Update the _EXPORT_DEV_PM_OPS() internal macro. It was not used anywhere
outside pm.h and pm_runtime.h, so it is safe to update it.
Before, this macro would take a few parameters to be used as sleep and
runtime callbacks. This made it unsuitable to use with different
callbacks, for instance the "noirq" ones.
It is now semantically different: instead of creating a conditionally
exported dev_pm_ops structure, it only contains part of the definition.
This macro should however never be used directly (hence the trailing
underscore). Instead, the following four macros are provided:
- EXPORT_DEV_PM_OPS(name)
- EXPORT_GPL_DEV_PM_OPS(name)
- EXPORT_NS_DEV_PM_OPS(name, ns)
- EXPORT_NS_GPL_DEV_PM_OPS(name, ns)
For instance, it is now possible to conditionally export noirq
suspend/resume PM functions like this:
EXPORT_GPL_DEV_PM_OPS(foo_pm_ops) = {
NOIRQ_SYSTEM_SLEEP_PM_OPS(suspend_fn, resume_fn)
};
The existing helper macros EXPORT_*_SIMPLE_DEV_PM_OPS() and
EXPORT_*_RUNTIME_DEV_PM_OPS() have been updated to use these new macros.
Signed-off-by: Paul Cercueil <paul@crapouillou.net>
Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Pull more power management updates from Rafael Wysocki:
"These update the ARM cpufreq drivers and fix up the CPPC cpufreq
driver after recent changes, update the OPP code and PM documentation
and add power sequences support to the system reboot and power off
code.
Specifics:
- Add Tegra234 cpufreq support (Sumit Gupta)
- Clean up and enhance the Mediatek cpufreq driver (Wan Jiabing,
Rex-BC Chen, and Jia-Wei Chang)
- Fix up the CPPC cpufreq driver after recent changes (Zheng Bin,
Pierre Gondois)
- Minor update to dt-binding for Qcom's opp-v2-kryo-cpu (Yassine
Oudjana)
- Use list iterator only inside the list_for_each_entry loop
(Xiaomeng Tong, and Jakob Koschel)
- New APIs related to finding OPP based on interconnect bandwidth
(Krzysztof Kozlowski)
- Fix the missing of_node_put() in _bandwidth_supported() (Dan
Carpenter)
- Cleanups (Krzysztof Kozlowski, and Viresh Kumar)
- Add Out of Band mode description to the intel-speed-select utility
documentation (Srinivas Pandruvada)
- Add power sequences support to the system reboot and power off code
and make related platform-specific changes for multiple platforms
(Dmitry Osipenko, Geert Uytterhoeven)"
* tag 'pm-5.19-rc1-2' of git://git.kernel.org/pub/scm/linux/kernel/git/rafael/linux-pm: (60 commits)
cpufreq: CPPC: Fix unused-function warning
cpufreq: CPPC: Fix build error without CONFIG_ACPI_CPPC_CPUFREQ_FIE
Documentation: admin-guide: PM: Add Out of Band mode
kernel/reboot: Change registration order of legacy power-off handler
m68k: virt: Switch to new sys-off handler API
kernel/reboot: Add devm_register_restart_handler()
kernel/reboot: Add devm_register_power_off_handler()
soc/tegra: pmc: Use sys-off handler API to power off Nexus 7 properly
reboot: Remove pm_power_off_prepare()
regulator: pfuze100: Use devm_register_sys_off_handler()
ACPI: power: Switch to sys-off handler API
memory: emif: Use kernel_can_power_off()
mips: Use do_kernel_power_off()
ia64: Use do_kernel_power_off()
x86: Use do_kernel_power_off()
sh: Use do_kernel_power_off()
m68k: Switch to new sys-off handler API
powerpc: Use do_kernel_power_off()
xen/x86: Use do_kernel_power_off()
parisc: Use do_kernel_power_off()
...
Pull cxl updates from Dan Williams:
"Compute Express Link (CXL) updates for this cycle.
The highlight is new driver-core infrastructure and CXL subsystem
changes for allowing lockdep to validate device_lock() usage. Thanks
to PeterZ for setting me straight on the current capabilities of the
lockdep API, and Greg acked it as well.
On the CXL ACPI side this update adds support for CXL _OSC so that
platform firmware knows that it is safe to still grant Linux native
control of PCIe hotplug and error handling in the presence of CXL
devices. A circular dependency problem was discovered between suspend
and CXL memory for cases where the suspend image might be stored in
CXL memory where that image also contains the PCI register state to
restore to re-enable the device. Disable suspend for now until an
architecture is defined to clarify that conflict.
Lastly a collection of reworks, fixes, and cleanups to the CXL
subsystem where support for snooping mailbox commands and properly
handling the "mem_enable" flow are the highlights.
Summary:
- Add driver-core infrastructure for lockdep validation of
device_lock(), and fixup a deadlock report that was previously
hidden behind the 'lockdep no validate' policy.
- Add CXL _OSC support for claiming native control of CXL hotplug and
error handling.
- Disable suspend in the presence of CXL memory unless and until a
protocol is identified for restoring PCI device context from memory
hosted on CXL PCI devices.
- Add support for snooping CXL mailbox commands to protect against
inopportune changes, like set-partition with the 'immediate' flag
set.
- Rework how the driver detects legacy CXL 1.1 configurations (CXL
DVSEC / 'mem_enable') before enabling new CXL 2.0 decode
configurations (CXL HDM Capability).
- Miscellaneous cleanups and fixes from -next exposure"
* tag 'cxl-for-5.19' of git://git.kernel.org/pub/scm/linux/kernel/git/cxl/cxl: (47 commits)
cxl/port: Enable HDM Capability after validating DVSEC Ranges
cxl/port: Reuse 'struct cxl_hdm' context for hdm init
cxl/port: Move endpoint HDM Decoder Capability init to port driver
cxl/pci: Drop @info argument to cxl_hdm_decode_init()
cxl/mem: Merge cxl_dvsec_ranges() and cxl_hdm_decode_init()
cxl/mem: Skip range enumeration if mem_enable clear
cxl/mem: Consolidate CXL DVSEC Range enumeration in the core
cxl/pci: Move cxl_await_media_ready() to the core
cxl/mem: Validate port connectivity before dvsec ranges
cxl/mem: Fix cxl_mem_probe() error exit
cxl/pci: Drop wait_for_valid() from cxl_await_media_ready()
cxl/pci: Consolidate wait_for_media() and wait_for_media_ready()
cxl/mem: Drop mem_enabled check from wait_for_media()
nvdimm: Fix firmware activation deadlock scenarios
device-core: Kill the lockdep_mutex
nvdimm: Drop nd_device_lock()
ACPI: NFIT: Drop nfit_device_lock()
nvdimm: Replace lockdep_mutex with local lock classes
cxl: Drop cxl_device_lock()
cxl/acpi: Add root device lockdep validation
...
All pm_power_off_prepare() users were converted to sys-off handler API.
Remove the obsolete global callback variable.
Signed-off-by: Dmitry Osipenko <dmitry.osipenko@collabora.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
The CXL specification claims S3 support at a hardware level, but at a
system software level there are some missing pieces. Section 9.4 (CXL
2.0) rightly claims that "CXL mem adapters may need aux power to retain
memory context across S3", but there is no enumeration mechanism for the
OS to determine if a given adapter has that support. Moreover the save
state and resume image for the system may inadvertantly end up in a CXL
device that needs to be restored before the save state is recoverable.
I.e. a circular dependency that is not resolvable without a third party
save-area.
Arrange for the cxl_mem driver to fail S3 attempts. This still nominaly
allows for suspend, but requires unbinding all CXL memory devices before
the suspend to ensure the typical DRAM flow is taken. The cxl_mem unbind
flow is intended to also tear down all CXL memory regions associated
with a given cxl_memdev.
It is reasonable to assume that any device participating in a System RAM
range published in the EFI memory map is covered by aux power and
save-area outside the device itself. So this restriction can be
minimized in the future once pre-existing region enumeration support
arrives, and perhaps a spec update to clarify if the EFI memory map is
sufficent for determining the range of devices managed by
platform-firmware for S3 support.
Per Rafael, if the CXL configuration prevents suspend then it should
fail early before tasks are frozen, and mem_sleep should stop showing
'mem' as an option [1]. Effectively CXL augments the platform suspend
->valid() op since, for example, the ACPI ops are not aware of the CXL /
PCI dependencies. Given the split role of platform firmware vs OS
provisioned CXL memory it is up to the cxl_mem driver to determine if
the CXL configuration has elements that platform firmware may not be
prepared to restore.
Link: https://lore.kernel.org/r/CAJZ5v0hGVN_=3iU8OLpHY3Ak35T5+JcBM-qs8SbojKrpd0VXsA@mail.gmail.com [1]
Cc: "Rafael J. Wysocki" <rafael@kernel.org>
Cc: Pavel Machek <pavel@ucw.cz>
Cc: Len Brown <len.brown@intel.com>
Reviewed-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Link: https://lore.kernel.org/r/165066828317.3907920.5690432272182042556.stgit@dwillia2-desk3.amr.corp.intel.com
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
As more drivers start to use namespaces, we need to have varients of these
useful macros that allow the export to be in a particular namespace.
Signed-off-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Cc: Paul Cercueil <paul@crapouillou.net>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Currently, suspend_report_result() prints only function information.
If any driver uses a common PM function, nobody knows who exactly
called the failing function.
A device pinter is needed to recognize the failing device.
For example:
PM: dpm_run_callback(): pnp_bus_suspend+0x0/0x10 returns 0
PM: dpm_run_callback(): pci_pm_suspend+0x0/0x150 returns 0
become after the change:
serial 00:05: PM: dpm_run_callback(): pnp_bus_suspend+0x0/0x10 returns 0
pci 0000:00:01.3: PM: dpm_run_callback(): pci_pm_suspend+0x0/0x150 returns 0
Signed-off-by: Youngjin Jang <yj84.jang@samsung.com>
[ rjw: Changelog edits ]
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
A lot of drivers create a dev_pm_ops struct with the system sleep
suspend/resume callbacks set to pm_runtime_force_suspend() and
pm_runtime_force_resume().
These drivers can now use the DEFINE_RUNTIME_DEV_PM_OPS() macro, which
will use pm_runtime_force_{suspend,resume}() as the system sleep
callbacks, while having the same dead code removal characteristic that
is already provided by DEFINE_SIMPLE_DEV_PM_OPS().
Signed-off-by: Paul Cercueil <paul@crapouillou.net>
Acked-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Reviewed-by: Ulf Hansson <ulf.hansson@linaro.org>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
These macros are defined conditionally, according to CONFIG_PM:
- if CONFIG_PM is enabled, these macros resolve to
DEFINE_SIMPLE_DEV_PM_OPS(), and the dev_pm_ops symbol will be
exported.
- if CONFIG_PM is disabled, these macros will result in a dummy static
dev_pm_ops to be created with the __maybe_unused flag. The dev_pm_ops
will then be discarded by the compiler, along with the provided
callback functions if they are not used anywhere else.
In the second case, the symbol is not exported, which should be
perfectly fine - users of the symbol should all use the pm_ptr() or
pm_sleep_ptr() macro, so the dev_pm_ops marked as "extern" in the
client's code will never be accessed.
Signed-off-by: Paul Cercueil <paul@crapouillou.net>
Acked-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Reviewed-by: Ulf Hansson <ulf.hansson@linaro.org>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Keep this macro in line with the other ones. This makes it possible to
use them in the cases where the underlying dev_pm_ops structure is
exported.
Restore the "static" qualifier in the two drivers where the
DEFINE_SIMPLE_DEV_PM_OPS macro was used.
Signed-off-by: Paul Cercueil <paul@crapouillou.net>
Acked-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Reviewed-by: Ulf Hansson <ulf.hansson@linaro.org>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
The deprecated UNIVERSAL_DEV_PM_OPS() macro uses the provided callbacks
for both runtime PM and system sleep, which is very likely to be a
mistake, as a system sleep can be triggered while a given device is
already PM-suspended, which would cause the suspend callback to be
called twice.
The amount of users of UNIVERSAL_DEV_PM_OPS() is also tiny (16
occurences) compared to the number of places where
SET_SYSTEM_SLEEP_PM_OPS() is used with pm_runtime_force_suspend() and
pm_runtime_force_resume(), which makes me think that none of these cases
are actually valid.
As the new macro DEFINE_UNIVERSAL_DEV_PM_OPS() which was introduced to
replace UNIVERSAL_DEV_PM_OPS() is currently unused, remove it before
someone starts to use it in yet another invalid case.
Signed-off-by: Paul Cercueil <paul@crapouillou.net>
Acked-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Reviewed-by: Ulf Hansson <ulf.hansson@linaro.org>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
In some cases (for example, during system-wide suspend and resume of
devices) it is useful to know whether or not runtime PM has ever been
enabled for a given device and, if so, what the runtime PM status of
it had been right before runtime PM was disabled for it last time.
For this reason, introduce a new struct dev_pm_info field called
last_status that will be used for capturing the runtime PM status of
the device when its power.disable_depth counter changes from 0 to 1.
The new field will be set to RPM_INVALID to start with and whenever
power.disable_depth changes from 1 to 0, so it will be valid only
when runtime PM of the device is currently disabled, but it has been
enabled at least once.
Immediately use power.last_status in rpm_resume() to make it handle
the case when PM runtime is disabled for the device, but its runtime
PM status is RPM_ACTIVE more consistently. Namely, make it return 1
if power.last_status is also equal to RPM_ACTIVE in that case (the
idea being that if the status was RPM_ACTIVE last time when
power.disable_depth was changing from 0 to 1 and it is still
RPM_ACTIVE, it can be assumed to reflect what happened to the device
last time when it was using runtime PM) and -EACCES otherwise.
Update the documentation to provide a description of last_status and
change the description of pm_runtime_resume() in it to reflect the
new behavior of rpm_active().
While at it, rearrange the code in pm_runtime_enable() to be more
straightforward and replace the WARN() macro in it with a pr_warn()
invocation which is less disruptive.
Link: https://lore.kernel.org/linux-pm/20211026222626.39222-1-ulf.hansson@linaro.org/t/#u
Reviewed-by: Ulf Hansson <ulf.hansson@linaro.org>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
This commit introduces the following macros:
SYSTEM_SLEEP_PM_OPS()
LATE_SYSTEM_SLEEP_PM_OPS()
NOIRQ_SYSTEM_SLEEP_PM_OPS()
RUNTIME_PM_OPS()
These new macros are very similar to their SET_*_PM_OPS() equivalent.
They however differ in the fact that the callbacks they set will always
be seen as referenced by the compiler. This means that the callback
functions don't need to be wrapped with a #ifdef CONFIG_PM guard, or
tagged with __maybe_unused, to prevent the compiler from complaining
about unused static symbols. The compiler will then simply evaluate at
compile time whether or not these symbols are dead code.
The callbacks that are only useful with CONFIG_PM_SLEEP is enabled, are
now also wrapped with a new pm_sleep_ptr() macro, which is inspired from
pm_ptr(). This is needed for drivers that use different callbacks for
sleep and runtime PM, to handle the case where CONFIG_PM is set and
CONFIG_PM_SLEEP is not.
This commit also deprecates the following macros:
SIMPLE_DEV_PM_OPS()
UNIVERSAL_DEV_PM_OPS()
And introduces the following macros:
DEFINE_SIMPLE_DEV_PM_OPS()
DEFINE_UNIVERSAL_DEV_PM_OPS()
These macros are similar to the functions they were created to replace,
with the following differences:
- They use the new macros introduced above, and as such always
reference the provided callback functions.
- They are not tagged with __maybe_unused. They are meant to be used
with pm_ptr() or pm_sleep_ptr() for DEFINE_UNIVERSAL_DEV_PM_OPS()
and DEFINE_SIMPLE_DEV_PM_OPS() respectively.
- They declare the symbol static, since every driver seems to do that
anyway; and if a non-static use-case is needed an indirection pointer
could be used.
The point of this change, is to progressively switch from a code model
where PM callbacks are all protected behind CONFIG_PM guards, to a code
model where the PM callbacks are always seen by the compiler, but
discarded if not used.
Signed-off-by: Paul Cercueil <paul@crapouillou.net>
Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
The pm_ptr() macro was previously conditionally defined, according to
the value of the CONFIG_PM option. This meant that the pointed structure
was either referenced (if CONFIG_PM was set), or never referenced (if
CONFIG_PM was not set), causing it to be detected as unused by the
compiler.
This worked fine, but required the __maybe_unused compiler attribute to
be used to every symbol pointed to by a pointer wrapped with pm_ptr().
We can do better. With this change, the pm_ptr() is now defined the
same, independently of the value of CONFIG_PM. It now uses the (?:)
ternary operator to conditionally resolve to its argument. Since the
condition is known at compile time, the compiler will then choose to
discard the unused symbols, which won't need to be tagged with
__maybe_unused anymore.
This pm_ptr() macro is usually used with pointers to dev_pm_ops
structures created with SIMPLE_DEV_PM_OPS() or similar macros. These do
use a __maybe_unused flag, which is now useless with this change, so it
later can be removed. However in the meantime it causes no harm, and all
the drivers still compile fine with the new pm_ptr() macro.
Signed-off-by: Paul Cercueil <paul@crapouillou.net>
Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Reviewed-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
As pm_runtime_need_not_resume() relies also on usage_count, it can return
a different value in pm_runtime_force_suspend() compared to when called in
pm_runtime_force_resume(). Different return values can happen if anything
calls PM runtime functions in between, and causes the parent child_count
to increase on every resume.
So far I've seen the issue only for omapdrm that does complicated things
with PM runtime calls during system suspend for legacy reasons:
omap_atomic_commit_tail() for omapdrm.0
dispc_runtime_get()
wakes up 58000000.dss as it's the dispc parent
dispc_runtime_resume()
rpm_resume() increases parent child_count
dispc_runtime_put() won't idle, PM runtime suspend blocked
pm_runtime_force_suspend() for 58000000.dss, !pm_runtime_need_not_resume()
__update_runtime_status()
system suspended
pm_runtime_force_resume() for 58000000.dss, pm_runtime_need_not_resume()
pm_runtime_enable() only called because of pm_runtime_need_not_resume()
omap_atomic_commit_tail() for omapdrm.0
dispc_runtime_get()
wakes up 58000000.dss as it's the dispc parent
dispc_runtime_resume()
rpm_resume() increases parent child_count
dispc_runtime_put() won't idle, PM runtime suspend blocked
...
rpm_suspend for 58000000.dss but parent child_count is now unbalanced
Let's fix the issue by adding a flag for needs_force_resume and use it in
pm_runtime_force_resume() instead of pm_runtime_need_not_resume().
Additionally omapdrm system suspend could be simplified later on to avoid
lots of unnecessary PM runtime calls and the complexity it adds. The
driver can just use internal functions that are shared between the PM
runtime and system suspend related functions.
Fixes: 4918e1f87c ("PM / runtime: Rework pm_runtime_force_suspend/resume()")
Signed-off-by: Tony Lindgren <tony@atomide.com>
Reviewed-by: Ulf Hansson <ulf.hansson@linaro.org>
Tested-by: Tomi Valkeinen <tomi.valkeinen@ideasonboard.com>
Cc: 4.16+ <stable@vger.kernel.org> # 4.16+
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
The clock API splits its interface into sleepable ant atomic contexts:
- clk_prepare/clk_unprepare for stuff that might sleep
- clk_enable_clk_disable for anything that may be done in atomic context
The code handling runtime PM for clocks only calls clk_disable() on
suspend requests, and clk_enable on resume requests. This means that
runtime PM with clock providers that only have the prepare/unprepare
methods implemented is basically useless.
Many clock implementations can't accommodate atomic contexts. This is
often the case when communication with the clock happens through another
subsystem like I2C or SCMI.
Let's make the clock PM code useful with such clocks by safely invoking
clk_prepare/clk_unprepare upon resume/suspend requests. Of course, when
such clocks are registered with the PM layer then pm_runtime_irq_safe()
can't be used, and neither pm_runtime_suspend() nor pm_runtime_resume()
may be invoked in atomic context.
For clocks that do implement the enable and disable methods then
everything just works as before.
A note on sparse:
According to https://lwn.net/Articles/109066/ there are things
that sparse can't cope with. In particular, pm_clk_op_lock() and
pm_clk_op_unlock() may or may not lock/unlock psd->lock depending on
some runtime condition. To work around that we tell it the lock is
always untaken for the purpose of static analisys.
Thanks to Naresh Kamboju for reporting issues with the initial patch.
Signed-off-by: Nicolas Pitre <npitre@baylibre.com>
Tested-by: Naresh Kamboju <naresh.kamboju@linaro.org>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Commit 8234f6734c ("PM-runtime: Switch autosuspend over to using
hrtimers") switched PM runtime autosuspend to use hrtimers and all
related time accounting in ns, but missed to update the timer_expires
data type in struct dev_pm_info to u64.
This causes the timer_expires value to be truncated on 32-bit
architectures when assignment is done from u64 values:
rpm_suspend()
|- dev->power.timer_expires = expires;
Fix it by changing the timer_expires type to u64.
Fixes: 8234f6734c ("PM-runtime: Switch autosuspend over to using hrtimers")
Signed-off-by: Grygorii Strashko <grygorii.strashko@ti.com>
Acked-by: Pavel Machek <pavel@ucw.cz>
Acked-by: Vincent Guittot <vincent.guittot@linaro.org>
Cc: 5.0+ <stable@vger.kernel.org> # 5.0+
[ rjw: Subject and changelog edits ]
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
This way, when the dev_pm_ops instance is not referenced anywhere, it
will simply be dropped by the compiler without a warning.
Signed-off-by: Paul Cercueil <paul@crapouillou.net>
Reviewed-by: Ulf Hansson <ulf.hansson@linaro.org>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
This macro is analogous to the infamous of_match_ptr(). If CONFIG_PM
is enabled, this macro will resolve to its argument, otherwise to NULL.
Signed-off-by: Paul Cercueil <paul@crapouillou.net>
Reviewed-by: Ulf Hansson <ulf.hansson@linaro.org>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Update the documentation of the driver flags for system-wide power
management to reflect the current code flows and be more consistent.
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Because all callers of dev_pm_smart_suspend_and_suspended use it only
for checking whether or not to skip driver suspend callbacks for a
device, rename it to dev_pm_skip_suspend() in analogy with
dev_pm_skip_resume().
No functional impact.
Suggested-by: Alan Stern <stern@rowland.harvard.edu>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Acked-by: Alan Stern <stern@rowland.harvard.edu>
Acked-by: Bjorn Helgaas <bhelgaas@google.com>
