Re: [Qemu-devel] [PATCH] virtio: add a pci driver to notify host the CPU_DEAD event

[Gleb Natapov <gleb@redhat.com>]

On Sun, Nov 27, 2011 at 10:47:43AM +0800, Liu Ping Fan wrote:
> From: Liu Ping Fan <pingfank@linux.vnet.ibm.com>
> 
> A driver for qemu device "cpustate". This driver catch the guest
> CPU_DEAD event, and notify host.
> 
And if you do eject properly via ACPI this driver is replaced by 3 lines
of ACPI code and works with older guests too.

> Signed-off-by: Liu Ping Fan <pingfank@linux.vnet.ibm.com>
> ---
>  drivers/virtio/Kconfig         |    6 ++
>  drivers/virtio/Makefile        |    1 +
>  drivers/virtio/cpustate_stub.c |  154 ++++++++++++++++++++++++++++++++++++++++
>  3 files changed, 161 insertions(+), 0 deletions(-)
>  create mode 100644 drivers/virtio/cpustate_stub.c
> 
> diff --git a/drivers/virtio/Kconfig b/drivers/virtio/Kconfig
> index 816ed08..96ad253 100644
> --- a/drivers/virtio/Kconfig
> +++ b/drivers/virtio/Kconfig
> @@ -46,4 +46,10 @@ config VIRTIO_BALLOON
>  
>   	 If unsure, say N.
>  
> + config VIRTIO_CPUSTATE
> + 	tristate "Driver to notify host the cpu dead event (EXPERIMENTAL)"
> + 	depends on EXPERIMENTAL
> + 	---help---
> + 	 This drivers provides support to notify host the cpu dead event.
> +
>  endmenu
> diff --git a/drivers/virtio/Makefile b/drivers/virtio/Makefile
> index 5a4c63c..06a5ecf 100644
> --- a/drivers/virtio/Makefile
> +++ b/drivers/virtio/Makefile
> @@ -3,3 +3,4 @@ obj-$(CONFIG_VIRTIO_RING) += virtio_ring.o
>  obj-$(CONFIG_VIRTIO_MMIO) += virtio_mmio.o
>  obj-$(CONFIG_VIRTIO_PCI) += virtio_pci.o
>  obj-$(CONFIG_VIRTIO_BALLOON) += virtio_balloon.o
> +obj-$(VIRTIO_CPUSTATE) += cpustate_stub.o
> diff --git a/drivers/virtio/cpustate_stub.c b/drivers/virtio/cpustate_stub.c
> new file mode 100644
> index 0000000..614da9d
> --- /dev/null
> +++ b/drivers/virtio/cpustate_stub.c
> @@ -0,0 +1,154 @@
> +/*
> + * PCI driver for qemu cpustate device. It notifies host the CPU_DEAD event
> + * in guest.
> + *
> + * Copyright IBM Corp. 2011
> + *
> + * This work is licensed under the terms of the GNU GPL, version 2 or later.
> + * See the COPYING file in the top-level directory.
> + *
> + */
> +#include <linux/module.h>
> +#include <linux/list.h>
> +#include <linux/cpu.h>
> +#include <linux/pci.h>
> +#include <linux/slab.h>
> +#include <linux/interrupt.h>
> +#include <linux/spinlock.h>
> +
> +#define PCI_DEVICE_ID_CPUSTATE 0x1010
> +
> +struct cpustate_stub_regs {
> +	unsigned int cpu_phyid;
> +	unsigned int event;
> +};
> +
> +struct cpustate_stub {
> +	struct work_struct work;
> +
> +	unsigned int cpu_phyid;
> +	unsigned int event;
> +
> +	struct cpustate_stub_regs __iomem *regs;
> +};
> +
> +static struct cpustate_stub *agent;
> +
> +static void cpustate_work(struct work_struct *work)
> +{
> +	struct cpustate_stub *stub = container_of(work,
> +					struct cpustate_stub, work);
> +	printk(KERN_INFO "%s,cpu_phyid=0x%x, event=0x%x\n",
> +		__func__, stub->cpu_phyid, stub->event);
> +	stub->regs->cpu_phyid = stub->cpu_phyid;
> +	stub->regs->event = stub->event;
> +	barrier();
> +}
> +
> +static int cpu_dead_callback(struct notifier_block *b, unsigned long action,
> +				 void *data)
> +{
> +	unsigned long cpu = (unsigned long)data;
> +	int cpu_phyid;
> +	switch (action) {
> +	case CPU_DEAD:{
> +		cpu_phyid = per_cpu(x86_cpu_to_apicid, cpu);
> +		agent->cpu_phyid = cpu_phyid;
> +		agent->event = CPU_DEAD;
> +		schedule_work(&agent->work);
> +		break;
> +	}
> +	default:
> +		break;
> +	}
> +	return NOTIFY_DONE;
> +}
> +
> +static struct notifier_block __cpuinitdata cpu_dead_notifier = {
> +	.notifier_call	= cpu_dead_callback,
> +	.priority	= 10,
> +};
> +
> +static int __devinit cpustate_probe(struct pci_dev *pci_dev,
> +	const struct pci_device_id *id)
> +{
> +	int ret = 0;
> +	agent = kzalloc(sizeof(struct cpustate_stub), GFP_KERNEL);
> +	if (agent == NULL) {
> +		ret = -1;
> +		goto fail;
> +	}
> +	/* enable the device */
> +	ret = pci_enable_device(pci_dev);
> +	if (ret) {
> +		printk(KERN_WARNING "%s, pci_enable_device fail,ret=0x%x\n",
> +			__func__, ret);
> +		goto fail;
> +	}
> +
> +	ret = pci_request_regions(pci_dev, "cpustate");
> +	if (ret) {
> +		printk(KERN_WARNING "%s, pci_request_regions fail,ret=0x%x\n",
> +			__func__, ret);
> +		goto out_enable_device;
> +	}
> +
> +	agent->regs = ioremap(pci_dev->resource[0].start,
> +			pci_dev->resource[0].end - pci_dev->resource[0].start);
> +	if (agent->regs == NULL) {
> +		printk(KERN_WARNING "%s, ioremap fail\n", __func__);
> +		goto out_req_regions;
> +	}
> +
> +	INIT_WORK(&agent->work, cpustate_work);
> +	register_cpu_notifier(&cpu_dead_notifier);
> +	printk(KERN_INFO "%s, success\n", __func__);
> +	return 0;
> +
> +out_req_regions:
> +	pci_release_regions(pci_dev);
> +out_enable_device:
> +	pci_disable_device(pci_dev);
> +	kfree(agent);
> +	agent = NULL;
> +fail:
> +	printk(KERN_WARNING "%s fail\n", __func__);
> +	return ret;
> +}
> +
> +static void __devexit cpustate_remove(struct pci_dev *pci_dev)
> +{
> +	unregister_cpu_notifier(&cpu_dead_notifier);
> +}
> +
> +/* Qumranet donated their vendor ID for devices 0x1000 thru 0x10FF. */
> +static DEFINE_PCI_DEVICE_TABLE(pci_cpustate_id_table) = {
> +	{ PCI_VENDOR_ID_IBM, PCI_DEVICE_ID_CPUSTATE,
> +		PCI_ANY_ID, PCI_ANY_ID,
> +		PCI_CLASS_SYSTEM_OTHER, 0,
> +		0 },
> +	{ 0 },
> +};
> +MODULE_DEVICE_TABLE(pci, pci_cpustate_id_table);
> +
> +static struct pci_driver pci_cpustate_driver = {
> +	.name		= "cpustate",
> +	.id_table	= pci_cpustate_id_table,
> +	.probe		= cpustate_probe,
> +	.remove		= __devexit_p(cpustate_remove),
> +};
> +
> +static int __init pci_cpustate_init(void)
> +{
> +	return pci_register_driver(&pci_cpustate_driver);
> +}
> +module_init(pci_cpustate_init);
> +
> +static void __exit pci_cpustate_exit(void)
> +{
> +	pci_unregister_driver(&pci_cpustate_driver);
> +}
> +module_exit(pci_cpustate_exit);
> +MODULE_DESCRIPTION("cpustate");
> +MODULE_LICENSE("GPL");
> +MODULE_VERSION("1");
> -- 
> 1.7.4.4
> 

--
			Gleb.