Revision eca6be566d47029f945a5f8e1c94d374e31df2ca authored by Sean Christopherson on 15 February 2019, 20:48:40 UTC, committed by Paolo Bonzini on 15 March 2019, 18:24:33 UTC
The series to add memcg accounting to KVM allocations[1] states:
There are many KVM kernel memory allocations which are tied to the
life of the VM process and should be charged to the VM process's
cgroup.
While it is correct to account KVM kernel allocations to the cgroup of
the process that created the VM, it's technically incorrect to state
that the KVM kernel memory allocations are tied to the life of the VM
process. This is because the VM itself, i.e. struct kvm, is not tied to
the life of the process which created it, rather it is tied to the life
of its associated file descriptor. In other words, kvm_destroy_vm() is
not invoked until fput() decrements its associated file's refcount to
zero. A simple example is to fork() in Qemu and have the child sleep
indefinitely; kvm_destroy_vm() isn't called until Qemu closes its file
descriptor *and* the rogue child is killed.
The allocations are guaranteed to be *accounted* to the process which
created the VM, but only because KVM's per-{VM,vCPU} ioctls reject the
ioctl() with -EIO if kvm->mm != current->mm. I.e. the child can keep
the VM "alive" but can't do anything useful with its reference.
Note that because 'struct kvm' also holds a reference to the mm_struct
of its owner, the above behavior also applies to userspace allocations.
Given that mucking with a VM's file descriptor can lead to subtle and
undesirable behavior, e.g. memcg charges persisting after a VM is shut
down, explicitly document a VM's lifecycle and its impact on the VM's
resources.
Alternatively, KVM could aggressively free resources when the creating
process exits, e.g. via mmu_notifier->release(). However, mmu_notifier
isn't guaranteed to be available, and freeing resources when the creator
exits is likely to be error prone and fragile as KVM would need to
ensure that it only freed resources that are truly out of reach. In
practice, the existing behavior shouldn't be problematic as a properly
configured system will prevent a child process from being moved out of
the appropriate cgroup hierarchy, i.e. prevent hiding the process from
the OOM killer, and will prevent an unprivileged user from being able to
to hold a reference to struct kvm via another method, e.g. debugfs.
[1]https://patchwork.kernel.org/patch/10806707/
Signed-off-by: Sean Christopherson <sean.j.christopherson@intel.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
1 parent c7a0e83
stackdepot.c
/*
* Generic stack depot for storing stack traces.
*
* Some debugging tools need to save stack traces of certain events which can
* be later presented to the user. For example, KASAN needs to safe alloc and
* free stacks for each object, but storing two stack traces per object
* requires too much memory (e.g. SLUB_DEBUG needs 256 bytes per object for
* that).
*
* Instead, stack depot maintains a hashtable of unique stacktraces. Since alloc
* and free stacks repeat a lot, we save about 100x space.
* Stacks are never removed from depot, so we store them contiguously one after
* another in a contiguos memory allocation.
*
* Author: Alexander Potapenko <glider@google.com>
* Copyright (C) 2016 Google, Inc.
*
* Based on code by Dmitry Chernenkov.
*
* This program is free software; you can redistribute it and/or
* modify it under the terms of the GNU General Public License
* version 2 as published by the Free Software Foundation.
*
* This program is distributed in the hope that it will be useful, but
* WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
* General Public License for more details.
*
*/
#include <linux/gfp.h>
#include <linux/jhash.h>
#include <linux/kernel.h>
#include <linux/mm.h>
#include <linux/percpu.h>
#include <linux/printk.h>
#include <linux/slab.h>
#include <linux/stacktrace.h>
#include <linux/stackdepot.h>
#include <linux/string.h>
#include <linux/types.h>
#define DEPOT_STACK_BITS (sizeof(depot_stack_handle_t) * 8)
#define STACK_ALLOC_NULL_PROTECTION_BITS 1
#define STACK_ALLOC_ORDER 2 /* 'Slab' size order for stack depot, 4 pages */
#define STACK_ALLOC_SIZE (1LL << (PAGE_SHIFT + STACK_ALLOC_ORDER))
#define STACK_ALLOC_ALIGN 4
#define STACK_ALLOC_OFFSET_BITS (STACK_ALLOC_ORDER + PAGE_SHIFT - \
STACK_ALLOC_ALIGN)
#define STACK_ALLOC_INDEX_BITS (DEPOT_STACK_BITS - \
STACK_ALLOC_NULL_PROTECTION_BITS - STACK_ALLOC_OFFSET_BITS)
#define STACK_ALLOC_SLABS_CAP 8192
#define STACK_ALLOC_MAX_SLABS \
(((1LL << (STACK_ALLOC_INDEX_BITS)) < STACK_ALLOC_SLABS_CAP) ? \
(1LL << (STACK_ALLOC_INDEX_BITS)) : STACK_ALLOC_SLABS_CAP)
/* The compact structure to store the reference to stacks. */
union handle_parts {
depot_stack_handle_t handle;
struct {
u32 slabindex : STACK_ALLOC_INDEX_BITS;
u32 offset : STACK_ALLOC_OFFSET_BITS;
u32 valid : STACK_ALLOC_NULL_PROTECTION_BITS;
};
};
struct stack_record {
struct stack_record *next; /* Link in the hashtable */
u32 hash; /* Hash in the hastable */
u32 size; /* Number of frames in the stack */
union handle_parts handle;
unsigned long entries[1]; /* Variable-sized array of entries. */
};
static void *stack_slabs[STACK_ALLOC_MAX_SLABS];
static int depot_index;
static int next_slab_inited;
static size_t depot_offset;
static DEFINE_SPINLOCK(depot_lock);
static bool init_stack_slab(void **prealloc)
{
if (!*prealloc)
return false;
/*
* This smp_load_acquire() pairs with smp_store_release() to
* |next_slab_inited| below and in depot_alloc_stack().
*/
if (smp_load_acquire(&next_slab_inited))
return true;
if (stack_slabs[depot_index] == NULL) {
stack_slabs[depot_index] = *prealloc;
} else {
stack_slabs[depot_index + 1] = *prealloc;
/*
* This smp_store_release pairs with smp_load_acquire() from
* |next_slab_inited| above and in depot_save_stack().
*/
smp_store_release(&next_slab_inited, 1);
}
*prealloc = NULL;
return true;
}
/* Allocation of a new stack in raw storage */
static struct stack_record *depot_alloc_stack(unsigned long *entries, int size,
u32 hash, void **prealloc, gfp_t alloc_flags)
{
int required_size = offsetof(struct stack_record, entries) +
sizeof(unsigned long) * size;
struct stack_record *stack;
required_size = ALIGN(required_size, 1 << STACK_ALLOC_ALIGN);
if (unlikely(depot_offset + required_size > STACK_ALLOC_SIZE)) {
if (unlikely(depot_index + 1 >= STACK_ALLOC_MAX_SLABS)) {
WARN_ONCE(1, "Stack depot reached limit capacity");
return NULL;
}
depot_index++;
depot_offset = 0;
/*
* smp_store_release() here pairs with smp_load_acquire() from
* |next_slab_inited| in depot_save_stack() and
* init_stack_slab().
*/
if (depot_index + 1 < STACK_ALLOC_MAX_SLABS)
smp_store_release(&next_slab_inited, 0);
}
init_stack_slab(prealloc);
if (stack_slabs[depot_index] == NULL)
return NULL;
stack = stack_slabs[depot_index] + depot_offset;
stack->hash = hash;
stack->size = size;
stack->handle.slabindex = depot_index;
stack->handle.offset = depot_offset >> STACK_ALLOC_ALIGN;
stack->handle.valid = 1;
memcpy(stack->entries, entries, size * sizeof(unsigned long));
depot_offset += required_size;
return stack;
}
#define STACK_HASH_ORDER 20
#define STACK_HASH_SIZE (1L << STACK_HASH_ORDER)
#define STACK_HASH_MASK (STACK_HASH_SIZE - 1)
#define STACK_HASH_SEED 0x9747b28c
static struct stack_record *stack_table[STACK_HASH_SIZE] = {
[0 ... STACK_HASH_SIZE - 1] = NULL
};
/* Calculate hash for a stack */
static inline u32 hash_stack(unsigned long *entries, unsigned int size)
{
return jhash2((u32 *)entries,
size * sizeof(unsigned long) / sizeof(u32),
STACK_HASH_SEED);
}
/* Use our own, non-instrumented version of memcmp().
*
* We actually don't care about the order, just the equality.
*/
static inline
int stackdepot_memcmp(const unsigned long *u1, const unsigned long *u2,
unsigned int n)
{
for ( ; n-- ; u1++, u2++) {
if (*u1 != *u2)
return 1;
}
return 0;
}
/* Find a stack that is equal to the one stored in entries in the hash */
static inline struct stack_record *find_stack(struct stack_record *bucket,
unsigned long *entries, int size,
u32 hash)
{
struct stack_record *found;
for (found = bucket; found; found = found->next) {
if (found->hash == hash &&
found->size == size &&
!stackdepot_memcmp(entries, found->entries, size))
return found;
}
return NULL;
}
void depot_fetch_stack(depot_stack_handle_t handle, struct stack_trace *trace)
{
union handle_parts parts = { .handle = handle };
void *slab = stack_slabs[parts.slabindex];
size_t offset = parts.offset << STACK_ALLOC_ALIGN;
struct stack_record *stack = slab + offset;
trace->nr_entries = trace->max_entries = stack->size;
trace->entries = stack->entries;
trace->skip = 0;
}
EXPORT_SYMBOL_GPL(depot_fetch_stack);
/**
* depot_save_stack - save stack in a stack depot.
* @trace - the stacktrace to save.
* @alloc_flags - flags for allocating additional memory if required.
*
* Returns the handle of the stack struct stored in depot.
*/
depot_stack_handle_t depot_save_stack(struct stack_trace *trace,
gfp_t alloc_flags)
{
u32 hash;
depot_stack_handle_t retval = 0;
struct stack_record *found = NULL, **bucket;
unsigned long flags;
struct page *page = NULL;
void *prealloc = NULL;
if (unlikely(trace->nr_entries == 0))
goto fast_exit;
hash = hash_stack(trace->entries, trace->nr_entries);
bucket = &stack_table[hash & STACK_HASH_MASK];
/*
* Fast path: look the stack trace up without locking.
* The smp_load_acquire() here pairs with smp_store_release() to
* |bucket| below.
*/
found = find_stack(smp_load_acquire(bucket), trace->entries,
trace->nr_entries, hash);
if (found)
goto exit;
/*
* Check if the current or the next stack slab need to be initialized.
* If so, allocate the memory - we won't be able to do that under the
* lock.
*
* The smp_load_acquire() here pairs with smp_store_release() to
* |next_slab_inited| in depot_alloc_stack() and init_stack_slab().
*/
if (unlikely(!smp_load_acquire(&next_slab_inited))) {
/*
* Zero out zone modifiers, as we don't have specific zone
* requirements. Keep the flags related to allocation in atomic
* contexts and I/O.
*/
alloc_flags &= ~GFP_ZONEMASK;
alloc_flags &= (GFP_ATOMIC | GFP_KERNEL);
alloc_flags |= __GFP_NOWARN;
page = alloc_pages(alloc_flags, STACK_ALLOC_ORDER);
if (page)
prealloc = page_address(page);
}
spin_lock_irqsave(&depot_lock, flags);
found = find_stack(*bucket, trace->entries, trace->nr_entries, hash);
if (!found) {
struct stack_record *new =
depot_alloc_stack(trace->entries, trace->nr_entries,
hash, &prealloc, alloc_flags);
if (new) {
new->next = *bucket;
/*
* This smp_store_release() pairs with
* smp_load_acquire() from |bucket| above.
*/
smp_store_release(bucket, new);
found = new;
}
} else if (prealloc) {
/*
* We didn't need to store this stack trace, but let's keep
* the preallocated memory for the future.
*/
WARN_ON(!init_stack_slab(&prealloc));
}
spin_unlock_irqrestore(&depot_lock, flags);
exit:
if (prealloc) {
/* Nobody used this memory, ok to free it. */
free_pages((unsigned long)prealloc, STACK_ALLOC_ORDER);
}
if (found)
retval = found->handle.handle;
fast_exit:
return retval;
}
EXPORT_SYMBOL_GPL(depot_save_stack);
Computing file changes ...
