2018-06-20 07:00:29 +08:00
|
|
|
// SPDX-License-Identifier: GPL-2.0
|
2010-04-24 01:18:08 +08:00
|
|
|
/*
|
|
|
|
* VMware Balloon driver.
|
|
|
|
*
|
2018-06-20 07:00:29 +08:00
|
|
|
* Copyright (C) 2000-2018, VMware, Inc. All Rights Reserved.
|
2010-04-24 01:18:08 +08:00
|
|
|
*
|
|
|
|
* This is VMware physical memory management driver for Linux. The driver
|
|
|
|
* acts like a "balloon" that can be inflated to reclaim physical pages by
|
|
|
|
* reserving them in the guest and invalidating them in the monitor,
|
|
|
|
* freeing up the underlying machine pages so they can be allocated to
|
|
|
|
* other guests. The balloon can also be deflated to allow the guest to
|
|
|
|
* use more physical memory. Higher level policies can control the sizes
|
|
|
|
* of balloons in VMs in order to manage physical memory resources.
|
|
|
|
*/
|
|
|
|
|
|
|
|
//#define DEBUG
|
|
|
|
#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
|
|
|
|
|
|
|
|
#include <linux/types.h>
|
2019-11-27 00:54:13 +08:00
|
|
|
#include <linux/io.h>
|
2010-04-24 01:18:08 +08:00
|
|
|
#include <linux/kernel.h>
|
|
|
|
#include <linux/mm.h>
|
2015-08-07 06:17:58 +08:00
|
|
|
#include <linux/vmalloc.h>
|
2010-04-24 01:18:08 +08:00
|
|
|
#include <linux/sched.h>
|
|
|
|
#include <linux/module.h>
|
|
|
|
#include <linux/workqueue.h>
|
|
|
|
#include <linux/debugfs.h>
|
|
|
|
#include <linux/seq_file.h>
|
2018-09-21 01:30:17 +08:00
|
|
|
#include <linux/rwsem.h>
|
|
|
|
#include <linux/slab.h>
|
vmw_balloon: rework the inflate and deflate loops
In preparation for supporting compaction and OOM notification, this
patch reworks the inflate/deflate loops. The main idea is to separate
the allocation, communication with the hypervisor, and the handling of
errors from each other. Doing will allow us to perform concurrent
inflation and deflation, excluding the actual communication with the
hypervisor.
To do so, we need to get rid of the remaining global state that is kept
in the balloon struct, specifically the refuse_list. When the VM
communicates with the hypervisor, it does not free or put back pages
to the balloon list and instead only moves the pages whose status
indicated failure into a refuse_list on the stack. Once the operation
completes, the inflation or deflation functions handle the list
appropriately.
As we do that, we can consolidate the communication with the hypervisor
for both the lock and unlock operations into a single function. We also
reuse the deflation function for popping the balloon.
As a preparation for preventing races, we hold a spinlock when the
communication actually takes place, and use atomic operations for
updating the balloon size. The balloon page list is still racy and will
be handled in the next patch.
Reviewed-by: Xavier Deguillard <xdeguillard@vmware.com>
Signed-off-by: Nadav Amit <namit@vmware.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2018-09-21 01:30:18 +08:00
|
|
|
#include <linux/spinlock.h>
|
2019-04-25 19:54:43 +08:00
|
|
|
#include <linux/balloon_compaction.h>
|
2015-08-07 06:18:01 +08:00
|
|
|
#include <linux/vmw_vmci_defs.h>
|
|
|
|
#include <linux/vmw_vmci_api.h>
|
2010-05-09 16:13:42 +08:00
|
|
|
#include <asm/hypervisor.h>
|
2010-04-24 01:18:08 +08:00
|
|
|
|
|
|
|
MODULE_AUTHOR("VMware, Inc.");
|
|
|
|
MODULE_DESCRIPTION("VMware Memory Control (Balloon) Driver");
|
|
|
|
MODULE_ALIAS("dmi:*:svnVMware*:*");
|
|
|
|
MODULE_ALIAS("vmware_vmmemctl");
|
|
|
|
MODULE_LICENSE("GPL");
|
|
|
|
|
2019-04-25 19:54:44 +08:00
|
|
|
static bool __read_mostly vmwballoon_shrinker_enable;
|
|
|
|
module_param(vmwballoon_shrinker_enable, bool, 0444);
|
|
|
|
MODULE_PARM_DESC(vmwballoon_shrinker_enable,
|
|
|
|
"Enable non-cooperative out-of-memory protection. Disabled by default as it may degrade performance.");
|
|
|
|
|
|
|
|
/* Delay in seconds after shrink before inflation. */
|
|
|
|
#define VMBALLOON_SHRINK_DELAY (5)
|
|
|
|
|
|
|
|
/* Maximum number of refused pages we accumulate during inflation cycle */
|
2010-06-05 05:14:52 +08:00
|
|
|
#define VMW_BALLOON_MAX_REFUSED 16
|
2010-04-24 01:18:08 +08:00
|
|
|
|
2019-04-25 19:54:43 +08:00
|
|
|
/* Magic number for the balloon mount-point */
|
|
|
|
#define BALLOON_VMW_MAGIC 0x0ba11007
|
|
|
|
|
2010-04-24 01:18:08 +08:00
|
|
|
/*
|
|
|
|
* Hypervisor communication port definitions.
|
|
|
|
*/
|
|
|
|
#define VMW_BALLOON_HV_PORT 0x5670
|
|
|
|
#define VMW_BALLOON_HV_MAGIC 0x456c6d6f
|
|
|
|
#define VMW_BALLOON_GUEST_ID 1 /* Linux */
|
|
|
|
|
2015-06-13 02:43:23 +08:00
|
|
|
enum vmwballoon_capabilities {
|
|
|
|
/*
|
|
|
|
* Bit 0 is reserved and not associated to any capability.
|
|
|
|
*/
|
2015-08-07 06:18:01 +08:00
|
|
|
VMW_BALLOON_BASIC_CMDS = (1 << 1),
|
|
|
|
VMW_BALLOON_BATCHED_CMDS = (1 << 2),
|
|
|
|
VMW_BALLOON_BATCHED_2M_CMDS = (1 << 3),
|
|
|
|
VMW_BALLOON_SIGNALLED_WAKEUP_CMD = (1 << 4),
|
2019-02-07 07:57:02 +08:00
|
|
|
VMW_BALLOON_64_BIT_TARGET = (1 << 5)
|
2015-06-13 02:43:23 +08:00
|
|
|
};
|
|
|
|
|
2019-02-07 07:57:02 +08:00
|
|
|
#define VMW_BALLOON_CAPABILITIES_COMMON (VMW_BALLOON_BASIC_CMDS \
|
2015-08-07 06:18:01 +08:00
|
|
|
| VMW_BALLOON_BATCHED_CMDS \
|
2015-08-07 06:18:01 +08:00
|
|
|
| VMW_BALLOON_BATCHED_2M_CMDS \
|
|
|
|
| VMW_BALLOON_SIGNALLED_WAKEUP_CMD)
|
2015-08-07 06:18:01 +08:00
|
|
|
|
2018-09-21 01:30:14 +08:00
|
|
|
#define VMW_BALLOON_2M_ORDER (PMD_SHIFT - PAGE_SHIFT)
|
2015-06-13 02:43:23 +08:00
|
|
|
|
2019-02-07 07:57:02 +08:00
|
|
|
/*
|
|
|
|
* 64-bit targets are only supported in 64-bit
|
|
|
|
*/
|
|
|
|
#ifdef CONFIG_64BIT
|
|
|
|
#define VMW_BALLOON_CAPABILITIES (VMW_BALLOON_CAPABILITIES_COMMON \
|
|
|
|
| VMW_BALLOON_64_BIT_TARGET)
|
|
|
|
#else
|
|
|
|
#define VMW_BALLOON_CAPABILITIES VMW_BALLOON_CAPABILITIES_COMMON
|
|
|
|
#endif
|
|
|
|
|
2018-09-21 01:30:17 +08:00
|
|
|
enum vmballoon_page_size_type {
|
|
|
|
VMW_BALLOON_4K_PAGE,
|
|
|
|
VMW_BALLOON_2M_PAGE,
|
|
|
|
VMW_BALLOON_LAST_SIZE = VMW_BALLOON_2M_PAGE
|
|
|
|
};
|
|
|
|
|
|
|
|
#define VMW_BALLOON_NUM_PAGE_SIZES (VMW_BALLOON_LAST_SIZE + 1)
|
|
|
|
|
vmw_balloon: rework the inflate and deflate loops
In preparation for supporting compaction and OOM notification, this
patch reworks the inflate/deflate loops. The main idea is to separate
the allocation, communication with the hypervisor, and the handling of
errors from each other. Doing will allow us to perform concurrent
inflation and deflation, excluding the actual communication with the
hypervisor.
To do so, we need to get rid of the remaining global state that is kept
in the balloon struct, specifically the refuse_list. When the VM
communicates with the hypervisor, it does not free or put back pages
to the balloon list and instead only moves the pages whose status
indicated failure into a refuse_list on the stack. Once the operation
completes, the inflation or deflation functions handle the list
appropriately.
As we do that, we can consolidate the communication with the hypervisor
for both the lock and unlock operations into a single function. We also
reuse the deflation function for popping the balloon.
As a preparation for preventing races, we hold a spinlock when the
communication actually takes place, and use atomic operations for
updating the balloon size. The balloon page list is still racy and will
be handled in the next patch.
Reviewed-by: Xavier Deguillard <xdeguillard@vmware.com>
Signed-off-by: Nadav Amit <namit@vmware.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2018-09-21 01:30:18 +08:00
|
|
|
static const char * const vmballoon_page_size_names[] = {
|
|
|
|
[VMW_BALLOON_4K_PAGE] = "4k",
|
|
|
|
[VMW_BALLOON_2M_PAGE] = "2M"
|
|
|
|
};
|
|
|
|
|
|
|
|
enum vmballoon_op {
|
|
|
|
VMW_BALLOON_INFLATE,
|
|
|
|
VMW_BALLOON_DEFLATE
|
|
|
|
};
|
|
|
|
|
2018-09-21 01:30:17 +08:00
|
|
|
enum vmballoon_op_stat_type {
|
|
|
|
VMW_BALLOON_OP_STAT,
|
|
|
|
VMW_BALLOON_OP_FAIL_STAT
|
|
|
|
};
|
|
|
|
|
|
|
|
#define VMW_BALLOON_OP_STAT_TYPES (VMW_BALLOON_OP_FAIL_STAT + 1)
|
|
|
|
|
|
|
|
/**
|
|
|
|
* enum vmballoon_cmd_type - backdoor commands.
|
|
|
|
*
|
|
|
|
* Availability of the commands is as followed:
|
|
|
|
*
|
|
|
|
* %VMW_BALLOON_CMD_START, %VMW_BALLOON_CMD_GET_TARGET and
|
|
|
|
* %VMW_BALLOON_CMD_GUEST_ID are always available.
|
|
|
|
*
|
|
|
|
* If the host reports %VMW_BALLOON_BASIC_CMDS are supported then
|
|
|
|
* %VMW_BALLOON_CMD_LOCK and %VMW_BALLOON_CMD_UNLOCK commands are available.
|
2015-08-07 06:17:58 +08:00
|
|
|
*
|
2018-09-21 01:30:17 +08:00
|
|
|
* If the host reports %VMW_BALLOON_BATCHED_CMDS are supported then
|
|
|
|
* %VMW_BALLOON_CMD_BATCHED_LOCK and VMW_BALLOON_CMD_BATCHED_UNLOCK commands
|
|
|
|
* are available.
|
2015-08-07 06:17:58 +08:00
|
|
|
*
|
2018-09-21 01:30:17 +08:00
|
|
|
* If the host reports %VMW_BALLOON_BATCHED_2M_CMDS are supported then
|
|
|
|
* %VMW_BALLOON_CMD_BATCHED_2M_LOCK and %VMW_BALLOON_CMD_BATCHED_2M_UNLOCK
|
|
|
|
* are supported.
|
|
|
|
*
|
|
|
|
* If the host reports VMW_BALLOON_SIGNALLED_WAKEUP_CMD is supported then
|
|
|
|
* VMW_BALLOON_CMD_VMCI_DOORBELL_SET command is supported.
|
|
|
|
*
|
|
|
|
* @VMW_BALLOON_CMD_START: Communicating supported version with the hypervisor.
|
|
|
|
* @VMW_BALLOON_CMD_GET_TARGET: Gets the balloon target size.
|
|
|
|
* @VMW_BALLOON_CMD_LOCK: Informs the hypervisor about a ballooned page.
|
|
|
|
* @VMW_BALLOON_CMD_UNLOCK: Informs the hypervisor about a page that is about
|
|
|
|
* to be deflated from the balloon.
|
|
|
|
* @VMW_BALLOON_CMD_GUEST_ID: Informs the hypervisor about the type of OS that
|
|
|
|
* runs in the VM.
|
|
|
|
* @VMW_BALLOON_CMD_BATCHED_LOCK: Inform the hypervisor about a batch of
|
|
|
|
* ballooned pages (up to 512).
|
|
|
|
* @VMW_BALLOON_CMD_BATCHED_UNLOCK: Inform the hypervisor about a batch of
|
|
|
|
* pages that are about to be deflated from the
|
|
|
|
* balloon (up to 512).
|
|
|
|
* @VMW_BALLOON_CMD_BATCHED_2M_LOCK: Similar to @VMW_BALLOON_CMD_BATCHED_LOCK
|
|
|
|
* for 2MB pages.
|
|
|
|
* @VMW_BALLOON_CMD_BATCHED_2M_UNLOCK: Similar to
|
|
|
|
* @VMW_BALLOON_CMD_BATCHED_UNLOCK for 2MB
|
|
|
|
* pages.
|
|
|
|
* @VMW_BALLOON_CMD_VMCI_DOORBELL_SET: A command to set doorbell notification
|
|
|
|
* that would be invoked when the balloon
|
|
|
|
* size changes.
|
|
|
|
* @VMW_BALLOON_CMD_LAST: Value of the last command.
|
2015-08-07 06:17:58 +08:00
|
|
|
*/
|
2018-09-21 01:30:17 +08:00
|
|
|
enum vmballoon_cmd_type {
|
|
|
|
VMW_BALLOON_CMD_START,
|
|
|
|
VMW_BALLOON_CMD_GET_TARGET,
|
|
|
|
VMW_BALLOON_CMD_LOCK,
|
|
|
|
VMW_BALLOON_CMD_UNLOCK,
|
|
|
|
VMW_BALLOON_CMD_GUEST_ID,
|
|
|
|
/* No command 5 */
|
|
|
|
VMW_BALLOON_CMD_BATCHED_LOCK = 6,
|
|
|
|
VMW_BALLOON_CMD_BATCHED_UNLOCK,
|
|
|
|
VMW_BALLOON_CMD_BATCHED_2M_LOCK,
|
|
|
|
VMW_BALLOON_CMD_BATCHED_2M_UNLOCK,
|
|
|
|
VMW_BALLOON_CMD_VMCI_DOORBELL_SET,
|
|
|
|
VMW_BALLOON_CMD_LAST = VMW_BALLOON_CMD_VMCI_DOORBELL_SET,
|
|
|
|
};
|
|
|
|
|
|
|
|
#define VMW_BALLOON_CMD_NUM (VMW_BALLOON_CMD_LAST + 1)
|
|
|
|
|
|
|
|
enum vmballoon_error_codes {
|
|
|
|
VMW_BALLOON_SUCCESS,
|
|
|
|
VMW_BALLOON_ERROR_CMD_INVALID,
|
|
|
|
VMW_BALLOON_ERROR_PPN_INVALID,
|
|
|
|
VMW_BALLOON_ERROR_PPN_LOCKED,
|
|
|
|
VMW_BALLOON_ERROR_PPN_UNLOCKED,
|
|
|
|
VMW_BALLOON_ERROR_PPN_PINNED,
|
|
|
|
VMW_BALLOON_ERROR_PPN_NOTNEEDED,
|
|
|
|
VMW_BALLOON_ERROR_RESET,
|
|
|
|
VMW_BALLOON_ERROR_BUSY
|
|
|
|
};
|
2015-06-13 02:43:23 +08:00
|
|
|
|
|
|
|
#define VMW_BALLOON_SUCCESS_WITH_CAPABILITIES (0x03000000)
|
|
|
|
|
2018-09-21 01:30:07 +08:00
|
|
|
#define VMW_BALLOON_CMD_WITH_TARGET_MASK \
|
|
|
|
((1UL << VMW_BALLOON_CMD_GET_TARGET) | \
|
|
|
|
(1UL << VMW_BALLOON_CMD_LOCK) | \
|
|
|
|
(1UL << VMW_BALLOON_CMD_UNLOCK) | \
|
|
|
|
(1UL << VMW_BALLOON_CMD_BATCHED_LOCK) | \
|
|
|
|
(1UL << VMW_BALLOON_CMD_BATCHED_UNLOCK) | \
|
|
|
|
(1UL << VMW_BALLOON_CMD_BATCHED_2M_LOCK) | \
|
|
|
|
(1UL << VMW_BALLOON_CMD_BATCHED_2M_UNLOCK))
|
|
|
|
|
2018-09-21 01:30:08 +08:00
|
|
|
static const char * const vmballoon_cmd_names[] = {
|
|
|
|
[VMW_BALLOON_CMD_START] = "start",
|
|
|
|
[VMW_BALLOON_CMD_GET_TARGET] = "target",
|
|
|
|
[VMW_BALLOON_CMD_LOCK] = "lock",
|
|
|
|
[VMW_BALLOON_CMD_UNLOCK] = "unlock",
|
|
|
|
[VMW_BALLOON_CMD_GUEST_ID] = "guestType",
|
|
|
|
[VMW_BALLOON_CMD_BATCHED_LOCK] = "batchLock",
|
|
|
|
[VMW_BALLOON_CMD_BATCHED_UNLOCK] = "batchUnlock",
|
|
|
|
[VMW_BALLOON_CMD_BATCHED_2M_LOCK] = "2m-lock",
|
|
|
|
[VMW_BALLOON_CMD_BATCHED_2M_UNLOCK] = "2m-unlock",
|
|
|
|
[VMW_BALLOON_CMD_VMCI_DOORBELL_SET] = "doorbellSet"
|
|
|
|
};
|
|
|
|
|
2018-09-21 01:30:17 +08:00
|
|
|
enum vmballoon_stat_page {
|
|
|
|
VMW_BALLOON_PAGE_STAT_ALLOC,
|
|
|
|
VMW_BALLOON_PAGE_STAT_ALLOC_FAIL,
|
|
|
|
VMW_BALLOON_PAGE_STAT_REFUSED_ALLOC,
|
|
|
|
VMW_BALLOON_PAGE_STAT_REFUSED_FREE,
|
|
|
|
VMW_BALLOON_PAGE_STAT_FREE,
|
|
|
|
VMW_BALLOON_PAGE_STAT_LAST = VMW_BALLOON_PAGE_STAT_FREE
|
2010-04-24 01:18:08 +08:00
|
|
|
};
|
|
|
|
|
2018-09-21 01:30:17 +08:00
|
|
|
#define VMW_BALLOON_PAGE_STAT_NUM (VMW_BALLOON_PAGE_STAT_LAST + 1)
|
|
|
|
|
|
|
|
enum vmballoon_stat_general {
|
|
|
|
VMW_BALLOON_STAT_TIMER,
|
|
|
|
VMW_BALLOON_STAT_DOORBELL,
|
2018-09-21 01:30:20 +08:00
|
|
|
VMW_BALLOON_STAT_RESET,
|
2019-04-25 19:54:44 +08:00
|
|
|
VMW_BALLOON_STAT_SHRINK,
|
|
|
|
VMW_BALLOON_STAT_SHRINK_FREE,
|
|
|
|
VMW_BALLOON_STAT_LAST = VMW_BALLOON_STAT_SHRINK_FREE
|
2018-09-21 01:30:17 +08:00
|
|
|
};
|
|
|
|
|
|
|
|
#define VMW_BALLOON_STAT_NUM (VMW_BALLOON_STAT_LAST + 1)
|
|
|
|
|
vmw_balloon: change batch/single lock abstractions
The current abstractions for batch vs single operations seem suboptimal
and complicate the implementation of additional features (OOM,
compaction).
The immediate problem of the current abstractions is that they cause
differences in how operations are handled when batching is on or off.
For example, the refused_alloc counter is not updated when batching is
on. These discrepancies are caused by code redundancies.
Instead, this patch presents three type of operations, according to
whether batching is on or off: (1) add page, (2) communication with the
hypervisor and (3) retrieving the status of a page.
To avoid the overhead of virtual functions, and since we do not expect
additional interfaces for communication with the hypervisor, we use
static keys instead of virtual functions.
Finally, while we are at it, change vmballoon_init_batching() to return
int instead of bool, to be consistent in the return type and avoid
potential coding errors.
Reviewed-by: Xavier Deguillard <xdeguillard@vmware.com>
Signed-off-by: Nadav Amit <namit@vmware.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2018-09-21 01:30:12 +08:00
|
|
|
static DEFINE_STATIC_KEY_TRUE(vmw_balloon_batching);
|
2018-09-21 01:30:17 +08:00
|
|
|
static DEFINE_STATIC_KEY_FALSE(balloon_stat_enabled);
|
2015-08-07 06:17:58 +08:00
|
|
|
|
vmw_balloon: rework the inflate and deflate loops
In preparation for supporting compaction and OOM notification, this
patch reworks the inflate/deflate loops. The main idea is to separate
the allocation, communication with the hypervisor, and the handling of
errors from each other. Doing will allow us to perform concurrent
inflation and deflation, excluding the actual communication with the
hypervisor.
To do so, we need to get rid of the remaining global state that is kept
in the balloon struct, specifically the refuse_list. When the VM
communicates with the hypervisor, it does not free or put back pages
to the balloon list and instead only moves the pages whose status
indicated failure into a refuse_list on the stack. Once the operation
completes, the inflation or deflation functions handle the list
appropriately.
As we do that, we can consolidate the communication with the hypervisor
for both the lock and unlock operations into a single function. We also
reuse the deflation function for popping the balloon.
As a preparation for preventing races, we hold a spinlock when the
communication actually takes place, and use atomic operations for
updating the balloon size. The balloon page list is still racy and will
be handled in the next patch.
Reviewed-by: Xavier Deguillard <xdeguillard@vmware.com>
Signed-off-by: Nadav Amit <namit@vmware.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2018-09-21 01:30:18 +08:00
|
|
|
struct vmballoon_ctl {
|
2010-04-24 01:18:08 +08:00
|
|
|
struct list_head pages;
|
|
|
|
struct list_head refused_pages;
|
2019-04-25 19:54:45 +08:00
|
|
|
struct list_head prealloc_pages;
|
2010-06-05 05:14:52 +08:00
|
|
|
unsigned int n_refused_pages;
|
vmw_balloon: rework the inflate and deflate loops
In preparation for supporting compaction and OOM notification, this
patch reworks the inflate/deflate loops. The main idea is to separate
the allocation, communication with the hypervisor, and the handling of
errors from each other. Doing will allow us to perform concurrent
inflation and deflation, excluding the actual communication with the
hypervisor.
To do so, we need to get rid of the remaining global state that is kept
in the balloon struct, specifically the refuse_list. When the VM
communicates with the hypervisor, it does not free or put back pages
to the balloon list and instead only moves the pages whose status
indicated failure into a refuse_list on the stack. Once the operation
completes, the inflation or deflation functions handle the list
appropriately.
As we do that, we can consolidate the communication with the hypervisor
for both the lock and unlock operations into a single function. We also
reuse the deflation function for popping the balloon.
As a preparation for preventing races, we hold a spinlock when the
communication actually takes place, and use atomic operations for
updating the balloon size. The balloon page list is still racy and will
be handled in the next patch.
Reviewed-by: Xavier Deguillard <xdeguillard@vmware.com>
Signed-off-by: Nadav Amit <namit@vmware.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2018-09-21 01:30:18 +08:00
|
|
|
unsigned int n_pages;
|
|
|
|
enum vmballoon_page_size_type page_size;
|
|
|
|
enum vmballoon_op op;
|
|
|
|
};
|
|
|
|
|
2018-09-21 01:30:10 +08:00
|
|
|
/**
|
|
|
|
* struct vmballoon_batch_entry - a batch entry for lock or unlock.
|
|
|
|
*
|
|
|
|
* @status: the status of the operation, which is written by the hypervisor.
|
|
|
|
* @reserved: reserved for future use. Must be set to zero.
|
|
|
|
* @pfn: the physical frame number of the page to be locked or unlocked.
|
|
|
|
*/
|
|
|
|
struct vmballoon_batch_entry {
|
|
|
|
u64 status : 5;
|
|
|
|
u64 reserved : PAGE_SHIFT - 5;
|
|
|
|
u64 pfn : 52;
|
|
|
|
} __packed;
|
|
|
|
|
2015-08-07 06:18:01 +08:00
|
|
|
struct vmballoon {
|
vmw_balloon: rework the inflate and deflate loops
In preparation for supporting compaction and OOM notification, this
patch reworks the inflate/deflate loops. The main idea is to separate
the allocation, communication with the hypervisor, and the handling of
errors from each other. Doing will allow us to perform concurrent
inflation and deflation, excluding the actual communication with the
hypervisor.
To do so, we need to get rid of the remaining global state that is kept
in the balloon struct, specifically the refuse_list. When the VM
communicates with the hypervisor, it does not free or put back pages
to the balloon list and instead only moves the pages whose status
indicated failure into a refuse_list on the stack. Once the operation
completes, the inflation or deflation functions handle the list
appropriately.
As we do that, we can consolidate the communication with the hypervisor
for both the lock and unlock operations into a single function. We also
reuse the deflation function for popping the balloon.
As a preparation for preventing races, we hold a spinlock when the
communication actually takes place, and use atomic operations for
updating the balloon size. The balloon page list is still racy and will
be handled in the next patch.
Reviewed-by: Xavier Deguillard <xdeguillard@vmware.com>
Signed-off-by: Nadav Amit <namit@vmware.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2018-09-21 01:30:18 +08:00
|
|
|
/**
|
|
|
|
* @max_page_size: maximum supported page size for ballooning.
|
|
|
|
*
|
|
|
|
* Protected by @conf_sem
|
|
|
|
*/
|
|
|
|
enum vmballoon_page_size_type max_page_size;
|
|
|
|
|
|
|
|
/**
|
|
|
|
* @size: balloon actual size in basic page size (frames).
|
|
|
|
*
|
|
|
|
* While we currently do not support size which is bigger than 32-bit,
|
|
|
|
* in preparation for future support, use 64-bits.
|
|
|
|
*/
|
|
|
|
atomic64_t size;
|
2010-04-24 01:18:08 +08:00
|
|
|
|
vmw_balloon: rework the inflate and deflate loops
In preparation for supporting compaction and OOM notification, this
patch reworks the inflate/deflate loops. The main idea is to separate
the allocation, communication with the hypervisor, and the handling of
errors from each other. Doing will allow us to perform concurrent
inflation and deflation, excluding the actual communication with the
hypervisor.
To do so, we need to get rid of the remaining global state that is kept
in the balloon struct, specifically the refuse_list. When the VM
communicates with the hypervisor, it does not free or put back pages
to the balloon list and instead only moves the pages whose status
indicated failure into a refuse_list on the stack. Once the operation
completes, the inflation or deflation functions handle the list
appropriately.
As we do that, we can consolidate the communication with the hypervisor
for both the lock and unlock operations into a single function. We also
reuse the deflation function for popping the balloon.
As a preparation for preventing races, we hold a spinlock when the
communication actually takes place, and use atomic operations for
updating the balloon size. The balloon page list is still racy and will
be handled in the next patch.
Reviewed-by: Xavier Deguillard <xdeguillard@vmware.com>
Signed-off-by: Nadav Amit <namit@vmware.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2018-09-21 01:30:18 +08:00
|
|
|
/**
|
|
|
|
* @target: balloon target size in basic page size (frames).
|
|
|
|
*
|
|
|
|
* We do not protect the target under the assumption that setting the
|
|
|
|
* value is always done through a single write. If this assumption ever
|
|
|
|
* breaks, we would have to use X_ONCE for accesses, and suffer the less
|
|
|
|
* optimized code. Although we may read stale target value if multiple
|
|
|
|
* accesses happen at once, the performance impact should be minor.
|
|
|
|
*/
|
|
|
|
unsigned long target;
|
2010-04-24 01:18:08 +08:00
|
|
|
|
vmw_balloon: rework the inflate and deflate loops
In preparation for supporting compaction and OOM notification, this
patch reworks the inflate/deflate loops. The main idea is to separate
the allocation, communication with the hypervisor, and the handling of
errors from each other. Doing will allow us to perform concurrent
inflation and deflation, excluding the actual communication with the
hypervisor.
To do so, we need to get rid of the remaining global state that is kept
in the balloon struct, specifically the refuse_list. When the VM
communicates with the hypervisor, it does not free or put back pages
to the balloon list and instead only moves the pages whose status
indicated failure into a refuse_list on the stack. Once the operation
completes, the inflation or deflation functions handle the list
appropriately.
As we do that, we can consolidate the communication with the hypervisor
for both the lock and unlock operations into a single function. We also
reuse the deflation function for popping the balloon.
As a preparation for preventing races, we hold a spinlock when the
communication actually takes place, and use atomic operations for
updating the balloon size. The balloon page list is still racy and will
be handled in the next patch.
Reviewed-by: Xavier Deguillard <xdeguillard@vmware.com>
Signed-off-by: Nadav Amit <namit@vmware.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2018-09-21 01:30:18 +08:00
|
|
|
/**
|
|
|
|
* @reset_required: reset flag
|
|
|
|
*
|
|
|
|
* Setting this flag may introduce races, but the code is expected to
|
|
|
|
* handle them gracefully. In the worst case, another operation will
|
|
|
|
* fail as reset did not take place. Clearing the flag is done while
|
|
|
|
* holding @conf_sem for write.
|
|
|
|
*/
|
2010-04-24 01:18:08 +08:00
|
|
|
bool reset_required;
|
|
|
|
|
vmw_balloon: rework the inflate and deflate loops
In preparation for supporting compaction and OOM notification, this
patch reworks the inflate/deflate loops. The main idea is to separate
the allocation, communication with the hypervisor, and the handling of
errors from each other. Doing will allow us to perform concurrent
inflation and deflation, excluding the actual communication with the
hypervisor.
To do so, we need to get rid of the remaining global state that is kept
in the balloon struct, specifically the refuse_list. When the VM
communicates with the hypervisor, it does not free or put back pages
to the balloon list and instead only moves the pages whose status
indicated failure into a refuse_list on the stack. Once the operation
completes, the inflation or deflation functions handle the list
appropriately.
As we do that, we can consolidate the communication with the hypervisor
for both the lock and unlock operations into a single function. We also
reuse the deflation function for popping the balloon.
As a preparation for preventing races, we hold a spinlock when the
communication actually takes place, and use atomic operations for
updating the balloon size. The balloon page list is still racy and will
be handled in the next patch.
Reviewed-by: Xavier Deguillard <xdeguillard@vmware.com>
Signed-off-by: Nadav Amit <namit@vmware.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2018-09-21 01:30:18 +08:00
|
|
|
/**
|
|
|
|
* @capabilities: hypervisor balloon capabilities.
|
|
|
|
*
|
|
|
|
* Protected by @conf_sem.
|
|
|
|
*/
|
2015-08-07 06:17:58 +08:00
|
|
|
unsigned long capabilities;
|
|
|
|
|
2018-09-21 01:30:10 +08:00
|
|
|
/**
|
|
|
|
* @batch_page: pointer to communication batch page.
|
|
|
|
*
|
|
|
|
* When batching is used, batch_page points to a page, which holds up to
|
|
|
|
* %VMW_BALLOON_BATCH_MAX_PAGES entries for locking or unlocking.
|
|
|
|
*/
|
|
|
|
struct vmballoon_batch_entry *batch_page;
|
|
|
|
|
vmw_balloon: rework the inflate and deflate loops
In preparation for supporting compaction and OOM notification, this
patch reworks the inflate/deflate loops. The main idea is to separate
the allocation, communication with the hypervisor, and the handling of
errors from each other. Doing will allow us to perform concurrent
inflation and deflation, excluding the actual communication with the
hypervisor.
To do so, we need to get rid of the remaining global state that is kept
in the balloon struct, specifically the refuse_list. When the VM
communicates with the hypervisor, it does not free or put back pages
to the balloon list and instead only moves the pages whose status
indicated failure into a refuse_list on the stack. Once the operation
completes, the inflation or deflation functions handle the list
appropriately.
As we do that, we can consolidate the communication with the hypervisor
for both the lock and unlock operations into a single function. We also
reuse the deflation function for popping the balloon.
As a preparation for preventing races, we hold a spinlock when the
communication actually takes place, and use atomic operations for
updating the balloon size. The balloon page list is still racy and will
be handled in the next patch.
Reviewed-by: Xavier Deguillard <xdeguillard@vmware.com>
Signed-off-by: Nadav Amit <namit@vmware.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2018-09-21 01:30:18 +08:00
|
|
|
/**
|
|
|
|
* @batch_max_pages: maximum pages that can be locked/unlocked.
|
|
|
|
*
|
|
|
|
* Indicates the number of pages that the hypervisor can lock or unlock
|
|
|
|
* at once, according to whether batching is enabled. If batching is
|
|
|
|
* disabled, only a single page can be locked/unlock on each operation.
|
|
|
|
*
|
|
|
|
* Protected by @conf_sem.
|
|
|
|
*/
|
2015-08-07 06:17:58 +08:00
|
|
|
unsigned int batch_max_pages;
|
vmw_balloon: rework the inflate and deflate loops
In preparation for supporting compaction and OOM notification, this
patch reworks the inflate/deflate loops. The main idea is to separate
the allocation, communication with the hypervisor, and the handling of
errors from each other. Doing will allow us to perform concurrent
inflation and deflation, excluding the actual communication with the
hypervisor.
To do so, we need to get rid of the remaining global state that is kept
in the balloon struct, specifically the refuse_list. When the VM
communicates with the hypervisor, it does not free or put back pages
to the balloon list and instead only moves the pages whose status
indicated failure into a refuse_list on the stack. Once the operation
completes, the inflation or deflation functions handle the list
appropriately.
As we do that, we can consolidate the communication with the hypervisor
for both the lock and unlock operations into a single function. We also
reuse the deflation function for popping the balloon.
As a preparation for preventing races, we hold a spinlock when the
communication actually takes place, and use atomic operations for
updating the balloon size. The balloon page list is still racy and will
be handled in the next patch.
Reviewed-by: Xavier Deguillard <xdeguillard@vmware.com>
Signed-off-by: Nadav Amit <namit@vmware.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2018-09-21 01:30:18 +08:00
|
|
|
|
|
|
|
/**
|
|
|
|
* @page: page to be locked/unlocked by the hypervisor
|
|
|
|
*
|
|
|
|
* @page is only used when batching is disabled and a single page is
|
|
|
|
* reclaimed on each iteration.
|
|
|
|
*
|
|
|
|
* Protected by @comm_lock.
|
|
|
|
*/
|
2015-08-07 06:17:58 +08:00
|
|
|
struct page *page;
|
|
|
|
|
2019-04-25 19:54:44 +08:00
|
|
|
/**
|
|
|
|
* @shrink_timeout: timeout until the next inflation.
|
|
|
|
*
|
|
|
|
* After an shrink event, indicates the time in jiffies after which
|
|
|
|
* inflation is allowed again. Can be written concurrently with reads,
|
|
|
|
* so must use READ_ONCE/WRITE_ONCE when accessing.
|
|
|
|
*/
|
|
|
|
unsigned long shrink_timeout;
|
|
|
|
|
2010-04-24 01:18:08 +08:00
|
|
|
/* statistics */
|
2018-09-21 01:30:17 +08:00
|
|
|
struct vmballoon_stats *stats;
|
2010-04-24 01:18:08 +08:00
|
|
|
|
2019-04-25 19:54:43 +08:00
|
|
|
/**
|
|
|
|
* @b_dev_info: balloon device information descriptor.
|
|
|
|
*/
|
|
|
|
struct balloon_dev_info b_dev_info;
|
|
|
|
|
2010-04-24 01:18:08 +08:00
|
|
|
struct delayed_work dwork;
|
2015-08-07 06:18:01 +08:00
|
|
|
|
2019-04-25 19:54:43 +08:00
|
|
|
/**
|
|
|
|
* @huge_pages - list of the inflated 2MB pages.
|
|
|
|
*
|
|
|
|
* Protected by @b_dev_info.pages_lock .
|
|
|
|
*/
|
|
|
|
struct list_head huge_pages;
|
|
|
|
|
vmw_balloon: rework the inflate and deflate loops
In preparation for supporting compaction and OOM notification, this
patch reworks the inflate/deflate loops. The main idea is to separate
the allocation, communication with the hypervisor, and the handling of
errors from each other. Doing will allow us to perform concurrent
inflation and deflation, excluding the actual communication with the
hypervisor.
To do so, we need to get rid of the remaining global state that is kept
in the balloon struct, specifically the refuse_list. When the VM
communicates with the hypervisor, it does not free or put back pages
to the balloon list and instead only moves the pages whose status
indicated failure into a refuse_list on the stack. Once the operation
completes, the inflation or deflation functions handle the list
appropriately.
As we do that, we can consolidate the communication with the hypervisor
for both the lock and unlock operations into a single function. We also
reuse the deflation function for popping the balloon.
As a preparation for preventing races, we hold a spinlock when the
communication actually takes place, and use atomic operations for
updating the balloon size. The balloon page list is still racy and will
be handled in the next patch.
Reviewed-by: Xavier Deguillard <xdeguillard@vmware.com>
Signed-off-by: Nadav Amit <namit@vmware.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2018-09-21 01:30:18 +08:00
|
|
|
/**
|
|
|
|
* @vmci_doorbell.
|
|
|
|
*
|
|
|
|
* Protected by @conf_sem.
|
|
|
|
*/
|
2015-08-07 06:18:01 +08:00
|
|
|
struct vmci_handle vmci_doorbell;
|
2018-09-21 01:30:17 +08:00
|
|
|
|
|
|
|
/**
|
|
|
|
* @conf_sem: semaphore to protect the configuration and the statistics.
|
|
|
|
*/
|
|
|
|
struct rw_semaphore conf_sem;
|
vmw_balloon: rework the inflate and deflate loops
In preparation for supporting compaction and OOM notification, this
patch reworks the inflate/deflate loops. The main idea is to separate
the allocation, communication with the hypervisor, and the handling of
errors from each other. Doing will allow us to perform concurrent
inflation and deflation, excluding the actual communication with the
hypervisor.
To do so, we need to get rid of the remaining global state that is kept
in the balloon struct, specifically the refuse_list. When the VM
communicates with the hypervisor, it does not free or put back pages
to the balloon list and instead only moves the pages whose status
indicated failure into a refuse_list on the stack. Once the operation
completes, the inflation or deflation functions handle the list
appropriately.
As we do that, we can consolidate the communication with the hypervisor
for both the lock and unlock operations into a single function. We also
reuse the deflation function for popping the balloon.
As a preparation for preventing races, we hold a spinlock when the
communication actually takes place, and use atomic operations for
updating the balloon size. The balloon page list is still racy and will
be handled in the next patch.
Reviewed-by: Xavier Deguillard <xdeguillard@vmware.com>
Signed-off-by: Nadav Amit <namit@vmware.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2018-09-21 01:30:18 +08:00
|
|
|
|
|
|
|
/**
|
|
|
|
* @comm_lock: lock to protect the communication with the host.
|
|
|
|
*
|
|
|
|
* Lock ordering: @conf_sem -> @comm_lock .
|
|
|
|
*/
|
|
|
|
spinlock_t comm_lock;
|
2019-04-25 19:54:44 +08:00
|
|
|
|
|
|
|
/**
|
|
|
|
* @shrinker: shrinker interface that is used to avoid over-inflation.
|
|
|
|
*/
|
2023-09-11 17:44:26 +08:00
|
|
|
struct shrinker *shrinker;
|
2010-04-24 01:18:08 +08:00
|
|
|
};
|
|
|
|
|
|
|
|
static struct vmballoon balloon;
|
|
|
|
|
2018-09-21 01:30:17 +08:00
|
|
|
struct vmballoon_stats {
|
|
|
|
/* timer / doorbell operations */
|
|
|
|
atomic64_t general_stat[VMW_BALLOON_STAT_NUM];
|
|
|
|
|
|
|
|
/* allocation statistics for huge and small pages */
|
|
|
|
atomic64_t
|
|
|
|
page_stat[VMW_BALLOON_PAGE_STAT_NUM][VMW_BALLOON_NUM_PAGE_SIZES];
|
|
|
|
|
|
|
|
/* Monitor operations: total operations, and failures */
|
|
|
|
atomic64_t ops[VMW_BALLOON_CMD_NUM][VMW_BALLOON_OP_STAT_TYPES];
|
|
|
|
};
|
|
|
|
|
|
|
|
static inline bool is_vmballoon_stats_on(void)
|
|
|
|
{
|
|
|
|
return IS_ENABLED(CONFIG_DEBUG_FS) &&
|
|
|
|
static_branch_unlikely(&balloon_stat_enabled);
|
|
|
|
}
|
|
|
|
|
|
|
|
static inline void vmballoon_stats_op_inc(struct vmballoon *b, unsigned int op,
|
|
|
|
enum vmballoon_op_stat_type type)
|
|
|
|
{
|
|
|
|
if (is_vmballoon_stats_on())
|
|
|
|
atomic64_inc(&b->stats->ops[op][type]);
|
|
|
|
}
|
|
|
|
|
|
|
|
static inline void vmballoon_stats_gen_inc(struct vmballoon *b,
|
|
|
|
enum vmballoon_stat_general stat)
|
|
|
|
{
|
|
|
|
if (is_vmballoon_stats_on())
|
|
|
|
atomic64_inc(&b->stats->general_stat[stat]);
|
|
|
|
}
|
|
|
|
|
|
|
|
static inline void vmballoon_stats_gen_add(struct vmballoon *b,
|
|
|
|
enum vmballoon_stat_general stat,
|
|
|
|
unsigned int val)
|
|
|
|
{
|
|
|
|
if (is_vmballoon_stats_on())
|
|
|
|
atomic64_add(val, &b->stats->general_stat[stat]);
|
|
|
|
}
|
|
|
|
|
|
|
|
static inline void vmballoon_stats_page_inc(struct vmballoon *b,
|
|
|
|
enum vmballoon_stat_page stat,
|
vmw_balloon: rework the inflate and deflate loops
In preparation for supporting compaction and OOM notification, this
patch reworks the inflate/deflate loops. The main idea is to separate
the allocation, communication with the hypervisor, and the handling of
errors from each other. Doing will allow us to perform concurrent
inflation and deflation, excluding the actual communication with the
hypervisor.
To do so, we need to get rid of the remaining global state that is kept
in the balloon struct, specifically the refuse_list. When the VM
communicates with the hypervisor, it does not free or put back pages
to the balloon list and instead only moves the pages whose status
indicated failure into a refuse_list on the stack. Once the operation
completes, the inflation or deflation functions handle the list
appropriately.
As we do that, we can consolidate the communication with the hypervisor
for both the lock and unlock operations into a single function. We also
reuse the deflation function for popping the balloon.
As a preparation for preventing races, we hold a spinlock when the
communication actually takes place, and use atomic operations for
updating the balloon size. The balloon page list is still racy and will
be handled in the next patch.
Reviewed-by: Xavier Deguillard <xdeguillard@vmware.com>
Signed-off-by: Nadav Amit <namit@vmware.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2018-09-21 01:30:18 +08:00
|
|
|
enum vmballoon_page_size_type size)
|
2018-09-21 01:30:17 +08:00
|
|
|
{
|
|
|
|
if (is_vmballoon_stats_on())
|
vmw_balloon: rework the inflate and deflate loops
In preparation for supporting compaction and OOM notification, this
patch reworks the inflate/deflate loops. The main idea is to separate
the allocation, communication with the hypervisor, and the handling of
errors from each other. Doing will allow us to perform concurrent
inflation and deflation, excluding the actual communication with the
hypervisor.
To do so, we need to get rid of the remaining global state that is kept
in the balloon struct, specifically the refuse_list. When the VM
communicates with the hypervisor, it does not free or put back pages
to the balloon list and instead only moves the pages whose status
indicated failure into a refuse_list on the stack. Once the operation
completes, the inflation or deflation functions handle the list
appropriately.
As we do that, we can consolidate the communication with the hypervisor
for both the lock and unlock operations into a single function. We also
reuse the deflation function for popping the balloon.
As a preparation for preventing races, we hold a spinlock when the
communication actually takes place, and use atomic operations for
updating the balloon size. The balloon page list is still racy and will
be handled in the next patch.
Reviewed-by: Xavier Deguillard <xdeguillard@vmware.com>
Signed-off-by: Nadav Amit <namit@vmware.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2018-09-21 01:30:18 +08:00
|
|
|
atomic64_inc(&b->stats->page_stat[stat][size]);
|
|
|
|
}
|
|
|
|
|
|
|
|
static inline void vmballoon_stats_page_add(struct vmballoon *b,
|
|
|
|
enum vmballoon_stat_page stat,
|
|
|
|
enum vmballoon_page_size_type size,
|
|
|
|
unsigned int val)
|
|
|
|
{
|
|
|
|
if (is_vmballoon_stats_on())
|
|
|
|
atomic64_add(val, &b->stats->page_stat[stat][size]);
|
2018-09-21 01:30:17 +08:00
|
|
|
}
|
|
|
|
|
2018-09-21 01:30:07 +08:00
|
|
|
static inline unsigned long
|
|
|
|
__vmballoon_cmd(struct vmballoon *b, unsigned long cmd, unsigned long arg1,
|
|
|
|
unsigned long arg2, unsigned long *result)
|
|
|
|
{
|
|
|
|
unsigned long status, dummy1, dummy2, dummy3, local_result;
|
|
|
|
|
2018-09-21 01:30:17 +08:00
|
|
|
vmballoon_stats_op_inc(b, cmd, VMW_BALLOON_OP_STAT);
|
2018-09-21 01:30:08 +08:00
|
|
|
|
2018-09-21 01:30:07 +08:00
|
|
|
asm volatile ("inl %%dx" :
|
|
|
|
"=a"(status),
|
|
|
|
"=c"(dummy1),
|
|
|
|
"=d"(dummy2),
|
|
|
|
"=b"(local_result),
|
|
|
|
"=S"(dummy3) :
|
|
|
|
"0"(VMW_BALLOON_HV_MAGIC),
|
|
|
|
"1"(cmd),
|
|
|
|
"2"(VMW_BALLOON_HV_PORT),
|
|
|
|
"3"(arg1),
|
|
|
|
"4"(arg2) :
|
|
|
|
"memory");
|
|
|
|
|
|
|
|
/* update the result if needed */
|
|
|
|
if (result)
|
|
|
|
*result = (cmd == VMW_BALLOON_CMD_START) ? dummy1 :
|
|
|
|
local_result;
|
|
|
|
|
|
|
|
/* update target when applicable */
|
|
|
|
if (status == VMW_BALLOON_SUCCESS &&
|
|
|
|
((1ul << cmd) & VMW_BALLOON_CMD_WITH_TARGET_MASK))
|
vmw_balloon: rework the inflate and deflate loops
In preparation for supporting compaction and OOM notification, this
patch reworks the inflate/deflate loops. The main idea is to separate
the allocation, communication with the hypervisor, and the handling of
errors from each other. Doing will allow us to perform concurrent
inflation and deflation, excluding the actual communication with the
hypervisor.
To do so, we need to get rid of the remaining global state that is kept
in the balloon struct, specifically the refuse_list. When the VM
communicates with the hypervisor, it does not free or put back pages
to the balloon list and instead only moves the pages whose status
indicated failure into a refuse_list on the stack. Once the operation
completes, the inflation or deflation functions handle the list
appropriately.
As we do that, we can consolidate the communication with the hypervisor
for both the lock and unlock operations into a single function. We also
reuse the deflation function for popping the balloon.
As a preparation for preventing races, we hold a spinlock when the
communication actually takes place, and use atomic operations for
updating the balloon size. The balloon page list is still racy and will
be handled in the next patch.
Reviewed-by: Xavier Deguillard <xdeguillard@vmware.com>
Signed-off-by: Nadav Amit <namit@vmware.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2018-09-21 01:30:18 +08:00
|
|
|
WRITE_ONCE(b->target, local_result);
|
2018-09-21 01:30:07 +08:00
|
|
|
|
2018-09-21 01:30:08 +08:00
|
|
|
if (status != VMW_BALLOON_SUCCESS &&
|
|
|
|
status != VMW_BALLOON_SUCCESS_WITH_CAPABILITIES) {
|
2018-09-21 01:30:17 +08:00
|
|
|
vmballoon_stats_op_inc(b, cmd, VMW_BALLOON_OP_FAIL_STAT);
|
2018-09-21 01:30:08 +08:00
|
|
|
pr_debug("%s: %s [0x%lx,0x%lx) failed, returned %ld\n",
|
|
|
|
__func__, vmballoon_cmd_names[cmd], arg1, arg2,
|
|
|
|
status);
|
|
|
|
}
|
|
|
|
|
2018-09-21 01:30:07 +08:00
|
|
|
/* mark reset required accordingly */
|
|
|
|
if (status == VMW_BALLOON_ERROR_RESET)
|
|
|
|
b->reset_required = true;
|
|
|
|
|
|
|
|
return status;
|
|
|
|
}
|
|
|
|
|
|
|
|
static __always_inline unsigned long
|
|
|
|
vmballoon_cmd(struct vmballoon *b, unsigned long cmd, unsigned long arg1,
|
|
|
|
unsigned long arg2)
|
|
|
|
{
|
|
|
|
unsigned long dummy;
|
|
|
|
|
|
|
|
return __vmballoon_cmd(b, cmd, arg1, arg2, &dummy);
|
|
|
|
}
|
|
|
|
|
2010-04-24 01:18:08 +08:00
|
|
|
/*
|
|
|
|
* Send "start" command to the host, communicating supported version
|
|
|
|
* of the protocol.
|
|
|
|
*/
|
2018-09-21 01:30:19 +08:00
|
|
|
static int vmballoon_send_start(struct vmballoon *b, unsigned long req_caps)
|
2010-04-24 01:18:08 +08:00
|
|
|
{
|
2018-09-21 01:30:07 +08:00
|
|
|
unsigned long status, capabilities;
|
2010-04-24 01:18:08 +08:00
|
|
|
|
2018-09-21 01:30:07 +08:00
|
|
|
status = __vmballoon_cmd(b, VMW_BALLOON_CMD_START, req_caps, 0,
|
|
|
|
&capabilities);
|
2015-08-07 06:17:58 +08:00
|
|
|
|
|
|
|
switch (status) {
|
|
|
|
case VMW_BALLOON_SUCCESS_WITH_CAPABILITIES:
|
|
|
|
b->capabilities = capabilities;
|
2015-08-07 06:18:01 +08:00
|
|
|
break;
|
2015-08-07 06:17:58 +08:00
|
|
|
case VMW_BALLOON_SUCCESS:
|
|
|
|
b->capabilities = VMW_BALLOON_BASIC_CMDS;
|
2015-08-07 06:18:01 +08:00
|
|
|
break;
|
|
|
|
default:
|
2018-09-21 01:30:19 +08:00
|
|
|
return -EIO;
|
2015-08-07 06:17:58 +08:00
|
|
|
}
|
2010-04-24 01:18:08 +08:00
|
|
|
|
2018-06-20 07:00:25 +08:00
|
|
|
/*
|
|
|
|
* 2MB pages are only supported with batching. If batching is for some
|
|
|
|
* reason disabled, do not use 2MB pages, since otherwise the legacy
|
|
|
|
* mechanism is used with 2MB pages, causing a failure.
|
|
|
|
*/
|
vmw_balloon: rework the inflate and deflate loops
In preparation for supporting compaction and OOM notification, this
patch reworks the inflate/deflate loops. The main idea is to separate
the allocation, communication with the hypervisor, and the handling of
errors from each other. Doing will allow us to perform concurrent
inflation and deflation, excluding the actual communication with the
hypervisor.
To do so, we need to get rid of the remaining global state that is kept
in the balloon struct, specifically the refuse_list. When the VM
communicates with the hypervisor, it does not free or put back pages
to the balloon list and instead only moves the pages whose status
indicated failure into a refuse_list on the stack. Once the operation
completes, the inflation or deflation functions handle the list
appropriately.
As we do that, we can consolidate the communication with the hypervisor
for both the lock and unlock operations into a single function. We also
reuse the deflation function for popping the balloon.
As a preparation for preventing races, we hold a spinlock when the
communication actually takes place, and use atomic operations for
updating the balloon size. The balloon page list is still racy and will
be handled in the next patch.
Reviewed-by: Xavier Deguillard <xdeguillard@vmware.com>
Signed-off-by: Nadav Amit <namit@vmware.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2018-09-21 01:30:18 +08:00
|
|
|
b->max_page_size = VMW_BALLOON_4K_PAGE;
|
2018-06-20 07:00:25 +08:00
|
|
|
if ((b->capabilities & VMW_BALLOON_BATCHED_2M_CMDS) &&
|
|
|
|
(b->capabilities & VMW_BALLOON_BATCHED_CMDS))
|
vmw_balloon: rework the inflate and deflate loops
In preparation for supporting compaction and OOM notification, this
patch reworks the inflate/deflate loops. The main idea is to separate
the allocation, communication with the hypervisor, and the handling of
errors from each other. Doing will allow us to perform concurrent
inflation and deflation, excluding the actual communication with the
hypervisor.
To do so, we need to get rid of the remaining global state that is kept
in the balloon struct, specifically the refuse_list. When the VM
communicates with the hypervisor, it does not free or put back pages
to the balloon list and instead only moves the pages whose status
indicated failure into a refuse_list on the stack. Once the operation
completes, the inflation or deflation functions handle the list
appropriately.
As we do that, we can consolidate the communication with the hypervisor
for both the lock and unlock operations into a single function. We also
reuse the deflation function for popping the balloon.
As a preparation for preventing races, we hold a spinlock when the
communication actually takes place, and use atomic operations for
updating the balloon size. The balloon page list is still racy and will
be handled in the next patch.
Reviewed-by: Xavier Deguillard <xdeguillard@vmware.com>
Signed-off-by: Nadav Amit <namit@vmware.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2018-09-21 01:30:18 +08:00
|
|
|
b->max_page_size = VMW_BALLOON_2M_PAGE;
|
|
|
|
|
2015-08-07 06:18:01 +08:00
|
|
|
|
2018-09-21 01:30:19 +08:00
|
|
|
return 0;
|
2010-04-24 01:18:08 +08:00
|
|
|
}
|
|
|
|
|
2018-09-21 01:30:19 +08:00
|
|
|
/**
|
|
|
|
* vmballoon_send_guest_id - communicate guest type to the host.
|
|
|
|
*
|
|
|
|
* @b: pointer to the balloon.
|
|
|
|
*
|
2010-04-24 01:18:08 +08:00
|
|
|
* Communicate guest type to the host so that it can adjust ballooning
|
|
|
|
* algorithm to the one most appropriate for the guest. This command
|
|
|
|
* is normally issued after sending "start" command and is part of
|
|
|
|
* standard reset sequence.
|
2018-09-21 01:30:19 +08:00
|
|
|
*
|
|
|
|
* Return: zero on success or appropriate error code.
|
2010-04-24 01:18:08 +08:00
|
|
|
*/
|
2018-09-21 01:30:19 +08:00
|
|
|
static int vmballoon_send_guest_id(struct vmballoon *b)
|
2010-04-24 01:18:08 +08:00
|
|
|
{
|
2018-09-21 01:30:07 +08:00
|
|
|
unsigned long status;
|
2010-04-24 01:18:08 +08:00
|
|
|
|
2018-09-21 01:30:07 +08:00
|
|
|
status = vmballoon_cmd(b, VMW_BALLOON_CMD_GUEST_ID,
|
|
|
|
VMW_BALLOON_GUEST_ID, 0);
|
2010-04-24 01:18:08 +08:00
|
|
|
|
2018-09-21 01:30:19 +08:00
|
|
|
return status == VMW_BALLOON_SUCCESS ? 0 : -EIO;
|
2010-04-24 01:18:08 +08:00
|
|
|
}
|
|
|
|
|
vmw_balloon: rework the inflate and deflate loops
In preparation for supporting compaction and OOM notification, this
patch reworks the inflate/deflate loops. The main idea is to separate
the allocation, communication with the hypervisor, and the handling of
errors from each other. Doing will allow us to perform concurrent
inflation and deflation, excluding the actual communication with the
hypervisor.
To do so, we need to get rid of the remaining global state that is kept
in the balloon struct, specifically the refuse_list. When the VM
communicates with the hypervisor, it does not free or put back pages
to the balloon list and instead only moves the pages whose status
indicated failure into a refuse_list on the stack. Once the operation
completes, the inflation or deflation functions handle the list
appropriately.
As we do that, we can consolidate the communication with the hypervisor
for both the lock and unlock operations into a single function. We also
reuse the deflation function for popping the balloon.
As a preparation for preventing races, we hold a spinlock when the
communication actually takes place, and use atomic operations for
updating the balloon size. The balloon page list is still racy and will
be handled in the next patch.
Reviewed-by: Xavier Deguillard <xdeguillard@vmware.com>
Signed-off-by: Nadav Amit <namit@vmware.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2018-09-21 01:30:18 +08:00
|
|
|
/**
|
|
|
|
* vmballoon_page_order() - return the order of the page
|
|
|
|
* @page_size: the size of the page.
|
|
|
|
*
|
|
|
|
* Return: the allocation order.
|
|
|
|
*/
|
|
|
|
static inline
|
|
|
|
unsigned int vmballoon_page_order(enum vmballoon_page_size_type page_size)
|
2015-08-07 06:18:01 +08:00
|
|
|
{
|
vmw_balloon: rework the inflate and deflate loops
In preparation for supporting compaction and OOM notification, this
patch reworks the inflate/deflate loops. The main idea is to separate
the allocation, communication with the hypervisor, and the handling of
errors from each other. Doing will allow us to perform concurrent
inflation and deflation, excluding the actual communication with the
hypervisor.
To do so, we need to get rid of the remaining global state that is kept
in the balloon struct, specifically the refuse_list. When the VM
communicates with the hypervisor, it does not free or put back pages
to the balloon list and instead only moves the pages whose status
indicated failure into a refuse_list on the stack. Once the operation
completes, the inflation or deflation functions handle the list
appropriately.
As we do that, we can consolidate the communication with the hypervisor
for both the lock and unlock operations into a single function. We also
reuse the deflation function for popping the balloon.
As a preparation for preventing races, we hold a spinlock when the
communication actually takes place, and use atomic operations for
updating the balloon size. The balloon page list is still racy and will
be handled in the next patch.
Reviewed-by: Xavier Deguillard <xdeguillard@vmware.com>
Signed-off-by: Nadav Amit <namit@vmware.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2018-09-21 01:30:18 +08:00
|
|
|
return page_size == VMW_BALLOON_2M_PAGE ? VMW_BALLOON_2M_ORDER : 0;
|
|
|
|
}
|
2015-08-07 06:18:01 +08:00
|
|
|
|
vmw_balloon: rework the inflate and deflate loops
In preparation for supporting compaction and OOM notification, this
patch reworks the inflate/deflate loops. The main idea is to separate
the allocation, communication with the hypervisor, and the handling of
errors from each other. Doing will allow us to perform concurrent
inflation and deflation, excluding the actual communication with the
hypervisor.
To do so, we need to get rid of the remaining global state that is kept
in the balloon struct, specifically the refuse_list. When the VM
communicates with the hypervisor, it does not free or put back pages
to the balloon list and instead only moves the pages whose status
indicated failure into a refuse_list on the stack. Once the operation
completes, the inflation or deflation functions handle the list
appropriately.
As we do that, we can consolidate the communication with the hypervisor
for both the lock and unlock operations into a single function. We also
reuse the deflation function for popping the balloon.
As a preparation for preventing races, we hold a spinlock when the
communication actually takes place, and use atomic operations for
updating the balloon size. The balloon page list is still racy and will
be handled in the next patch.
Reviewed-by: Xavier Deguillard <xdeguillard@vmware.com>
Signed-off-by: Nadav Amit <namit@vmware.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2018-09-21 01:30:18 +08:00
|
|
|
/**
|
|
|
|
* vmballoon_page_in_frames() - returns the number of frames in a page.
|
|
|
|
* @page_size: the size of the page.
|
|
|
|
*
|
|
|
|
* Return: the number of 4k frames.
|
|
|
|
*/
|
|
|
|
static inline unsigned int
|
|
|
|
vmballoon_page_in_frames(enum vmballoon_page_size_type page_size)
|
|
|
|
{
|
|
|
|
return 1 << vmballoon_page_order(page_size);
|
2015-08-07 06:18:01 +08:00
|
|
|
}
|
|
|
|
|
2019-03-06 07:42:41 +08:00
|
|
|
/**
|
|
|
|
* vmballoon_mark_page_offline() - mark a page as offline
|
|
|
|
* @page: pointer for the page.
|
|
|
|
* @page_size: the size of the page.
|
|
|
|
*/
|
|
|
|
static void
|
|
|
|
vmballoon_mark_page_offline(struct page *page,
|
|
|
|
enum vmballoon_page_size_type page_size)
|
|
|
|
{
|
|
|
|
int i;
|
|
|
|
|
|
|
|
for (i = 0; i < vmballoon_page_in_frames(page_size); i++)
|
|
|
|
__SetPageOffline(page + i);
|
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
|
|
|
* vmballoon_mark_page_online() - mark a page as online
|
|
|
|
* @page: pointer for the page.
|
|
|
|
* @page_size: the size of the page.
|
|
|
|
*/
|
|
|
|
static void
|
|
|
|
vmballoon_mark_page_online(struct page *page,
|
|
|
|
enum vmballoon_page_size_type page_size)
|
|
|
|
{
|
|
|
|
int i;
|
|
|
|
|
|
|
|
for (i = 0; i < vmballoon_page_in_frames(page_size); i++)
|
|
|
|
__ClearPageOffline(page + i);
|
|
|
|
}
|
|
|
|
|
2018-09-21 01:30:16 +08:00
|
|
|
/**
|
|
|
|
* vmballoon_send_get_target() - Retrieve desired balloon size from the host.
|
|
|
|
*
|
|
|
|
* @b: pointer to the balloon.
|
|
|
|
*
|
|
|
|
* Return: zero on success, EINVAL if limit does not fit in 32-bit, as required
|
|
|
|
* by the host-guest protocol and EIO if an error occurred in communicating with
|
|
|
|
* the host.
|
2010-04-24 01:18:08 +08:00
|
|
|
*/
|
2018-09-21 01:30:16 +08:00
|
|
|
static int vmballoon_send_get_target(struct vmballoon *b)
|
2010-04-24 01:18:08 +08:00
|
|
|
{
|
|
|
|
unsigned long status;
|
|
|
|
unsigned long limit;
|
|
|
|
|
2018-12-28 16:34:29 +08:00
|
|
|
limit = totalram_pages();
|
2010-04-24 01:18:08 +08:00
|
|
|
|
2019-02-07 07:57:02 +08:00
|
|
|
/* Ensure limit fits in 32-bits if 64-bit targets are not supported */
|
|
|
|
if (!(b->capabilities & VMW_BALLOON_64_BIT_TARGET) &&
|
|
|
|
limit != (u32)limit)
|
2018-09-21 01:30:16 +08:00
|
|
|
return -EINVAL;
|
2010-04-24 01:18:08 +08:00
|
|
|
|
2018-09-21 01:30:07 +08:00
|
|
|
status = vmballoon_cmd(b, VMW_BALLOON_CMD_GET_TARGET, limit, 0);
|
|
|
|
|
2018-09-21 01:30:16 +08:00
|
|
|
return status == VMW_BALLOON_SUCCESS ? 0 : -EIO;
|
2010-04-24 01:18:08 +08:00
|
|
|
}
|
|
|
|
|
vmw_balloon: rework the inflate and deflate loops
In preparation for supporting compaction and OOM notification, this
patch reworks the inflate/deflate loops. The main idea is to separate
the allocation, communication with the hypervisor, and the handling of
errors from each other. Doing will allow us to perform concurrent
inflation and deflation, excluding the actual communication with the
hypervisor.
To do so, we need to get rid of the remaining global state that is kept
in the balloon struct, specifically the refuse_list. When the VM
communicates with the hypervisor, it does not free or put back pages
to the balloon list and instead only moves the pages whose status
indicated failure into a refuse_list on the stack. Once the operation
completes, the inflation or deflation functions handle the list
appropriately.
As we do that, we can consolidate the communication with the hypervisor
for both the lock and unlock operations into a single function. We also
reuse the deflation function for popping the balloon.
As a preparation for preventing races, we hold a spinlock when the
communication actually takes place, and use atomic operations for
updating the balloon size. The balloon page list is still racy and will
be handled in the next patch.
Reviewed-by: Xavier Deguillard <xdeguillard@vmware.com>
Signed-off-by: Nadav Amit <namit@vmware.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2018-09-21 01:30:18 +08:00
|
|
|
/**
|
|
|
|
* vmballoon_alloc_page_list - allocates a list of pages.
|
|
|
|
*
|
|
|
|
* @b: pointer to the balloon.
|
|
|
|
* @ctl: pointer for the %struct vmballoon_ctl, which defines the operation.
|
|
|
|
* @req_n_pages: the number of requested pages.
|
|
|
|
*
|
|
|
|
* Tries to allocate @req_n_pages. Add them to the list of balloon pages in
|
|
|
|
* @ctl.pages and updates @ctl.n_pages to reflect the number of pages.
|
|
|
|
*
|
|
|
|
* Return: zero on success or error code otherwise.
|
|
|
|
*/
|
|
|
|
static int vmballoon_alloc_page_list(struct vmballoon *b,
|
|
|
|
struct vmballoon_ctl *ctl,
|
|
|
|
unsigned int req_n_pages)
|
2015-08-07 06:18:01 +08:00
|
|
|
{
|
vmw_balloon: rework the inflate and deflate loops
In preparation for supporting compaction and OOM notification, this
patch reworks the inflate/deflate loops. The main idea is to separate
the allocation, communication with the hypervisor, and the handling of
errors from each other. Doing will allow us to perform concurrent
inflation and deflation, excluding the actual communication with the
hypervisor.
To do so, we need to get rid of the remaining global state that is kept
in the balloon struct, specifically the refuse_list. When the VM
communicates with the hypervisor, it does not free or put back pages
to the balloon list and instead only moves the pages whose status
indicated failure into a refuse_list on the stack. Once the operation
completes, the inflation or deflation functions handle the list
appropriately.
As we do that, we can consolidate the communication with the hypervisor
for both the lock and unlock operations into a single function. We also
reuse the deflation function for popping the balloon.
As a preparation for preventing races, we hold a spinlock when the
communication actually takes place, and use atomic operations for
updating the balloon size. The balloon page list is still racy and will
be handled in the next patch.
Reviewed-by: Xavier Deguillard <xdeguillard@vmware.com>
Signed-off-by: Nadav Amit <namit@vmware.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2018-09-21 01:30:18 +08:00
|
|
|
struct page *page;
|
|
|
|
unsigned int i;
|
2015-08-07 06:18:01 +08:00
|
|
|
|
vmw_balloon: rework the inflate and deflate loops
In preparation for supporting compaction and OOM notification, this
patch reworks the inflate/deflate loops. The main idea is to separate
the allocation, communication with the hypervisor, and the handling of
errors from each other. Doing will allow us to perform concurrent
inflation and deflation, excluding the actual communication with the
hypervisor.
To do so, we need to get rid of the remaining global state that is kept
in the balloon struct, specifically the refuse_list. When the VM
communicates with the hypervisor, it does not free or put back pages
to the balloon list and instead only moves the pages whose status
indicated failure into a refuse_list on the stack. Once the operation
completes, the inflation or deflation functions handle the list
appropriately.
As we do that, we can consolidate the communication with the hypervisor
for both the lock and unlock operations into a single function. We also
reuse the deflation function for popping the balloon.
As a preparation for preventing races, we hold a spinlock when the
communication actually takes place, and use atomic operations for
updating the balloon size. The balloon page list is still racy and will
be handled in the next patch.
Reviewed-by: Xavier Deguillard <xdeguillard@vmware.com>
Signed-off-by: Nadav Amit <namit@vmware.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2018-09-21 01:30:18 +08:00
|
|
|
for (i = 0; i < req_n_pages; i++) {
|
2019-04-25 19:54:45 +08:00
|
|
|
/*
|
|
|
|
* First check if we happen to have pages that were allocated
|
|
|
|
* before. This happens when 2MB page rejected during inflation
|
|
|
|
* by the hypervisor, and then split into 4KB pages.
|
|
|
|
*/
|
|
|
|
if (!list_empty(&ctl->prealloc_pages)) {
|
|
|
|
page = list_first_entry(&ctl->prealloc_pages,
|
|
|
|
struct page, lru);
|
|
|
|
list_del(&page->lru);
|
|
|
|
} else {
|
|
|
|
if (ctl->page_size == VMW_BALLOON_2M_PAGE)
|
|
|
|
page = alloc_pages(__GFP_HIGHMEM|__GFP_NOWARN|
|
2019-04-25 19:54:43 +08:00
|
|
|
__GFP_NOMEMALLOC, VMW_BALLOON_2M_ORDER);
|
2019-04-25 19:54:45 +08:00
|
|
|
else
|
|
|
|
page = balloon_page_alloc();
|
2015-08-07 06:18:01 +08:00
|
|
|
|
2019-04-25 19:54:45 +08:00
|
|
|
vmballoon_stats_page_inc(b, VMW_BALLOON_PAGE_STAT_ALLOC,
|
|
|
|
ctl->page_size);
|
|
|
|
}
|
vmw_balloon: rework the inflate and deflate loops
In preparation for supporting compaction and OOM notification, this
patch reworks the inflate/deflate loops. The main idea is to separate
the allocation, communication with the hypervisor, and the handling of
errors from each other. Doing will allow us to perform concurrent
inflation and deflation, excluding the actual communication with the
hypervisor.
To do so, we need to get rid of the remaining global state that is kept
in the balloon struct, specifically the refuse_list. When the VM
communicates with the hypervisor, it does not free or put back pages
to the balloon list and instead only moves the pages whose status
indicated failure into a refuse_list on the stack. Once the operation
completes, the inflation or deflation functions handle the list
appropriately.
As we do that, we can consolidate the communication with the hypervisor
for both the lock and unlock operations into a single function. We also
reuse the deflation function for popping the balloon.
As a preparation for preventing races, we hold a spinlock when the
communication actually takes place, and use atomic operations for
updating the balloon size. The balloon page list is still racy and will
be handled in the next patch.
Reviewed-by: Xavier Deguillard <xdeguillard@vmware.com>
Signed-off-by: Nadav Amit <namit@vmware.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2018-09-21 01:30:18 +08:00
|
|
|
|
|
|
|
if (page) {
|
|
|
|
/* Success. Add the page to the list and continue. */
|
|
|
|
list_add(&page->lru, &ctl->pages);
|
|
|
|
continue;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* Allocation failed. Update statistics and stop. */
|
|
|
|
vmballoon_stats_page_inc(b, VMW_BALLOON_PAGE_STAT_ALLOC_FAIL,
|
|
|
|
ctl->page_size);
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
|
|
|
|
ctl->n_pages = i;
|
|
|
|
|
|
|
|
return req_n_pages == ctl->n_pages ? 0 : -ENOMEM;
|
2015-08-07 06:18:01 +08:00
|
|
|
}
|
|
|
|
|
vmw_balloon: rework the inflate and deflate loops
In preparation for supporting compaction and OOM notification, this
patch reworks the inflate/deflate loops. The main idea is to separate
the allocation, communication with the hypervisor, and the handling of
errors from each other. Doing will allow us to perform concurrent
inflation and deflation, excluding the actual communication with the
hypervisor.
To do so, we need to get rid of the remaining global state that is kept
in the balloon struct, specifically the refuse_list. When the VM
communicates with the hypervisor, it does not free or put back pages
to the balloon list and instead only moves the pages whose status
indicated failure into a refuse_list on the stack. Once the operation
completes, the inflation or deflation functions handle the list
appropriately.
As we do that, we can consolidate the communication with the hypervisor
for both the lock and unlock operations into a single function. We also
reuse the deflation function for popping the balloon.
As a preparation for preventing races, we hold a spinlock when the
communication actually takes place, and use atomic operations for
updating the balloon size. The balloon page list is still racy and will
be handled in the next patch.
Reviewed-by: Xavier Deguillard <xdeguillard@vmware.com>
Signed-off-by: Nadav Amit <namit@vmware.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2018-09-21 01:30:18 +08:00
|
|
|
/**
|
|
|
|
* vmballoon_handle_one_result - Handle lock/unlock result for a single page.
|
|
|
|
*
|
|
|
|
* @b: pointer for %struct vmballoon.
|
|
|
|
* @page: pointer for the page whose result should be handled.
|
|
|
|
* @page_size: size of the page.
|
|
|
|
* @status: status of the operation as provided by the hypervisor.
|
2010-04-24 01:18:08 +08:00
|
|
|
*/
|
vmw_balloon: rework the inflate and deflate loops
In preparation for supporting compaction and OOM notification, this
patch reworks the inflate/deflate loops. The main idea is to separate
the allocation, communication with the hypervisor, and the handling of
errors from each other. Doing will allow us to perform concurrent
inflation and deflation, excluding the actual communication with the
hypervisor.
To do so, we need to get rid of the remaining global state that is kept
in the balloon struct, specifically the refuse_list. When the VM
communicates with the hypervisor, it does not free or put back pages
to the balloon list and instead only moves the pages whose status
indicated failure into a refuse_list on the stack. Once the operation
completes, the inflation or deflation functions handle the list
appropriately.
As we do that, we can consolidate the communication with the hypervisor
for both the lock and unlock operations into a single function. We also
reuse the deflation function for popping the balloon.
As a preparation for preventing races, we hold a spinlock when the
communication actually takes place, and use atomic operations for
updating the balloon size. The balloon page list is still racy and will
be handled in the next patch.
Reviewed-by: Xavier Deguillard <xdeguillard@vmware.com>
Signed-off-by: Nadav Amit <namit@vmware.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2018-09-21 01:30:18 +08:00
|
|
|
static int vmballoon_handle_one_result(struct vmballoon *b, struct page *page,
|
|
|
|
enum vmballoon_page_size_type page_size,
|
|
|
|
unsigned long status)
|
2010-04-24 01:18:08 +08:00
|
|
|
{
|
vmw_balloon: rework the inflate and deflate loops
In preparation for supporting compaction and OOM notification, this
patch reworks the inflate/deflate loops. The main idea is to separate
the allocation, communication with the hypervisor, and the handling of
errors from each other. Doing will allow us to perform concurrent
inflation and deflation, excluding the actual communication with the
hypervisor.
To do so, we need to get rid of the remaining global state that is kept
in the balloon struct, specifically the refuse_list. When the VM
communicates with the hypervisor, it does not free or put back pages
to the balloon list and instead only moves the pages whose status
indicated failure into a refuse_list on the stack. Once the operation
completes, the inflation or deflation functions handle the list
appropriately.
As we do that, we can consolidate the communication with the hypervisor
for both the lock and unlock operations into a single function. We also
reuse the deflation function for popping the balloon.
As a preparation for preventing races, we hold a spinlock when the
communication actually takes place, and use atomic operations for
updating the balloon size. The balloon page list is still racy and will
be handled in the next patch.
Reviewed-by: Xavier Deguillard <xdeguillard@vmware.com>
Signed-off-by: Nadav Amit <namit@vmware.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2018-09-21 01:30:18 +08:00
|
|
|
/* On success do nothing. The page is already on the balloon list. */
|
|
|
|
if (likely(status == VMW_BALLOON_SUCCESS))
|
|
|
|
return 0;
|
2010-04-24 01:18:08 +08:00
|
|
|
|
vmw_balloon: rework the inflate and deflate loops
In preparation for supporting compaction and OOM notification, this
patch reworks the inflate/deflate loops. The main idea is to separate
the allocation, communication with the hypervisor, and the handling of
errors from each other. Doing will allow us to perform concurrent
inflation and deflation, excluding the actual communication with the
hypervisor.
To do so, we need to get rid of the remaining global state that is kept
in the balloon struct, specifically the refuse_list. When the VM
communicates with the hypervisor, it does not free or put back pages
to the balloon list and instead only moves the pages whose status
indicated failure into a refuse_list on the stack. Once the operation
completes, the inflation or deflation functions handle the list
appropriately.
As we do that, we can consolidate the communication with the hypervisor
for both the lock and unlock operations into a single function. We also
reuse the deflation function for popping the balloon.
As a preparation for preventing races, we hold a spinlock when the
communication actually takes place, and use atomic operations for
updating the balloon size. The balloon page list is still racy and will
be handled in the next patch.
Reviewed-by: Xavier Deguillard <xdeguillard@vmware.com>
Signed-off-by: Nadav Amit <namit@vmware.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2018-09-21 01:30:18 +08:00
|
|
|
pr_debug("%s: failed comm pfn %lx status %lu page_size %s\n", __func__,
|
|
|
|
page_to_pfn(page), status,
|
|
|
|
vmballoon_page_size_names[page_size]);
|
|
|
|
|
|
|
|
/* Error occurred */
|
|
|
|
vmballoon_stats_page_inc(b, VMW_BALLOON_PAGE_STAT_REFUSED_ALLOC,
|
|
|
|
page_size);
|
|
|
|
|
|
|
|
return -EIO;
|
2010-04-24 01:18:08 +08:00
|
|
|
}
|
|
|
|
|
vmw_balloon: change batch/single lock abstractions
The current abstractions for batch vs single operations seem suboptimal
and complicate the implementation of additional features (OOM,
compaction).
The immediate problem of the current abstractions is that they cause
differences in how operations are handled when batching is on or off.
For example, the refused_alloc counter is not updated when batching is
on. These discrepancies are caused by code redundancies.
Instead, this patch presents three type of operations, according to
whether batching is on or off: (1) add page, (2) communication with the
hypervisor and (3) retrieving the status of a page.
To avoid the overhead of virtual functions, and since we do not expect
additional interfaces for communication with the hypervisor, we use
static keys instead of virtual functions.
Finally, while we are at it, change vmballoon_init_batching() to return
int instead of bool, to be consistent in the return type and avoid
potential coding errors.
Reviewed-by: Xavier Deguillard <xdeguillard@vmware.com>
Signed-off-by: Nadav Amit <namit@vmware.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2018-09-21 01:30:12 +08:00
|
|
|
/**
|
|
|
|
* vmballoon_status_page - returns the status of (un)lock operation
|
|
|
|
*
|
|
|
|
* @b: pointer to the balloon.
|
|
|
|
* @idx: index for the page for which the operation is performed.
|
|
|
|
* @p: pointer to where the page struct is returned.
|
|
|
|
*
|
|
|
|
* Following a lock or unlock operation, returns the status of the operation for
|
|
|
|
* an individual page. Provides the page that the operation was performed on on
|
|
|
|
* the @page argument.
|
|
|
|
*
|
|
|
|
* Returns: The status of a lock or unlock operation for an individual page.
|
2010-04-24 01:18:08 +08:00
|
|
|
*/
|
vmw_balloon: change batch/single lock abstractions
The current abstractions for batch vs single operations seem suboptimal
and complicate the implementation of additional features (OOM,
compaction).
The immediate problem of the current abstractions is that they cause
differences in how operations are handled when batching is on or off.
For example, the refused_alloc counter is not updated when batching is
on. These discrepancies are caused by code redundancies.
Instead, this patch presents three type of operations, according to
whether batching is on or off: (1) add page, (2) communication with the
hypervisor and (3) retrieving the status of a page.
To avoid the overhead of virtual functions, and since we do not expect
additional interfaces for communication with the hypervisor, we use
static keys instead of virtual functions.
Finally, while we are at it, change vmballoon_init_batching() to return
int instead of bool, to be consistent in the return type and avoid
potential coding errors.
Reviewed-by: Xavier Deguillard <xdeguillard@vmware.com>
Signed-off-by: Nadav Amit <namit@vmware.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2018-09-21 01:30:12 +08:00
|
|
|
static unsigned long vmballoon_status_page(struct vmballoon *b, int idx,
|
|
|
|
struct page **p)
|
2010-04-24 01:18:08 +08:00
|
|
|
{
|
vmw_balloon: change batch/single lock abstractions
The current abstractions for batch vs single operations seem suboptimal
and complicate the implementation of additional features (OOM,
compaction).
The immediate problem of the current abstractions is that they cause
differences in how operations are handled when batching is on or off.
For example, the refused_alloc counter is not updated when batching is
on. These discrepancies are caused by code redundancies.
Instead, this patch presents three type of operations, according to
whether batching is on or off: (1) add page, (2) communication with the
hypervisor and (3) retrieving the status of a page.
To avoid the overhead of virtual functions, and since we do not expect
additional interfaces for communication with the hypervisor, we use
static keys instead of virtual functions.
Finally, while we are at it, change vmballoon_init_batching() to return
int instead of bool, to be consistent in the return type and avoid
potential coding errors.
Reviewed-by: Xavier Deguillard <xdeguillard@vmware.com>
Signed-off-by: Nadav Amit <namit@vmware.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2018-09-21 01:30:12 +08:00
|
|
|
if (static_branch_likely(&vmw_balloon_batching)) {
|
|
|
|
/* batching mode */
|
|
|
|
*p = pfn_to_page(b->batch_page[idx].pfn);
|
|
|
|
return b->batch_page[idx].status;
|
|
|
|
}
|
2018-09-21 01:30:07 +08:00
|
|
|
|
vmw_balloon: change batch/single lock abstractions
The current abstractions for batch vs single operations seem suboptimal
and complicate the implementation of additional features (OOM,
compaction).
The immediate problem of the current abstractions is that they cause
differences in how operations are handled when batching is on or off.
For example, the refused_alloc counter is not updated when batching is
on. These discrepancies are caused by code redundancies.
Instead, this patch presents three type of operations, according to
whether batching is on or off: (1) add page, (2) communication with the
hypervisor and (3) retrieving the status of a page.
To avoid the overhead of virtual functions, and since we do not expect
additional interfaces for communication with the hypervisor, we use
static keys instead of virtual functions.
Finally, while we are at it, change vmballoon_init_batching() to return
int instead of bool, to be consistent in the return type and avoid
potential coding errors.
Reviewed-by: Xavier Deguillard <xdeguillard@vmware.com>
Signed-off-by: Nadav Amit <namit@vmware.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2018-09-21 01:30:12 +08:00
|
|
|
/* non-batching mode */
|
|
|
|
*p = b->page;
|
2010-04-24 01:18:08 +08:00
|
|
|
|
vmw_balloon: change batch/single lock abstractions
The current abstractions for batch vs single operations seem suboptimal
and complicate the implementation of additional features (OOM,
compaction).
The immediate problem of the current abstractions is that they cause
differences in how operations are handled when batching is on or off.
For example, the refused_alloc counter is not updated when batching is
on. These discrepancies are caused by code redundancies.
Instead, this patch presents three type of operations, according to
whether batching is on or off: (1) add page, (2) communication with the
hypervisor and (3) retrieving the status of a page.
To avoid the overhead of virtual functions, and since we do not expect
additional interfaces for communication with the hypervisor, we use
static keys instead of virtual functions.
Finally, while we are at it, change vmballoon_init_batching() to return
int instead of bool, to be consistent in the return type and avoid
potential coding errors.
Reviewed-by: Xavier Deguillard <xdeguillard@vmware.com>
Signed-off-by: Nadav Amit <namit@vmware.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2018-09-21 01:30:12 +08:00
|
|
|
/*
|
|
|
|
* If a failure occurs, the indication will be provided in the status
|
|
|
|
* of the entire operation, which is considered before the individual
|
|
|
|
* page status. So for non-batching mode, the indication is always of
|
|
|
|
* success.
|
|
|
|
*/
|
|
|
|
return VMW_BALLOON_SUCCESS;
|
|
|
|
}
|
2010-04-24 01:18:08 +08:00
|
|
|
|
vmw_balloon: change batch/single lock abstractions
The current abstractions for batch vs single operations seem suboptimal
and complicate the implementation of additional features (OOM,
compaction).
The immediate problem of the current abstractions is that they cause
differences in how operations are handled when batching is on or off.
For example, the refused_alloc counter is not updated when batching is
on. These discrepancies are caused by code redundancies.
Instead, this patch presents three type of operations, according to
whether batching is on or off: (1) add page, (2) communication with the
hypervisor and (3) retrieving the status of a page.
To avoid the overhead of virtual functions, and since we do not expect
additional interfaces for communication with the hypervisor, we use
static keys instead of virtual functions.
Finally, while we are at it, change vmballoon_init_batching() to return
int instead of bool, to be consistent in the return type and avoid
potential coding errors.
Reviewed-by: Xavier Deguillard <xdeguillard@vmware.com>
Signed-off-by: Nadav Amit <namit@vmware.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2018-09-21 01:30:12 +08:00
|
|
|
/**
|
|
|
|
* vmballoon_lock_op - notifies the host about inflated/deflated pages.
|
|
|
|
* @b: pointer to the balloon.
|
|
|
|
* @num_pages: number of inflated/deflated pages.
|
vmw_balloon: rework the inflate and deflate loops
In preparation for supporting compaction and OOM notification, this
patch reworks the inflate/deflate loops. The main idea is to separate
the allocation, communication with the hypervisor, and the handling of
errors from each other. Doing will allow us to perform concurrent
inflation and deflation, excluding the actual communication with the
hypervisor.
To do so, we need to get rid of the remaining global state that is kept
in the balloon struct, specifically the refuse_list. When the VM
communicates with the hypervisor, it does not free or put back pages
to the balloon list and instead only moves the pages whose status
indicated failure into a refuse_list on the stack. Once the operation
completes, the inflation or deflation functions handle the list
appropriately.
As we do that, we can consolidate the communication with the hypervisor
for both the lock and unlock operations into a single function. We also
reuse the deflation function for popping the balloon.
As a preparation for preventing races, we hold a spinlock when the
communication actually takes place, and use atomic operations for
updating the balloon size. The balloon page list is still racy and will
be handled in the next patch.
Reviewed-by: Xavier Deguillard <xdeguillard@vmware.com>
Signed-off-by: Nadav Amit <namit@vmware.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2018-09-21 01:30:18 +08:00
|
|
|
* @page_size: size of the page.
|
|
|
|
* @op: the type of operation (lock or unlock).
|
vmw_balloon: change batch/single lock abstractions
The current abstractions for batch vs single operations seem suboptimal
and complicate the implementation of additional features (OOM,
compaction).
The immediate problem of the current abstractions is that they cause
differences in how operations are handled when batching is on or off.
For example, the refused_alloc counter is not updated when batching is
on. These discrepancies are caused by code redundancies.
Instead, this patch presents three type of operations, according to
whether batching is on or off: (1) add page, (2) communication with the
hypervisor and (3) retrieving the status of a page.
To avoid the overhead of virtual functions, and since we do not expect
additional interfaces for communication with the hypervisor, we use
static keys instead of virtual functions.
Finally, while we are at it, change vmballoon_init_batching() to return
int instead of bool, to be consistent in the return type and avoid
potential coding errors.
Reviewed-by: Xavier Deguillard <xdeguillard@vmware.com>
Signed-off-by: Nadav Amit <namit@vmware.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2018-09-21 01:30:12 +08:00
|
|
|
*
|
|
|
|
* Notify the host about page(s) that were ballooned (or removed from the
|
|
|
|
* balloon) so that host can use it without fear that guest will need it (or
|
|
|
|
* stop using them since the VM does). Host may reject some pages, we need to
|
|
|
|
* check the return value and maybe submit a different page. The pages that are
|
|
|
|
* inflated/deflated are pointed by @b->page.
|
|
|
|
*
|
|
|
|
* Return: result as provided by the hypervisor.
|
|
|
|
*/
|
|
|
|
static unsigned long vmballoon_lock_op(struct vmballoon *b,
|
|
|
|
unsigned int num_pages,
|
vmw_balloon: rework the inflate and deflate loops
In preparation for supporting compaction and OOM notification, this
patch reworks the inflate/deflate loops. The main idea is to separate
the allocation, communication with the hypervisor, and the handling of
errors from each other. Doing will allow us to perform concurrent
inflation and deflation, excluding the actual communication with the
hypervisor.
To do so, we need to get rid of the remaining global state that is kept
in the balloon struct, specifically the refuse_list. When the VM
communicates with the hypervisor, it does not free or put back pages
to the balloon list and instead only moves the pages whose status
indicated failure into a refuse_list on the stack. Once the operation
completes, the inflation or deflation functions handle the list
appropriately.
As we do that, we can consolidate the communication with the hypervisor
for both the lock and unlock operations into a single function. We also
reuse the deflation function for popping the balloon.
As a preparation for preventing races, we hold a spinlock when the
communication actually takes place, and use atomic operations for
updating the balloon size. The balloon page list is still racy and will
be handled in the next patch.
Reviewed-by: Xavier Deguillard <xdeguillard@vmware.com>
Signed-off-by: Nadav Amit <namit@vmware.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2018-09-21 01:30:18 +08:00
|
|
|
enum vmballoon_page_size_type page_size,
|
|
|
|
enum vmballoon_op op)
|
vmw_balloon: change batch/single lock abstractions
The current abstractions for batch vs single operations seem suboptimal
and complicate the implementation of additional features (OOM,
compaction).
The immediate problem of the current abstractions is that they cause
differences in how operations are handled when batching is on or off.
For example, the refused_alloc counter is not updated when batching is
on. These discrepancies are caused by code redundancies.
Instead, this patch presents three type of operations, according to
whether batching is on or off: (1) add page, (2) communication with the
hypervisor and (3) retrieving the status of a page.
To avoid the overhead of virtual functions, and since we do not expect
additional interfaces for communication with the hypervisor, we use
static keys instead of virtual functions.
Finally, while we are at it, change vmballoon_init_batching() to return
int instead of bool, to be consistent in the return type and avoid
potential coding errors.
Reviewed-by: Xavier Deguillard <xdeguillard@vmware.com>
Signed-off-by: Nadav Amit <namit@vmware.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2018-09-21 01:30:12 +08:00
|
|
|
{
|
|
|
|
unsigned long cmd, pfn;
|
|
|
|
|
vmw_balloon: rework the inflate and deflate loops
In preparation for supporting compaction and OOM notification, this
patch reworks the inflate/deflate loops. The main idea is to separate
the allocation, communication with the hypervisor, and the handling of
errors from each other. Doing will allow us to perform concurrent
inflation and deflation, excluding the actual communication with the
hypervisor.
To do so, we need to get rid of the remaining global state that is kept
in the balloon struct, specifically the refuse_list. When the VM
communicates with the hypervisor, it does not free or put back pages
to the balloon list and instead only moves the pages whose status
indicated failure into a refuse_list on the stack. Once the operation
completes, the inflation or deflation functions handle the list
appropriately.
As we do that, we can consolidate the communication with the hypervisor
for both the lock and unlock operations into a single function. We also
reuse the deflation function for popping the balloon.
As a preparation for preventing races, we hold a spinlock when the
communication actually takes place, and use atomic operations for
updating the balloon size. The balloon page list is still racy and will
be handled in the next patch.
Reviewed-by: Xavier Deguillard <xdeguillard@vmware.com>
Signed-off-by: Nadav Amit <namit@vmware.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2018-09-21 01:30:18 +08:00
|
|
|
lockdep_assert_held(&b->comm_lock);
|
|
|
|
|
vmw_balloon: change batch/single lock abstractions
The current abstractions for batch vs single operations seem suboptimal
and complicate the implementation of additional features (OOM,
compaction).
The immediate problem of the current abstractions is that they cause
differences in how operations are handled when batching is on or off.
For example, the refused_alloc counter is not updated when batching is
on. These discrepancies are caused by code redundancies.
Instead, this patch presents three type of operations, according to
whether batching is on or off: (1) add page, (2) communication with the
hypervisor and (3) retrieving the status of a page.
To avoid the overhead of virtual functions, and since we do not expect
additional interfaces for communication with the hypervisor, we use
static keys instead of virtual functions.
Finally, while we are at it, change vmballoon_init_batching() to return
int instead of bool, to be consistent in the return type and avoid
potential coding errors.
Reviewed-by: Xavier Deguillard <xdeguillard@vmware.com>
Signed-off-by: Nadav Amit <namit@vmware.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2018-09-21 01:30:12 +08:00
|
|
|
if (static_branch_likely(&vmw_balloon_batching)) {
|
vmw_balloon: rework the inflate and deflate loops
In preparation for supporting compaction and OOM notification, this
patch reworks the inflate/deflate loops. The main idea is to separate
the allocation, communication with the hypervisor, and the handling of
errors from each other. Doing will allow us to perform concurrent
inflation and deflation, excluding the actual communication with the
hypervisor.
To do so, we need to get rid of the remaining global state that is kept
in the balloon struct, specifically the refuse_list. When the VM
communicates with the hypervisor, it does not free or put back pages
to the balloon list and instead only moves the pages whose status
indicated failure into a refuse_list on the stack. Once the operation
completes, the inflation or deflation functions handle the list
appropriately.
As we do that, we can consolidate the communication with the hypervisor
for both the lock and unlock operations into a single function. We also
reuse the deflation function for popping the balloon.
As a preparation for preventing races, we hold a spinlock when the
communication actually takes place, and use atomic operations for
updating the balloon size. The balloon page list is still racy and will
be handled in the next patch.
Reviewed-by: Xavier Deguillard <xdeguillard@vmware.com>
Signed-off-by: Nadav Amit <namit@vmware.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2018-09-21 01:30:18 +08:00
|
|
|
if (op == VMW_BALLOON_INFLATE)
|
|
|
|
cmd = page_size == VMW_BALLOON_2M_PAGE ?
|
|
|
|
VMW_BALLOON_CMD_BATCHED_2M_LOCK :
|
|
|
|
VMW_BALLOON_CMD_BATCHED_LOCK;
|
vmw_balloon: change batch/single lock abstractions
The current abstractions for batch vs single operations seem suboptimal
and complicate the implementation of additional features (OOM,
compaction).
The immediate problem of the current abstractions is that they cause
differences in how operations are handled when batching is on or off.
For example, the refused_alloc counter is not updated when batching is
on. These discrepancies are caused by code redundancies.
Instead, this patch presents three type of operations, according to
whether batching is on or off: (1) add page, (2) communication with the
hypervisor and (3) retrieving the status of a page.
To avoid the overhead of virtual functions, and since we do not expect
additional interfaces for communication with the hypervisor, we use
static keys instead of virtual functions.
Finally, while we are at it, change vmballoon_init_batching() to return
int instead of bool, to be consistent in the return type and avoid
potential coding errors.
Reviewed-by: Xavier Deguillard <xdeguillard@vmware.com>
Signed-off-by: Nadav Amit <namit@vmware.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2018-09-21 01:30:12 +08:00
|
|
|
else
|
vmw_balloon: rework the inflate and deflate loops
In preparation for supporting compaction and OOM notification, this
patch reworks the inflate/deflate loops. The main idea is to separate
the allocation, communication with the hypervisor, and the handling of
errors from each other. Doing will allow us to perform concurrent
inflation and deflation, excluding the actual communication with the
hypervisor.
To do so, we need to get rid of the remaining global state that is kept
in the balloon struct, specifically the refuse_list. When the VM
communicates with the hypervisor, it does not free or put back pages
to the balloon list and instead only moves the pages whose status
indicated failure into a refuse_list on the stack. Once the operation
completes, the inflation or deflation functions handle the list
appropriately.
As we do that, we can consolidate the communication with the hypervisor
for both the lock and unlock operations into a single function. We also
reuse the deflation function for popping the balloon.
As a preparation for preventing races, we hold a spinlock when the
communication actually takes place, and use atomic operations for
updating the balloon size. The balloon page list is still racy and will
be handled in the next patch.
Reviewed-by: Xavier Deguillard <xdeguillard@vmware.com>
Signed-off-by: Nadav Amit <namit@vmware.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2018-09-21 01:30:18 +08:00
|
|
|
cmd = page_size == VMW_BALLOON_2M_PAGE ?
|
|
|
|
VMW_BALLOON_CMD_BATCHED_2M_UNLOCK :
|
|
|
|
VMW_BALLOON_CMD_BATCHED_UNLOCK;
|
vmw_balloon: change batch/single lock abstractions
The current abstractions for batch vs single operations seem suboptimal
and complicate the implementation of additional features (OOM,
compaction).
The immediate problem of the current abstractions is that they cause
differences in how operations are handled when batching is on or off.
For example, the refused_alloc counter is not updated when batching is
on. These discrepancies are caused by code redundancies.
Instead, this patch presents three type of operations, according to
whether batching is on or off: (1) add page, (2) communication with the
hypervisor and (3) retrieving the status of a page.
To avoid the overhead of virtual functions, and since we do not expect
additional interfaces for communication with the hypervisor, we use
static keys instead of virtual functions.
Finally, while we are at it, change vmballoon_init_batching() to return
int instead of bool, to be consistent in the return type and avoid
potential coding errors.
Reviewed-by: Xavier Deguillard <xdeguillard@vmware.com>
Signed-off-by: Nadav Amit <namit@vmware.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2018-09-21 01:30:12 +08:00
|
|
|
|
|
|
|
pfn = PHYS_PFN(virt_to_phys(b->batch_page));
|
|
|
|
} else {
|
vmw_balloon: rework the inflate and deflate loops
In preparation for supporting compaction and OOM notification, this
patch reworks the inflate/deflate loops. The main idea is to separate
the allocation, communication with the hypervisor, and the handling of
errors from each other. Doing will allow us to perform concurrent
inflation and deflation, excluding the actual communication with the
hypervisor.
To do so, we need to get rid of the remaining global state that is kept
in the balloon struct, specifically the refuse_list. When the VM
communicates with the hypervisor, it does not free or put back pages
to the balloon list and instead only moves the pages whose status
indicated failure into a refuse_list on the stack. Once the operation
completes, the inflation or deflation functions handle the list
appropriately.
As we do that, we can consolidate the communication with the hypervisor
for both the lock and unlock operations into a single function. We also
reuse the deflation function for popping the balloon.
As a preparation for preventing races, we hold a spinlock when the
communication actually takes place, and use atomic operations for
updating the balloon size. The balloon page list is still racy and will
be handled in the next patch.
Reviewed-by: Xavier Deguillard <xdeguillard@vmware.com>
Signed-off-by: Nadav Amit <namit@vmware.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2018-09-21 01:30:18 +08:00
|
|
|
cmd = op == VMW_BALLOON_INFLATE ? VMW_BALLOON_CMD_LOCK :
|
|
|
|
VMW_BALLOON_CMD_UNLOCK;
|
vmw_balloon: change batch/single lock abstractions
The current abstractions for batch vs single operations seem suboptimal
and complicate the implementation of additional features (OOM,
compaction).
The immediate problem of the current abstractions is that they cause
differences in how operations are handled when batching is on or off.
For example, the refused_alloc counter is not updated when batching is
on. These discrepancies are caused by code redundancies.
Instead, this patch presents three type of operations, according to
whether batching is on or off: (1) add page, (2) communication with the
hypervisor and (3) retrieving the status of a page.
To avoid the overhead of virtual functions, and since we do not expect
additional interfaces for communication with the hypervisor, we use
static keys instead of virtual functions.
Finally, while we are at it, change vmballoon_init_batching() to return
int instead of bool, to be consistent in the return type and avoid
potential coding errors.
Reviewed-by: Xavier Deguillard <xdeguillard@vmware.com>
Signed-off-by: Nadav Amit <namit@vmware.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2018-09-21 01:30:12 +08:00
|
|
|
pfn = page_to_pfn(b->page);
|
|
|
|
|
|
|
|
/* In non-batching mode, PFNs must fit in 32-bit */
|
|
|
|
if (unlikely(pfn != (u32)pfn))
|
|
|
|
return VMW_BALLOON_ERROR_PPN_INVALID;
|
2015-06-13 02:43:22 +08:00
|
|
|
}
|
2010-04-24 01:18:08 +08:00
|
|
|
|
vmw_balloon: change batch/single lock abstractions
The current abstractions for batch vs single operations seem suboptimal
and complicate the implementation of additional features (OOM,
compaction).
The immediate problem of the current abstractions is that they cause
differences in how operations are handled when batching is on or off.
For example, the refused_alloc counter is not updated when batching is
on. These discrepancies are caused by code redundancies.
Instead, this patch presents three type of operations, according to
whether batching is on or off: (1) add page, (2) communication with the
hypervisor and (3) retrieving the status of a page.
To avoid the overhead of virtual functions, and since we do not expect
additional interfaces for communication with the hypervisor, we use
static keys instead of virtual functions.
Finally, while we are at it, change vmballoon_init_batching() to return
int instead of bool, to be consistent in the return type and avoid
potential coding errors.
Reviewed-by: Xavier Deguillard <xdeguillard@vmware.com>
Signed-off-by: Nadav Amit <namit@vmware.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2018-09-21 01:30:12 +08:00
|
|
|
return vmballoon_cmd(b, cmd, pfn, num_pages);
|
2010-04-24 01:18:08 +08:00
|
|
|
}
|
|
|
|
|
vmw_balloon: rework the inflate and deflate loops
In preparation for supporting compaction and OOM notification, this
patch reworks the inflate/deflate loops. The main idea is to separate
the allocation, communication with the hypervisor, and the handling of
errors from each other. Doing will allow us to perform concurrent
inflation and deflation, excluding the actual communication with the
hypervisor.
To do so, we need to get rid of the remaining global state that is kept
in the balloon struct, specifically the refuse_list. When the VM
communicates with the hypervisor, it does not free or put back pages
to the balloon list and instead only moves the pages whose status
indicated failure into a refuse_list on the stack. Once the operation
completes, the inflation or deflation functions handle the list
appropriately.
As we do that, we can consolidate the communication with the hypervisor
for both the lock and unlock operations into a single function. We also
reuse the deflation function for popping the balloon.
As a preparation for preventing races, we hold a spinlock when the
communication actually takes place, and use atomic operations for
updating the balloon size. The balloon page list is still racy and will
be handled in the next patch.
Reviewed-by: Xavier Deguillard <xdeguillard@vmware.com>
Signed-off-by: Nadav Amit <namit@vmware.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2018-09-21 01:30:18 +08:00
|
|
|
/**
|
|
|
|
* vmballoon_add_page - adds a page towards lock/unlock operation.
|
|
|
|
*
|
|
|
|
* @b: pointer to the balloon.
|
|
|
|
* @idx: index of the page to be ballooned in this batch.
|
|
|
|
* @p: pointer to the page that is about to be ballooned.
|
|
|
|
*
|
|
|
|
* Adds the page to be ballooned. Must be called while holding @comm_lock.
|
|
|
|
*/
|
|
|
|
static void vmballoon_add_page(struct vmballoon *b, unsigned int idx,
|
|
|
|
struct page *p)
|
|
|
|
{
|
|
|
|
lockdep_assert_held(&b->comm_lock);
|
|
|
|
|
|
|
|
if (static_branch_likely(&vmw_balloon_batching))
|
|
|
|
b->batch_page[idx] = (struct vmballoon_batch_entry)
|
|
|
|
{ .pfn = page_to_pfn(p) };
|
|
|
|
else
|
|
|
|
b->page = p;
|
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
|
|
|
* vmballoon_lock - lock or unlock a batch of pages.
|
|
|
|
*
|
|
|
|
* @b: pointer to the balloon.
|
|
|
|
* @ctl: pointer for the %struct vmballoon_ctl, which defines the operation.
|
|
|
|
*
|
|
|
|
* Notifies the host of about ballooned pages (after inflation or deflation,
|
|
|
|
* according to @ctl). If the host rejects the page put it on the
|
|
|
|
* @ctl refuse list. These refused page are then released when moving to the
|
|
|
|
* next size of pages.
|
|
|
|
*
|
|
|
|
* Note that we neither free any @page here nor put them back on the ballooned
|
|
|
|
* pages list. Instead we queue it for later processing. We do that for several
|
|
|
|
* reasons. First, we do not want to free the page under the lock. Second, it
|
|
|
|
* allows us to unify the handling of lock and unlock. In the inflate case, the
|
|
|
|
* caller will check if there are too many refused pages and release them.
|
|
|
|
* Although it is not identical to the past behavior, it should not affect
|
|
|
|
* performance.
|
|
|
|
*/
|
|
|
|
static int vmballoon_lock(struct vmballoon *b, struct vmballoon_ctl *ctl)
|
2015-08-07 06:17:58 +08:00
|
|
|
{
|
vmw_balloon: change batch/single lock abstractions
The current abstractions for batch vs single operations seem suboptimal
and complicate the implementation of additional features (OOM,
compaction).
The immediate problem of the current abstractions is that they cause
differences in how operations are handled when batching is on or off.
For example, the refused_alloc counter is not updated when batching is
on. These discrepancies are caused by code redundancies.
Instead, this patch presents three type of operations, according to
whether batching is on or off: (1) add page, (2) communication with the
hypervisor and (3) retrieving the status of a page.
To avoid the overhead of virtual functions, and since we do not expect
additional interfaces for communication with the hypervisor, we use
static keys instead of virtual functions.
Finally, while we are at it, change vmballoon_init_batching() to return
int instead of bool, to be consistent in the return type and avoid
potential coding errors.
Reviewed-by: Xavier Deguillard <xdeguillard@vmware.com>
Signed-off-by: Nadav Amit <namit@vmware.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2018-09-21 01:30:12 +08:00
|
|
|
unsigned long batch_status;
|
vmw_balloon: rework the inflate and deflate loops
In preparation for supporting compaction and OOM notification, this
patch reworks the inflate/deflate loops. The main idea is to separate
the allocation, communication with the hypervisor, and the handling of
errors from each other. Doing will allow us to perform concurrent
inflation and deflation, excluding the actual communication with the
hypervisor.
To do so, we need to get rid of the remaining global state that is kept
in the balloon struct, specifically the refuse_list. When the VM
communicates with the hypervisor, it does not free or put back pages
to the balloon list and instead only moves the pages whose status
indicated failure into a refuse_list on the stack. Once the operation
completes, the inflation or deflation functions handle the list
appropriately.
As we do that, we can consolidate the communication with the hypervisor
for both the lock and unlock operations into a single function. We also
reuse the deflation function for popping the balloon.
As a preparation for preventing races, we hold a spinlock when the
communication actually takes place, and use atomic operations for
updating the balloon size. The balloon page list is still racy and will
be handled in the next patch.
Reviewed-by: Xavier Deguillard <xdeguillard@vmware.com>
Signed-off-by: Nadav Amit <namit@vmware.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2018-09-21 01:30:18 +08:00
|
|
|
struct page *page;
|
|
|
|
unsigned int i, num_pages;
|
|
|
|
|
|
|
|
num_pages = ctl->n_pages;
|
|
|
|
if (num_pages == 0)
|
|
|
|
return 0;
|
2015-08-07 06:17:58 +08:00
|
|
|
|
vmw_balloon: rework the inflate and deflate loops
In preparation for supporting compaction and OOM notification, this
patch reworks the inflate/deflate loops. The main idea is to separate
the allocation, communication with the hypervisor, and the handling of
errors from each other. Doing will allow us to perform concurrent
inflation and deflation, excluding the actual communication with the
hypervisor.
To do so, we need to get rid of the remaining global state that is kept
in the balloon struct, specifically the refuse_list. When the VM
communicates with the hypervisor, it does not free or put back pages
to the balloon list and instead only moves the pages whose status
indicated failure into a refuse_list on the stack. Once the operation
completes, the inflation or deflation functions handle the list
appropriately.
As we do that, we can consolidate the communication with the hypervisor
for both the lock and unlock operations into a single function. We also
reuse the deflation function for popping the balloon.
As a preparation for preventing races, we hold a spinlock when the
communication actually takes place, and use atomic operations for
updating the balloon size. The balloon page list is still racy and will
be handled in the next patch.
Reviewed-by: Xavier Deguillard <xdeguillard@vmware.com>
Signed-off-by: Nadav Amit <namit@vmware.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2018-09-21 01:30:18 +08:00
|
|
|
/* communication with the host is done under the communication lock */
|
|
|
|
spin_lock(&b->comm_lock);
|
2015-08-07 06:17:58 +08:00
|
|
|
|
vmw_balloon: rework the inflate and deflate loops
In preparation for supporting compaction and OOM notification, this
patch reworks the inflate/deflate loops. The main idea is to separate
the allocation, communication with the hypervisor, and the handling of
errors from each other. Doing will allow us to perform concurrent
inflation and deflation, excluding the actual communication with the
hypervisor.
To do so, we need to get rid of the remaining global state that is kept
in the balloon struct, specifically the refuse_list. When the VM
communicates with the hypervisor, it does not free or put back pages
to the balloon list and instead only moves the pages whose status
indicated failure into a refuse_list on the stack. Once the operation
completes, the inflation or deflation functions handle the list
appropriately.
As we do that, we can consolidate the communication with the hypervisor
for both the lock and unlock operations into a single function. We also
reuse the deflation function for popping the balloon.
As a preparation for preventing races, we hold a spinlock when the
communication actually takes place, and use atomic operations for
updating the balloon size. The balloon page list is still racy and will
be handled in the next patch.
Reviewed-by: Xavier Deguillard <xdeguillard@vmware.com>
Signed-off-by: Nadav Amit <namit@vmware.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2018-09-21 01:30:18 +08:00
|
|
|
i = 0;
|
|
|
|
list_for_each_entry(page, &ctl->pages, lru)
|
|
|
|
vmballoon_add_page(b, i++, page);
|
|
|
|
|
|
|
|
batch_status = vmballoon_lock_op(b, ctl->n_pages, ctl->page_size,
|
|
|
|
ctl->op);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Iterate over the pages in the provided list. Since we are changing
|
|
|
|
* @ctl->n_pages we are saving the original value in @num_pages and
|
|
|
|
* use this value to bound the loop.
|
|
|
|
*/
|
2015-08-07 06:17:58 +08:00
|
|
|
for (i = 0; i < num_pages; i++) {
|
vmw_balloon: change batch/single lock abstractions
The current abstractions for batch vs single operations seem suboptimal
and complicate the implementation of additional features (OOM,
compaction).
The immediate problem of the current abstractions is that they cause
differences in how operations are handled when batching is on or off.
For example, the refused_alloc counter is not updated when batching is
on. These discrepancies are caused by code redundancies.
Instead, this patch presents three type of operations, according to
whether batching is on or off: (1) add page, (2) communication with the
hypervisor and (3) retrieving the status of a page.
To avoid the overhead of virtual functions, and since we do not expect
additional interfaces for communication with the hypervisor, we use
static keys instead of virtual functions.
Finally, while we are at it, change vmballoon_init_batching() to return
int instead of bool, to be consistent in the return type and avoid
potential coding errors.
Reviewed-by: Xavier Deguillard <xdeguillard@vmware.com>
Signed-off-by: Nadav Amit <namit@vmware.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2018-09-21 01:30:12 +08:00
|
|
|
unsigned long status;
|
2015-08-07 06:17:58 +08:00
|
|
|
|
vmw_balloon: rework the inflate and deflate loops
In preparation for supporting compaction and OOM notification, this
patch reworks the inflate/deflate loops. The main idea is to separate
the allocation, communication with the hypervisor, and the handling of
errors from each other. Doing will allow us to perform concurrent
inflation and deflation, excluding the actual communication with the
hypervisor.
To do so, we need to get rid of the remaining global state that is kept
in the balloon struct, specifically the refuse_list. When the VM
communicates with the hypervisor, it does not free or put back pages
to the balloon list and instead only moves the pages whose status
indicated failure into a refuse_list on the stack. Once the operation
completes, the inflation or deflation functions handle the list
appropriately.
As we do that, we can consolidate the communication with the hypervisor
for both the lock and unlock operations into a single function. We also
reuse the deflation function for popping the balloon.
As a preparation for preventing races, we hold a spinlock when the
communication actually takes place, and use atomic operations for
updating the balloon size. The balloon page list is still racy and will
be handled in the next patch.
Reviewed-by: Xavier Deguillard <xdeguillard@vmware.com>
Signed-off-by: Nadav Amit <namit@vmware.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2018-09-21 01:30:18 +08:00
|
|
|
status = vmballoon_status_page(b, i, &page);
|
vmw_balloon: change batch/single lock abstractions
The current abstractions for batch vs single operations seem suboptimal
and complicate the implementation of additional features (OOM,
compaction).
The immediate problem of the current abstractions is that they cause
differences in how operations are handled when batching is on or off.
For example, the refused_alloc counter is not updated when batching is
on. These discrepancies are caused by code redundancies.
Instead, this patch presents three type of operations, according to
whether batching is on or off: (1) add page, (2) communication with the
hypervisor and (3) retrieving the status of a page.
To avoid the overhead of virtual functions, and since we do not expect
additional interfaces for communication with the hypervisor, we use
static keys instead of virtual functions.
Finally, while we are at it, change vmballoon_init_batching() to return
int instead of bool, to be consistent in the return type and avoid
potential coding errors.
Reviewed-by: Xavier Deguillard <xdeguillard@vmware.com>
Signed-off-by: Nadav Amit <namit@vmware.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2018-09-21 01:30:12 +08:00
|
|
|
|
|
|
|
/*
|
|
|
|
* Failure of the whole batch overrides a single operation
|
|
|
|
* results.
|
|
|
|
*/
|
|
|
|
if (batch_status != VMW_BALLOON_SUCCESS)
|
|
|
|
status = batch_status;
|
2015-08-07 06:17:58 +08:00
|
|
|
|
vmw_balloon: rework the inflate and deflate loops
In preparation for supporting compaction and OOM notification, this
patch reworks the inflate/deflate loops. The main idea is to separate
the allocation, communication with the hypervisor, and the handling of
errors from each other. Doing will allow us to perform concurrent
inflation and deflation, excluding the actual communication with the
hypervisor.
To do so, we need to get rid of the remaining global state that is kept
in the balloon struct, specifically the refuse_list. When the VM
communicates with the hypervisor, it does not free or put back pages
to the balloon list and instead only moves the pages whose status
indicated failure into a refuse_list on the stack. Once the operation
completes, the inflation or deflation functions handle the list
appropriately.
As we do that, we can consolidate the communication with the hypervisor
for both the lock and unlock operations into a single function. We also
reuse the deflation function for popping the balloon.
As a preparation for preventing races, we hold a spinlock when the
communication actually takes place, and use atomic operations for
updating the balloon size. The balloon page list is still racy and will
be handled in the next patch.
Reviewed-by: Xavier Deguillard <xdeguillard@vmware.com>
Signed-off-by: Nadav Amit <namit@vmware.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2018-09-21 01:30:18 +08:00
|
|
|
/* Continue if no error happened */
|
|
|
|
if (!vmballoon_handle_one_result(b, page, ctl->page_size,
|
|
|
|
status))
|
vmw_balloon: change batch/single lock abstractions
The current abstractions for batch vs single operations seem suboptimal
and complicate the implementation of additional features (OOM,
compaction).
The immediate problem of the current abstractions is that they cause
differences in how operations are handled when batching is on or off.
For example, the refused_alloc counter is not updated when batching is
on. These discrepancies are caused by code redundancies.
Instead, this patch presents three type of operations, according to
whether batching is on or off: (1) add page, (2) communication with the
hypervisor and (3) retrieving the status of a page.
To avoid the overhead of virtual functions, and since we do not expect
additional interfaces for communication with the hypervisor, we use
static keys instead of virtual functions.
Finally, while we are at it, change vmballoon_init_batching() to return
int instead of bool, to be consistent in the return type and avoid
potential coding errors.
Reviewed-by: Xavier Deguillard <xdeguillard@vmware.com>
Signed-off-by: Nadav Amit <namit@vmware.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2018-09-21 01:30:12 +08:00
|
|
|
continue;
|
|
|
|
|
2018-09-21 01:30:13 +08:00
|
|
|
/*
|
vmw_balloon: rework the inflate and deflate loops
In preparation for supporting compaction and OOM notification, this
patch reworks the inflate/deflate loops. The main idea is to separate
the allocation, communication with the hypervisor, and the handling of
errors from each other. Doing will allow us to perform concurrent
inflation and deflation, excluding the actual communication with the
hypervisor.
To do so, we need to get rid of the remaining global state that is kept
in the balloon struct, specifically the refuse_list. When the VM
communicates with the hypervisor, it does not free or put back pages
to the balloon list and instead only moves the pages whose status
indicated failure into a refuse_list on the stack. Once the operation
completes, the inflation or deflation functions handle the list
appropriately.
As we do that, we can consolidate the communication with the hypervisor
for both the lock and unlock operations into a single function. We also
reuse the deflation function for popping the balloon.
As a preparation for preventing races, we hold a spinlock when the
communication actually takes place, and use atomic operations for
updating the balloon size. The balloon page list is still racy and will
be handled in the next patch.
Reviewed-by: Xavier Deguillard <xdeguillard@vmware.com>
Signed-off-by: Nadav Amit <namit@vmware.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2018-09-21 01:30:18 +08:00
|
|
|
* Error happened. Move the pages to the refused list and update
|
|
|
|
* the pages number.
|
2018-09-21 01:30:13 +08:00
|
|
|
*/
|
vmw_balloon: rework the inflate and deflate loops
In preparation for supporting compaction and OOM notification, this
patch reworks the inflate/deflate loops. The main idea is to separate
the allocation, communication with the hypervisor, and the handling of
errors from each other. Doing will allow us to perform concurrent
inflation and deflation, excluding the actual communication with the
hypervisor.
To do so, we need to get rid of the remaining global state that is kept
in the balloon struct, specifically the refuse_list. When the VM
communicates with the hypervisor, it does not free or put back pages
to the balloon list and instead only moves the pages whose status
indicated failure into a refuse_list on the stack. Once the operation
completes, the inflation or deflation functions handle the list
appropriately.
As we do that, we can consolidate the communication with the hypervisor
for both the lock and unlock operations into a single function. We also
reuse the deflation function for popping the balloon.
As a preparation for preventing races, we hold a spinlock when the
communication actually takes place, and use atomic operations for
updating the balloon size. The balloon page list is still racy and will
be handled in the next patch.
Reviewed-by: Xavier Deguillard <xdeguillard@vmware.com>
Signed-off-by: Nadav Amit <namit@vmware.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2018-09-21 01:30:18 +08:00
|
|
|
list_move(&page->lru, &ctl->refused_pages);
|
|
|
|
ctl->n_pages--;
|
|
|
|
ctl->n_refused_pages++;
|
2015-08-07 06:17:58 +08:00
|
|
|
}
|
|
|
|
|
vmw_balloon: rework the inflate and deflate loops
In preparation for supporting compaction and OOM notification, this
patch reworks the inflate/deflate loops. The main idea is to separate
the allocation, communication with the hypervisor, and the handling of
errors from each other. Doing will allow us to perform concurrent
inflation and deflation, excluding the actual communication with the
hypervisor.
To do so, we need to get rid of the remaining global state that is kept
in the balloon struct, specifically the refuse_list. When the VM
communicates with the hypervisor, it does not free or put back pages
to the balloon list and instead only moves the pages whose status
indicated failure into a refuse_list on the stack. Once the operation
completes, the inflation or deflation functions handle the list
appropriately.
As we do that, we can consolidate the communication with the hypervisor
for both the lock and unlock operations into a single function. We also
reuse the deflation function for popping the balloon.
As a preparation for preventing races, we hold a spinlock when the
communication actually takes place, and use atomic operations for
updating the balloon size. The balloon page list is still racy and will
be handled in the next patch.
Reviewed-by: Xavier Deguillard <xdeguillard@vmware.com>
Signed-off-by: Nadav Amit <namit@vmware.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2018-09-21 01:30:18 +08:00
|
|
|
spin_unlock(&b->comm_lock);
|
|
|
|
|
vmw_balloon: change batch/single lock abstractions
The current abstractions for batch vs single operations seem suboptimal
and complicate the implementation of additional features (OOM,
compaction).
The immediate problem of the current abstractions is that they cause
differences in how operations are handled when batching is on or off.
For example, the refused_alloc counter is not updated when batching is
on. These discrepancies are caused by code redundancies.
Instead, this patch presents three type of operations, according to
whether batching is on or off: (1) add page, (2) communication with the
hypervisor and (3) retrieving the status of a page.
To avoid the overhead of virtual functions, and since we do not expect
additional interfaces for communication with the hypervisor, we use
static keys instead of virtual functions.
Finally, while we are at it, change vmballoon_init_batching() to return
int instead of bool, to be consistent in the return type and avoid
potential coding errors.
Reviewed-by: Xavier Deguillard <xdeguillard@vmware.com>
Signed-off-by: Nadav Amit <namit@vmware.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2018-09-21 01:30:12 +08:00
|
|
|
return batch_status == VMW_BALLOON_SUCCESS ? 0 : -EIO;
|
2015-08-07 06:17:58 +08:00
|
|
|
}
|
|
|
|
|
vmw_balloon: rework the inflate and deflate loops
In preparation for supporting compaction and OOM notification, this
patch reworks the inflate/deflate loops. The main idea is to separate
the allocation, communication with the hypervisor, and the handling of
errors from each other. Doing will allow us to perform concurrent
inflation and deflation, excluding the actual communication with the
hypervisor.
To do so, we need to get rid of the remaining global state that is kept
in the balloon struct, specifically the refuse_list. When the VM
communicates with the hypervisor, it does not free or put back pages
to the balloon list and instead only moves the pages whose status
indicated failure into a refuse_list on the stack. Once the operation
completes, the inflation or deflation functions handle the list
appropriately.
As we do that, we can consolidate the communication with the hypervisor
for both the lock and unlock operations into a single function. We also
reuse the deflation function for popping the balloon.
As a preparation for preventing races, we hold a spinlock when the
communication actually takes place, and use atomic operations for
updating the balloon size. The balloon page list is still racy and will
be handled in the next patch.
Reviewed-by: Xavier Deguillard <xdeguillard@vmware.com>
Signed-off-by: Nadav Amit <namit@vmware.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2018-09-21 01:30:18 +08:00
|
|
|
/**
|
|
|
|
* vmballoon_release_page_list() - Releases a page list
|
|
|
|
*
|
|
|
|
* @page_list: list of pages to release.
|
|
|
|
* @n_pages: pointer to the number of pages.
|
|
|
|
* @page_size: whether the pages in the list are 2MB (or else 4KB).
|
|
|
|
*
|
|
|
|
* Releases the list of pages and zeros the number of pages.
|
2010-04-24 01:18:08 +08:00
|
|
|
*/
|
vmw_balloon: rework the inflate and deflate loops
In preparation for supporting compaction and OOM notification, this
patch reworks the inflate/deflate loops. The main idea is to separate
the allocation, communication with the hypervisor, and the handling of
errors from each other. Doing will allow us to perform concurrent
inflation and deflation, excluding the actual communication with the
hypervisor.
To do so, we need to get rid of the remaining global state that is kept
in the balloon struct, specifically the refuse_list. When the VM
communicates with the hypervisor, it does not free or put back pages
to the balloon list and instead only moves the pages whose status
indicated failure into a refuse_list on the stack. Once the operation
completes, the inflation or deflation functions handle the list
appropriately.
As we do that, we can consolidate the communication with the hypervisor
for both the lock and unlock operations into a single function. We also
reuse the deflation function for popping the balloon.
As a preparation for preventing races, we hold a spinlock when the
communication actually takes place, and use atomic operations for
updating the balloon size. The balloon page list is still racy and will
be handled in the next patch.
Reviewed-by: Xavier Deguillard <xdeguillard@vmware.com>
Signed-off-by: Nadav Amit <namit@vmware.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2018-09-21 01:30:18 +08:00
|
|
|
static void vmballoon_release_page_list(struct list_head *page_list,
|
|
|
|
int *n_pages,
|
|
|
|
enum vmballoon_page_size_type page_size)
|
2010-04-24 01:18:08 +08:00
|
|
|
{
|
vmw_balloon: rework the inflate and deflate loops
In preparation for supporting compaction and OOM notification, this
patch reworks the inflate/deflate loops. The main idea is to separate
the allocation, communication with the hypervisor, and the handling of
errors from each other. Doing will allow us to perform concurrent
inflation and deflation, excluding the actual communication with the
hypervisor.
To do so, we need to get rid of the remaining global state that is kept
in the balloon struct, specifically the refuse_list. When the VM
communicates with the hypervisor, it does not free or put back pages
to the balloon list and instead only moves the pages whose status
indicated failure into a refuse_list on the stack. Once the operation
completes, the inflation or deflation functions handle the list
appropriately.
As we do that, we can consolidate the communication with the hypervisor
for both the lock and unlock operations into a single function. We also
reuse the deflation function for popping the balloon.
As a preparation for preventing races, we hold a spinlock when the
communication actually takes place, and use atomic operations for
updating the balloon size. The balloon page list is still racy and will
be handled in the next patch.
Reviewed-by: Xavier Deguillard <xdeguillard@vmware.com>
Signed-off-by: Nadav Amit <namit@vmware.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2018-09-21 01:30:18 +08:00
|
|
|
struct page *page, *tmp;
|
vmw_balloon: change batch/single lock abstractions
The current abstractions for batch vs single operations seem suboptimal
and complicate the implementation of additional features (OOM,
compaction).
The immediate problem of the current abstractions is that they cause
differences in how operations are handled when batching is on or off.
For example, the refused_alloc counter is not updated when batching is
on. These discrepancies are caused by code redundancies.
Instead, this patch presents three type of operations, according to
whether batching is on or off: (1) add page, (2) communication with the
hypervisor and (3) retrieving the status of a page.
To avoid the overhead of virtual functions, and since we do not expect
additional interfaces for communication with the hypervisor, we use
static keys instead of virtual functions.
Finally, while we are at it, change vmballoon_init_batching() to return
int instead of bool, to be consistent in the return type and avoid
potential coding errors.
Reviewed-by: Xavier Deguillard <xdeguillard@vmware.com>
Signed-off-by: Nadav Amit <namit@vmware.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2018-09-21 01:30:12 +08:00
|
|
|
|
vmw_balloon: rework the inflate and deflate loops
In preparation for supporting compaction and OOM notification, this
patch reworks the inflate/deflate loops. The main idea is to separate
the allocation, communication with the hypervisor, and the handling of
errors from each other. Doing will allow us to perform concurrent
inflation and deflation, excluding the actual communication with the
hypervisor.
To do so, we need to get rid of the remaining global state that is kept
in the balloon struct, specifically the refuse_list. When the VM
communicates with the hypervisor, it does not free or put back pages
to the balloon list and instead only moves the pages whose status
indicated failure into a refuse_list on the stack. Once the operation
completes, the inflation or deflation functions handle the list
appropriately.
As we do that, we can consolidate the communication with the hypervisor
for both the lock and unlock operations into a single function. We also
reuse the deflation function for popping the balloon.
As a preparation for preventing races, we hold a spinlock when the
communication actually takes place, and use atomic operations for
updating the balloon size. The balloon page list is still racy and will
be handled in the next patch.
Reviewed-by: Xavier Deguillard <xdeguillard@vmware.com>
Signed-off-by: Nadav Amit <namit@vmware.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2018-09-21 01:30:18 +08:00
|
|
|
list_for_each_entry_safe(page, tmp, page_list, lru) {
|
|
|
|
list_del(&page->lru);
|
|
|
|
__free_pages(page, vmballoon_page_order(page_size));
|
2015-08-07 06:17:58 +08:00
|
|
|
}
|
|
|
|
|
2019-04-25 19:54:45 +08:00
|
|
|
if (n_pages)
|
|
|
|
*n_pages = 0;
|
2015-08-07 06:17:58 +08:00
|
|
|
}
|
|
|
|
|
vmw_balloon: rework the inflate and deflate loops
In preparation for supporting compaction and OOM notification, this
patch reworks the inflate/deflate loops. The main idea is to separate
the allocation, communication with the hypervisor, and the handling of
errors from each other. Doing will allow us to perform concurrent
inflation and deflation, excluding the actual communication with the
hypervisor.
To do so, we need to get rid of the remaining global state that is kept
in the balloon struct, specifically the refuse_list. When the VM
communicates with the hypervisor, it does not free or put back pages
to the balloon list and instead only moves the pages whose status
indicated failure into a refuse_list on the stack. Once the operation
completes, the inflation or deflation functions handle the list
appropriately.
As we do that, we can consolidate the communication with the hypervisor
for both the lock and unlock operations into a single function. We also
reuse the deflation function for popping the balloon.
As a preparation for preventing races, we hold a spinlock when the
communication actually takes place, and use atomic operations for
updating the balloon size. The balloon page list is still racy and will
be handled in the next patch.
Reviewed-by: Xavier Deguillard <xdeguillard@vmware.com>
Signed-off-by: Nadav Amit <namit@vmware.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2018-09-21 01:30:18 +08:00
|
|
|
|
2010-04-24 01:18:08 +08:00
|
|
|
/*
|
|
|
|
* Release pages that were allocated while attempting to inflate the
|
|
|
|
* balloon but were refused by the host for one reason or another.
|
|
|
|
*/
|
2015-08-07 06:18:01 +08:00
|
|
|
static void vmballoon_release_refused_pages(struct vmballoon *b,
|
vmw_balloon: rework the inflate and deflate loops
In preparation for supporting compaction and OOM notification, this
patch reworks the inflate/deflate loops. The main idea is to separate
the allocation, communication with the hypervisor, and the handling of
errors from each other. Doing will allow us to perform concurrent
inflation and deflation, excluding the actual communication with the
hypervisor.
To do so, we need to get rid of the remaining global state that is kept
in the balloon struct, specifically the refuse_list. When the VM
communicates with the hypervisor, it does not free or put back pages
to the balloon list and instead only moves the pages whose status
indicated failure into a refuse_list on the stack. Once the operation
completes, the inflation or deflation functions handle the list
appropriately.
As we do that, we can consolidate the communication with the hypervisor
for both the lock and unlock operations into a single function. We also
reuse the deflation function for popping the balloon.
As a preparation for preventing races, we hold a spinlock when the
communication actually takes place, and use atomic operations for
updating the balloon size. The balloon page list is still racy and will
be handled in the next patch.
Reviewed-by: Xavier Deguillard <xdeguillard@vmware.com>
Signed-off-by: Nadav Amit <namit@vmware.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2018-09-21 01:30:18 +08:00
|
|
|
struct vmballoon_ctl *ctl)
|
2010-04-24 01:18:08 +08:00
|
|
|
{
|
vmw_balloon: rework the inflate and deflate loops
In preparation for supporting compaction and OOM notification, this
patch reworks the inflate/deflate loops. The main idea is to separate
the allocation, communication with the hypervisor, and the handling of
errors from each other. Doing will allow us to perform concurrent
inflation and deflation, excluding the actual communication with the
hypervisor.
To do so, we need to get rid of the remaining global state that is kept
in the balloon struct, specifically the refuse_list. When the VM
communicates with the hypervisor, it does not free or put back pages
to the balloon list and instead only moves the pages whose status
indicated failure into a refuse_list on the stack. Once the operation
completes, the inflation or deflation functions handle the list
appropriately.
As we do that, we can consolidate the communication with the hypervisor
for both the lock and unlock operations into a single function. We also
reuse the deflation function for popping the balloon.
As a preparation for preventing races, we hold a spinlock when the
communication actually takes place, and use atomic operations for
updating the balloon size. The balloon page list is still racy and will
be handled in the next patch.
Reviewed-by: Xavier Deguillard <xdeguillard@vmware.com>
Signed-off-by: Nadav Amit <namit@vmware.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2018-09-21 01:30:18 +08:00
|
|
|
vmballoon_stats_page_inc(b, VMW_BALLOON_PAGE_STAT_REFUSED_FREE,
|
|
|
|
ctl->page_size);
|
2010-04-24 01:18:08 +08:00
|
|
|
|
vmw_balloon: rework the inflate and deflate loops
In preparation for supporting compaction and OOM notification, this
patch reworks the inflate/deflate loops. The main idea is to separate
the allocation, communication with the hypervisor, and the handling of
errors from each other. Doing will allow us to perform concurrent
inflation and deflation, excluding the actual communication with the
hypervisor.
To do so, we need to get rid of the remaining global state that is kept
in the balloon struct, specifically the refuse_list. When the VM
communicates with the hypervisor, it does not free or put back pages
to the balloon list and instead only moves the pages whose status
indicated failure into a refuse_list on the stack. Once the operation
completes, the inflation or deflation functions handle the list
appropriately.
As we do that, we can consolidate the communication with the hypervisor
for both the lock and unlock operations into a single function. We also
reuse the deflation function for popping the balloon.
As a preparation for preventing races, we hold a spinlock when the
communication actually takes place, and use atomic operations for
updating the balloon size. The balloon page list is still racy and will
be handled in the next patch.
Reviewed-by: Xavier Deguillard <xdeguillard@vmware.com>
Signed-off-by: Nadav Amit <namit@vmware.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2018-09-21 01:30:18 +08:00
|
|
|
vmballoon_release_page_list(&ctl->refused_pages, &ctl->n_refused_pages,
|
|
|
|
ctl->page_size);
|
2015-08-07 06:17:58 +08:00
|
|
|
}
|
|
|
|
|
2018-09-21 01:30:15 +08:00
|
|
|
/**
|
|
|
|
* vmballoon_change - retrieve the required balloon change
|
|
|
|
*
|
|
|
|
* @b: pointer for the balloon.
|
|
|
|
*
|
|
|
|
* Return: the required change for the balloon size. A positive number
|
|
|
|
* indicates inflation, a negative number indicates a deflation.
|
|
|
|
*/
|
|
|
|
static int64_t vmballoon_change(struct vmballoon *b)
|
|
|
|
{
|
|
|
|
int64_t size, target;
|
|
|
|
|
vmw_balloon: rework the inflate and deflate loops
In preparation for supporting compaction and OOM notification, this
patch reworks the inflate/deflate loops. The main idea is to separate
the allocation, communication with the hypervisor, and the handling of
errors from each other. Doing will allow us to perform concurrent
inflation and deflation, excluding the actual communication with the
hypervisor.
To do so, we need to get rid of the remaining global state that is kept
in the balloon struct, specifically the refuse_list. When the VM
communicates with the hypervisor, it does not free or put back pages
to the balloon list and instead only moves the pages whose status
indicated failure into a refuse_list on the stack. Once the operation
completes, the inflation or deflation functions handle the list
appropriately.
As we do that, we can consolidate the communication with the hypervisor
for both the lock and unlock operations into a single function. We also
reuse the deflation function for popping the balloon.
As a preparation for preventing races, we hold a spinlock when the
communication actually takes place, and use atomic operations for
updating the balloon size. The balloon page list is still racy and will
be handled in the next patch.
Reviewed-by: Xavier Deguillard <xdeguillard@vmware.com>
Signed-off-by: Nadav Amit <namit@vmware.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2018-09-21 01:30:18 +08:00
|
|
|
size = atomic64_read(&b->size);
|
|
|
|
target = READ_ONCE(b->target);
|
2018-09-21 01:30:15 +08:00
|
|
|
|
|
|
|
/*
|
|
|
|
* We must cast first because of int sizes
|
|
|
|
* Otherwise we might get huge positives instead of negatives
|
|
|
|
*/
|
|
|
|
|
|
|
|
if (b->reset_required)
|
|
|
|
return 0;
|
|
|
|
|
|
|
|
/* consider a 2MB slack on deflate, unless the balloon is emptied */
|
vmw_balloon: rework the inflate and deflate loops
In preparation for supporting compaction and OOM notification, this
patch reworks the inflate/deflate loops. The main idea is to separate
the allocation, communication with the hypervisor, and the handling of
errors from each other. Doing will allow us to perform concurrent
inflation and deflation, excluding the actual communication with the
hypervisor.
To do so, we need to get rid of the remaining global state that is kept
in the balloon struct, specifically the refuse_list. When the VM
communicates with the hypervisor, it does not free or put back pages
to the balloon list and instead only moves the pages whose status
indicated failure into a refuse_list on the stack. Once the operation
completes, the inflation or deflation functions handle the list
appropriately.
As we do that, we can consolidate the communication with the hypervisor
for both the lock and unlock operations into a single function. We also
reuse the deflation function for popping the balloon.
As a preparation for preventing races, we hold a spinlock when the
communication actually takes place, and use atomic operations for
updating the balloon size. The balloon page list is still racy and will
be handled in the next patch.
Reviewed-by: Xavier Deguillard <xdeguillard@vmware.com>
Signed-off-by: Nadav Amit <namit@vmware.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2018-09-21 01:30:18 +08:00
|
|
|
if (target < size && target != 0 &&
|
|
|
|
size - target < vmballoon_page_in_frames(VMW_BALLOON_2M_PAGE))
|
2018-09-21 01:30:15 +08:00
|
|
|
return 0;
|
|
|
|
|
2019-04-25 19:54:44 +08:00
|
|
|
/* If an out-of-memory recently occurred, inflation is disallowed. */
|
|
|
|
if (target > size && time_before(jiffies, READ_ONCE(b->shrink_timeout)))
|
|
|
|
return 0;
|
|
|
|
|
2018-09-21 01:30:15 +08:00
|
|
|
return target - size;
|
|
|
|
}
|
|
|
|
|
vmw_balloon: rework the inflate and deflate loops
In preparation for supporting compaction and OOM notification, this
patch reworks the inflate/deflate loops. The main idea is to separate
the allocation, communication with the hypervisor, and the handling of
errors from each other. Doing will allow us to perform concurrent
inflation and deflation, excluding the actual communication with the
hypervisor.
To do so, we need to get rid of the remaining global state that is kept
in the balloon struct, specifically the refuse_list. When the VM
communicates with the hypervisor, it does not free or put back pages
to the balloon list and instead only moves the pages whose status
indicated failure into a refuse_list on the stack. Once the operation
completes, the inflation or deflation functions handle the list
appropriately.
As we do that, we can consolidate the communication with the hypervisor
for both the lock and unlock operations into a single function. We also
reuse the deflation function for popping the balloon.
As a preparation for preventing races, we hold a spinlock when the
communication actually takes place, and use atomic operations for
updating the balloon size. The balloon page list is still racy and will
be handled in the next patch.
Reviewed-by: Xavier Deguillard <xdeguillard@vmware.com>
Signed-off-by: Nadav Amit <namit@vmware.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2018-09-21 01:30:18 +08:00
|
|
|
/**
|
|
|
|
* vmballoon_enqueue_page_list() - Enqueues list of pages after inflation.
|
|
|
|
*
|
|
|
|
* @b: pointer to balloon.
|
|
|
|
* @pages: list of pages to enqueue.
|
|
|
|
* @n_pages: pointer to number of pages in list. The value is zeroed.
|
|
|
|
* @page_size: whether the pages are 2MB or 4KB pages.
|
|
|
|
*
|
|
|
|
* Enqueues the provides list of pages in the ballooned page list, clears the
|
|
|
|
* list and zeroes the number of pages that was provided.
|
|
|
|
*/
|
|
|
|
static void vmballoon_enqueue_page_list(struct vmballoon *b,
|
|
|
|
struct list_head *pages,
|
|
|
|
unsigned int *n_pages,
|
|
|
|
enum vmballoon_page_size_type page_size)
|
|
|
|
{
|
2019-04-25 19:54:43 +08:00
|
|
|
unsigned long flags;
|
2019-08-21 00:01:21 +08:00
|
|
|
struct page *page;
|
2019-04-25 19:54:43 +08:00
|
|
|
|
|
|
|
if (page_size == VMW_BALLOON_4K_PAGE) {
|
|
|
|
balloon_page_list_enqueue(&b->b_dev_info, pages);
|
|
|
|
} else {
|
|
|
|
/*
|
|
|
|
* Keep the huge pages in a local list which is not available
|
|
|
|
* for the balloon compaction mechanism.
|
|
|
|
*/
|
|
|
|
spin_lock_irqsave(&b->b_dev_info.pages_lock, flags);
|
2019-08-21 00:01:21 +08:00
|
|
|
|
|
|
|
list_for_each_entry(page, pages, lru) {
|
|
|
|
vmballoon_mark_page_offline(page, VMW_BALLOON_2M_PAGE);
|
|
|
|
}
|
|
|
|
|
2019-04-25 19:54:43 +08:00
|
|
|
list_splice_init(pages, &b->huge_pages);
|
|
|
|
__count_vm_events(BALLOON_INFLATE, *n_pages *
|
|
|
|
vmballoon_page_in_frames(VMW_BALLOON_2M_PAGE));
|
|
|
|
spin_unlock_irqrestore(&b->b_dev_info.pages_lock, flags);
|
|
|
|
}
|
vmw_balloon: rework the inflate and deflate loops
In preparation for supporting compaction and OOM notification, this
patch reworks the inflate/deflate loops. The main idea is to separate
the allocation, communication with the hypervisor, and the handling of
errors from each other. Doing will allow us to perform concurrent
inflation and deflation, excluding the actual communication with the
hypervisor.
To do so, we need to get rid of the remaining global state that is kept
in the balloon struct, specifically the refuse_list. When the VM
communicates with the hypervisor, it does not free or put back pages
to the balloon list and instead only moves the pages whose status
indicated failure into a refuse_list on the stack. Once the operation
completes, the inflation or deflation functions handle the list
appropriately.
As we do that, we can consolidate the communication with the hypervisor
for both the lock and unlock operations into a single function. We also
reuse the deflation function for popping the balloon.
As a preparation for preventing races, we hold a spinlock when the
communication actually takes place, and use atomic operations for
updating the balloon size. The balloon page list is still racy and will
be handled in the next patch.
Reviewed-by: Xavier Deguillard <xdeguillard@vmware.com>
Signed-off-by: Nadav Amit <namit@vmware.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2018-09-21 01:30:18 +08:00
|
|
|
|
|
|
|
*n_pages = 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
|
|
|
* vmballoon_dequeue_page_list() - Dequeues page lists for deflation.
|
|
|
|
*
|
|
|
|
* @b: pointer to balloon.
|
|
|
|
* @pages: list of pages to enqueue.
|
|
|
|
* @n_pages: pointer to number of pages in list. The value is zeroed.
|
|
|
|
* @page_size: whether the pages are 2MB or 4KB pages.
|
|
|
|
* @n_req_pages: the number of requested pages.
|
|
|
|
*
|
|
|
|
* Dequeues the number of requested pages from the balloon for deflation. The
|
|
|
|
* number of dequeued pages may be lower, if not enough pages in the requested
|
|
|
|
* size are available.
|
|
|
|
*/
|
|
|
|
static void vmballoon_dequeue_page_list(struct vmballoon *b,
|
|
|
|
struct list_head *pages,
|
|
|
|
unsigned int *n_pages,
|
|
|
|
enum vmballoon_page_size_type page_size,
|
|
|
|
unsigned int n_req_pages)
|
|
|
|
{
|
|
|
|
struct page *page, *tmp;
|
|
|
|
unsigned int i = 0;
|
2019-04-25 19:54:43 +08:00
|
|
|
unsigned long flags;
|
vmw_balloon: rework the inflate and deflate loops
In preparation for supporting compaction and OOM notification, this
patch reworks the inflate/deflate loops. The main idea is to separate
the allocation, communication with the hypervisor, and the handling of
errors from each other. Doing will allow us to perform concurrent
inflation and deflation, excluding the actual communication with the
hypervisor.
To do so, we need to get rid of the remaining global state that is kept
in the balloon struct, specifically the refuse_list. When the VM
communicates with the hypervisor, it does not free or put back pages
to the balloon list and instead only moves the pages whose status
indicated failure into a refuse_list on the stack. Once the operation
completes, the inflation or deflation functions handle the list
appropriately.
As we do that, we can consolidate the communication with the hypervisor
for both the lock and unlock operations into a single function. We also
reuse the deflation function for popping the balloon.
As a preparation for preventing races, we hold a spinlock when the
communication actually takes place, and use atomic operations for
updating the balloon size. The balloon page list is still racy and will
be handled in the next patch.
Reviewed-by: Xavier Deguillard <xdeguillard@vmware.com>
Signed-off-by: Nadav Amit <namit@vmware.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2018-09-21 01:30:18 +08:00
|
|
|
|
2019-04-25 19:54:43 +08:00
|
|
|
/* In the case of 4k pages, use the compaction infrastructure */
|
|
|
|
if (page_size == VMW_BALLOON_4K_PAGE) {
|
|
|
|
*n_pages = balloon_page_list_dequeue(&b->b_dev_info, pages,
|
|
|
|
n_req_pages);
|
|
|
|
return;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* 2MB pages */
|
|
|
|
spin_lock_irqsave(&b->b_dev_info.pages_lock, flags);
|
|
|
|
list_for_each_entry_safe(page, tmp, &b->huge_pages, lru) {
|
2019-08-21 00:01:21 +08:00
|
|
|
vmballoon_mark_page_online(page, VMW_BALLOON_2M_PAGE);
|
|
|
|
|
vmw_balloon: rework the inflate and deflate loops
In preparation for supporting compaction and OOM notification, this
patch reworks the inflate/deflate loops. The main idea is to separate
the allocation, communication with the hypervisor, and the handling of
errors from each other. Doing will allow us to perform concurrent
inflation and deflation, excluding the actual communication with the
hypervisor.
To do so, we need to get rid of the remaining global state that is kept
in the balloon struct, specifically the refuse_list. When the VM
communicates with the hypervisor, it does not free or put back pages
to the balloon list and instead only moves the pages whose status
indicated failure into a refuse_list on the stack. Once the operation
completes, the inflation or deflation functions handle the list
appropriately.
As we do that, we can consolidate the communication with the hypervisor
for both the lock and unlock operations into a single function. We also
reuse the deflation function for popping the balloon.
As a preparation for preventing races, we hold a spinlock when the
communication actually takes place, and use atomic operations for
updating the balloon size. The balloon page list is still racy and will
be handled in the next patch.
Reviewed-by: Xavier Deguillard <xdeguillard@vmware.com>
Signed-off-by: Nadav Amit <namit@vmware.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2018-09-21 01:30:18 +08:00
|
|
|
list_move(&page->lru, pages);
|
|
|
|
if (++i == n_req_pages)
|
|
|
|
break;
|
|
|
|
}
|
2019-04-25 19:54:43 +08:00
|
|
|
|
|
|
|
__count_vm_events(BALLOON_DEFLATE,
|
|
|
|
i * vmballoon_page_in_frames(VMW_BALLOON_2M_PAGE));
|
|
|
|
spin_unlock_irqrestore(&b->b_dev_info.pages_lock, flags);
|
vmw_balloon: rework the inflate and deflate loops
In preparation for supporting compaction and OOM notification, this
patch reworks the inflate/deflate loops. The main idea is to separate
the allocation, communication with the hypervisor, and the handling of
errors from each other. Doing will allow us to perform concurrent
inflation and deflation, excluding the actual communication with the
hypervisor.
To do so, we need to get rid of the remaining global state that is kept
in the balloon struct, specifically the refuse_list. When the VM
communicates with the hypervisor, it does not free or put back pages
to the balloon list and instead only moves the pages whose status
indicated failure into a refuse_list on the stack. Once the operation
completes, the inflation or deflation functions handle the list
appropriately.
As we do that, we can consolidate the communication with the hypervisor
for both the lock and unlock operations into a single function. We also
reuse the deflation function for popping the balloon.
As a preparation for preventing races, we hold a spinlock when the
communication actually takes place, and use atomic operations for
updating the balloon size. The balloon page list is still racy and will
be handled in the next patch.
Reviewed-by: Xavier Deguillard <xdeguillard@vmware.com>
Signed-off-by: Nadav Amit <namit@vmware.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2018-09-21 01:30:18 +08:00
|
|
|
*n_pages = i;
|
|
|
|
}
|
|
|
|
|
2019-04-25 19:54:45 +08:00
|
|
|
/**
|
|
|
|
* vmballoon_split_refused_pages() - Split the 2MB refused pages to 4k.
|
|
|
|
*
|
|
|
|
* If inflation of 2MB pages was denied by the hypervisor, it is likely to be
|
|
|
|
* due to one or few 4KB pages. These 2MB pages may keep being allocated and
|
|
|
|
* then being refused. To prevent this case, this function splits the refused
|
|
|
|
* pages into 4KB pages and adds them into @prealloc_pages list.
|
|
|
|
*
|
|
|
|
* @ctl: pointer for the %struct vmballoon_ctl, which defines the operation.
|
|
|
|
*/
|
|
|
|
static void vmballoon_split_refused_pages(struct vmballoon_ctl *ctl)
|
|
|
|
{
|
|
|
|
struct page *page, *tmp;
|
|
|
|
unsigned int i, order;
|
|
|
|
|
|
|
|
order = vmballoon_page_order(ctl->page_size);
|
|
|
|
|
|
|
|
list_for_each_entry_safe(page, tmp, &ctl->refused_pages, lru) {
|
|
|
|
list_del(&page->lru);
|
|
|
|
split_page(page, order);
|
|
|
|
for (i = 0; i < (1 << order); i++)
|
|
|
|
list_add(&page[i].lru, &ctl->prealloc_pages);
|
|
|
|
}
|
|
|
|
ctl->n_refused_pages = 0;
|
|
|
|
}
|
|
|
|
|
vmw_balloon: rework the inflate and deflate loops
In preparation for supporting compaction and OOM notification, this
patch reworks the inflate/deflate loops. The main idea is to separate
the allocation, communication with the hypervisor, and the handling of
errors from each other. Doing will allow us to perform concurrent
inflation and deflation, excluding the actual communication with the
hypervisor.
To do so, we need to get rid of the remaining global state that is kept
in the balloon struct, specifically the refuse_list. When the VM
communicates with the hypervisor, it does not free or put back pages
to the balloon list and instead only moves the pages whose status
indicated failure into a refuse_list on the stack. Once the operation
completes, the inflation or deflation functions handle the list
appropriately.
As we do that, we can consolidate the communication with the hypervisor
for both the lock and unlock operations into a single function. We also
reuse the deflation function for popping the balloon.
As a preparation for preventing races, we hold a spinlock when the
communication actually takes place, and use atomic operations for
updating the balloon size. The balloon page list is still racy and will
be handled in the next patch.
Reviewed-by: Xavier Deguillard <xdeguillard@vmware.com>
Signed-off-by: Nadav Amit <namit@vmware.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2018-09-21 01:30:18 +08:00
|
|
|
/**
|
|
|
|
* vmballoon_inflate() - Inflate the balloon towards its target size.
|
|
|
|
*
|
|
|
|
* @b: pointer to the balloon.
|
2010-04-24 01:18:08 +08:00
|
|
|
*/
|
|
|
|
static void vmballoon_inflate(struct vmballoon *b)
|
|
|
|
{
|
vmw_balloon: rework the inflate and deflate loops
In preparation for supporting compaction and OOM notification, this
patch reworks the inflate/deflate loops. The main idea is to separate
the allocation, communication with the hypervisor, and the handling of
errors from each other. Doing will allow us to perform concurrent
inflation and deflation, excluding the actual communication with the
hypervisor.
To do so, we need to get rid of the remaining global state that is kept
in the balloon struct, specifically the refuse_list. When the VM
communicates with the hypervisor, it does not free or put back pages
to the balloon list and instead only moves the pages whose status
indicated failure into a refuse_list on the stack. Once the operation
completes, the inflation or deflation functions handle the list
appropriately.
As we do that, we can consolidate the communication with the hypervisor
for both the lock and unlock operations into a single function. We also
reuse the deflation function for popping the balloon.
As a preparation for preventing races, we hold a spinlock when the
communication actually takes place, and use atomic operations for
updating the balloon size. The balloon page list is still racy and will
be handled in the next patch.
Reviewed-by: Xavier Deguillard <xdeguillard@vmware.com>
Signed-off-by: Nadav Amit <namit@vmware.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2018-09-21 01:30:18 +08:00
|
|
|
int64_t to_inflate_frames;
|
|
|
|
struct vmballoon_ctl ctl = {
|
|
|
|
.pages = LIST_HEAD_INIT(ctl.pages),
|
|
|
|
.refused_pages = LIST_HEAD_INIT(ctl.refused_pages),
|
2019-04-25 19:54:45 +08:00
|
|
|
.prealloc_pages = LIST_HEAD_INIT(ctl.prealloc_pages),
|
vmw_balloon: rework the inflate and deflate loops
In preparation for supporting compaction and OOM notification, this
patch reworks the inflate/deflate loops. The main idea is to separate
the allocation, communication with the hypervisor, and the handling of
errors from each other. Doing will allow us to perform concurrent
inflation and deflation, excluding the actual communication with the
hypervisor.
To do so, we need to get rid of the remaining global state that is kept
in the balloon struct, specifically the refuse_list. When the VM
communicates with the hypervisor, it does not free or put back pages
to the balloon list and instead only moves the pages whose status
indicated failure into a refuse_list on the stack. Once the operation
completes, the inflation or deflation functions handle the list
appropriately.
As we do that, we can consolidate the communication with the hypervisor
for both the lock and unlock operations into a single function. We also
reuse the deflation function for popping the balloon.
As a preparation for preventing races, we hold a spinlock when the
communication actually takes place, and use atomic operations for
updating the balloon size. The balloon page list is still racy and will
be handled in the next patch.
Reviewed-by: Xavier Deguillard <xdeguillard@vmware.com>
Signed-off-by: Nadav Amit <namit@vmware.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2018-09-21 01:30:18 +08:00
|
|
|
.page_size = b->max_page_size,
|
|
|
|
.op = VMW_BALLOON_INFLATE
|
|
|
|
};
|
2010-04-24 01:18:08 +08:00
|
|
|
|
vmw_balloon: rework the inflate and deflate loops
In preparation for supporting compaction and OOM notification, this
patch reworks the inflate/deflate loops. The main idea is to separate
the allocation, communication with the hypervisor, and the handling of
errors from each other. Doing will allow us to perform concurrent
inflation and deflation, excluding the actual communication with the
hypervisor.
To do so, we need to get rid of the remaining global state that is kept
in the balloon struct, specifically the refuse_list. When the VM
communicates with the hypervisor, it does not free or put back pages
to the balloon list and instead only moves the pages whose status
indicated failure into a refuse_list on the stack. Once the operation
completes, the inflation or deflation functions handle the list
appropriately.
As we do that, we can consolidate the communication with the hypervisor
for both the lock and unlock operations into a single function. We also
reuse the deflation function for popping the balloon.
As a preparation for preventing races, we hold a spinlock when the
communication actually takes place, and use atomic operations for
updating the balloon size. The balloon page list is still racy and will
be handled in the next patch.
Reviewed-by: Xavier Deguillard <xdeguillard@vmware.com>
Signed-off-by: Nadav Amit <namit@vmware.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2018-09-21 01:30:18 +08:00
|
|
|
while ((to_inflate_frames = vmballoon_change(b)) > 0) {
|
|
|
|
unsigned int to_inflate_pages, page_in_frames;
|
|
|
|
int alloc_error, lock_error = 0;
|
2010-04-24 01:18:08 +08:00
|
|
|
|
vmw_balloon: rework the inflate and deflate loops
In preparation for supporting compaction and OOM notification, this
patch reworks the inflate/deflate loops. The main idea is to separate
the allocation, communication with the hypervisor, and the handling of
errors from each other. Doing will allow us to perform concurrent
inflation and deflation, excluding the actual communication with the
hypervisor.
To do so, we need to get rid of the remaining global state that is kept
in the balloon struct, specifically the refuse_list. When the VM
communicates with the hypervisor, it does not free or put back pages
to the balloon list and instead only moves the pages whose status
indicated failure into a refuse_list on the stack. Once the operation
completes, the inflation or deflation functions handle the list
appropriately.
As we do that, we can consolidate the communication with the hypervisor
for both the lock and unlock operations into a single function. We also
reuse the deflation function for popping the balloon.
As a preparation for preventing races, we hold a spinlock when the
communication actually takes place, and use atomic operations for
updating the balloon size. The balloon page list is still racy and will
be handled in the next patch.
Reviewed-by: Xavier Deguillard <xdeguillard@vmware.com>
Signed-off-by: Nadav Amit <namit@vmware.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2018-09-21 01:30:18 +08:00
|
|
|
VM_BUG_ON(!list_empty(&ctl.pages));
|
|
|
|
VM_BUG_ON(ctl.n_pages != 0);
|
2010-04-24 01:18:08 +08:00
|
|
|
|
vmw_balloon: rework the inflate and deflate loops
In preparation for supporting compaction and OOM notification, this
patch reworks the inflate/deflate loops. The main idea is to separate
the allocation, communication with the hypervisor, and the handling of
errors from each other. Doing will allow us to perform concurrent
inflation and deflation, excluding the actual communication with the
hypervisor.
To do so, we need to get rid of the remaining global state that is kept
in the balloon struct, specifically the refuse_list. When the VM
communicates with the hypervisor, it does not free or put back pages
to the balloon list and instead only moves the pages whose status
indicated failure into a refuse_list on the stack. Once the operation
completes, the inflation or deflation functions handle the list
appropriately.
As we do that, we can consolidate the communication with the hypervisor
for both the lock and unlock operations into a single function. We also
reuse the deflation function for popping the balloon.
As a preparation for preventing races, we hold a spinlock when the
communication actually takes place, and use atomic operations for
updating the balloon size. The balloon page list is still racy and will
be handled in the next patch.
Reviewed-by: Xavier Deguillard <xdeguillard@vmware.com>
Signed-off-by: Nadav Amit <namit@vmware.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2018-09-21 01:30:18 +08:00
|
|
|
page_in_frames = vmballoon_page_in_frames(ctl.page_size);
|
2010-04-24 01:18:08 +08:00
|
|
|
|
vmw_balloon: rework the inflate and deflate loops
In preparation for supporting compaction and OOM notification, this
patch reworks the inflate/deflate loops. The main idea is to separate
the allocation, communication with the hypervisor, and the handling of
errors from each other. Doing will allow us to perform concurrent
inflation and deflation, excluding the actual communication with the
hypervisor.
To do so, we need to get rid of the remaining global state that is kept
in the balloon struct, specifically the refuse_list. When the VM
communicates with the hypervisor, it does not free or put back pages
to the balloon list and instead only moves the pages whose status
indicated failure into a refuse_list on the stack. Once the operation
completes, the inflation or deflation functions handle the list
appropriately.
As we do that, we can consolidate the communication with the hypervisor
for both the lock and unlock operations into a single function. We also
reuse the deflation function for popping the balloon.
As a preparation for preventing races, we hold a spinlock when the
communication actually takes place, and use atomic operations for
updating the balloon size. The balloon page list is still racy and will
be handled in the next patch.
Reviewed-by: Xavier Deguillard <xdeguillard@vmware.com>
Signed-off-by: Nadav Amit <namit@vmware.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2018-09-21 01:30:18 +08:00
|
|
|
to_inflate_pages = min_t(unsigned long, b->batch_max_pages,
|
|
|
|
DIV_ROUND_UP_ULL(to_inflate_frames,
|
|
|
|
page_in_frames));
|
|
|
|
|
|
|
|
/* Start by allocating */
|
|
|
|
alloc_error = vmballoon_alloc_page_list(b, &ctl,
|
|
|
|
to_inflate_pages);
|
|
|
|
|
|
|
|
/* Actually lock the pages by telling the hypervisor */
|
|
|
|
lock_error = vmballoon_lock(b, &ctl);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* If an error indicates that something serious went wrong,
|
|
|
|
* stop the inflation.
|
|
|
|
*/
|
|
|
|
if (lock_error)
|
2018-09-21 01:30:11 +08:00
|
|
|
break;
|
2010-04-24 01:18:08 +08:00
|
|
|
|
vmw_balloon: rework the inflate and deflate loops
In preparation for supporting compaction and OOM notification, this
patch reworks the inflate/deflate loops. The main idea is to separate
the allocation, communication with the hypervisor, and the handling of
errors from each other. Doing will allow us to perform concurrent
inflation and deflation, excluding the actual communication with the
hypervisor.
To do so, we need to get rid of the remaining global state that is kept
in the balloon struct, specifically the refuse_list. When the VM
communicates with the hypervisor, it does not free or put back pages
to the balloon list and instead only moves the pages whose status
indicated failure into a refuse_list on the stack. Once the operation
completes, the inflation or deflation functions handle the list
appropriately.
As we do that, we can consolidate the communication with the hypervisor
for both the lock and unlock operations into a single function. We also
reuse the deflation function for popping the balloon.
As a preparation for preventing races, we hold a spinlock when the
communication actually takes place, and use atomic operations for
updating the balloon size. The balloon page list is still racy and will
be handled in the next patch.
Reviewed-by: Xavier Deguillard <xdeguillard@vmware.com>
Signed-off-by: Nadav Amit <namit@vmware.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2018-09-21 01:30:18 +08:00
|
|
|
/* Update the balloon size */
|
|
|
|
atomic64_add(ctl.n_pages * page_in_frames, &b->size);
|
2018-09-21 01:30:13 +08:00
|
|
|
|
vmw_balloon: rework the inflate and deflate loops
In preparation for supporting compaction and OOM notification, this
patch reworks the inflate/deflate loops. The main idea is to separate
the allocation, communication with the hypervisor, and the handling of
errors from each other. Doing will allow us to perform concurrent
inflation and deflation, excluding the actual communication with the
hypervisor.
To do so, we need to get rid of the remaining global state that is kept
in the balloon struct, specifically the refuse_list. When the VM
communicates with the hypervisor, it does not free or put back pages
to the balloon list and instead only moves the pages whose status
indicated failure into a refuse_list on the stack. Once the operation
completes, the inflation or deflation functions handle the list
appropriately.
As we do that, we can consolidate the communication with the hypervisor
for both the lock and unlock operations into a single function. We also
reuse the deflation function for popping the balloon.
As a preparation for preventing races, we hold a spinlock when the
communication actually takes place, and use atomic operations for
updating the balloon size. The balloon page list is still racy and will
be handled in the next patch.
Reviewed-by: Xavier Deguillard <xdeguillard@vmware.com>
Signed-off-by: Nadav Amit <namit@vmware.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2018-09-21 01:30:18 +08:00
|
|
|
vmballoon_enqueue_page_list(b, &ctl.pages, &ctl.n_pages,
|
|
|
|
ctl.page_size);
|
2018-09-21 01:30:07 +08:00
|
|
|
|
vmw_balloon: rework the inflate and deflate loops
In preparation for supporting compaction and OOM notification, this
patch reworks the inflate/deflate loops. The main idea is to separate
the allocation, communication with the hypervisor, and the handling of
errors from each other. Doing will allow us to perform concurrent
inflation and deflation, excluding the actual communication with the
hypervisor.
To do so, we need to get rid of the remaining global state that is kept
in the balloon struct, specifically the refuse_list. When the VM
communicates with the hypervisor, it does not free or put back pages
to the balloon list and instead only moves the pages whose status
indicated failure into a refuse_list on the stack. Once the operation
completes, the inflation or deflation functions handle the list
appropriately.
As we do that, we can consolidate the communication with the hypervisor
for both the lock and unlock operations into a single function. We also
reuse the deflation function for popping the balloon.
As a preparation for preventing races, we hold a spinlock when the
communication actually takes place, and use atomic operations for
updating the balloon size. The balloon page list is still racy and will
be handled in the next patch.
Reviewed-by: Xavier Deguillard <xdeguillard@vmware.com>
Signed-off-by: Nadav Amit <namit@vmware.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2018-09-21 01:30:18 +08:00
|
|
|
/*
|
|
|
|
* If allocation failed or the number of refused pages exceeds
|
|
|
|
* the maximum allowed, move to the next page size.
|
|
|
|
*/
|
|
|
|
if (alloc_error ||
|
|
|
|
ctl.n_refused_pages >= VMW_BALLOON_MAX_REFUSED) {
|
|
|
|
if (ctl.page_size == VMW_BALLOON_4K_PAGE)
|
|
|
|
break;
|
2018-09-21 01:30:13 +08:00
|
|
|
|
|
|
|
/*
|
2019-04-25 19:54:45 +08:00
|
|
|
* Split the refused pages to 4k. This will also empty
|
|
|
|
* the refused pages list.
|
2018-09-21 01:30:13 +08:00
|
|
|
*/
|
2019-04-25 19:54:45 +08:00
|
|
|
vmballoon_split_refused_pages(&ctl);
|
vmw_balloon: rework the inflate and deflate loops
In preparation for supporting compaction and OOM notification, this
patch reworks the inflate/deflate loops. The main idea is to separate
the allocation, communication with the hypervisor, and the handling of
errors from each other. Doing will allow us to perform concurrent
inflation and deflation, excluding the actual communication with the
hypervisor.
To do so, we need to get rid of the remaining global state that is kept
in the balloon struct, specifically the refuse_list. When the VM
communicates with the hypervisor, it does not free or put back pages
to the balloon list and instead only moves the pages whose status
indicated failure into a refuse_list on the stack. Once the operation
completes, the inflation or deflation functions handle the list
appropriately.
As we do that, we can consolidate the communication with the hypervisor
for both the lock and unlock operations into a single function. We also
reuse the deflation function for popping the balloon.
As a preparation for preventing races, we hold a spinlock when the
communication actually takes place, and use atomic operations for
updating the balloon size. The balloon page list is still racy and will
be handled in the next patch.
Reviewed-by: Xavier Deguillard <xdeguillard@vmware.com>
Signed-off-by: Nadav Amit <namit@vmware.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2018-09-21 01:30:18 +08:00
|
|
|
ctl.page_size--;
|
2015-08-07 06:17:58 +08:00
|
|
|
}
|
2015-06-13 02:43:22 +08:00
|
|
|
|
2015-08-07 06:18:01 +08:00
|
|
|
cond_resched();
|
2010-04-24 01:18:08 +08:00
|
|
|
}
|
|
|
|
|
vmw_balloon: rework the inflate and deflate loops
In preparation for supporting compaction and OOM notification, this
patch reworks the inflate/deflate loops. The main idea is to separate
the allocation, communication with the hypervisor, and the handling of
errors from each other. Doing will allow us to perform concurrent
inflation and deflation, excluding the actual communication with the
hypervisor.
To do so, we need to get rid of the remaining global state that is kept
in the balloon struct, specifically the refuse_list. When the VM
communicates with the hypervisor, it does not free or put back pages
to the balloon list and instead only moves the pages whose status
indicated failure into a refuse_list on the stack. Once the operation
completes, the inflation or deflation functions handle the list
appropriately.
As we do that, we can consolidate the communication with the hypervisor
for both the lock and unlock operations into a single function. We also
reuse the deflation function for popping the balloon.
As a preparation for preventing races, we hold a spinlock when the
communication actually takes place, and use atomic operations for
updating the balloon size. The balloon page list is still racy and will
be handled in the next patch.
Reviewed-by: Xavier Deguillard <xdeguillard@vmware.com>
Signed-off-by: Nadav Amit <namit@vmware.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2018-09-21 01:30:18 +08:00
|
|
|
/*
|
|
|
|
* Release pages that were allocated while attempting to inflate the
|
|
|
|
* balloon but were refused by the host for one reason or another,
|
|
|
|
* and update the statistics.
|
|
|
|
*/
|
|
|
|
if (ctl.n_refused_pages != 0)
|
|
|
|
vmballoon_release_refused_pages(b, &ctl);
|
2019-04-25 19:54:45 +08:00
|
|
|
|
|
|
|
vmballoon_release_page_list(&ctl.prealloc_pages, NULL, ctl.page_size);
|
2010-04-24 01:18:08 +08:00
|
|
|
}
|
|
|
|
|
vmw_balloon: rework the inflate and deflate loops
In preparation for supporting compaction and OOM notification, this
patch reworks the inflate/deflate loops. The main idea is to separate
the allocation, communication with the hypervisor, and the handling of
errors from each other. Doing will allow us to perform concurrent
inflation and deflation, excluding the actual communication with the
hypervisor.
To do so, we need to get rid of the remaining global state that is kept
in the balloon struct, specifically the refuse_list. When the VM
communicates with the hypervisor, it does not free or put back pages
to the balloon list and instead only moves the pages whose status
indicated failure into a refuse_list on the stack. Once the operation
completes, the inflation or deflation functions handle the list
appropriately.
As we do that, we can consolidate the communication with the hypervisor
for both the lock and unlock operations into a single function. We also
reuse the deflation function for popping the balloon.
As a preparation for preventing races, we hold a spinlock when the
communication actually takes place, and use atomic operations for
updating the balloon size. The balloon page list is still racy and will
be handled in the next patch.
Reviewed-by: Xavier Deguillard <xdeguillard@vmware.com>
Signed-off-by: Nadav Amit <namit@vmware.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2018-09-21 01:30:18 +08:00
|
|
|
/**
|
|
|
|
* vmballoon_deflate() - Decrease the size of the balloon.
|
|
|
|
*
|
|
|
|
* @b: pointer to the balloon
|
|
|
|
* @n_frames: the number of frames to deflate. If zero, automatically
|
|
|
|
* calculated according to the target size.
|
|
|
|
* @coordinated: whether to coordinate with the host
|
|
|
|
*
|
2010-04-24 01:18:08 +08:00
|
|
|
* Decrease the size of the balloon allowing guest to use more memory.
|
vmw_balloon: rework the inflate and deflate loops
In preparation for supporting compaction and OOM notification, this
patch reworks the inflate/deflate loops. The main idea is to separate
the allocation, communication with the hypervisor, and the handling of
errors from each other. Doing will allow us to perform concurrent
inflation and deflation, excluding the actual communication with the
hypervisor.
To do so, we need to get rid of the remaining global state that is kept
in the balloon struct, specifically the refuse_list. When the VM
communicates with the hypervisor, it does not free or put back pages
to the balloon list and instead only moves the pages whose status
indicated failure into a refuse_list on the stack. Once the operation
completes, the inflation or deflation functions handle the list
appropriately.
As we do that, we can consolidate the communication with the hypervisor
for both the lock and unlock operations into a single function. We also
reuse the deflation function for popping the balloon.
As a preparation for preventing races, we hold a spinlock when the
communication actually takes place, and use atomic operations for
updating the balloon size. The balloon page list is still racy and will
be handled in the next patch.
Reviewed-by: Xavier Deguillard <xdeguillard@vmware.com>
Signed-off-by: Nadav Amit <namit@vmware.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2018-09-21 01:30:18 +08:00
|
|
|
*
|
|
|
|
* Return: The number of deflated frames (i.e., basic page size units)
|
2010-04-24 01:18:08 +08:00
|
|
|
*/
|
vmw_balloon: rework the inflate and deflate loops
In preparation for supporting compaction and OOM notification, this
patch reworks the inflate/deflate loops. The main idea is to separate
the allocation, communication with the hypervisor, and the handling of
errors from each other. Doing will allow us to perform concurrent
inflation and deflation, excluding the actual communication with the
hypervisor.
To do so, we need to get rid of the remaining global state that is kept
in the balloon struct, specifically the refuse_list. When the VM
communicates with the hypervisor, it does not free or put back pages
to the balloon list and instead only moves the pages whose status
indicated failure into a refuse_list on the stack. Once the operation
completes, the inflation or deflation functions handle the list
appropriately.
As we do that, we can consolidate the communication with the hypervisor
for both the lock and unlock operations into a single function. We also
reuse the deflation function for popping the balloon.
As a preparation for preventing races, we hold a spinlock when the
communication actually takes place, and use atomic operations for
updating the balloon size. The balloon page list is still racy and will
be handled in the next patch.
Reviewed-by: Xavier Deguillard <xdeguillard@vmware.com>
Signed-off-by: Nadav Amit <namit@vmware.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2018-09-21 01:30:18 +08:00
|
|
|
static unsigned long vmballoon_deflate(struct vmballoon *b, uint64_t n_frames,
|
|
|
|
bool coordinated)
|
2010-04-24 01:18:08 +08:00
|
|
|
{
|
vmw_balloon: rework the inflate and deflate loops
In preparation for supporting compaction and OOM notification, this
patch reworks the inflate/deflate loops. The main idea is to separate
the allocation, communication with the hypervisor, and the handling of
errors from each other. Doing will allow us to perform concurrent
inflation and deflation, excluding the actual communication with the
hypervisor.
To do so, we need to get rid of the remaining global state that is kept
in the balloon struct, specifically the refuse_list. When the VM
communicates with the hypervisor, it does not free or put back pages
to the balloon list and instead only moves the pages whose status
indicated failure into a refuse_list on the stack. Once the operation
completes, the inflation or deflation functions handle the list
appropriately.
As we do that, we can consolidate the communication with the hypervisor
for both the lock and unlock operations into a single function. We also
reuse the deflation function for popping the balloon.
As a preparation for preventing races, we hold a spinlock when the
communication actually takes place, and use atomic operations for
updating the balloon size. The balloon page list is still racy and will
be handled in the next patch.
Reviewed-by: Xavier Deguillard <xdeguillard@vmware.com>
Signed-off-by: Nadav Amit <namit@vmware.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2018-09-21 01:30:18 +08:00
|
|
|
unsigned long deflated_frames = 0;
|
|
|
|
unsigned long tried_frames = 0;
|
|
|
|
struct vmballoon_ctl ctl = {
|
|
|
|
.pages = LIST_HEAD_INIT(ctl.pages),
|
|
|
|
.refused_pages = LIST_HEAD_INIT(ctl.refused_pages),
|
|
|
|
.page_size = VMW_BALLOON_4K_PAGE,
|
|
|
|
.op = VMW_BALLOON_DEFLATE
|
|
|
|
};
|
2010-04-24 01:18:08 +08:00
|
|
|
|
|
|
|
/* free pages to reach target */
|
vmw_balloon: rework the inflate and deflate loops
In preparation for supporting compaction and OOM notification, this
patch reworks the inflate/deflate loops. The main idea is to separate
the allocation, communication with the hypervisor, and the handling of
errors from each other. Doing will allow us to perform concurrent
inflation and deflation, excluding the actual communication with the
hypervisor.
To do so, we need to get rid of the remaining global state that is kept
in the balloon struct, specifically the refuse_list. When the VM
communicates with the hypervisor, it does not free or put back pages
to the balloon list and instead only moves the pages whose status
indicated failure into a refuse_list on the stack. Once the operation
completes, the inflation or deflation functions handle the list
appropriately.
As we do that, we can consolidate the communication with the hypervisor
for both the lock and unlock operations into a single function. We also
reuse the deflation function for popping the balloon.
As a preparation for preventing races, we hold a spinlock when the
communication actually takes place, and use atomic operations for
updating the balloon size. The balloon page list is still racy and will
be handled in the next patch.
Reviewed-by: Xavier Deguillard <xdeguillard@vmware.com>
Signed-off-by: Nadav Amit <namit@vmware.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2018-09-21 01:30:18 +08:00
|
|
|
while (true) {
|
|
|
|
unsigned int to_deflate_pages, n_unlocked_frames;
|
|
|
|
unsigned int page_in_frames;
|
|
|
|
int64_t to_deflate_frames;
|
|
|
|
bool deflated_all;
|
|
|
|
|
|
|
|
page_in_frames = vmballoon_page_in_frames(ctl.page_size);
|
|
|
|
|
|
|
|
VM_BUG_ON(!list_empty(&ctl.pages));
|
|
|
|
VM_BUG_ON(ctl.n_pages);
|
|
|
|
VM_BUG_ON(!list_empty(&ctl.refused_pages));
|
|
|
|
VM_BUG_ON(ctl.n_refused_pages);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* If we were requested a specific number of frames, we try to
|
|
|
|
* deflate this number of frames. Otherwise, deflation is
|
|
|
|
* performed according to the target and balloon size.
|
|
|
|
*/
|
|
|
|
to_deflate_frames = n_frames ? n_frames - tried_frames :
|
|
|
|
-vmballoon_change(b);
|
|
|
|
|
|
|
|
/* break if no work to do */
|
|
|
|
if (to_deflate_frames <= 0)
|
|
|
|
break;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Calculate the number of frames based on current page size,
|
|
|
|
* but limit the deflated frames to a single chunk
|
|
|
|
*/
|
|
|
|
to_deflate_pages = min_t(unsigned long, b->batch_max_pages,
|
|
|
|
DIV_ROUND_UP_ULL(to_deflate_frames,
|
|
|
|
page_in_frames));
|
|
|
|
|
|
|
|
/* First take the pages from the balloon pages. */
|
|
|
|
vmballoon_dequeue_page_list(b, &ctl.pages, &ctl.n_pages,
|
|
|
|
ctl.page_size, to_deflate_pages);
|
2015-08-07 06:17:58 +08:00
|
|
|
|
vmw_balloon: rework the inflate and deflate loops
In preparation for supporting compaction and OOM notification, this
patch reworks the inflate/deflate loops. The main idea is to separate
the allocation, communication with the hypervisor, and the handling of
errors from each other. Doing will allow us to perform concurrent
inflation and deflation, excluding the actual communication with the
hypervisor.
To do so, we need to get rid of the remaining global state that is kept
in the balloon struct, specifically the refuse_list. When the VM
communicates with the hypervisor, it does not free or put back pages
to the balloon list and instead only moves the pages whose status
indicated failure into a refuse_list on the stack. Once the operation
completes, the inflation or deflation functions handle the list
appropriately.
As we do that, we can consolidate the communication with the hypervisor
for both the lock and unlock operations into a single function. We also
reuse the deflation function for popping the balloon.
As a preparation for preventing races, we hold a spinlock when the
communication actually takes place, and use atomic operations for
updating the balloon size. The balloon page list is still racy and will
be handled in the next patch.
Reviewed-by: Xavier Deguillard <xdeguillard@vmware.com>
Signed-off-by: Nadav Amit <namit@vmware.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2018-09-21 01:30:18 +08:00
|
|
|
/*
|
|
|
|
* Before pages are moving to the refused list, count their
|
|
|
|
* frames as frames that we tried to deflate.
|
|
|
|
*/
|
|
|
|
tried_frames += ctl.n_pages * page_in_frames;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Unlock the pages by communicating with the hypervisor if the
|
|
|
|
* communication is coordinated (i.e., not pop). We ignore the
|
|
|
|
* return code. Instead we check if all the pages we manage to
|
|
|
|
* unlock all the pages. If we failed, we will move to the next
|
|
|
|
* page size, and would eventually try again later.
|
|
|
|
*/
|
|
|
|
if (coordinated)
|
|
|
|
vmballoon_lock(b, &ctl);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Check if we deflated enough. We will move to the next page
|
|
|
|
* size if we did not manage to do so. This calculation takes
|
|
|
|
* place now, as once the pages are released, the number of
|
|
|
|
* pages is zeroed.
|
|
|
|
*/
|
|
|
|
deflated_all = (ctl.n_pages == to_deflate_pages);
|
2015-08-07 06:18:01 +08:00
|
|
|
|
vmw_balloon: rework the inflate and deflate loops
In preparation for supporting compaction and OOM notification, this
patch reworks the inflate/deflate loops. The main idea is to separate
the allocation, communication with the hypervisor, and the handling of
errors from each other. Doing will allow us to perform concurrent
inflation and deflation, excluding the actual communication with the
hypervisor.
To do so, we need to get rid of the remaining global state that is kept
in the balloon struct, specifically the refuse_list. When the VM
communicates with the hypervisor, it does not free or put back pages
to the balloon list and instead only moves the pages whose status
indicated failure into a refuse_list on the stack. Once the operation
completes, the inflation or deflation functions handle the list
appropriately.
As we do that, we can consolidate the communication with the hypervisor
for both the lock and unlock operations into a single function. We also
reuse the deflation function for popping the balloon.
As a preparation for preventing races, we hold a spinlock when the
communication actually takes place, and use atomic operations for
updating the balloon size. The balloon page list is still racy and will
be handled in the next patch.
Reviewed-by: Xavier Deguillard <xdeguillard@vmware.com>
Signed-off-by: Nadav Amit <namit@vmware.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2018-09-21 01:30:18 +08:00
|
|
|
/* Update local and global counters */
|
|
|
|
n_unlocked_frames = ctl.n_pages * page_in_frames;
|
|
|
|
atomic64_sub(n_unlocked_frames, &b->size);
|
|
|
|
deflated_frames += n_unlocked_frames;
|
2010-04-24 01:18:08 +08:00
|
|
|
|
vmw_balloon: rework the inflate and deflate loops
In preparation for supporting compaction and OOM notification, this
patch reworks the inflate/deflate loops. The main idea is to separate
the allocation, communication with the hypervisor, and the handling of
errors from each other. Doing will allow us to perform concurrent
inflation and deflation, excluding the actual communication with the
hypervisor.
To do so, we need to get rid of the remaining global state that is kept
in the balloon struct, specifically the refuse_list. When the VM
communicates with the hypervisor, it does not free or put back pages
to the balloon list and instead only moves the pages whose status
indicated failure into a refuse_list on the stack. Once the operation
completes, the inflation or deflation functions handle the list
appropriately.
As we do that, we can consolidate the communication with the hypervisor
for both the lock and unlock operations into a single function. We also
reuse the deflation function for popping the balloon.
As a preparation for preventing races, we hold a spinlock when the
communication actually takes place, and use atomic operations for
updating the balloon size. The balloon page list is still racy and will
be handled in the next patch.
Reviewed-by: Xavier Deguillard <xdeguillard@vmware.com>
Signed-off-by: Nadav Amit <namit@vmware.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2018-09-21 01:30:18 +08:00
|
|
|
vmballoon_stats_page_add(b, VMW_BALLOON_PAGE_STAT_FREE,
|
|
|
|
ctl.page_size, ctl.n_pages);
|
2015-08-07 06:18:01 +08:00
|
|
|
|
vmw_balloon: rework the inflate and deflate loops
In preparation for supporting compaction and OOM notification, this
patch reworks the inflate/deflate loops. The main idea is to separate
the allocation, communication with the hypervisor, and the handling of
errors from each other. Doing will allow us to perform concurrent
inflation and deflation, excluding the actual communication with the
hypervisor.
To do so, we need to get rid of the remaining global state that is kept
in the balloon struct, specifically the refuse_list. When the VM
communicates with the hypervisor, it does not free or put back pages
to the balloon list and instead only moves the pages whose status
indicated failure into a refuse_list on the stack. Once the operation
completes, the inflation or deflation functions handle the list
appropriately.
As we do that, we can consolidate the communication with the hypervisor
for both the lock and unlock operations into a single function. We also
reuse the deflation function for popping the balloon.
As a preparation for preventing races, we hold a spinlock when the
communication actually takes place, and use atomic operations for
updating the balloon size. The balloon page list is still racy and will
be handled in the next patch.
Reviewed-by: Xavier Deguillard <xdeguillard@vmware.com>
Signed-off-by: Nadav Amit <namit@vmware.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2018-09-21 01:30:18 +08:00
|
|
|
/* free the ballooned pages */
|
|
|
|
vmballoon_release_page_list(&ctl.pages, &ctl.n_pages,
|
|
|
|
ctl.page_size);
|
|
|
|
|
|
|
|
/* Return the refused pages to the ballooned list. */
|
|
|
|
vmballoon_enqueue_page_list(b, &ctl.refused_pages,
|
|
|
|
&ctl.n_refused_pages,
|
|
|
|
ctl.page_size);
|
|
|
|
|
|
|
|
/* If we failed to unlock all the pages, move to next size. */
|
|
|
|
if (!deflated_all) {
|
|
|
|
if (ctl.page_size == b->max_page_size)
|
|
|
|
break;
|
|
|
|
ctl.page_size++;
|
2015-08-07 06:18:01 +08:00
|
|
|
}
|
2010-04-24 01:18:08 +08:00
|
|
|
|
vmw_balloon: rework the inflate and deflate loops
In preparation for supporting compaction and OOM notification, this
patch reworks the inflate/deflate loops. The main idea is to separate
the allocation, communication with the hypervisor, and the handling of
errors from each other. Doing will allow us to perform concurrent
inflation and deflation, excluding the actual communication with the
hypervisor.
To do so, we need to get rid of the remaining global state that is kept
in the balloon struct, specifically the refuse_list. When the VM
communicates with the hypervisor, it does not free or put back pages
to the balloon list and instead only moves the pages whose status
indicated failure into a refuse_list on the stack. Once the operation
completes, the inflation or deflation functions handle the list
appropriately.
As we do that, we can consolidate the communication with the hypervisor
for both the lock and unlock operations into a single function. We also
reuse the deflation function for popping the balloon.
As a preparation for preventing races, we hold a spinlock when the
communication actually takes place, and use atomic operations for
updating the balloon size. The balloon page list is still racy and will
be handled in the next patch.
Reviewed-by: Xavier Deguillard <xdeguillard@vmware.com>
Signed-off-by: Nadav Amit <namit@vmware.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2018-09-21 01:30:18 +08:00
|
|
|
cond_resched();
|
2015-08-07 06:18:01 +08:00
|
|
|
}
|
vmw_balloon: rework the inflate and deflate loops
In preparation for supporting compaction and OOM notification, this
patch reworks the inflate/deflate loops. The main idea is to separate
the allocation, communication with the hypervisor, and the handling of
errors from each other. Doing will allow us to perform concurrent
inflation and deflation, excluding the actual communication with the
hypervisor.
To do so, we need to get rid of the remaining global state that is kept
in the balloon struct, specifically the refuse_list. When the VM
communicates with the hypervisor, it does not free or put back pages
to the balloon list and instead only moves the pages whose status
indicated failure into a refuse_list on the stack. Once the operation
completes, the inflation or deflation functions handle the list
appropriately.
As we do that, we can consolidate the communication with the hypervisor
for both the lock and unlock operations into a single function. We also
reuse the deflation function for popping the balloon.
As a preparation for preventing races, we hold a spinlock when the
communication actually takes place, and use atomic operations for
updating the balloon size. The balloon page list is still racy and will
be handled in the next patch.
Reviewed-by: Xavier Deguillard <xdeguillard@vmware.com>
Signed-off-by: Nadav Amit <namit@vmware.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2018-09-21 01:30:18 +08:00
|
|
|
|
|
|
|
return deflated_frames;
|
2015-08-07 06:17:58 +08:00
|
|
|
}
|
|
|
|
|
vmw_balloon: change batch/single lock abstractions
The current abstractions for batch vs single operations seem suboptimal
and complicate the implementation of additional features (OOM,
compaction).
The immediate problem of the current abstractions is that they cause
differences in how operations are handled when batching is on or off.
For example, the refused_alloc counter is not updated when batching is
on. These discrepancies are caused by code redundancies.
Instead, this patch presents three type of operations, according to
whether batching is on or off: (1) add page, (2) communication with the
hypervisor and (3) retrieving the status of a page.
To avoid the overhead of virtual functions, and since we do not expect
additional interfaces for communication with the hypervisor, we use
static keys instead of virtual functions.
Finally, while we are at it, change vmballoon_init_batching() to return
int instead of bool, to be consistent in the return type and avoid
potential coding errors.
Reviewed-by: Xavier Deguillard <xdeguillard@vmware.com>
Signed-off-by: Nadav Amit <namit@vmware.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2018-09-21 01:30:12 +08:00
|
|
|
/**
|
|
|
|
* vmballoon_deinit_batching - disables batching mode.
|
|
|
|
*
|
|
|
|
* @b: pointer to &struct vmballoon.
|
|
|
|
*
|
|
|
|
* Disables batching, by deallocating the page for communication with the
|
|
|
|
* hypervisor and disabling the static key to indicate that batching is off.
|
|
|
|
*/
|
|
|
|
static void vmballoon_deinit_batching(struct vmballoon *b)
|
|
|
|
{
|
|
|
|
free_page((unsigned long)b->batch_page);
|
|
|
|
b->batch_page = NULL;
|
|
|
|
static_branch_disable(&vmw_balloon_batching);
|
|
|
|
b->batch_max_pages = 1;
|
|
|
|
}
|
2015-08-07 06:17:58 +08:00
|
|
|
|
vmw_balloon: change batch/single lock abstractions
The current abstractions for batch vs single operations seem suboptimal
and complicate the implementation of additional features (OOM,
compaction).
The immediate problem of the current abstractions is that they cause
differences in how operations are handled when batching is on or off.
For example, the refused_alloc counter is not updated when batching is
on. These discrepancies are caused by code redundancies.
Instead, this patch presents three type of operations, according to
whether batching is on or off: (1) add page, (2) communication with the
hypervisor and (3) retrieving the status of a page.
To avoid the overhead of virtual functions, and since we do not expect
additional interfaces for communication with the hypervisor, we use
static keys instead of virtual functions.
Finally, while we are at it, change vmballoon_init_batching() to return
int instead of bool, to be consistent in the return type and avoid
potential coding errors.
Reviewed-by: Xavier Deguillard <xdeguillard@vmware.com>
Signed-off-by: Nadav Amit <namit@vmware.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2018-09-21 01:30:12 +08:00
|
|
|
/**
|
|
|
|
* vmballoon_init_batching - enable batching mode.
|
|
|
|
*
|
|
|
|
* @b: pointer to &struct vmballoon.
|
|
|
|
*
|
|
|
|
* Enables batching, by allocating a page for communication with the hypervisor
|
|
|
|
* and enabling the static_key to use batching.
|
|
|
|
*
|
|
|
|
* Return: zero on success or an appropriate error-code.
|
|
|
|
*/
|
|
|
|
static int vmballoon_init_batching(struct vmballoon *b)
|
2015-08-07 06:17:58 +08:00
|
|
|
{
|
vmw_balloon: fixing double free when batching mode is off
The balloon.page field is used for two different purposes if batching is
on or off. If batching is on, the field point to the page which is used
to communicate with with the hypervisor. If it is off, balloon.page
points to the page that is about to be (un)locked.
Unfortunately, this dual-purpose of the field introduced a bug: when the
balloon is popped (e.g., when the machine is reset or the balloon driver
is explicitly removed), the balloon driver frees, unconditionally, the
page that is held in balloon.page. As a result, if batching is
disabled, this leads to double freeing the last page that is sent to the
hypervisor.
The following error occurs during rmmod when kernel checkers are on, and
the balloon is not empty:
[ 42.307653] ------------[ cut here ]------------
[ 42.307657] Kernel BUG at ffffffffba1e4b28 [verbose debug info unavailable]
[ 42.307720] invalid opcode: 0000 [#1] SMP DEBUG_PAGEALLOC
[ 42.312512] Modules linked in: vmw_vsock_vmci_transport vsock ppdev joydev vmw_balloon(-) input_leds serio_raw vmw_vmci parport_pc shpchp parport i2c_piix4 nfit mac_hid autofs4 vmwgfx drm_kms_helper hid_generic syscopyarea sysfillrect usbhid sysimgblt fb_sys_fops hid ttm mptspi scsi_transport_spi ahci mptscsih drm psmouse vmxnet3 libahci mptbase pata_acpi
[ 42.312766] CPU: 10 PID: 1527 Comm: rmmod Not tainted 4.12.0+ #5
[ 42.312803] Hardware name: VMware, Inc. VMware Virtual Platform/440BX Desktop Reference Platform, BIOS 6.00 09/30/2016
[ 42.313042] task: ffff9bf9680f8000 task.stack: ffffbfefc1638000
[ 42.313290] RIP: 0010:__free_pages+0x38/0x40
[ 42.313510] RSP: 0018:ffffbfefc163be98 EFLAGS: 00010246
[ 42.313731] RAX: 000000000000003e RBX: ffffffffc02b9720 RCX: 0000000000000006
[ 42.313972] RDX: 0000000000000000 RSI: 0000000000000000 RDI: ffff9bf97e08e0a0
[ 42.314201] RBP: ffffbfefc163be98 R08: 0000000000000000 R09: 0000000000000000
[ 42.314435] R10: 0000000000000000 R11: 0000000000000000 R12: ffffffffc02b97e4
[ 42.314505] R13: ffffffffc02b9748 R14: ffffffffc02b9728 R15: 0000000000000200
[ 42.314550] FS: 00007f3af5fec700(0000) GS:ffff9bf97e080000(0000) knlGS:0000000000000000
[ 42.314599] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[ 42.314635] CR2: 00007f44f6f4ab24 CR3: 00000003a7d12000 CR4: 00000000000006e0
[ 42.314864] Call Trace:
[ 42.315774] vmballoon_pop+0x102/0x130 [vmw_balloon]
[ 42.315816] vmballoon_exit+0x42/0xd64 [vmw_balloon]
[ 42.315853] SyS_delete_module+0x1e2/0x250
[ 42.315891] entry_SYSCALL_64_fastpath+0x23/0xc2
[ 42.315924] RIP: 0033:0x7f3af5b0e8e7
[ 42.315949] RSP: 002b:00007fffe6ce0148 EFLAGS: 00000206 ORIG_RAX: 00000000000000b0
[ 42.315996] RAX: ffffffffffffffda RBX: 000055be676401e0 RCX: 00007f3af5b0e8e7
[ 42.316951] RDX: 000000000000000a RSI: 0000000000000800 RDI: 000055be67640248
[ 42.317887] RBP: 0000000000000003 R08: 0000000000000000 R09: 1999999999999999
[ 42.318845] R10: 0000000000000883 R11: 0000000000000206 R12: 00007fffe6cdf130
[ 42.319755] R13: 0000000000000000 R14: 0000000000000000 R15: 000055be676401e0
[ 42.320606] Code: c0 74 1c f0 ff 4f 1c 74 02 5d c3 85 f6 74 07 e8 0f d8 ff ff 5d c3 31 f6 e8 c6 fb ff ff 5d c3 48 c7 c6 c8 0f c5 ba e8 58 be 02 00 <0f> 0b 66 0f 1f 44 00 00 66 66 66 66 90 48 85 ff 75 01 c3 55 48
[ 42.323462] RIP: __free_pages+0x38/0x40 RSP: ffffbfefc163be98
[ 42.325735] ---[ end trace 872e008e33f81508 ]---
To solve the bug, we eliminate the dual purpose of balloon.page.
Fixes: f220a80f0c2e ("VMware balloon: add batching to the vmw_balloon.")
Cc: stable@vger.kernel.org
Reported-by: Oleksandr Natalenko <onatalen@redhat.com>
Signed-off-by: Gil Kupfer <gilkup@gmail.com>
Signed-off-by: Nadav Amit <namit@vmware.com>
Reviewed-by: Xavier Deguillard <xdeguillard@vmware.com>
Tested-by: Oleksandr Natalenko <oleksandr@redhat.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2018-06-01 15:47:47 +08:00
|
|
|
struct page *page;
|
2015-08-07 06:17:58 +08:00
|
|
|
|
vmw_balloon: fixing double free when batching mode is off
The balloon.page field is used for two different purposes if batching is
on or off. If batching is on, the field point to the page which is used
to communicate with with the hypervisor. If it is off, balloon.page
points to the page that is about to be (un)locked.
Unfortunately, this dual-purpose of the field introduced a bug: when the
balloon is popped (e.g., when the machine is reset or the balloon driver
is explicitly removed), the balloon driver frees, unconditionally, the
page that is held in balloon.page. As a result, if batching is
disabled, this leads to double freeing the last page that is sent to the
hypervisor.
The following error occurs during rmmod when kernel checkers are on, and
the balloon is not empty:
[ 42.307653] ------------[ cut here ]------------
[ 42.307657] Kernel BUG at ffffffffba1e4b28 [verbose debug info unavailable]
[ 42.307720] invalid opcode: 0000 [#1] SMP DEBUG_PAGEALLOC
[ 42.312512] Modules linked in: vmw_vsock_vmci_transport vsock ppdev joydev vmw_balloon(-) input_leds serio_raw vmw_vmci parport_pc shpchp parport i2c_piix4 nfit mac_hid autofs4 vmwgfx drm_kms_helper hid_generic syscopyarea sysfillrect usbhid sysimgblt fb_sys_fops hid ttm mptspi scsi_transport_spi ahci mptscsih drm psmouse vmxnet3 libahci mptbase pata_acpi
[ 42.312766] CPU: 10 PID: 1527 Comm: rmmod Not tainted 4.12.0+ #5
[ 42.312803] Hardware name: VMware, Inc. VMware Virtual Platform/440BX Desktop Reference Platform, BIOS 6.00 09/30/2016
[ 42.313042] task: ffff9bf9680f8000 task.stack: ffffbfefc1638000
[ 42.313290] RIP: 0010:__free_pages+0x38/0x40
[ 42.313510] RSP: 0018:ffffbfefc163be98 EFLAGS: 00010246
[ 42.313731] RAX: 000000000000003e RBX: ffffffffc02b9720 RCX: 0000000000000006
[ 42.313972] RDX: 0000000000000000 RSI: 0000000000000000 RDI: ffff9bf97e08e0a0
[ 42.314201] RBP: ffffbfefc163be98 R08: 0000000000000000 R09: 0000000000000000
[ 42.314435] R10: 0000000000000000 R11: 0000000000000000 R12: ffffffffc02b97e4
[ 42.314505] R13: ffffffffc02b9748 R14: ffffffffc02b9728 R15: 0000000000000200
[ 42.314550] FS: 00007f3af5fec700(0000) GS:ffff9bf97e080000(0000) knlGS:0000000000000000
[ 42.314599] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[ 42.314635] CR2: 00007f44f6f4ab24 CR3: 00000003a7d12000 CR4: 00000000000006e0
[ 42.314864] Call Trace:
[ 42.315774] vmballoon_pop+0x102/0x130 [vmw_balloon]
[ 42.315816] vmballoon_exit+0x42/0xd64 [vmw_balloon]
[ 42.315853] SyS_delete_module+0x1e2/0x250
[ 42.315891] entry_SYSCALL_64_fastpath+0x23/0xc2
[ 42.315924] RIP: 0033:0x7f3af5b0e8e7
[ 42.315949] RSP: 002b:00007fffe6ce0148 EFLAGS: 00000206 ORIG_RAX: 00000000000000b0
[ 42.315996] RAX: ffffffffffffffda RBX: 000055be676401e0 RCX: 00007f3af5b0e8e7
[ 42.316951] RDX: 000000000000000a RSI: 0000000000000800 RDI: 000055be67640248
[ 42.317887] RBP: 0000000000000003 R08: 0000000000000000 R09: 1999999999999999
[ 42.318845] R10: 0000000000000883 R11: 0000000000000206 R12: 00007fffe6cdf130
[ 42.319755] R13: 0000000000000000 R14: 0000000000000000 R15: 000055be676401e0
[ 42.320606] Code: c0 74 1c f0 ff 4f 1c 74 02 5d c3 85 f6 74 07 e8 0f d8 ff ff 5d c3 31 f6 e8 c6 fb ff ff 5d c3 48 c7 c6 c8 0f c5 ba e8 58 be 02 00 <0f> 0b 66 0f 1f 44 00 00 66 66 66 66 90 48 85 ff 75 01 c3 55 48
[ 42.323462] RIP: __free_pages+0x38/0x40 RSP: ffffbfefc163be98
[ 42.325735] ---[ end trace 872e008e33f81508 ]---
To solve the bug, we eliminate the dual purpose of balloon.page.
Fixes: f220a80f0c2e ("VMware balloon: add batching to the vmw_balloon.")
Cc: stable@vger.kernel.org
Reported-by: Oleksandr Natalenko <onatalen@redhat.com>
Signed-off-by: Gil Kupfer <gilkup@gmail.com>
Signed-off-by: Nadav Amit <namit@vmware.com>
Reviewed-by: Xavier Deguillard <xdeguillard@vmware.com>
Tested-by: Oleksandr Natalenko <oleksandr@redhat.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2018-06-01 15:47:47 +08:00
|
|
|
page = alloc_page(GFP_KERNEL | __GFP_ZERO);
|
|
|
|
if (!page)
|
vmw_balloon: change batch/single lock abstractions
The current abstractions for batch vs single operations seem suboptimal
and complicate the implementation of additional features (OOM,
compaction).
The immediate problem of the current abstractions is that they cause
differences in how operations are handled when batching is on or off.
For example, the refused_alloc counter is not updated when batching is
on. These discrepancies are caused by code redundancies.
Instead, this patch presents three type of operations, according to
whether batching is on or off: (1) add page, (2) communication with the
hypervisor and (3) retrieving the status of a page.
To avoid the overhead of virtual functions, and since we do not expect
additional interfaces for communication with the hypervisor, we use
static keys instead of virtual functions.
Finally, while we are at it, change vmballoon_init_batching() to return
int instead of bool, to be consistent in the return type and avoid
potential coding errors.
Reviewed-by: Xavier Deguillard <xdeguillard@vmware.com>
Signed-off-by: Nadav Amit <namit@vmware.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2018-09-21 01:30:12 +08:00
|
|
|
return -ENOMEM;
|
2015-08-07 06:17:58 +08:00
|
|
|
|
vmw_balloon: fixing double free when batching mode is off
The balloon.page field is used for two different purposes if batching is
on or off. If batching is on, the field point to the page which is used
to communicate with with the hypervisor. If it is off, balloon.page
points to the page that is about to be (un)locked.
Unfortunately, this dual-purpose of the field introduced a bug: when the
balloon is popped (e.g., when the machine is reset or the balloon driver
is explicitly removed), the balloon driver frees, unconditionally, the
page that is held in balloon.page. As a result, if batching is
disabled, this leads to double freeing the last page that is sent to the
hypervisor.
The following error occurs during rmmod when kernel checkers are on, and
the balloon is not empty:
[ 42.307653] ------------[ cut here ]------------
[ 42.307657] Kernel BUG at ffffffffba1e4b28 [verbose debug info unavailable]
[ 42.307720] invalid opcode: 0000 [#1] SMP DEBUG_PAGEALLOC
[ 42.312512] Modules linked in: vmw_vsock_vmci_transport vsock ppdev joydev vmw_balloon(-) input_leds serio_raw vmw_vmci parport_pc shpchp parport i2c_piix4 nfit mac_hid autofs4 vmwgfx drm_kms_helper hid_generic syscopyarea sysfillrect usbhid sysimgblt fb_sys_fops hid ttm mptspi scsi_transport_spi ahci mptscsih drm psmouse vmxnet3 libahci mptbase pata_acpi
[ 42.312766] CPU: 10 PID: 1527 Comm: rmmod Not tainted 4.12.0+ #5
[ 42.312803] Hardware name: VMware, Inc. VMware Virtual Platform/440BX Desktop Reference Platform, BIOS 6.00 09/30/2016
[ 42.313042] task: ffff9bf9680f8000 task.stack: ffffbfefc1638000
[ 42.313290] RIP: 0010:__free_pages+0x38/0x40
[ 42.313510] RSP: 0018:ffffbfefc163be98 EFLAGS: 00010246
[ 42.313731] RAX: 000000000000003e RBX: ffffffffc02b9720 RCX: 0000000000000006
[ 42.313972] RDX: 0000000000000000 RSI: 0000000000000000 RDI: ffff9bf97e08e0a0
[ 42.314201] RBP: ffffbfefc163be98 R08: 0000000000000000 R09: 0000000000000000
[ 42.314435] R10: 0000000000000000 R11: 0000000000000000 R12: ffffffffc02b97e4
[ 42.314505] R13: ffffffffc02b9748 R14: ffffffffc02b9728 R15: 0000000000000200
[ 42.314550] FS: 00007f3af5fec700(0000) GS:ffff9bf97e080000(0000) knlGS:0000000000000000
[ 42.314599] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[ 42.314635] CR2: 00007f44f6f4ab24 CR3: 00000003a7d12000 CR4: 00000000000006e0
[ 42.314864] Call Trace:
[ 42.315774] vmballoon_pop+0x102/0x130 [vmw_balloon]
[ 42.315816] vmballoon_exit+0x42/0xd64 [vmw_balloon]
[ 42.315853] SyS_delete_module+0x1e2/0x250
[ 42.315891] entry_SYSCALL_64_fastpath+0x23/0xc2
[ 42.315924] RIP: 0033:0x7f3af5b0e8e7
[ 42.315949] RSP: 002b:00007fffe6ce0148 EFLAGS: 00000206 ORIG_RAX: 00000000000000b0
[ 42.315996] RAX: ffffffffffffffda RBX: 000055be676401e0 RCX: 00007f3af5b0e8e7
[ 42.316951] RDX: 000000000000000a RSI: 0000000000000800 RDI: 000055be67640248
[ 42.317887] RBP: 0000000000000003 R08: 0000000000000000 R09: 1999999999999999
[ 42.318845] R10: 0000000000000883 R11: 0000000000000206 R12: 00007fffe6cdf130
[ 42.319755] R13: 0000000000000000 R14: 0000000000000000 R15: 000055be676401e0
[ 42.320606] Code: c0 74 1c f0 ff 4f 1c 74 02 5d c3 85 f6 74 07 e8 0f d8 ff ff 5d c3 31 f6 e8 c6 fb ff ff 5d c3 48 c7 c6 c8 0f c5 ba e8 58 be 02 00 <0f> 0b 66 0f 1f 44 00 00 66 66 66 66 90 48 85 ff 75 01 c3 55 48
[ 42.323462] RIP: __free_pages+0x38/0x40 RSP: ffffbfefc163be98
[ 42.325735] ---[ end trace 872e008e33f81508 ]---
To solve the bug, we eliminate the dual purpose of balloon.page.
Fixes: f220a80f0c2e ("VMware balloon: add batching to the vmw_balloon.")
Cc: stable@vger.kernel.org
Reported-by: Oleksandr Natalenko <onatalen@redhat.com>
Signed-off-by: Gil Kupfer <gilkup@gmail.com>
Signed-off-by: Nadav Amit <namit@vmware.com>
Reviewed-by: Xavier Deguillard <xdeguillard@vmware.com>
Tested-by: Oleksandr Natalenko <oleksandr@redhat.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2018-06-01 15:47:47 +08:00
|
|
|
b->batch_page = page_address(page);
|
vmw_balloon: change batch/single lock abstractions
The current abstractions for batch vs single operations seem suboptimal
and complicate the implementation of additional features (OOM,
compaction).
The immediate problem of the current abstractions is that they cause
differences in how operations are handled when batching is on or off.
For example, the refused_alloc counter is not updated when batching is
on. These discrepancies are caused by code redundancies.
Instead, this patch presents three type of operations, according to
whether batching is on or off: (1) add page, (2) communication with the
hypervisor and (3) retrieving the status of a page.
To avoid the overhead of virtual functions, and since we do not expect
additional interfaces for communication with the hypervisor, we use
static keys instead of virtual functions.
Finally, while we are at it, change vmballoon_init_batching() to return
int instead of bool, to be consistent in the return type and avoid
potential coding errors.
Reviewed-by: Xavier Deguillard <xdeguillard@vmware.com>
Signed-off-by: Nadav Amit <namit@vmware.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2018-09-21 01:30:12 +08:00
|
|
|
b->batch_max_pages = PAGE_SIZE / sizeof(struct vmballoon_batch_entry);
|
|
|
|
|
|
|
|
static_branch_enable(&vmw_balloon_batching);
|
|
|
|
|
|
|
|
return 0;
|
2015-08-07 06:17:58 +08:00
|
|
|
}
|
|
|
|
|
2015-08-07 06:18:01 +08:00
|
|
|
/*
|
|
|
|
* Receive notification and resize balloon
|
|
|
|
*/
|
|
|
|
static void vmballoon_doorbell(void *client_data)
|
|
|
|
{
|
|
|
|
struct vmballoon *b = client_data;
|
|
|
|
|
2018-09-21 01:30:17 +08:00
|
|
|
vmballoon_stats_gen_inc(b, VMW_BALLOON_STAT_DOORBELL);
|
2015-08-07 06:18:01 +08:00
|
|
|
|
|
|
|
mod_delayed_work(system_freezable_wq, &b->dwork, 0);
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Clean up vmci doorbell
|
|
|
|
*/
|
|
|
|
static void vmballoon_vmci_cleanup(struct vmballoon *b)
|
|
|
|
{
|
2018-09-21 01:30:07 +08:00
|
|
|
vmballoon_cmd(b, VMW_BALLOON_CMD_VMCI_DOORBELL_SET,
|
|
|
|
VMCI_INVALID_ID, VMCI_INVALID_ID);
|
2015-08-07 06:18:01 +08:00
|
|
|
|
|
|
|
if (!vmci_handle_is_invalid(b->vmci_doorbell)) {
|
|
|
|
vmci_doorbell_destroy(b->vmci_doorbell);
|
|
|
|
b->vmci_doorbell = VMCI_INVALID_HANDLE;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2018-09-21 01:30:19 +08:00
|
|
|
/**
|
|
|
|
* vmballoon_vmci_init - Initialize vmci doorbell.
|
|
|
|
*
|
|
|
|
* @b: pointer to the balloon.
|
|
|
|
*
|
|
|
|
* Return: zero on success or when wakeup command not supported. Error-code
|
|
|
|
* otherwise.
|
|
|
|
*
|
|
|
|
* Initialize vmci doorbell, to get notified as soon as balloon changes.
|
2015-08-07 06:18:01 +08:00
|
|
|
*/
|
|
|
|
static int vmballoon_vmci_init(struct vmballoon *b)
|
|
|
|
{
|
2018-09-21 01:30:07 +08:00
|
|
|
unsigned long error;
|
2015-08-07 06:18:01 +08:00
|
|
|
|
2018-06-20 07:00:26 +08:00
|
|
|
if ((b->capabilities & VMW_BALLOON_SIGNALLED_WAKEUP_CMD) == 0)
|
|
|
|
return 0;
|
2015-08-07 06:18:01 +08:00
|
|
|
|
2018-06-20 07:00:26 +08:00
|
|
|
error = vmci_doorbell_create(&b->vmci_doorbell, VMCI_FLAG_DELAYED_CB,
|
|
|
|
VMCI_PRIVILEGE_FLAG_RESTRICTED,
|
|
|
|
vmballoon_doorbell, b);
|
2015-08-07 06:18:01 +08:00
|
|
|
|
2018-06-20 07:00:26 +08:00
|
|
|
if (error != VMCI_SUCCESS)
|
|
|
|
goto fail;
|
|
|
|
|
2018-09-21 01:30:07 +08:00
|
|
|
error = __vmballoon_cmd(b, VMW_BALLOON_CMD_VMCI_DOORBELL_SET,
|
|
|
|
b->vmci_doorbell.context,
|
|
|
|
b->vmci_doorbell.resource, NULL);
|
2018-06-20 07:00:26 +08:00
|
|
|
|
|
|
|
if (error != VMW_BALLOON_SUCCESS)
|
|
|
|
goto fail;
|
2015-08-07 06:18:01 +08:00
|
|
|
|
|
|
|
return 0;
|
2018-06-20 07:00:26 +08:00
|
|
|
fail:
|
|
|
|
vmballoon_vmci_cleanup(b);
|
|
|
|
return -EIO;
|
2015-08-07 06:18:01 +08:00
|
|
|
}
|
|
|
|
|
vmw_balloon: rework the inflate and deflate loops
In preparation for supporting compaction and OOM notification, this
patch reworks the inflate/deflate loops. The main idea is to separate
the allocation, communication with the hypervisor, and the handling of
errors from each other. Doing will allow us to perform concurrent
inflation and deflation, excluding the actual communication with the
hypervisor.
To do so, we need to get rid of the remaining global state that is kept
in the balloon struct, specifically the refuse_list. When the VM
communicates with the hypervisor, it does not free or put back pages
to the balloon list and instead only moves the pages whose status
indicated failure into a refuse_list on the stack. Once the operation
completes, the inflation or deflation functions handle the list
appropriately.
As we do that, we can consolidate the communication with the hypervisor
for both the lock and unlock operations into a single function. We also
reuse the deflation function for popping the balloon.
As a preparation for preventing races, we hold a spinlock when the
communication actually takes place, and use atomic operations for
updating the balloon size. The balloon page list is still racy and will
be handled in the next patch.
Reviewed-by: Xavier Deguillard <xdeguillard@vmware.com>
Signed-off-by: Nadav Amit <namit@vmware.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2018-09-21 01:30:18 +08:00
|
|
|
/**
|
|
|
|
* vmballoon_pop - Quickly release all pages allocate for the balloon.
|
|
|
|
*
|
|
|
|
* @b: pointer to the balloon.
|
|
|
|
*
|
|
|
|
* This function is called when host decides to "reset" balloon for one reason
|
|
|
|
* or another. Unlike normal "deflate" we do not (shall not) notify host of the
|
|
|
|
* pages being released.
|
|
|
|
*/
|
|
|
|
static void vmballoon_pop(struct vmballoon *b)
|
|
|
|
{
|
|
|
|
unsigned long size;
|
|
|
|
|
|
|
|
while ((size = atomic64_read(&b->size)))
|
|
|
|
vmballoon_deflate(b, size, false);
|
|
|
|
}
|
|
|
|
|
2015-08-07 06:17:58 +08:00
|
|
|
/*
|
|
|
|
* Perform standard reset sequence by popping the balloon (in case it
|
|
|
|
* is not empty) and then restarting protocol. This operation normally
|
|
|
|
* happens when host responds with VMW_BALLOON_ERROR_RESET to a command.
|
|
|
|
*/
|
|
|
|
static void vmballoon_reset(struct vmballoon *b)
|
|
|
|
{
|
2015-08-07 06:18:01 +08:00
|
|
|
int error;
|
|
|
|
|
2018-09-21 01:30:17 +08:00
|
|
|
down_write(&b->conf_sem);
|
|
|
|
|
2015-08-07 06:18:01 +08:00
|
|
|
vmballoon_vmci_cleanup(b);
|
|
|
|
|
2015-08-07 06:17:58 +08:00
|
|
|
/* free all pages, skipping monitor unlock */
|
|
|
|
vmballoon_pop(b);
|
|
|
|
|
2018-09-21 01:30:19 +08:00
|
|
|
if (vmballoon_send_start(b, VMW_BALLOON_CAPABILITIES))
|
2019-02-12 02:45:45 +08:00
|
|
|
goto unlock;
|
2015-08-07 06:17:58 +08:00
|
|
|
|
|
|
|
if ((b->capabilities & VMW_BALLOON_BATCHED_CMDS) != 0) {
|
vmw_balloon: change batch/single lock abstractions
The current abstractions for batch vs single operations seem suboptimal
and complicate the implementation of additional features (OOM,
compaction).
The immediate problem of the current abstractions is that they cause
differences in how operations are handled when batching is on or off.
For example, the refused_alloc counter is not updated when batching is
on. These discrepancies are caused by code redundancies.
Instead, this patch presents three type of operations, according to
whether batching is on or off: (1) add page, (2) communication with the
hypervisor and (3) retrieving the status of a page.
To avoid the overhead of virtual functions, and since we do not expect
additional interfaces for communication with the hypervisor, we use
static keys instead of virtual functions.
Finally, while we are at it, change vmballoon_init_batching() to return
int instead of bool, to be consistent in the return type and avoid
potential coding errors.
Reviewed-by: Xavier Deguillard <xdeguillard@vmware.com>
Signed-off-by: Nadav Amit <namit@vmware.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2018-09-21 01:30:12 +08:00
|
|
|
if (vmballoon_init_batching(b)) {
|
2015-08-07 06:17:58 +08:00
|
|
|
/*
|
|
|
|
* We failed to initialize batching, inform the monitor
|
|
|
|
* about it by sending a null capability.
|
|
|
|
*
|
|
|
|
* The guest will retry in one second.
|
|
|
|
*/
|
|
|
|
vmballoon_send_start(b, 0);
|
2019-02-12 02:45:45 +08:00
|
|
|
goto unlock;
|
2015-08-07 06:17:58 +08:00
|
|
|
}
|
|
|
|
} else if ((b->capabilities & VMW_BALLOON_BASIC_CMDS) != 0) {
|
vmw_balloon: change batch/single lock abstractions
The current abstractions for batch vs single operations seem suboptimal
and complicate the implementation of additional features (OOM,
compaction).
The immediate problem of the current abstractions is that they cause
differences in how operations are handled when batching is on or off.
For example, the refused_alloc counter is not updated when batching is
on. These discrepancies are caused by code redundancies.
Instead, this patch presents three type of operations, according to
whether batching is on or off: (1) add page, (2) communication with the
hypervisor and (3) retrieving the status of a page.
To avoid the overhead of virtual functions, and since we do not expect
additional interfaces for communication with the hypervisor, we use
static keys instead of virtual functions.
Finally, while we are at it, change vmballoon_init_batching() to return
int instead of bool, to be consistent in the return type and avoid
potential coding errors.
Reviewed-by: Xavier Deguillard <xdeguillard@vmware.com>
Signed-off-by: Nadav Amit <namit@vmware.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2018-09-21 01:30:12 +08:00
|
|
|
vmballoon_deinit_batching(b);
|
2015-08-07 06:17:58 +08:00
|
|
|
}
|
|
|
|
|
2018-09-21 01:30:20 +08:00
|
|
|
vmballoon_stats_gen_inc(b, VMW_BALLOON_STAT_RESET);
|
2015-08-07 06:17:58 +08:00
|
|
|
b->reset_required = false;
|
2015-08-07 06:18:01 +08:00
|
|
|
|
|
|
|
error = vmballoon_vmci_init(b);
|
|
|
|
if (error)
|
2022-03-23 01:00:52 +08:00
|
|
|
pr_err_once("failed to initialize vmci doorbell\n");
|
2015-08-07 06:18:01 +08:00
|
|
|
|
2018-09-21 01:30:19 +08:00
|
|
|
if (vmballoon_send_guest_id(b))
|
2022-03-23 01:00:52 +08:00
|
|
|
pr_err_once("failed to send guest ID to the host\n");
|
2018-09-21 01:30:17 +08:00
|
|
|
|
2019-02-12 02:45:45 +08:00
|
|
|
unlock:
|
2018-09-21 01:30:17 +08:00
|
|
|
up_write(&b->conf_sem);
|
2010-04-24 01:18:08 +08:00
|
|
|
}
|
|
|
|
|
2018-09-21 01:30:15 +08:00
|
|
|
/**
|
|
|
|
* vmballoon_work - periodic balloon worker for reset, inflation and deflation.
|
|
|
|
*
|
|
|
|
* @work: pointer to the &work_struct which is provided by the workqueue.
|
|
|
|
*
|
|
|
|
* Resets the protocol if needed, gets the new size and adjusts balloon as
|
|
|
|
* needed. Repeat in 1 sec.
|
2010-04-24 01:18:08 +08:00
|
|
|
*/
|
|
|
|
static void vmballoon_work(struct work_struct *work)
|
|
|
|
{
|
|
|
|
struct delayed_work *dwork = to_delayed_work(work);
|
|
|
|
struct vmballoon *b = container_of(dwork, struct vmballoon, dwork);
|
2018-09-21 01:30:15 +08:00
|
|
|
int64_t change = 0;
|
2010-04-24 01:18:08 +08:00
|
|
|
|
|
|
|
if (b->reset_required)
|
|
|
|
vmballoon_reset(b);
|
|
|
|
|
2018-09-21 01:30:17 +08:00
|
|
|
down_read(&b->conf_sem);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Update the stats while holding the semaphore to ensure that
|
|
|
|
* @stats_enabled is consistent with whether the stats are actually
|
|
|
|
* enabled
|
|
|
|
*/
|
|
|
|
vmballoon_stats_gen_inc(b, VMW_BALLOON_STAT_TIMER);
|
|
|
|
|
2018-09-21 01:30:16 +08:00
|
|
|
if (!vmballoon_send_get_target(b))
|
2018-09-21 01:30:15 +08:00
|
|
|
change = vmballoon_change(b);
|
|
|
|
|
|
|
|
if (change != 0) {
|
vmw_balloon: rework the inflate and deflate loops
In preparation for supporting compaction and OOM notification, this
patch reworks the inflate/deflate loops. The main idea is to separate
the allocation, communication with the hypervisor, and the handling of
errors from each other. Doing will allow us to perform concurrent
inflation and deflation, excluding the actual communication with the
hypervisor.
To do so, we need to get rid of the remaining global state that is kept
in the balloon struct, specifically the refuse_list. When the VM
communicates with the hypervisor, it does not free or put back pages
to the balloon list and instead only moves the pages whose status
indicated failure into a refuse_list on the stack. Once the operation
completes, the inflation or deflation functions handle the list
appropriately.
As we do that, we can consolidate the communication with the hypervisor
for both the lock and unlock operations into a single function. We also
reuse the deflation function for popping the balloon.
As a preparation for preventing races, we hold a spinlock when the
communication actually takes place, and use atomic operations for
updating the balloon size. The balloon page list is still racy and will
be handled in the next patch.
Reviewed-by: Xavier Deguillard <xdeguillard@vmware.com>
Signed-off-by: Nadav Amit <namit@vmware.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2018-09-21 01:30:18 +08:00
|
|
|
pr_debug("%s - size: %llu, target %lu\n", __func__,
|
|
|
|
atomic64_read(&b->size), READ_ONCE(b->target));
|
2010-04-24 01:18:08 +08:00
|
|
|
|
2018-09-21 01:30:15 +08:00
|
|
|
if (change > 0)
|
2010-04-24 01:18:08 +08:00
|
|
|
vmballoon_inflate(b);
|
2018-09-21 01:30:15 +08:00
|
|
|
else /* (change < 0) */
|
vmw_balloon: rework the inflate and deflate loops
In preparation for supporting compaction and OOM notification, this
patch reworks the inflate/deflate loops. The main idea is to separate
the allocation, communication with the hypervisor, and the handling of
errors from each other. Doing will allow us to perform concurrent
inflation and deflation, excluding the actual communication with the
hypervisor.
To do so, we need to get rid of the remaining global state that is kept
in the balloon struct, specifically the refuse_list. When the VM
communicates with the hypervisor, it does not free or put back pages
to the balloon list and instead only moves the pages whose status
indicated failure into a refuse_list on the stack. Once the operation
completes, the inflation or deflation functions handle the list
appropriately.
As we do that, we can consolidate the communication with the hypervisor
for both the lock and unlock operations into a single function. We also
reuse the deflation function for popping the balloon.
As a preparation for preventing races, we hold a spinlock when the
communication actually takes place, and use atomic operations for
updating the balloon size. The balloon page list is still racy and will
be handled in the next patch.
Reviewed-by: Xavier Deguillard <xdeguillard@vmware.com>
Signed-off-by: Nadav Amit <namit@vmware.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2018-09-21 01:30:18 +08:00
|
|
|
vmballoon_deflate(b, 0, true);
|
2010-04-24 01:18:08 +08:00
|
|
|
}
|
|
|
|
|
2018-09-21 01:30:17 +08:00
|
|
|
up_read(&b->conf_sem);
|
|
|
|
|
2011-07-27 07:08:56 +08:00
|
|
|
/*
|
|
|
|
* We are using a freezable workqueue so that balloon operations are
|
|
|
|
* stopped while the system transitions to/from sleep/hibernation.
|
|
|
|
*/
|
|
|
|
queue_delayed_work(system_freezable_wq,
|
|
|
|
dwork, round_jiffies_relative(HZ));
|
2018-09-21 01:30:17 +08:00
|
|
|
|
2010-04-24 01:18:08 +08:00
|
|
|
}
|
|
|
|
|
2019-04-25 19:54:44 +08:00
|
|
|
/**
|
|
|
|
* vmballoon_shrinker_scan() - deflate the balloon due to memory pressure.
|
|
|
|
* @shrinker: pointer to the balloon shrinker.
|
|
|
|
* @sc: page reclaim information.
|
|
|
|
*
|
|
|
|
* Returns: number of pages that were freed during deflation.
|
|
|
|
*/
|
|
|
|
static unsigned long vmballoon_shrinker_scan(struct shrinker *shrinker,
|
|
|
|
struct shrink_control *sc)
|
|
|
|
{
|
|
|
|
struct vmballoon *b = &balloon;
|
|
|
|
unsigned long deflated_frames;
|
|
|
|
|
|
|
|
pr_debug("%s - size: %llu", __func__, atomic64_read(&b->size));
|
|
|
|
|
|
|
|
vmballoon_stats_gen_inc(b, VMW_BALLOON_STAT_SHRINK);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* If the lock is also contended for read, we cannot easily reclaim and
|
|
|
|
* we bail out.
|
|
|
|
*/
|
|
|
|
if (!down_read_trylock(&b->conf_sem))
|
|
|
|
return 0;
|
|
|
|
|
|
|
|
deflated_frames = vmballoon_deflate(b, sc->nr_to_scan, true);
|
|
|
|
|
|
|
|
vmballoon_stats_gen_add(b, VMW_BALLOON_STAT_SHRINK_FREE,
|
|
|
|
deflated_frames);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Delay future inflation for some time to mitigate the situations in
|
|
|
|
* which balloon continuously grows and shrinks. Use WRITE_ONCE() since
|
|
|
|
* the access is asynchronous.
|
|
|
|
*/
|
|
|
|
WRITE_ONCE(b->shrink_timeout, jiffies + HZ * VMBALLOON_SHRINK_DELAY);
|
|
|
|
|
|
|
|
up_read(&b->conf_sem);
|
|
|
|
|
|
|
|
return deflated_frames;
|
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
|
|
|
* vmballoon_shrinker_count() - return the number of ballooned pages.
|
|
|
|
* @shrinker: pointer to the balloon shrinker.
|
|
|
|
* @sc: page reclaim information.
|
|
|
|
*
|
|
|
|
* Returns: number of 4k pages that are allocated for the balloon and can
|
|
|
|
* therefore be reclaimed under pressure.
|
|
|
|
*/
|
|
|
|
static unsigned long vmballoon_shrinker_count(struct shrinker *shrinker,
|
|
|
|
struct shrink_control *sc)
|
|
|
|
{
|
|
|
|
struct vmballoon *b = &balloon;
|
|
|
|
|
|
|
|
return atomic64_read(&b->size);
|
|
|
|
}
|
|
|
|
|
|
|
|
static void vmballoon_unregister_shrinker(struct vmballoon *b)
|
|
|
|
{
|
2023-09-11 17:44:26 +08:00
|
|
|
shrinker_free(b->shrinker);
|
|
|
|
b->shrinker = NULL;
|
2019-04-25 19:54:44 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
static int vmballoon_register_shrinker(struct vmballoon *b)
|
|
|
|
{
|
|
|
|
/* Do nothing if the shrinker is not enabled */
|
|
|
|
if (!vmwballoon_shrinker_enable)
|
|
|
|
return 0;
|
|
|
|
|
2023-09-11 17:44:26 +08:00
|
|
|
b->shrinker = shrinker_alloc(0, "vmw-balloon");
|
|
|
|
if (!b->shrinker)
|
|
|
|
return -ENOMEM;
|
2019-04-25 19:54:44 +08:00
|
|
|
|
2023-09-11 17:44:26 +08:00
|
|
|
b->shrinker->scan_objects = vmballoon_shrinker_scan;
|
|
|
|
b->shrinker->count_objects = vmballoon_shrinker_count;
|
|
|
|
b->shrinker->private_data = b;
|
2019-04-25 19:54:44 +08:00
|
|
|
|
2023-09-11 17:44:26 +08:00
|
|
|
shrinker_register(b->shrinker);
|
2019-04-25 19:54:44 +08:00
|
|
|
|
2023-09-11 17:44:26 +08:00
|
|
|
return 0;
|
2019-04-25 19:54:44 +08:00
|
|
|
}
|
|
|
|
|
2010-04-24 01:18:08 +08:00
|
|
|
/*
|
|
|
|
* DEBUGFS Interface
|
|
|
|
*/
|
|
|
|
#ifdef CONFIG_DEBUG_FS
|
|
|
|
|
2018-09-21 01:30:17 +08:00
|
|
|
static const char * const vmballoon_stat_page_names[] = {
|
|
|
|
[VMW_BALLOON_PAGE_STAT_ALLOC] = "alloc",
|
|
|
|
[VMW_BALLOON_PAGE_STAT_ALLOC_FAIL] = "allocFail",
|
|
|
|
[VMW_BALLOON_PAGE_STAT_REFUSED_ALLOC] = "errAlloc",
|
|
|
|
[VMW_BALLOON_PAGE_STAT_REFUSED_FREE] = "errFree",
|
|
|
|
[VMW_BALLOON_PAGE_STAT_FREE] = "free"
|
|
|
|
};
|
|
|
|
|
|
|
|
static const char * const vmballoon_stat_names[] = {
|
|
|
|
[VMW_BALLOON_STAT_TIMER] = "timer",
|
2018-09-21 01:30:20 +08:00
|
|
|
[VMW_BALLOON_STAT_DOORBELL] = "doorbell",
|
|
|
|
[VMW_BALLOON_STAT_RESET] = "reset",
|
2019-04-25 19:54:44 +08:00
|
|
|
[VMW_BALLOON_STAT_SHRINK] = "shrink",
|
|
|
|
[VMW_BALLOON_STAT_SHRINK_FREE] = "shrinkFree"
|
2018-09-21 01:30:17 +08:00
|
|
|
};
|
|
|
|
|
|
|
|
static int vmballoon_enable_stats(struct vmballoon *b)
|
|
|
|
{
|
|
|
|
int r = 0;
|
|
|
|
|
|
|
|
down_write(&b->conf_sem);
|
|
|
|
|
|
|
|
/* did we somehow race with another reader which enabled stats? */
|
|
|
|
if (b->stats)
|
|
|
|
goto out;
|
|
|
|
|
|
|
|
b->stats = kzalloc(sizeof(*b->stats), GFP_KERNEL);
|
|
|
|
|
|
|
|
if (!b->stats) {
|
|
|
|
/* allocation failed */
|
|
|
|
r = -ENOMEM;
|
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
static_key_enable(&balloon_stat_enabled.key);
|
|
|
|
out:
|
|
|
|
up_write(&b->conf_sem);
|
|
|
|
return r;
|
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
|
|
|
* vmballoon_debug_show - shows statistics of balloon operations.
|
|
|
|
* @f: pointer to the &struct seq_file.
|
|
|
|
* @offset: ignored.
|
|
|
|
*
|
|
|
|
* Provides the statistics that can be accessed in vmmemctl in the debugfs.
|
|
|
|
* To avoid the overhead - mainly that of memory - of collecting the statistics,
|
|
|
|
* we only collect statistics after the first time the counters are read.
|
|
|
|
*
|
|
|
|
* Return: zero on success or an error code.
|
|
|
|
*/
|
2010-04-24 01:18:08 +08:00
|
|
|
static int vmballoon_debug_show(struct seq_file *f, void *offset)
|
|
|
|
{
|
|
|
|
struct vmballoon *b = f->private;
|
2018-09-21 01:30:17 +08:00
|
|
|
int i, j;
|
|
|
|
|
|
|
|
/* enables stats if they are disabled */
|
|
|
|
if (!b->stats) {
|
|
|
|
int r = vmballoon_enable_stats(b);
|
|
|
|
|
|
|
|
if (r)
|
|
|
|
return r;
|
|
|
|
}
|
2010-04-24 01:18:08 +08:00
|
|
|
|
2015-08-07 06:18:00 +08:00
|
|
|
/* format capabilities info */
|
vmw_balloon: rework the inflate and deflate loops
In preparation for supporting compaction and OOM notification, this
patch reworks the inflate/deflate loops. The main idea is to separate
the allocation, communication with the hypervisor, and the handling of
errors from each other. Doing will allow us to perform concurrent
inflation and deflation, excluding the actual communication with the
hypervisor.
To do so, we need to get rid of the remaining global state that is kept
in the balloon struct, specifically the refuse_list. When the VM
communicates with the hypervisor, it does not free or put back pages
to the balloon list and instead only moves the pages whose status
indicated failure into a refuse_list on the stack. Once the operation
completes, the inflation or deflation functions handle the list
appropriately.
As we do that, we can consolidate the communication with the hypervisor
for both the lock and unlock operations into a single function. We also
reuse the deflation function for popping the balloon.
As a preparation for preventing races, we hold a spinlock when the
communication actually takes place, and use atomic operations for
updating the balloon size. The balloon page list is still racy and will
be handled in the next patch.
Reviewed-by: Xavier Deguillard <xdeguillard@vmware.com>
Signed-off-by: Nadav Amit <namit@vmware.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2018-09-21 01:30:18 +08:00
|
|
|
seq_printf(f, "%-22s: %#16x\n", "balloon capabilities",
|
2018-09-21 01:30:17 +08:00
|
|
|
VMW_BALLOON_CAPABILITIES);
|
vmw_balloon: rework the inflate and deflate loops
In preparation for supporting compaction and OOM notification, this
patch reworks the inflate/deflate loops. The main idea is to separate
the allocation, communication with the hypervisor, and the handling of
errors from each other. Doing will allow us to perform concurrent
inflation and deflation, excluding the actual communication with the
hypervisor.
To do so, we need to get rid of the remaining global state that is kept
in the balloon struct, specifically the refuse_list. When the VM
communicates with the hypervisor, it does not free or put back pages
to the balloon list and instead only moves the pages whose status
indicated failure into a refuse_list on the stack. Once the operation
completes, the inflation or deflation functions handle the list
appropriately.
As we do that, we can consolidate the communication with the hypervisor
for both the lock and unlock operations into a single function. We also
reuse the deflation function for popping the balloon.
As a preparation for preventing races, we hold a spinlock when the
communication actually takes place, and use atomic operations for
updating the balloon size. The balloon page list is still racy and will
be handled in the next patch.
Reviewed-by: Xavier Deguillard <xdeguillard@vmware.com>
Signed-off-by: Nadav Amit <namit@vmware.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2018-09-21 01:30:18 +08:00
|
|
|
seq_printf(f, "%-22s: %#16lx\n", "used capabilities", b->capabilities);
|
2018-09-21 01:30:17 +08:00
|
|
|
seq_printf(f, "%-22s: %16s\n", "is resetting",
|
|
|
|
b->reset_required ? "y" : "n");
|
2015-08-07 06:18:00 +08:00
|
|
|
|
2010-04-24 01:18:08 +08:00
|
|
|
/* format size info */
|
vmw_balloon: rework the inflate and deflate loops
In preparation for supporting compaction and OOM notification, this
patch reworks the inflate/deflate loops. The main idea is to separate
the allocation, communication with the hypervisor, and the handling of
errors from each other. Doing will allow us to perform concurrent
inflation and deflation, excluding the actual communication with the
hypervisor.
To do so, we need to get rid of the remaining global state that is kept
in the balloon struct, specifically the refuse_list. When the VM
communicates with the hypervisor, it does not free or put back pages
to the balloon list and instead only moves the pages whose status
indicated failure into a refuse_list on the stack. Once the operation
completes, the inflation or deflation functions handle the list
appropriately.
As we do that, we can consolidate the communication with the hypervisor
for both the lock and unlock operations into a single function. We also
reuse the deflation function for popping the balloon.
As a preparation for preventing races, we hold a spinlock when the
communication actually takes place, and use atomic operations for
updating the balloon size. The balloon page list is still racy and will
be handled in the next patch.
Reviewed-by: Xavier Deguillard <xdeguillard@vmware.com>
Signed-off-by: Nadav Amit <namit@vmware.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2018-09-21 01:30:18 +08:00
|
|
|
seq_printf(f, "%-22s: %16lu\n", "target", READ_ONCE(b->target));
|
|
|
|
seq_printf(f, "%-22s: %16llu\n", "current", atomic64_read(&b->size));
|
2010-04-24 01:18:08 +08:00
|
|
|
|
2018-09-21 01:30:08 +08:00
|
|
|
for (i = 0; i < VMW_BALLOON_CMD_NUM; i++) {
|
|
|
|
if (vmballoon_cmd_names[i] == NULL)
|
|
|
|
continue;
|
|
|
|
|
2018-09-21 01:30:17 +08:00
|
|
|
seq_printf(f, "%-22s: %16llu (%llu failed)\n",
|
|
|
|
vmballoon_cmd_names[i],
|
|
|
|
atomic64_read(&b->stats->ops[i][VMW_BALLOON_OP_STAT]),
|
|
|
|
atomic64_read(&b->stats->ops[i][VMW_BALLOON_OP_FAIL_STAT]));
|
2018-09-21 01:30:08 +08:00
|
|
|
}
|
|
|
|
|
2018-09-21 01:30:17 +08:00
|
|
|
for (i = 0; i < VMW_BALLOON_STAT_NUM; i++)
|
|
|
|
seq_printf(f, "%-22s: %16llu\n",
|
|
|
|
vmballoon_stat_names[i],
|
|
|
|
atomic64_read(&b->stats->general_stat[i]));
|
|
|
|
|
|
|
|
for (i = 0; i < VMW_BALLOON_PAGE_STAT_NUM; i++) {
|
|
|
|
for (j = 0; j < VMW_BALLOON_NUM_PAGE_SIZES; j++)
|
|
|
|
seq_printf(f, "%-18s(%s): %16llu\n",
|
|
|
|
vmballoon_stat_page_names[i],
|
|
|
|
vmballoon_page_size_names[j],
|
|
|
|
atomic64_read(&b->stats->page_stat[i][j]));
|
|
|
|
}
|
2010-04-24 01:18:08 +08:00
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2018-12-02 01:05:30 +08:00
|
|
|
DEFINE_SHOW_ATTRIBUTE(vmballoon_debug);
|
2010-04-24 01:18:08 +08:00
|
|
|
|
2019-06-12 02:55:28 +08:00
|
|
|
static void __init vmballoon_debugfs_init(struct vmballoon *b)
|
2010-04-24 01:18:08 +08:00
|
|
|
{
|
2021-02-16 23:12:09 +08:00
|
|
|
debugfs_create_file("vmmemctl", S_IRUGO, NULL, b,
|
|
|
|
&vmballoon_debug_fops);
|
2010-04-24 01:18:08 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
static void __exit vmballoon_debugfs_exit(struct vmballoon *b)
|
|
|
|
{
|
2018-09-21 01:30:17 +08:00
|
|
|
static_key_disable(&balloon_stat_enabled.key);
|
2023-02-02 22:11:00 +08:00
|
|
|
debugfs_lookup_and_remove("vmmemctl", NULL);
|
2018-09-21 01:30:17 +08:00
|
|
|
kfree(b->stats);
|
|
|
|
b->stats = NULL;
|
2010-04-24 01:18:08 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
#else
|
|
|
|
|
2019-06-12 02:55:28 +08:00
|
|
|
static inline void vmballoon_debugfs_init(struct vmballoon *b)
|
2010-04-24 01:18:08 +08:00
|
|
|
{
|
|
|
|
}
|
|
|
|
|
|
|
|
static inline void vmballoon_debugfs_exit(struct vmballoon *b)
|
|
|
|
{
|
|
|
|
}
|
|
|
|
|
|
|
|
#endif /* CONFIG_DEBUG_FS */
|
|
|
|
|
2019-04-25 19:54:43 +08:00
|
|
|
|
|
|
|
#ifdef CONFIG_BALLOON_COMPACTION
|
|
|
|
/**
|
|
|
|
* vmballoon_migratepage() - migrates a balloon page.
|
|
|
|
* @b_dev_info: balloon device information descriptor.
|
|
|
|
* @newpage: the page to which @page should be migrated.
|
|
|
|
* @page: a ballooned page that should be migrated.
|
|
|
|
* @mode: migration mode, ignored.
|
|
|
|
*
|
|
|
|
* This function is really open-coded, but that is according to the interface
|
|
|
|
* that balloon_compaction provides.
|
|
|
|
*
|
|
|
|
* Return: zero on success, -EAGAIN when migration cannot be performed
|
|
|
|
* momentarily, and -EBUSY if migration failed and should be retried
|
|
|
|
* with that specific page.
|
|
|
|
*/
|
|
|
|
static int vmballoon_migratepage(struct balloon_dev_info *b_dev_info,
|
|
|
|
struct page *newpage, struct page *page,
|
|
|
|
enum migrate_mode mode)
|
|
|
|
{
|
|
|
|
unsigned long status, flags;
|
|
|
|
struct vmballoon *b;
|
|
|
|
int ret;
|
|
|
|
|
|
|
|
b = container_of(b_dev_info, struct vmballoon, b_dev_info);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* If the semaphore is taken, there is ongoing configuration change
|
|
|
|
* (i.e., balloon reset), so try again.
|
|
|
|
*/
|
|
|
|
if (!down_read_trylock(&b->conf_sem))
|
|
|
|
return -EAGAIN;
|
|
|
|
|
|
|
|
spin_lock(&b->comm_lock);
|
|
|
|
/*
|
|
|
|
* We must start by deflating and not inflating, as otherwise the
|
|
|
|
* hypervisor may tell us that it has enough memory and the new page is
|
|
|
|
* not needed. Since the old page is isolated, we cannot use the list
|
|
|
|
* interface to unlock it, as the LRU field is used for isolation.
|
|
|
|
* Instead, we use the native interface directly.
|
|
|
|
*/
|
|
|
|
vmballoon_add_page(b, 0, page);
|
|
|
|
status = vmballoon_lock_op(b, 1, VMW_BALLOON_4K_PAGE,
|
|
|
|
VMW_BALLOON_DEFLATE);
|
|
|
|
|
|
|
|
if (status == VMW_BALLOON_SUCCESS)
|
|
|
|
status = vmballoon_status_page(b, 0, &page);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* If a failure happened, let the migration mechanism know that it
|
|
|
|
* should not retry.
|
|
|
|
*/
|
|
|
|
if (status != VMW_BALLOON_SUCCESS) {
|
|
|
|
spin_unlock(&b->comm_lock);
|
|
|
|
ret = -EBUSY;
|
|
|
|
goto out_unlock;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* The page is isolated, so it is safe to delete it without holding
|
|
|
|
* @pages_lock . We keep holding @comm_lock since we will need it in a
|
|
|
|
* second.
|
|
|
|
*/
|
|
|
|
balloon_page_delete(page);
|
|
|
|
|
|
|
|
put_page(page);
|
|
|
|
|
|
|
|
/* Inflate */
|
|
|
|
vmballoon_add_page(b, 0, newpage);
|
|
|
|
status = vmballoon_lock_op(b, 1, VMW_BALLOON_4K_PAGE,
|
|
|
|
VMW_BALLOON_INFLATE);
|
|
|
|
|
|
|
|
if (status == VMW_BALLOON_SUCCESS)
|
|
|
|
status = vmballoon_status_page(b, 0, &newpage);
|
|
|
|
|
|
|
|
spin_unlock(&b->comm_lock);
|
|
|
|
|
|
|
|
if (status != VMW_BALLOON_SUCCESS) {
|
|
|
|
/*
|
|
|
|
* A failure happened. While we can deflate the page we just
|
|
|
|
* inflated, this deflation can also encounter an error. Instead
|
|
|
|
* we will decrease the size of the balloon to reflect the
|
|
|
|
* change and report failure.
|
|
|
|
*/
|
|
|
|
atomic64_dec(&b->size);
|
|
|
|
ret = -EBUSY;
|
|
|
|
} else {
|
|
|
|
/*
|
|
|
|
* Success. Take a reference for the page, and we will add it to
|
|
|
|
* the list after acquiring the lock.
|
|
|
|
*/
|
|
|
|
get_page(newpage);
|
|
|
|
ret = MIGRATEPAGE_SUCCESS;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* Update the balloon list under the @pages_lock */
|
|
|
|
spin_lock_irqsave(&b->b_dev_info.pages_lock, flags);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* On inflation success, we already took a reference for the @newpage.
|
|
|
|
* If we succeed just insert it to the list and update the statistics
|
|
|
|
* under the lock.
|
|
|
|
*/
|
|
|
|
if (ret == MIGRATEPAGE_SUCCESS) {
|
|
|
|
balloon_page_insert(&b->b_dev_info, newpage);
|
|
|
|
__count_vm_event(BALLOON_MIGRATE);
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* We deflated successfully, so regardless to the inflation success, we
|
|
|
|
* need to reduce the number of isolated_pages.
|
|
|
|
*/
|
|
|
|
b->b_dev_info.isolated_pages--;
|
|
|
|
spin_unlock_irqrestore(&b->b_dev_info.pages_lock, flags);
|
|
|
|
|
|
|
|
out_unlock:
|
|
|
|
up_read(&b->conf_sem);
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
|
|
|
* vmballoon_compaction_init() - initialized compaction for the balloon.
|
|
|
|
*
|
|
|
|
* @b: pointer to the balloon.
|
|
|
|
*
|
|
|
|
* If during the initialization a failure occurred, this function does not
|
|
|
|
* perform cleanup. The caller must call vmballoon_compaction_deinit() in this
|
|
|
|
* case.
|
|
|
|
*
|
|
|
|
* Return: zero on success or error code on failure.
|
|
|
|
*/
|
2022-06-08 03:38:48 +08:00
|
|
|
static __init void vmballoon_compaction_init(struct vmballoon *b)
|
2019-04-25 19:54:43 +08:00
|
|
|
{
|
|
|
|
b->b_dev_info.migratepage = vmballoon_migratepage;
|
|
|
|
}
|
|
|
|
|
|
|
|
#else /* CONFIG_BALLOON_COMPACTION */
|
2022-06-08 03:38:48 +08:00
|
|
|
static inline void vmballoon_compaction_init(struct vmballoon *b)
|
2019-04-25 19:54:43 +08:00
|
|
|
{
|
|
|
|
}
|
|
|
|
#endif /* CONFIG_BALLOON_COMPACTION */
|
|
|
|
|
2010-04-24 01:18:08 +08:00
|
|
|
static int __init vmballoon_init(void)
|
|
|
|
{
|
|
|
|
int error;
|
vmw_balloon: rework the inflate and deflate loops
In preparation for supporting compaction and OOM notification, this
patch reworks the inflate/deflate loops. The main idea is to separate
the allocation, communication with the hypervisor, and the handling of
errors from each other. Doing will allow us to perform concurrent
inflation and deflation, excluding the actual communication with the
hypervisor.
To do so, we need to get rid of the remaining global state that is kept
in the balloon struct, specifically the refuse_list. When the VM
communicates with the hypervisor, it does not free or put back pages
to the balloon list and instead only moves the pages whose status
indicated failure into a refuse_list on the stack. Once the operation
completes, the inflation or deflation functions handle the list
appropriately.
As we do that, we can consolidate the communication with the hypervisor
for both the lock and unlock operations into a single function. We also
reuse the deflation function for popping the balloon.
As a preparation for preventing races, we hold a spinlock when the
communication actually takes place, and use atomic operations for
updating the balloon size. The balloon page list is still racy and will
be handled in the next patch.
Reviewed-by: Xavier Deguillard <xdeguillard@vmware.com>
Signed-off-by: Nadav Amit <namit@vmware.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2018-09-21 01:30:18 +08:00
|
|
|
|
2010-04-24 01:18:08 +08:00
|
|
|
/*
|
|
|
|
* Check if we are running on VMware's hypervisor and bail out
|
|
|
|
* if we are not.
|
|
|
|
*/
|
2017-11-09 21:27:36 +08:00
|
|
|
if (x86_hyper_type != X86_HYPER_VMWARE)
|
2010-04-24 01:18:08 +08:00
|
|
|
return -ENODEV;
|
|
|
|
|
|
|
|
INIT_DELAYED_WORK(&balloon.dwork, vmballoon_work);
|
|
|
|
|
2019-04-25 19:54:44 +08:00
|
|
|
error = vmballoon_register_shrinker(&balloon);
|
|
|
|
if (error)
|
2023-09-11 17:44:26 +08:00
|
|
|
return error;
|
2019-04-25 19:54:44 +08:00
|
|
|
|
2019-04-25 19:54:43 +08:00
|
|
|
/*
|
|
|
|
* Initialization of compaction must be done after the call to
|
|
|
|
* balloon_devinfo_init() .
|
|
|
|
*/
|
|
|
|
balloon_devinfo_init(&balloon.b_dev_info);
|
2022-06-08 03:38:48 +08:00
|
|
|
vmballoon_compaction_init(&balloon);
|
2019-04-25 19:54:43 +08:00
|
|
|
|
|
|
|
INIT_LIST_HEAD(&balloon.huge_pages);
|
vmw_balloon: rework the inflate and deflate loops
In preparation for supporting compaction and OOM notification, this
patch reworks the inflate/deflate loops. The main idea is to separate
the allocation, communication with the hypervisor, and the handling of
errors from each other. Doing will allow us to perform concurrent
inflation and deflation, excluding the actual communication with the
hypervisor.
To do so, we need to get rid of the remaining global state that is kept
in the balloon struct, specifically the refuse_list. When the VM
communicates with the hypervisor, it does not free or put back pages
to the balloon list and instead only moves the pages whose status
indicated failure into a refuse_list on the stack. Once the operation
completes, the inflation or deflation functions handle the list
appropriately.
As we do that, we can consolidate the communication with the hypervisor
for both the lock and unlock operations into a single function. We also
reuse the deflation function for popping the balloon.
As a preparation for preventing races, we hold a spinlock when the
communication actually takes place, and use atomic operations for
updating the balloon size. The balloon page list is still racy and will
be handled in the next patch.
Reviewed-by: Xavier Deguillard <xdeguillard@vmware.com>
Signed-off-by: Nadav Amit <namit@vmware.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2018-09-21 01:30:18 +08:00
|
|
|
spin_lock_init(&balloon.comm_lock);
|
2018-09-21 01:30:17 +08:00
|
|
|
init_rwsem(&balloon.conf_sem);
|
2015-08-07 06:18:01 +08:00
|
|
|
balloon.vmci_doorbell = VMCI_INVALID_HANDLE;
|
2015-08-07 06:18:01 +08:00
|
|
|
balloon.batch_page = NULL;
|
|
|
|
balloon.page = NULL;
|
|
|
|
balloon.reset_required = true;
|
|
|
|
|
2011-07-27 07:08:56 +08:00
|
|
|
queue_delayed_work(system_freezable_wq, &balloon.dwork, 0);
|
2010-04-24 01:18:08 +08:00
|
|
|
|
2019-07-13 03:24:03 +08:00
|
|
|
vmballoon_debugfs_init(&balloon);
|
|
|
|
|
2010-04-24 01:18:08 +08:00
|
|
|
return 0;
|
|
|
|
}
|
2018-06-20 07:00:27 +08:00
|
|
|
|
|
|
|
/*
|
|
|
|
* Using late_initcall() instead of module_init() allows the balloon to use the
|
|
|
|
* VMCI doorbell even when the balloon is built into the kernel. Otherwise the
|
|
|
|
* VMCI is probed only after the balloon is initialized. If the balloon is used
|
|
|
|
* as a module, late_initcall() is equivalent to module_init().
|
|
|
|
*/
|
|
|
|
late_initcall(vmballoon_init);
|
2010-04-24 01:18:08 +08:00
|
|
|
|
|
|
|
static void __exit vmballoon_exit(void)
|
|
|
|
{
|
2019-04-25 19:54:44 +08:00
|
|
|
vmballoon_unregister_shrinker(&balloon);
|
2015-08-07 06:18:01 +08:00
|
|
|
vmballoon_vmci_cleanup(&balloon);
|
2010-04-24 01:18:08 +08:00
|
|
|
cancel_delayed_work_sync(&balloon.dwork);
|
|
|
|
|
|
|
|
vmballoon_debugfs_exit(&balloon);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Deallocate all reserved memory, and reset connection with monitor.
|
|
|
|
* Reset connection before deallocating memory to avoid potential for
|
|
|
|
* additional spurious resets from guest touching deallocated pages.
|
|
|
|
*/
|
2015-08-07 06:18:01 +08:00
|
|
|
vmballoon_send_start(&balloon, 0);
|
2010-04-24 01:18:08 +08:00
|
|
|
vmballoon_pop(&balloon);
|
|
|
|
}
|
|
|
|
module_exit(vmballoon_exit);
|