IF YOU WOULD LIKE TO GET AN ACCOUNT, please write an
email to Administrator. User accounts are meant only to access repo
and report issues and/or generate pull requests.
This is a purpose-specific Git hosting for
BaseALT
projects. Thank you for your understanding!
Только зарегистрированные пользователи имеют доступ к сервису!
Для получения аккаунта, обратитесь к администратору.
Commit 318a19718261 (device property: refactor built-in properties
support) went way too far and brought a union aliasing. Partially
revert it here to get rid of union aliasing.
Note, all Apple properties are considered as u8 arrays. To get a value
of any of them the caller must use device_property_read_u8_array().
What's union aliasing?
~~~~~~~~~~~~~~~~~~~~~~
The C99 standard in section 6.2.5 paragraph 20 defines union type as
"an overlapping nonempty set of member objects". It also states in
section 6.7.2.1 paragraph 14 that "the value of at most one of the
members can be stored in a union object at any time'.
Union aliasing is a type punning mechanism using union members to store
as one type and read back as another.
Why it's not good?
~~~~~~~~~~~~~~~~~~
Section 6.2.6.1 paragraph 6 says that a union object may not be a trap
representation, although its member objects may be.
Meanwhile annex J.1 says that "the value of a union member other than
the last one stored into" is unspecified [removed in C11].
In TC3, a footnote is added which specifies that accessing a member of a
union other than the last one stored causes "the object representation"
to be re-interpreted in the new type and specifically refers to this as
"type punning". This conflicts to some degree with Annex J.1.
While it's working in Linux with GCC, the use of union members to do
type punning is not clear area in the C standard and might lead to
unspecified behaviour.
More information is available in this [1] blog post.
[1]: https://davmac.wordpress.com/2010/02/26/c99-revisited/
Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Acked-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Reviewed-by: Mika Westerberg <mika.westerberg@linux.intel.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Capability to attach an existing clk to a MMIO regmap was
introduced in 4.17rc1.
However, when using attached clk, regmap does not do the clk_get.
Therefore it should not do the clk_put when freeing the MMIO
regmap context.
There does not appear to be any users of attached clocks yet
so this would be a good time to make this change before anything
depends on the existing behaviour.
Signed-off-by: James Kelly <jamespeterkelly@gmail.com>
Acked-by: Maxime Ripard <maxime.ripard@bootlin.com>
Signed-off-by: Mark Brown <broonie@kernel.org>
As dev_pm_domain_attach() isn't the only way to assign PM domain pointers
to devices, clearly we must allow a device to have the pointer already
being assigned. For this reason, return 0 instead of -EEXIST.
Reported-by: Krzysztof Kozlowski <krzk@kernel.org>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
Tested-by: Tested-by: Krzysztof Kozlowski <krzk@kernel.org>
Tested-by: Tony Lindgren <tony@atomide.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
The limitation of being able to check only for -EPROBE_DEFER from
dev_pm_domain_attach() has been removed. Hence let's respect all error
codes and bail out accordingly.
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
Acked-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
The callers of dev_pm_domain_attach() currently checks the returned error
code for -EPROBE_DEFER and needs to ignore other error codes. This is an
unnecessary limitation, which also leads to a rather strange behaviour in
the error path.
Address this limitation, by changing the return codes from
acpi_dev_pm_attach() and genpd_dev_pm_attach(). More precisely, let them
return 0, when no PM domain is needed for the device and then return 1, in
case the device was successfully attached to its PM domain. In this way,
dev_pm_domain_attach(), gets a better understanding of what happens in the
attach attempts and also allowing its caller to better act on real errors
codes.
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Instead of checking if an existing PM domain pointer has been assigned in
genpd_dev_pm_attach() and acpi_dev_pm_attach(), move the check to the
common path in dev_pm_domain_attach(), thus potentially avoid one
unnecessary check.
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
The driver core together with the PM core, nowadays deals with deferring
all probes during the device system sleep phases. Therefore genpd no longer
need to care about this situation, so let's drop the corresponding code.
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
The parsing of the Samsung specific DT binding is gone, but the comment in
the function header remained. Let's drop the comment to avoid confusions.
Fixes: 001d50c9a14f (PM / Domains: Remove obsolete "samsung,power-domain" check)
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
In case the PM domain fails to be powered on in genpd_dev_pm_attach(), it
returns -EPROBE_DEFER, but keeping the device attached to its PM domain.
This leads to problems when the next attempt to attach is re-tried. More
precisely, in that situation an -EEXIST error code is returned, because the
device already has its PM domain pointer assigned, from the first attempt.
Now, because of the sloppy error handling by the existing callers of
dev_pm_domain_attach(), probing is allowed to continue when -EEXIST is
returned. However, in such case there are no guarantees that the PM domain
is powered on by genpd, which may lead to hangs when buses/drivers tried to
access their devices.
Let's fix this behaviour, simply by detaching the device when powering on
fails in genpd_dev_pm_attach().
Cc: v4.11+ <stable@vger.kernel.org> # v4.11+
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
__printf is useful to verify format and arguments. Remove the following
warning (with W=1):
drivers/base/core.c:2435:2: warning: function might be possible candidate for ‘gnu_printf’ format attribute [-Wsuggest-attribute=format]
Signed-off-by: Mathieu Malaterre <malat@debian.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
if device_register() returned an error. Always use put_device()
to give up the initialized reference and release allocated memory.
Signed-off-by: Arvind Yadav <arvind.yadav.cs@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Currently the firmware loader only exposes one silent path for querying
optional firmware, and that is firmware_request_direct(). This function
also disables the sysfs fallback mechanism, which might not always be the
desired behaviour [0].
This patch introduces a variations of request_firmware() that enable the
caller to disable the undesired warning messages but enables the sysfs
fallback mechanism. This is equivalent to adding FW_OPT_NO_WARN to the
old behaviour.
[0]: https://git.kernel.org/linus/c0cc00f250e1
Signed-off-by: Andres Rodriguez <andresx7@gmail.com>
Reviewed-by: Kees Cook <keescook@chromium.org>
Acked-by: Luis R. Rodriguez <mcgrof@kernel.org>
[mcgrof: used the old API calls as the full rename is not done yet, and
add the caller for when FW_LOADER is disabled, enhance documentation ]
Signed-off-by: Luis R. Rodriguez <mcgrof@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
If we resort to using the sysfs fallback mechanism we don't print
the filename. This can be deceiving given we could have a series of
callers intertwined and it'd be unclear exactly for what firmware
this was meant for.
Additionally, although we don't currently use FW_OPT_NO_WARN when
dealing with the fallback mechanism, we will soon, so just respect
its use consistently.
And even if you *don't* want to print always on failure, you may
want to print when debugging so enable dynamic debug print when
FW_OPT_NO_WARN is used.
Reviewed-by: Kees Cook <keescook@chromium.org>
Signed-off-by: Luis R. Rodriguez <mcgrof@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
This will make it easier to track and easier to understand
what components and features are part of the FW_LOADER. There
are some components related to firmware which have *nothing* to
do with the FW_LOADER, souch as PREVENT_FIRMWARE_BUILD.
Reviewed-by: Kees Cook <keescook@chromium.org>
Signed-off-by: Luis R. Rodriguez <mcgrof@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
As per checkpatch using help is preferred over ---help---.
Signed-off-by: Luis R. Rodriguez <mcgrof@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
If you try to read FW_LOADER today it speaks of old riddles and
unless you have been following development closely you will lose
track of what is what. Even the documentation for PREVENT_FIRMWARE_BUILD
is a bit fuzzy and how it fits into this big picture.
Give the FW_LOADER kconfig documentation some love with more up to
date developments and recommendations. While at it, wrap the FW_LOADER
code into its own menu to compartmentalize and make it clearer which
components really are part of the FW_LOADER. This should also make
it easier to later move these kconfig entries into the firmware_loader/
directory later.
This also now recommends using firmwared [0] for folks left needing a
uevent handler in userspace for the sysfs firmware fallback mechanis
given udev's uevent firmware mechanism was ripped out a while ago.
[0] https://github.com/teg/firmwared
Reviewed-by: Kees Cook <keescook@chromium.org>
Signed-off-by: Luis R. Rodriguez <mcgrof@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
This also sets the expecations for future fallback interfaces, even
if they are not exported.
Reviewed-by: Kees Cook <keescook@chromium.org>
Signed-off-by: Luis R. Rodriguez <mcgrof@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
This is done since this call is now exposed through kernel-doc,
and since this also paves the way for different future types of
fallback mechanims.
Signed-off-by: Andres Rodriguez <andresx7@gmail.com>
Reviewed-by: Kees Cook <keescook@chromium.org>
Acked-by: Luis R. Rodriguez <mcgrof@kernel.org>
[mcgrof: small coding style changes]
Signed-off-by: Luis R. Rodriguez <mcgrof@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The kernel-doc spec dictates a function name ends in ().
Signed-off-by: Andres Rodriguez <andresx7@gmail.com>
Reviewed-by: Kees Cook <keescook@chromium.org>
Acked-by: Randy Dunlap <rdunlap@infradead.org>
Acked-by: Luis R. Rodriguez <mcgrof@kernel.org>
[mcgrof: adjust since the wide API rename is not yet merged]
Signed-off-by: Luis R. Rodriguez <mcgrof@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
This should let us associate enum kdoc to these values.
While at it, kdocify the fw_opt.
Signed-off-by: Andres Rodriguez <andresx7@gmail.com>
Reviewed-by: Kees Cook <keescook@chromium.org>
Acked-by: Luis R. Rodriguez <mcgrof@kernel.org>
[mcgrof: coding style fixes, merge kdoc with enum move]
Signed-off-by: Luis R. Rodriguez <mcgrof@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
syzbot is hitting WARN() at kernfs_add_one() [1].
This is because kernfs_create_link() is confused by previous device_add()
call which continued without setting dev->kobj.parent field when
get_device_parent() failed by memory allocation fault injection.
Fix this by propagating the error from class_dir_create_and_add() to
the calllers of get_device_parent().
[1] https://syzkaller.appspot.com/bug?id=fae0fb607989ea744526d1c082a5b8de6529116f
Signed-off-by: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp>
Reported-by: syzbot <syzbot+df47f81c226b31d89fb1@syzkaller.appspotmail.com>
Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Cc: stable <stable@vger.kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Nobody would be insane enough to try and use level triggered
MSIs on PCI, but let's make sure it doesn't happen. Also,
let's mandate that the irqchip backing the platform MSI domain
is providing the IRQCHIP_SUPPORTS_LEVEL_MSI flag.
Signed-off-by: Marc Zyngier <marc.zyngier@arm.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Cc: Rob Herring <robh@kernel.org>
Cc: Jason Cooper <jason@lakedaemon.net>
Cc: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Cc: Srinivas Kandagatla <srinivas.kandagatla@linaro.org>
Cc: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
Cc: Miquel Raynal <miquel.raynal@bootlin.com>
Link: https://lkml.kernel.org/r/20180508121438.11301-3-marc.zyngier@arm.com
The inline versions of rpm_sysfs_remove() and wakeup_sysfs_add|remove(),
are not being used while CONFIG_PM is unset, hence let's drop them.
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
The functions pm_qos_sysfs_add|remove() are available as inline functions
only while CONFIG_PM is unset, but are not being used. Likely they are a
leftover from an earlier cleanup in the past, anyway let's drop them.
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
The inline versions of dev_pm_arm|disarm_wake_irq() and
dev_pm_enable|disable_wake_irq_check() are not being used while CONFIG_PM
is unset, hence let's drop them.
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
Acked-by: Tony Lindgren <tony@atomide.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
The inline versions of device_wakeup_arm|disarm_wake_irqs(), which are
available while when CONFIG_PM is set and CONFIG_PM_SLEEP unset, are not
being used, hence let's drop them.
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
Acked-by: Tony Lindgren <tony@atomide.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
When wakelock support was added, the wakeup_source_add() function
was updated to set the last_time value of the wakeup source. This
has the unintended side effect of producing confusing output from
pm_print_active_wakeup_sources() when a wakeup source is added
prior to a sleep that is blocked by a different wakeup source.
The function pm_print_active_wakeup_sources() will search for the
most recently active wakeup source when no active source is found.
If a wakeup source is added after a different wakeup source blocks
the system from going to sleep it may have a later last_time value
than the blocking source and be output as the last active wakeup
source even if it has never actually been active.
It looks to me like the change to wakeup_source_add() was made to
prevent the wakelock garbage collection from accidentally dropping
a wakelock during the narrow window between adding the wakelock to
the wakelock list in wakelock_lookup_add() and the activation of
the wakeup source in pm_wake_lock().
This commit changes the behavior so that only the last_time of the
wakeup source used by a wakelock is initialized prior to adding it
to the wakeup source list. This preserves the meaning of the
last_time value as the last time the wakeup source was active and
allows a wakeup source that has never been active to have a
last_time value of 0.
Fixes: b86ff9820fd5 (PM / Sleep: Add user space interface for manipulating wakeup sources, v3)
Signed-off-by: Doug Berger <opendmb@gmail.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
single_open() interface requires that the whole output must
fit into a single buffer. This will lead to timeout when
system memory is not in a good situation.
This patch use seq_open() to show wakeup stats. This method
need only one page, so timeout will not be observed.
Signed-off-by: Ganesh Mahendran <opensource.ganesh@gmail.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
When we print diagnostic messages about suspend/resume, we have a device
pointer, so use dev_printk() to match other device-related things. Add the
function name, similar to initcall_debug output. E.g.,
- calling 0000:01:00.0+ @ 998, parent: 0000:00:1c.0
+ pci 0000:01:00.0: calling <something> @ 998, parent: 0000:00:1c.0
I wondered if this would break scripts/bootgraph.pl, but I don't think it
will because bootgraph.pl doesn't add any timing information to $start{}
after it sees "Write protecting the" or "Freeing unused kernel memory".
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
initcall_debug_report() always called ktime_get(), even if we didn't
need the result.
Change it so we only call it when we're going to use the result, and
change initcall_debug_start() to follow the same style.
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Commit e8bca479c3f2 (PM / sleep: trace events for device PM callbacks)
removed the only uses of "state" and "info" from initcall_debug_report().
Remove the now-unused arguments completely.
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
This implements of_genpd_opp_to_performance_state() which can be used
from the device drivers or the OPP core to find the performance state
encoded in the "required-opps" property of a node. Normally this would
be called only once for each OPP of the device for which the OPP table
of the device is getting generated.
Different platforms may encode the performance state differently using
the OPP table (they may simply return value of opp-hz or opp-microvolt,
or apply some algorithm on top of those values) and so a new callback
->opp_to_performance_state() is implemented to allow platform specific
drivers to convert the power domain OPP to a performance state value.
Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org>
Acked-by: Ulf Hansson <ulf.hansson@linaro.org>
The generic power domains can have an OPP table for themselves now, and
phandle of their OPP nodes can be used by the devices powered by the
domain. In order for the OPP core to translate requirements between the
devices and their power domains, both need to have an OPP table in
kernel.
Parse the OPP table for power domains
if they have their
set_performance_state() callback set.
With this patch, an OPP table would be created for the genpd in kernel
based on the OPP table present in DT, if the genpd have its
set_performance_state() callback set.
Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org>
Acked-by: Ulf Hansson <ulf.hansson@linaro.org>
The power-domain core would be using the OPP core going forward and the
OPP core has the basic requirement of a device structure for its working.
Add a struct device to the genpd structure. This doesn't register the
device with device core as the "dev" pointer is mostly used by the OPP
core as a cookie for now and registering the device is not mandatory.
Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org>
Acked-by: Ulf Hansson <ulf.hansson@linaro.org>
With each bus implementing its own DMA configuration callback, there is no
need for bus to explicitly set the force_dma flag. Modify the
of_dma_configure function to accept an input parameter which specifies if
implicit DMA configuration is required when it is not described by the
firmware.
Signed-off-by: Nipun Gupta <nipun.gupta@nxp.com>
Acked-by: Bjorn Helgaas <bhelgaas@google.com> # PCI parts
Reviewed-by: Rob Herring <robh@kernel.org>
[hch: tweaked the changelog a bit]
Signed-off-by: Christoph Hellwig <hch@lst.de>
ACPI/OF support for configuration of DMA is a bus specific aspect, and
thus should be configured by the bus. Introduces a 'dma_configure' bus
method so that busses can control their DMA capabilities.
Also update the PCI, Platform, ACPI and host1x buses to use the new
method.
Suggested-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Nipun Gupta <nipun.gupta@nxp.com>
Acked-by: Bjorn Helgaas <bhelgaas@google.com> # PCI parts
Acked-by: Thierry Reding <treding@nvidia.com>
Reviewed-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
[hch: simplified host1x_dma_configure based on a comment from Thierry,
rewrote changelog]
Signed-off-by: Christoph Hellwig <hch@lst.de>
Add the sysfs file for the new vulerability. It does not do much except
show the words 'Vulnerable' for recent x86 cores.
Intel cores prior to family 6 are known not to be vulnerable, and so are
some Atoms and some Xeon Phi.
It assumes that older Cyrix, Centaur, etc. cores are immune.
Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Reviewed-by: Borislav Petkov <bp@suse.de>
Reviewed-by: Ingo Molnar <mingo@kernel.org>
Here are some small driver core and firmware fixes for 4.17-rc3
There's a kobject WARN() removal to make syzkaller a lot happier about
some "normal" error paths that it keeps hitting, which should reduce the
number of false-positives we have been getting recently.
There's also some fimware test and documentation fixes, and the
coredump() function signature change that needed to happen after -rc1
before drivers started to take advantage of it.
All of these have been in linux-next with no reported issues.
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
-----BEGIN PGP SIGNATURE-----
iG0EABECAC0WIQT0tgzFv3jCIUoxPcsxR9QN2y37KQUCWuMxrw8cZ3JlZ0Brcm9h
aC5jb20ACgkQMUfUDdst+ymFpQCg1JM62/W8e6mQ4vdZNQmAzgMKMEMAniOMcVRX
/oDWXp64mYwJu+GTxnIJ
=+9Gk
-----END PGP SIGNATURE-----
Merge tag 'driver-core-4.17-rc3' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/driver-core
Pull driver core fixes from Greg Kroah-Hartman:
"Here are some small driver core and firmware fixes for 4.17-rc3
There's a kobject WARN() removal to make syzkaller a lot happier about
some "normal" error paths that it keeps hitting, which should reduce
the number of false-positives we have been getting recently.
There's also some fimware test and documentation fixes, and the
coredump() function signature change that needed to happen after -rc1
before drivers started to take advantage of it.
All of these have been in linux-next with no reported issues"
* tag 'driver-core-4.17-rc3' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/driver-core:
firmware: some documentation fixes
selftests:firmware: fixes a call to a wrong function name
kobject: don't use WARN for registration failures
firmware: Fix firmware documentation for recent file renames
test_firmware: fix setting old custom fw path back on exit, second try
test_firmware: Install all scripts
drivers: change struct device_driver::coredump() return type to void
Including:
- Fixup outdated kernel-doc paths
- Slightly too short title underline
- Some typos
Signed-off-by: Andres Rodriguez <andresx7@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
When bridge and its endpoint is enumerated the devices are added to the
dpm list. Afterward, the bridge defers probe when IOMMU is not ready.
This causes the bridge to be moved to the end of the dpm list when
deferred probe kicks in. The order of the dpm list for bridge and
endpoint is reversed.
Add reordering code to move the bridge and its children and consumers to
the end of the pm list so the order for suspend and resume is not altered.
The code also move device and its children and consumers to the tail of
device_kset list if it is registered.
Signed-off-by: Toan Le <toanle@apm.com>
Signed-off-by: Feng Kan <fkan@apm.com>
Reviewed-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Postpone calling virt_to_page() translation on memory locations not
guaranteed to be backed by a struct page. Try first to map memory from
the device coherent memory pool, then perform translation if that fails.
On some architectures, specifically SH when configured with the SPARSEMEM
memory model, assuming a struct page is always assigned to a memory
address lead to unexpected hangs during the virtual to page address
translation. This patch fixes that specific issue but applies in the
general case too.
Suggested-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Jacopo Mondi <jacopo+renesas@jmondi.org>
Reviewed-by: Robin Murphy <robin.murphy@arm.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
The use of "correctly mapped" here is misleading, since it can give the
wrong expectation in the case that the memory *should* have been mapped
from the per-device pool, but doing so failed for other reasons.
Signed-off-by: Robin Murphy <robin.murphy@arm.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
__highest_present_section_nr is a more strict boundary than
NR_MEM_SECTIONS. So checking __highest_present_section_nr directly is
enough.
Link: http://lkml.kernel.org/r/20180330032044.21647-1-richard.weiyang@gmail.com
Signed-off-by: Wei Yang <richard.weiyang@gmail.com>
Reviewed-by: Andrew Morton <akpm@linux-foundation.org>
Cc: Dave Hansen <dave.hansen@linux.intel.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: David Rientjes <rientjes@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
During memory hotplugging we traverse struct pages three times:
1. memset(0) in sparse_add_one_section()
2. loop in __add_section() to set do: set_page_node(page, nid); and
SetPageReserved(page);
3. loop in memmap_init_zone() to call __init_single_pfn()
This patch removes the first two loops, and leaves only loop 3. All
struct pages are initialized in one place, the same as it is done during
boot.
The benefits:
- We improve memory hotplug performance because we are not evicting the
cache several times and also reduce loop branching overhead.
- Remove condition from hotpath in __init_single_pfn(), that was added
in order to fix the problem that was reported by Bharata in the above
email thread, thus also improve performance during normal boot.
- Make memory hotplug more similar to the boot memory initialization
path because we zero and initialize struct pages only in one
function.
- Simplifies memory hotplug struct page initialization code, and thus
enables future improvements, such as multi-threading the
initialization of struct pages in order to improve hotplug
performance even further on larger machines.
[pasha.tatashin@oracle.com: v5]
Link: http://lkml.kernel.org/r/20180228030308.1116-7-pasha.tatashin@oracle.com
Link: http://lkml.kernel.org/r/20180215165920.8570-7-pasha.tatashin@oracle.com
Signed-off-by: Pavel Tatashin <pasha.tatashin@oracle.com>
Reviewed-by: Ingo Molnar <mingo@kernel.org>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Baoquan He <bhe@redhat.com>
Cc: Bharata B Rao <bharata@linux.vnet.ibm.com>
Cc: Daniel Jordan <daniel.m.jordan@oracle.com>
Cc: Dan Williams <dan.j.williams@intel.com>
Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Cc: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Cc: Mel Gorman <mgorman@techsingularity.net>
Cc: Steven Sistare <steven.sistare@oracle.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Vlastimil Babka <vbabka@suse.cz>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
During memory hotplugging the probe routine will leave struct pages
uninitialized, the same as it is currently done during boot. Therefore,
we do not want to access the inside of struct pages before
__init_single_page() is called during onlining.
Because during hotplug we know that pages in one memory block belong to
the same numa node, we can skip the checking. We should keep checking
for the boot case.
[pasha.tatashin@oracle.com: s/register_new_memory()/hotplug_memory_register()]
Link: http://lkml.kernel.org/r/20180228030308.1116-6-pasha.tatashin@oracle.com
Link: http://lkml.kernel.org/r/20180215165920.8570-6-pasha.tatashin@oracle.com
Signed-off-by: Pavel Tatashin <pasha.tatashin@oracle.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Reviewed-by: Ingo Molnar <mingo@kernel.org>
Cc: Baoquan He <bhe@redhat.com>
Cc: Bharata B Rao <bharata@linux.vnet.ibm.com>
Cc: Daniel Jordan <daniel.m.jordan@oracle.com>
Cc: Dan Williams <dan.j.williams@intel.com>
Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Cc: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Cc: Mel Gorman <mgorman@techsingularity.net>
Cc: Steven Sistare <steven.sistare@oracle.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Vlastimil Babka <vbabka@suse.cz>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
When memory is hotplugged pages_correctly_reserved() is called to verify
that the added memory is present, this routine traverses through every
struct page and verifies that PageReserved() is set. This is a slow
operation especially if a large amount of memory is added.
Instead of checking every page, it is enough to simply check that the
section is present, has mapping (struct page array is allocated), and
the mapping is online.
In addition, we should not excpect that probe routine sets flags in
struct page, as the struct pages have not yet been initialized. The
initialization should be done in __init_single_page(), the same as
during boot.
Link: http://lkml.kernel.org/r/20180215165920.8570-5-pasha.tatashin@oracle.com
Signed-off-by: Pavel Tatashin <pasha.tatashin@oracle.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Baoquan He <bhe@redhat.com>
Cc: Bharata B Rao <bharata@linux.vnet.ibm.com>
Cc: Daniel Jordan <daniel.m.jordan@oracle.com>
Cc: Dan Williams <dan.j.williams@intel.com>
Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Cc: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Cc: Mel Gorman <mgorman@techsingularity.net>
Cc: Steven Sistare <steven.sistare@oracle.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Vlastimil Babka <vbabka@suse.cz>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
