen: implement save/restore
the problem is that drivers/xen/manage.c is built unconditionally if
CONFIG_XEN is enabled and makes use of xen_suspend(), but
drivers/char/hvc_xen.c, where the xen_suspend() method is implemented,
is only build if CONFIG_HVC_XEN=y as well.
i have solved this by providing a NOP implementation for xen_suspend()
in the !CONFIG_HVC_XEN case.
Signed-off-by: Ingo Molnar <mingo@elte.hu>
The commit
commit 4b82b27770
Author: Cyrill Gorcunov <gorcunov@gmail.com>
Date: Sat May 24 19:36:35 2008 +0400
set nmi_watchdog to NMI_IO_APIC as by default. This causes hangs on some
machines with buggy watchdogs. Fix it - i.e. restore old behaviour.
Thanks to Sitsofe Wheeler and Adrian Bunk for catching the problem
and Maciej W. Rozycki for explanation what is going on there.
Signed-off-by: Cyrill Gorcunov <gorcunov@gmail.com>
CC: Maciej W. Rozycki <macro@linux-mips.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
> +#define ARCH_KMALLOC_MINALIGN (sizeof(long) * 2)
> +#define ARCH_SLAB_MINALIGN (sizeof(long) * 2)
This doesn't work if SLAB is selected and slab debugging is enabled as
these are passed to the preprocessor, and the preprocessor doesn't
understand sizeof.
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
* 'for-linus' of git://git.kernel.dk/linux-2.6-block:
cfq-iosched: fix RCU problem in cfq_cic_lookup()
block: make blktrace use per-cpu buffers for message notes
Added in elevator switch message to blktrace stream
Added in MESSAGE notes for blktraces
block: reorder cfq_queue to save space on 64bit builds
block: Move the second call to get_request to the end of the loop
splice: handle try_to_release_page() failure
splice: fix sendfile() issue with relay
Specify the minimum slab/kmalloc alignment to be 8 bytes. This fixes a
crash when SLOB is selected as the memory allocator. The FRV arch needs
this so that it can use the load- and store-double instructions without
faulting. By default SLOB sets the minimum to be 4 bytes.
Signed-off-by: David Howells <dhowells@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Fix a typo in the header guard of asm/ipc.h.
Signed-off-by: Vegard Nossum <vegard.nossum@gmail.com>
Signed-off-by: David Howells <dhowells@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Currently it uses a single static char array, but that risks
being corrupted when multiple users issue message notes at the
same time. Make the buffers dynamically allocated when the trace
is setup and make them per-cpu instead.
The default max message size of 1k is also very large, the
interface is mainly for small text notes. So shrink it to 128 bytes.
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Allows messages to be inserted into blktrace streams.
Signed-off-by: Alan D. Brunelle <alan.brunelle@hp.com>
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Problem: An application violating the architectural rules regarding
operation dependencies and having specific Register Stack Engine (RSE)
state at the time of the violation, may result in an illegal operation
fault and invalid RSE state. Such faults may initiate a cascade of
repeated illegal operation faults within OS interruption handlers.
The specific behavior is OS dependent.
Implication: An application causing an illegal operation fault with
specific RSE state may result in a series of illegal operation faults
and an eventual OS stack overflow condition.
Workaround: OS interruption handlers that switch to kernel backing
store implement a check for invalid RSE state to avoid the series
of illegal operation faults.
The core of the workaround is the RSE_WORKAROUND code sequence
inserted into each invocation of the SAVE_MIN_WITH_COVER and
SAVE_MIN_WITH_COVER_R19 macros. This sequence includes hard-coded
constants that depend on the number of stacked physical registers
being 96. The rest of this patch consists of code to disable this
workaround should this not be the case (with the presumption that
if a future Itanium processor increases the number of registers, it
would also remove the need for this patch).
Move the start of the RBS up to a mod32 boundary to avoid some
corner cases.
The dispatch_illegal_op_fault code outgrew the spot it was
squatting in when built with this patch and CONFIG_VIRT_CPU_ACCOUNTING=y
Move it out to the end of the ivt.
Signed-off-by: Tony Luck <tony.luck@intel.com>
Hook into the device model to make sure that timekeeping's resume handler
is called. This deals with our clocksource's non-monotonicity over the
save/restore. Explicitly call clock_has_changed() to make sure that
all the timers get retriggered properly.
Signed-off-by: Jeremy Fitzhardinge <jeremy.fitzhardinge@citrix.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
This patch implements Xen save/restore and migration.
Saving is triggered via xenbus, which is polled in
drivers/xen/manage.c. When a suspend request comes in, the kernel
prepares itself for saving by:
1 - Freeze all processes. This is primarily to prevent any
partially-completed pagetable updates from confusing the suspend
process. If CONFIG_PREEMPT isn't defined, then this isn't necessary.
2 - Suspend xenbus and other devices
3 - Stop_machine, to make sure all the other vcpus are quiescent. The
Xen tools require the domain to run its save off vcpu0.
4 - Within the stop_machine state, it pins any unpinned pgds (under
construction or destruction), performs canonicalizes various other
pieces of state (mostly converting mfns to pfns), and finally
5 - Suspend the domain
Restore reverses the steps used to save the domain, ending when all
the frozen processes are thawed.
Signed-off-by: Jeremy Fitzhardinge <jeremy.fitzhardinge@citrix.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Add code to:
1. Deal with the console page being canonicalized. During save, the
console's mfn in the start_info structure is canonicalized to a pfn.
In order to deal with that, we always use a copy of the pfn and
indirect off that all the time. However, we fall back to using the
mfn if the pfn hasn't been initialized yet.
2. Restore the console event channel, and rebind it to the existing irq.
Signed-off-by: Jeremy Fitzhardinge <jeremy.fitzhardinge@citrix.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Add rebind_evtchn_irq(), which will rebind an device driver's existing
irq to a new event channel on restore. Since the new event channel
will be masked and bound to vcpu0, we update the state accordingly and
unmask the irq once everything is set up.
Signed-off-by: Jeremy Fitzhardinge <jeremy.fitzhardinge@citrix.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Add a config option to set the max size of a Xen domain. This is used
to scale the size of the physical-to-machine array; it ends up using
around 1 page/GByte, so there's no reason to be very restrictive.
For a 32-bit guest, the default value of 8GB is probably sufficient;
there's not much point in giving a 32-bit machine much more memory
than that.
Signed-off-by: Jeremy Fitzhardinge <jeremy.fitzhardinge@citrix.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
We now support the use of memory hotplug, so the physical to machine
page mapping structure must be dynamic. This is implemented as a
two-level radix tree structure, which allows us to efficiently
incrementally allocate memory for the p2m table as new pages are
added.
Signed-off-by: Jeremy Fitzhardinge <jeremy.fitzhardinge@citrix.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Add xen handles realted definitions for xen memory which ia64/xen needs.
Pointer argumsnts for ia64/xen hypercall are passed in pseudo physical
address (guest physical address) so that it is required to convert
guest kernel virtual address into pseudo physical address.
The xen guest handle represents such arguments.
Signed-off-by: Isaku Yamahata <yamahata@valinux.co.jp>
Signed-off-by: Jeremy Fitzhardinge <jeremy.fitzhardinge@citrix.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
The pvfb backend indicates dynamic mode support by creating node
feature_resize with a non-zero value in its xenstore directory.
xen-fbfront sends a resize notification event on mode change. Fully
backwards compatible both ways.
Framebuffer size and initial resolution can be controlled through
kernel parameter xen_fbfront.video. The backend enforces a separate
size limit, which it advertises in node videoram in its xenstore
directory.
xen-kbdfront gets the maximum screen resolution from nodes width and
height in the backend's xenstore directory instead of hardcoding it.
Additional goodie: support for larger framebuffers (512M on a 64-bit
system with 4K pages).
Changing the number of bits per pixels dynamically is not supported,
yet.
Ported from
http://xenbits.xensource.com/linux-2.6.18-xen.hg?rev/92f7b3144f41http://xenbits.xensource.com/linux-2.6.18-xen.hg?rev/bfc040135633
Signed-off-by: Pat Campbell <plc@novell.com>
Signed-off-by: Markus Armbruster <armbru@redhat.com>
Signed-off-by: Jeremy Fitzhardinge <jeremy.fitzhardinge@citrix.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Without console= arguments on the kernel command line, the first
console to register becomes enabled and the preferred console (the one
behind /dev/console). This is normally tty (assuming
CONFIG_VT_CONSOLE is enabled, which it commonly is).
This is okay as long tty is a useful console. But unless we have the
PV framebuffer, and it is enabled for this domain, tty0 in domU is
merely a dummy. In that case, we want the preferred console to be the
Xen console hvc0, and we want it without having to fiddle with the
kernel command line. Commit b8c2d3dfbc
did that for us.
Since we now have the PV framebuffer, we want to enable and prefer tty
again, but only when PVFB is enabled. But even then we still want to
enable the Xen console as well.
Problem: when tty registers, we can't yet know whether the PVFB is
enabled. By the time we can know (xenstore is up), the console setup
game is over.
Solution: enable console tty by default, but keep hvc as the preferred
console. Change the preferred console to tty when PVFB probes
successfully, unless we've been given console kernel parameters.
Signed-off-by: Markus Armbruster <armbru@redhat.com>
Signed-off-by: Jeremy Fitzhardinge <jeremy.fitzhardinge@citrix.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Add pte_flags() to extract the flags from a pte. This is a special
case of pte_val() which is only guaranteed to return the pte's flags
correctly; the page number may be corrupted or missing.
The intent is to allow paravirt implementations to return pte flags
without having to do any translation of the page number (most notably,
Xen).
Signed-off-by: Jeremy Fitzhardinge <jeremy.fitzhardinge@citrix.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Use the new sched_op hypercall, mainly because xenner doesn't support
the old one.
Signed-off-by: Jeremy Fitzhardinge <jeremy.fitzhardinge@citrix.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Xen will trap and emulate clts, but its better to use a hypercall.
Also, xenner doesn't handle clts.
Signed-off-by: Jeremy Fitzhardinge <jeremy.fitzhardinge@citrix.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Add a couple of functions which can write directly to the Xen console
for debugging. This output ends up on the host's dom0 console
(assuming it allows the domain to write there).
Signed-off-by: Jeremy Fitzhardinge <jeremy.fitzhardinge@citrix.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Rather than making _PAGE_* constants signed, and then relying on
sign-extension to make sure that masks derived from them are wide
enough, just explicitly type them pteval_t. This guarantees that they
and any derived values are the right size for the current pte format.
The reliance on sign extension is fragile, and invokes some very
subtle corners of the C type system.
Signed-off-by: Jeremy Fitzhardinge <jeremy.fitzhardinge@citrix.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
gcc 4.0 fails to allocate %eax for the pattern operand in the rep
store instructions used by memset; force it to do so by declaring a
register variable.
Signed-off-by: H. Peter Anvin <hpa@zytor.com>
Based on Roland's patch. This approach was suggested by Austin Clements
from the very beginning, and then by Linus.
As Austin pointed out, the execing task can be killed by SI_TIMER signal
because exec flushes the signal handlers, but doesn't discard the pending
signals generated by posix timers. Perhaps not a bug, but people find this
surprising. See http://bugzilla.kernel.org/show_bug.cgi?id=10460
Signed-off-by: Oleg Nesterov <oleg@tv-sign.ru>
Cc: Austin Clements <amdragon+kernelbugzilla@mit.edu>
Cc: Roland McGrath <roland@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
* git://git.kernel.org/pub/scm/linux/kernel/git/davem/net-2.6: (52 commits)
vlan: Use bitmask of feature flags instead of seperate feature bits
fmvj18x_cs: add NextCom NC5310 rev B support
xirc2ps_cs: re-initialize the multicast address in do_reset
3C509: rx_bytes should not be increased when alloc_skb failed
NETFRONT: Use __skb_queue_purge()
VIRTIO: Use __skb_queue_purge()
phylib: do EXPORT_SYMBOL on get_phy_id
netlink: Fix nla_parse_nested_compat() to call nla_parse() directly
WAN: protect HDLC proto list while insmod/rmmod
drivers/net/fs_enet: remove null pointer dereference
S2io: Version update for napi and MSI-X patches
S2io: Added napi support when MSIX is enabled.
S2io: Move all the transmit completions to a single msi-x (alarm) vector
drivers/net/ehea - remove unnecessary memset after kzalloc
au1000_eth: remove useless check
Blackfin EMAC Driver: Removed duplicated include <linux/ethtool.h>
cpmac bugfixes and enhancements
e1000e: use resource_size_t, not unsigned long, for phys addrs
net/usb: add support for Apple USB Ethernet Adapter
uli526x: add support for netpoll
...
If a journal checksum error is detected, the ext4 filesystem will call
ext4_error(), and the mount will either continue, become a read-only
mount, or cause a kernel panic based on the superblock flags
indicating the user's preference of what to do in case of filesystem
corruption being detected.
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
Align i2c_device_id.driver_data to 8 bytes to not fail on crossbuilds.
(Added in d2653e9273.)
Signed-off-by: Jiri Slaby <jirislaby@gmail.com>
Signed-off-by: Jean Delvare <khali@linux-fr.org>
global_reg_snapshot shouldn't be visible in our userspace headers.
Signed-off-by: Adrian Bunk <bunk@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Commit d167a518 "[PATCH] x86_64: x86_64 version of the smp alternative patch."
has left the LOCK macro in include/asm-x86_64/atomic.h, which is now
include/asm-x86/atomic_64.h. Its scope should be local to the file, other
architectures don't provide it, I couldn't find an in-tree user of it and
allyesconfig, allmodconfig and allnoconfig build fine without it, so this patch
removes it.
Signed-off-by: Sven Wegener <sven.wegener@stealer.net>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
By slightly changing 32bit mode die_nmi() we may unify the
interface and make it common for both (32/64bit) modes
Signed-off-by: Cyrill Gorcunov <gorcunov@gmail.com>
Cc: hpa@zytor.com
Cc: mingo@redhat.com
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Fernando Luis Vazquez Cao <fernando@oss.ntt.co.jp>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
and make e820_mark_nosave_regions to take limit_pfn to use max_low_pfn
for 32bit and end_pfn for 64bit
Signed-off-by: Yinghai Lu <yhlu.kernel@gmail.com>
Cc: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>