Fixed indentation for print messages and code.
Signed-off-by: Tomas Melin <tomas.melin@iki.fi>
Acked-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Combined two if statements making the code cleaner, removing one level
of indentation.
Signed-off-by: Tomas Melin <tomas.melin@iki.fi>
Acked-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Removed useless comments and cleaned up text.
Signed-off-by: Tomas Melin <tomas.melin@iki.fi>
Acked-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Simplifies function logic by assuming that n_chan >1 if not <=1.
Removes one level of indentation.
Signed-off-by: Tomas Melin <tomas.melin@iki.fi>
Acked-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
zspage_order defines how many pages are needed to make a zspage.
So _order_ is rather awkward naming. It already deceive Jonathan
- http://lwn.net/Articles/477067/
" For each size, the code calculates an optimum number of pages (up to 16)"
Let's change from _order_ to _pages_ and some function names.
Signed-off-by: Minchan Kim <minchan@kernel.org>
Acked-by: Nitin Gupta <ngupta@vflare.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
This test is always true so it means we revalidate the length every
time, which generates more network traffic. When it is SEEK_SET or
SEEK_CUR, then we don't need to revalidate.
Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Reviewed-by: Jeff Layton <jlayton@redhat.com>
Signed-off-by: Steve French <sfrench@us.ibm.com>
twl6040 has three power supply source:
VBAT needs to be connected to VBAT, VIO, and V2V1.
Add regulator support for the VIO, V2V1 supplies.
Initially handle the two supply together with bulk commands.
Signed-off-by: Peter Ujfalusi <peter.ujfalusi@ti.com>
Reviewed-by: Mark Brown <broonie@opensource.wolfsonmicro.com>
Signed-off-by: Tero Kristo <t-kristo@ti.com>
Signed-off-by: Samuel Ortiz <sameo@linux.intel.com>
Function rename to ensure that the functionality of nfs_unlock_request()
mirrors that of nfs_lock_request(). Then let nfs_unlock_and_release_request()
do the work of what used to be called nfs_unlock_request()...
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
Cc: Fred Isaman <iisaman@netapp.com>
We only have two places where we need to grab a reference when trying
to lock the nfs_page. We're better off making that explicit.
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
Cc: Fred Isaman <iisaman@netapp.com>
We now hold a reference to the nfs_page across the calls to
nfs_set_page_writeback and nfs_end_page_writeback, and that
means we already have a reference to the struct page.
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
Cc: Fred Isaman <iisaman@netapp.com>
We have to unlock the nfs_page before we call nfs_end_page_writeback
to avoid races with functions that expect the page to be unlocked
when PG_locked and PG_writeback are not set.
The problem is that nfs_unlock_request also releases the nfs_page,
causing a deadlock if the release of the nfs_open_context
triggers an iput() while the PG_writeback flag is still set...
The solution is to separate the unlocking and release of the nfs_page,
so that we can do the former before nfs_end_page_writeback and the
latter after.
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
Cc: Fred Isaman <iisaman@netapp.com>
Things have slowed down a lot for us, but we have five more fixes for
omap and kirkwood below. Three are for boards setup issues, two are
SoC-level fixes.
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.11 (GNU/Linux)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=GX6x
-----END PGP SIGNATURE-----
Merge tag 'fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/arm/arm-soc
Pull ARM: SoC fixes from Olof Johansson:
"Things have slowed down a lot for us, but we have five more fixes for
omap and kirkwood below. Three are for boards setup issues, two are
SoC-level fixes."
* tag 'fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/arm/arm-soc:
ARM: OMAP: igep0020: fix smsc911x dummy regulator id
ARM: orion5x: Fix GPIO enable bits for MPP9
ARM: kirkwood: add missing kexec.h include
ARM: OMAP: Revert "ARM: OMAP: ctrl: Fix CONTROL_DSIPHY register fields"
ARM: OMAP1: Amstrad Delta: Fix wrong IRQ base in FIQ handler
This is a last minute bug fix that was only just noticed since the code
path that's being exercised here is one that is fairly rarely used. The
changelog for the change itself is extremely clear and the code itself
is obvious to inspection so should be pretty safe.
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.12 (GNU/Linux)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=xcal
-----END PGP SIGNATURE-----
Merge tag 'regmap-3.4' of git://git.kernel.org/pub/scm/linux/kernel/git/broonie/regmap
Pull last minute regman bug fix from Mark Brown:
"This is a last minute bug fix that was only just noticed since the
code path that's being exercised here is one that is fairly rarely
used. The changelog for the change itself is extremely clear and the
code itself is obvious to inspection so should be pretty safe."
* tag 'regmap-3.4' of git://git.kernel.org/pub/scm/linux/kernel/git/broonie/regmap:
regmap: fix possible memory corruption in regmap_bulk_read()
The supply is connected to the DSS DO-D5 pins and is thus needed for
both serial and parallel display interfaces on the igep0030 as well as
the igep0020.
If the igep0030 module isn't connected to a display, no DSI or DPI
display will be specified in board code, and the DSS driver won't enable
to VPLL2 regulator anyway.
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Acked-by: Enric Balletbo i Serra <eballetbo@gmail.com>
Tested-by: Enric Balletbo i Serra <eballetbo@gmail.com>
Signed-off-by: Tony Lindgren <tony@atomide.com>
Pull KVM fixes from Avi Kivity:
"Two asynchronous page fault fixes (one guest, one host), a powerpc
page refcount fix, and an ia64 build fix."
* git://git.kernel.org/pub/scm/virt/kvm/kvm:
KVM: ia64: fix build due to typo
KVM: PPC: Book3S HV: Fix refcounting of hugepages
KVM: Do not take reference to mm during async #PF
KVM: ensure async PF event wakes up vcpu from halt
Pull powerpc fixes from Benjamin Herrenschmidt:
"Here are a couple of last minute fixes for 3.4 for regressions
introduced by my rewrite of the lazy irq masking code."
* 'merge' of git://git.kernel.org/pub/scm/linux/kernel/git/benh/powerpc:
powerpc/irq: Make alignment & program interrupt behave the same
powerpc/irq: Fix bug with new lazy IRQ handling code
The INTC core is using a Kconfig option for TI816x which doesn't
exist. Convert it to use TI81XX.
Cc: Hemant Pedanekar <hemantp@ti.com>
Signed-off-by: Kevin Hilman <khilman@ti.com>
Acked-by: Vaibhav Hiremath <hvaibhav@ti.com>
Signed-off-by: Tony Lindgren <tony@atomide.com>
Each irq_chip for the main interrupt controller has offsets set for irq
masking registers, which added to respective base results in a pointer
to appropriate hardware register. However this is not correct for
INTC_CONTROL as there is only one INTC_CONTROL register. This does not
cause problems because generic ack code is never called, but remove
this assignment to avoid confusion.
Signed-off-by: Grazvydas Ignotas <notasas@gmail.com>
Signed-off-by: Tony Lindgren <tony@atomide.com>
Add statics to board-omap4-panda.c's internal functions and data
structures to remove sparse warnings:
arch/arm/mach-omap2/board-omap4panda.c:234:29: warning: symbol
'omap_panda_wlan_data' was not declared. Should it be static?
arch/arm/mach-omap2/board-omap4panda.c:441:24: warning: symbol
'omap4_panda_dvi_device' was not declared. Should it be static?
arch/arm/mach-omap2/board-omap4panda.c:451:12: warning: symbol
'omap4_panda_dvi_init' was not declared. Should it be static?
arch/arm/mach-omap2/board-omap4panda.c:512:13: warning: symbol
'omap4_panda_display_init' was not declared. Should it be static?
Signed-off-by: Tomi Valkeinen <tomi.valkeinen@ti.com>
Signed-off-by: Tony Lindgren <tony@atomide.com>
Looks like the register offsets are incorrect in the OMAP mailbox code
(arch/arm/mach-omap2/mailbox.c) for the OMAP4_MAILBOX_IRQ* macros. The
discrepancy is with p.224 of TI document SPRUGX9 and p3891 of SWPU231K.
Acked-by: Vaibhav Bedia <vaibhav.bedia@ti.com>
Signed-off-by: Henry Chan <enli.chan@gmail.com>
Signed-off-by: Tony Lindgren <tony@atomide.com>
arch/arm/plat-omap/devices.c: In function 'omap_dsp_reserve_sdram_memblock':
arch/arm/plat-omap/devices.c:170: warning: format '%x' expects type 'unsigned int', but argument 3 has type 'phys_addr_t'
arch/arm/mach-omap2/dsp.c: In function 'omap_dsp_init':
arch/arm/mach-omap2/dsp.c:60: warning: format '%x' expects type 'unsigned int', but argument 3 has type 'phys_addr_t'
arch/arm/mach-omap2/dsp.c:60: warning: format '%x' expects type 'unsigned int', but argument 4 has type 'phys_addr_t'
Signed-off-by: Felipe Contreras <felipe.contreras@gmail.com>
Signed-off-by: Tony Lindgren <tony@atomide.com>
We need to read the real register values
Signed-off-by: Brian Austin <brian.austin@cirrus.com>
Signed-off-by: Mark Brown <broonie@opensource.wolfsonmicro.com>
perf stat on PPC currently fails to run:
$ perf stat -- sleep 1
Error: open_counter returned with 6 (No such device or address). /bin/dmesg may provide additional information.
Fatal: Not all events could be opened.
The problem is that until 2.6.37 (behavior changed with commit b0a873e)
perf on PPC returns ENXIO when hw_perf_event_init() fails. With this
patch we get the expected behavior:
$ perf stat -v -- sleep 1
cycles event is not supported by the kernel.
stalled-cycles-frontend event is not supported by the kernel.
stalled-cycles-backend event is not supported by the kernel.
instructions event is not supported by the kernel.
branches event is not supported by the kernel.
branch-misses event is not supported by the kernel.
...
Signed-off-by: David Ahern <dsahern@gmail.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/1336490956-57145-1-git-send-email-dsahern@gmail.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Kmemleak tracks the percpu allocations via a specific API and the
originally allocated areas must be removed from kmemleak (via
kmemleak_free). The code was already doing this for SMP systems.
Reported-by: Sami Liedes <sami.liedes@iki.fi>
Cc: Tejun Heo <tj@kernel.org>
Cc: Christoph Lameter <cl@linux-foundation.org>
Signed-off-by: Catalin Marinas <catalin.marinas@arm.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
pcpu_embed_first_chunk() allocates memory for each node, copies percpu
data and frees unused portions of it before proceeding to the next
group. This assumes that allocations for different nodes doesn't
overlap; however, depending on memory topology, the bootmem allocator
may end up allocating memory from a different node than the requested
one which may overlap with the portion freed from one of the previous
percpu areas. This leads to percpu groups for different nodes
overlapping which is a serious bug.
This patch separates out copy & partial free from the allocation loop
such that all allocations are complete before partial frees happen.
This also fixes overlapping frees which could happen on allocation
failure path - out_free_areas path frees whole groups but the groups
could have portions freed at that point.
Signed-off-by: Tejun Heo <tj@kernel.org>
Cc: stable@vger.kernel.org
Reported-by: "Pavel V. Panteleev" <pp_84@mail.ru>
Tested-by: "Pavel V. Panteleev" <pp_84@mail.ru>
LKML-Reference: <E1SNhwY-0007ui-V7.pp_84-mail-ru@f220.mail.ru>
Current OMAP code supports couple of clocksource options based
on compilation flag (CONFIG_OMAP_32K_TIMER). The 32KHz sync-timer
and a gptimer which can run on 32KHz or system clock (e.g 38.4 MHz).
So there can be 3 options -
1. 32KHz sync-timer
2. Sys_clock based (e.g 13/19.2/26/38.4 MHz) gptimer
3. 32KHz based gptimer.
The optional gptimer based clocksource was added so that it can
give the high precision than sync-timer, so expected usage was 2
and not 3.
Unfortunately option 2, clocksource doesn't meet the requirement of
free-running clock as per clocksource need. It stops in low power states
when sys_clock is cut. That makes gptimer based clocksource option
useless for OMAP2/3/4 devices with sys_clock as a clock input.
So, in order to use option 2, deeper idle state MUST be disabled.
Option 3 will still work but it is no better than 32K sync-timer
based clocksource.
We must support both sync timer and gptimer based clocksource as
some OMAP based derivative SoCs like AM33XX does not have the
sync timer.
Considering above, make sync-timer and gptimer clocksource runtime
selectable so that both OMAP and AMXXXX continue to use the same code.
And, in order to precisely configure/setup sched_clock for given
clocksource, decision has to be made early enough in boot sequence.
So, the solution is,
Use standard kernel parameter ("clocksource=") to override
default 32k_sync-timer, in addition to this, we also use hwmod database
lookup mechanism, through which at run-time we can identify availability
of 32k-sync timer on the device, else fall back to gptimer.
Also, moved low-level SoC specific init code to respective files,
(mach-omap1/timer32k.c and mach-omap2/timer.c)
Signed-off-by: Vaibhav Hiremath <hvaibhav@ti.com>
Signed-off-by: Felipe Balbi <balbi@ti.com>
Reviewed-by: Santosh Shilimkar <santosh.shilimkar@ti.com>
Acked-by: Kevin Hilman <khilman@ti.com>
Tested-by: Kevin Hilman <khilman@ti.com>
Cc: Benoit Cousson <b-cousson@ti.com>
Cc: Paul Walmsley <paul@pwsan.com>
Cc: Tarun Kanti DebBarma <tarun.kanti@ti.com>
Cc: Ming Lei <tom.leiming@gmail.com>
Signed-off-by: Tony Lindgren <tony@atomide.com>
Depending on the bootloader, passing command-line arguments
with spaces may have issues. Some bootloaders doesn't seem
to pass along the quotes, passing only 'gp' part of the string,
which leads to wrong override configuration.
The only affected kernel parameter configuration for OMAP family
is "clocksource=", used to override kernel clocksource.
So this patch changes "gp timer" => "gp_timer", for clockevent,
clocksource and timer irq_handler.
Signed-off-by: Vaibhav Hiremath <hvaibhav@ti.com>
Acked-by: Kevin Hilman <khilman@ti.com>
Tested-by: Kevin Hilman <khilman@ti.com>
Cc: Santosh Shilimkar <santosh.shilimkar@ti.com>
Cc: Benoit Cousson <b-cousson@ti.com>
Cc: Paul Walmsley <paul@pwsan.com>
Signed-off-by: Tony Lindgren <tony@atomide.com>
On OMAP1, omap_32k_timer_init() function always returns "true",
irrespective of whether error occurred while initializing 32k sync
counter as a kernel clocksource or not and execution will never
fallback to mpu_timer clocksource init code.
This patch adds check for return value from function
omap_init_clocksource_32k(), and fallback to omap_mpu_timer_init()
in case of failure/error from omap_init_clocksource_32k().
Signed-off-by: Vaibhav Hiremath <hvaibhav@ti.com>
Acked-by: Kevin Hilman <khilman@ti.com>
Tested-by: Kevin Hilman <khilman@ti.com>
Cc: Paul Walmsley <paul@pwsan.com>
Cc: Benoit Cousson <b-cousson@ti.com>
Signed-off-by: Tony Lindgren <tony@atomide.com>
fix for a longstanding watchdog timer integration problem.
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.12 (GNU/Linux)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=gBFx
-----END PGP SIGNATURE-----
Merge tag 'omap-devel-c-for-3.5' of git://git.kernel.org/pub/scm/linux/kernel/git/pjw/omap-pending into devel-hwmod-data
Some OMAP IP block data additions for 3.5, along with a
fix for a longstanding watchdog timer integration problem.
/ssd/git/drm-core-next/drivers/gpu/drm/radeon/radeon_fence.c: In function ‘radeon_debugfs_fence_info’:
/ssd/git/drm-core-next/drivers/gpu/drm/radeon/radeon_fence.c:606:7: warning: format ‘%lx’ expects argument of type ‘long unsigned int’, but argument 3 has type ‘long long int’ [-Wformat]
Signed-off-by: Dave Airlie <airlied@redhat.com>
For eMMC cards that has been initialized from a bootloader,
the VCC voltage supply must not be cut in an uncontrolled
manner, without first sending SLEEP or POWEROFF_NOTIFY.
The regulator_init_complete late initcall, may cut the VCC
regulator if it's reference counter is zero. To be able to
prevent the regulator from being cut, mmc_start_host, which
should execute at device init and thus before late init,
calls mmc_power_up. Then the host driver is able to increase
the reference to the regulator.
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
Reviewed-by: Mark Brown <broonie@opensource.wolfsonmicro.com>
Signed-off-by: Chris Ball <cjb@laptop.org>
No need to malloc it any more.
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
Signed-off-by: Christian König <deathsimple@vodafone.de>
Signed-off-by: Dave Airlie <airlied@redhat.com>
If we don't store local data into global variables
it isn't necessary to lock anything.
v2: rebased on new SA interface
Signed-off-by: Christian König <deathsimple@vodafone.de>
Signed-off-by: Dave Airlie <airlied@redhat.com>
It never really belonged there in the first place.
Signed-off-by: Christian König <deathsimple@vodafone.de>
Signed-off-by: Dave Airlie <airlied@redhat.com>
We can now protected the semaphore ram by a
fence, so free it immediately.
Signed-off-by: Christian König <deathsimple@vodafone.de>
Signed-off-by: Dave Airlie <airlied@redhat.com>
It isn't necessary any more and the suballocator seems to perform
even better.
Signed-off-by: Christian König <deathsimple@vodafone.de>
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
Directly use the suballocator to get small chunks of memory.
It's equally fast and doesn't crash when we encounter a GPU reset.
v2: rebased on new SA interface.
Signed-off-by: Christian König <deathsimple@vodafone.de>
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
A startover with a new idea for a multiple ring allocator.
Should perform as well as a normal ring allocator as long
as only one ring does somthing, but falls back to a more
complex algorithm if more complex things start to happen.
We store the last allocated bo in last, we always try to allocate
after the last allocated bo. Principle is that in a linear GPU ring
progression was is after last is the oldest bo we allocated and thus
the first one that should no longer be in use by the GPU.
If it's not the case we skip over the bo after last to the closest
done bo if such one exist. If none exist and we are not asked to
block we report failure to allocate.
If we are asked to block we wait on all the oldest fence of all
rings. We just wait for any of those fence to complete.
v2: We need to be able to let hole point to the list_head, otherwise
try free will never free the first allocation of the list. Also
stop calling radeon_fence_signalled more than necessary.
v3: Don't free allocations without considering them as a hole,
otherwise we might lose holes. Also return ENOMEM instead of ENOENT
when running out of fences to wait for. Limit the number of holes
we try for each ring to 3.
Signed-off-by: Christian König <deathsimple@vodafone.de>
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
Use one wait queue for all rings. When one ring progress, other
likely does to and we are not expecting to have a lot of waiter
anyway.
Also add a fence_wait_any that will wait until the first fence
in the fence array (one fence per ring) is signaled. This allow
to wait on all rings.
v2: some minor cleanups and improvements.
Signed-off-by: Christian König <deathsimple@vodafone.de>
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
