This one isn't related to previous patch. If online cpus are
below NUM_INVALIDATE_TLB_VECTORS, we don't need the lock. The
comments in the code declares we don't need the check, but a hot
lock still needs an atomic operation and expensive, so add the
check here.
Uses nr_cpu_ids here as suggested by Eric Dumazet.
Signed-off-by: Shaohua Li <shaohua.li@intel.com>
Acked-by: Eric Dumazet <eric.dumazet@gmail.com>
Cc: Andi Kleen <andi@firstfloor.org>
LKML-Reference: <1295232730.1949.710.camel@sli10-conroe>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Make the maxium TLB invalidate vectors depend on NR_CPUS linearly,
with a maximum of 32 vectors.
We currently only have 8 vectors for TLB invalidation and that is clearly
inadequate. If we have a lot of CPUs, the CPUs need share the 8 vectors and
tlbstate_lock is used to protect them. flush_tlb_page() is
heavily used in page reclaim, which will cause a lot of lock
contention for tlbstate_lock.
Andi Kleen suggested increasing the vectors number to 32, which should be
good for current typical systems to reduce the tlbstate_lock contention.
My test system has 4 sockets and 64G memory, and 64 CPUs. My
workload creates 64 processes. Each process mmap reads a big
empty sparse file. The total size of the files are 2*total_mem,
so this will cause a lot of page reclaim.
Below is the result I get from perf call-graph profiling:
without the patch:
------------------
24.25% usemem [kernel] [k] _raw_spin_lock
|
--- _raw_spin_lock
|
|--42.15%-- native_flush_tlb_others
with the patch:
------------------
14.96% usemem [kernel] [k] _raw_spin_lock
|
--- _raw_spin_lock
|--13.89%-- native_flush_tlb_others
So this heavily reduces the tlbstate_lock contention.
Suggested-by: Andi Kleen <andi@firstfloor.org>
Signed-off-by: Shaohua Li <shaohua.li@intel.com>
Cc: Eric Dumazet <eric.dumazet@gmail.com>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <1295232727.1949.709.camel@sli10-conroe>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Add up to 32 invalidate_interrupt handlers. How many handlers are
added depends on NUM_INVALIDATE_TLB_VECTORS. So if
NUM_INVALIDATE_TLB_VECTORS is smaller than 32, we reduce code
size.
Signed-off-by: Shaohua Li <shaohua.li@intel.com>
Cc: Andi Kleen <andi@firstfloor.org>
Cc: Eric Dumazet <eric.dumazet@gmail.com>
LKML-Reference: <1295232725.1949.708.camel@sli10-conroe>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Cleanup the vector usage and make them continuous if possible.
Signed-off-by: Shaohua Li <shaohua.li@intel.com>
Cc: Andi Kleen <andi@firstfloor.org>
Cc: Eric Dumazet <eric.dumazet@gmail.com>
LKML-Reference: <1295232722.1949.707.camel@sli10-conroe>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
We use it in non __cpuinit code now too so drop marker.
Signed-off-by: Borislav Petkov <borislav.petkov@amd.com>
LKML-Reference: <20110211171754.GA21047@aftab>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Conflicts:
arch/x86/mm/numa_64.c
Merge reason: fix the conflict, update to latest -rc and pick up this
dependent fix from Yinghai:
e6d2e2b2b1: memblock: don't adjust size in memblock_find_base()
Signed-off-by: Ingo Molnar <mingo@elte.hu>
This allows the cache/processor/fault glue to be more easily used
from assembler code. Tested on Assabet and Tegra 2.
Tested-by: Colin Cross <ccross@android.com>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
commit a3c08e5d(x86: Convert irq_chip access to new functions)
accidentally zapped desc = irq_to_desc(irq); in the vector loop.
So we lock some random irq descriptor.
Add it back.
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Cc: <stable@kernel.org> # .37
* 's5p-fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/kgene/linux-samsung:
ARM: SAMSUNG: Ensure struct sys_device is declared in plat/pm.h
ARM: S5PV310: Cleanup System MMU
ARM: S5PV310: Add support System MMU on SMDKV310
The ptrace debug information register was advertising breakpoint and
watchpoint resources for unsupported debug architectures. This meant
that setting breakpoints on these architectures would appear to succeed,
although they would never fire in reality.
This patch fixes the breakpoint slot probing so that it returns 0 when
running on an unsupported debug architecture.
Reported-by: Ulrich Weigand <ulrich.weigand@de.ibm.com>
Signed-off-by: Will Deacon <will.deacon@arm.com>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
Reading baseline CP14 registers, other than DBGDIDR, when the OS Lock
is set leads to UNPREDICTABLE behaviour.
This patch ensures that we clear the OS lock before accessing anything
other than the DBGDIDR, thereby avoiding this behaviour.
Signed-off-by: Will Deacon <will.deacon@arm.com>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
locomo_info isn't actually used as a platform_data on collie platform:
arm/mach-sa1100/collie.c:237: warning: ‘locomo_info’ defined but not used
So locomo driver doesn't setup IRQs correctly. Pass locomo_info to the
driver.
Signed-off-by: Dmitry Eremin-Solenikov <dbaryshkov@gmail.com>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
rmk says: "You might as well make OABI_COMPAT depend on !THUMB2_KERNEL.
OABI userland is useless without FPA support."
nwfpe doesn't work with Thumb-2 anyway and will probably never get
ported, so I can't argue with that.
This patch implements the dependency change.
Signed-off-by: Dave Martin <dave.martin@linaro.org>
Acked-by: Nicolas Pitre <nicolas.pitre@linaro.org>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
Both xen_hvm_init_shared_info and xen_build_mfn_list_list can be
called at resume time as well as at start of day but only reference
__init functions (extend_brk) at start of day. Hence annotate with
__ref.
WARNING: arch/x86/built-in.o(.text+0x4f1): Section mismatch in reference
from the function xen_hvm_init_shared_info() to the function
.init.text:extend_brk()
The function xen_hvm_init_shared_info() references
the function __init extend_brk().
This is often because xen_hvm_init_shared_info lacks a __init
annotation or the annotation of extend_brk is wrong.
xen_hvm_init_shared_info calls extend_brk() iff !shared_info_page and
initialises shared_info_page with the result. This happens at start of
day only.
WARNING: arch/x86/built-in.o(.text+0x599b): Section mismatch in reference
from the function xen_build_mfn_list_list() to the function
.init.text:extend_brk()
The function xen_build_mfn_list_list() references
the function __init extend_brk().
This is often because xen_build_mfn_list_list lacks a __init
annotation or the annotation of extend_brk is wrong.
(this warning occurs multiple times)
xen_build_mfn_list_list only calls extend_brk() at boot time, while
building the initial mfn list list
Signed-off-by: Ian Campbell <ian.campbell@citrix.com>
Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
CC arch/x86/xen/p2m.o
arch/x86/xen/p2m.c: In function 'm2p_remove_override':
arch/x86/xen/p2m.c:460: warning: 'address' may be used uninitialized in this function
arch/x86/xen/p2m.c: In function 'm2p_add_override':
arch/x86/xen/p2m.c:426: warning: 'address' may be used uninitialized in this function
In actual fact address is inialised in one "if (!PageHighMem(page))"
statement and used in a second and so is always initialised before
use.
Signed-off-by: Ian Campbell <ian.campbell@citrix.com>
Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
Fixes 2 small regressions of recent iomux changes:
- current MX51_PAD_UART2_TXD__UART2_TXD declaration overwrites
IOMUXC_UART2_IPP_UART_RXD_MUX_SELECT_INPUT register (0x09ec) and prevent
UART2 Rx from working properly.
(Tested on my custom i.MX51 board where UART2 is used as console)
- current MX51_PAD_USBH1_DATA2__UART2_TXD declaration also has the same problem.
(Not tested)
Signed-off-by: Julien Boibessot <julien.boibessot@armadeus.com>
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>
The arguments to these callbacks were changed in
e981a30 (ARM: mx3: irq_data conversion.)
but the comments were not adapted.
Cc: Lennert Buytenhek <buytenh@secretlab.ca>
Signed-off-by: Uwe Kleine-König <u.kleine-koenig@pengutronix.de>
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>
mxc_board_init is too generic to be useful. Additionally change some
mxc_timer to ${machine}_timer, too.
Signed-off-by: Uwe Kleine-König <u.kleine-koenig@pengutronix.de>
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>
mach-aaec2000 is no longer actively maintained and is only receiving
fixups to remain building with other kernel updates.
Cc: Bellido Nicolas <ml@acolin.be>
Signed-off-by: Jamie Iles <jamie@jamieiles.com>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
Add a function to set the SCU low-power mode for SMP CPUs. This
centralizes this functionality rather than having to expose the
SCU register definitions to each platform.
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
The mxc-ehci driver calls SoC specific phy initialization right after
calling board specific initialization. To offer greater flexibility for
boards to setup the phy and to get rid of some unnecessary flags in
platform data this patch lets the boards call the SoC specific phy
initialization and remove it from the driver.
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>
Currently we have a mxc_initialize_usb_hw which is called on every
i.MX SoC. This function dispatches the different SoC types, which
is quite ugly. This patch moves the SoC specific USB initialization
to their correspondive mach directories.
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>
Adding one single defconfig for the tegra family of boards, to over time
cover the superset of supported platform and drivers.
Signed-off-by: Olof Johansson <olof@lixom.net>
Signed-off-by: Colin Cross <ccross@android.com>
Sometimes, due to high interrupt latency in the continuous mode
of DMA transfer, the half buffer complete interrupt is handled
after DMA has transferred the full buffer. When this is detected,
stop DMA immediately and restart with the next buffer if the next
buffer is ready.
originally fixed by Victor(Weiguo) Pan <wpan@nvidia.com>
In place of using the simple spin_lock()/spi_unlock() in the
interrupt thread, using the spin_lock_irqsave() and
spin_unlock_irqrestore(). The lock is shared between the normal
process context and interrupt context.
originally fixed by Laxman Dewangan (ldewangan@nvidia.com)
The use of shadow registers caused memory corruption at physical
address 0 because the enable bit was not shadowed, and assuming it
needed to be set would enable an unconfigured dma block. Most of the
register accesses don't need to know the previous state of the
registers, and the few places that do need to modify only a few bits
in the registers are the same ones that were sometimes incorrectly
setting the enable bit. This patch convert tegra_dma_update_hardware
to set the entire register, and the other users to read-modify-write,
and drops the shadow registers completely.
Also fixes missing locking in tegra_dma_allocate_channel
Signed-off-by: Colin Cross <ccross@android.com>
On Tegra, calling clk_set_rate on the CPU clock may call into the
regulator API. If the regulator driver that controls the CPU
voltage rail has been suspended, this can lead to attempted
communication with a hardware block that has already been turned
off.
Adds a SUSPEND_PREPARE notification hook to drop the frequency to
the lowest possible during suspend.
Also adds 216MHz (off of PLLP) as the lowest CPU frequency, which
allows PLLX to be turned off.
Signed-off-by: Colin Cross <ccross@android.com>
Adds gart, hdmi, avp, host1x, and pwm controllers to mach/iomap.h
Signed-off-by: Gary King <gking@nvidia.com>
Signed-off-by: Colin Cross <ccross@android.com>
Previously we were relying on it being pulled in by other headers for
the prototype of s3c24xx_irq_suspend() and s3c24xx_irq_resume().
Signed-off-by: Mark Brown <broonie@opensource.wolfsonmicro.com>
Signed-off-by: Kukjin Kim <kgene.kim@samsung.com>
This patch cleans following up.
- Moved definition of System MMU IPNUM into mach/sysmmu.h
- Removed useless SYSMMU_DEBUG configuration
- Removed useless header file plat/sysmmu.h
Signed-off-by: Kukjin Kim <kgene.kim@samsung.com>
The 's5pv310_device_sysmmu' is used on SMDKV310. But since it is not
compiled now, there is a build error. To fix this compilation error,
S5PV310_DEV_SYSMMU needs to be selected for SMDKV310 board.
This patch enables System MMU support on SMDKV310.
Signed-off-by: Thomas Abraham <thomas.ab@samsung.com>
[kgene.kim@samsung.com: Adding description]
Signed-off-by: Kukjin Kim <kgene.kim@samsung.com>
* 'tty-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/tty-2.6:
serial: bfin_5xx: split uart RX lock from uart port lock to avoid deadlock
68360serial: Plumb in rs_360_get_icount()
n_gsm: copy mtu over when configuring via ioctl interface
virtio: console: Move file back to drivers/char/
Convert all the parisc driver interrupt handlers (dino, eisa, gsc,
iosapic and superio) as well as the cpu interrupts. Prepare
show_interrupts for GENERIC_HARDIRQS_NO_DEPRECATED and finally selects
that Kconfig option
[jejb: compile and testing fixes]
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>
The content for ALT_SMP() in the definition of WFE() expands to 6
bytes (IT cc ; WFEcc.W), which breaks the assumptions of the fixup
code, leading to lockups when the affected code gets run.
This patch works around the problem by explicitly using an
IT + WFEcc.N pair.
Signed-off-by: Dave Martin <dave.martin@linaro.org>
Acked-by: Will Deacon <will.deacon@arm.com>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
Since tail is the previous fp - 1, we need to compare the new fp with tail + 1
to ensure that we don't end up passing in the same tail again, in order to
avoid a potential infinite loop in the perf interrupt handler (which has been
observed to occur). A similar fix seems to be needed in the OProfile code.
Acked-by: Will Deacon <will.deacon@arm.com>
Signed-off-by: Rabin Vincent <rabin.vincent@stericsson.com>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
With certain configurations, we inline the unlock functions in modules,
which results in SMP alternatives being created in modules. We need to
fix those up when loading a module to prevent undefined instruction
faults.
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
This used to be done in .map_io which is supposed to only setup the memory
mapping.
Signed-off-by: Uwe Kleine-König <u.kleine-koenig@pengutronix.de>
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>
