Re: [RFC PATCH v1 1/3] mm: teach mm by current context info to not doI/O during memory allocation

[Ming Lei]

On Wed, Oct 17, 2012 at 4:19 AM, Andrew Morton
<akpm@xxxxxxxxxxxxxxxxxxxx> wrote:
>
> The patch seems reasonable to me.  I'd like to see some examples of
> these resume-time callsite which are performing the GFP_KERNEL
> allocations, please.  You have found some kernel bugs, so those should
> be fully described.

OK, there are two examples of GFP_KERNEL allocation in subsystem
runtime resume path:

1), almost all devices in some pci platform
acpi_os_allocate
	<-acpi_ut_allocate
		<-ACPI_ALLOCATE_ZEROED
			<-acpi_evaluate_object
				<-__acpi_bus_set_power
					<-acpi_bus_set_power
						<-acpi_pci_set_power_state
							<-platform_pci_set_power_state
								<-pci_platform_power_transition
									<-__pci_complete_power_transition
										<-pci_set_power_state
											<-pci_restore_standard_config
												<-pci_pm_runtime_resume

2), all devices in usb subsystem
usb_get_status
	<-finish_port_resume
		<-usb_port_resume
			<-generic_resume
				<-usb_resume_device
					<-usb_resume_both
						<-usb_runtime_resume	

I also have many examples in which GFP_KERNEL allocation
is involved in runtime resume path of individual drivers.

The above two examples just show how difficult to solve the problem
by traditional way, :-)

Also as pointed by Oliver, network driver need memory allocation with
no io in iSCSI runtime resume situation too.

Thanks,
--
Ming Lei
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/